Comparative Study of MPPT Techniques for Solar

Photovoltaic System
Shraiya Pant R. P. Saini
Department of Hydro and Renewable Energy Department of Hydro and Renewable Energy
Indian Institute of Technology Roorkee Indian Institute of Technology Roorkee
Roorkee, India Roorkee, India
[email protected] [email protected]

Abstract— Solar photovoltaics (PV) has proven itself to be the The various MPPT techniques differ from each other in
most dependable source for harvesting solar energy. The output several aspects such as: effectiveness, complexity, number of
of solar PV system relies upon solar radiation and temperature sensors required, ease of hardware implementation, speed of
which vary throughout the day. This causes variations in the convergence, etc. [3].
maximum power point (MPP) on the solar PV output The most widely used MPPT method is P&O, on
characteristics curve. Therefore, various Maximum Power account of its ease of implementation, simplicity and
Point Tracking (MPPT) techniques are used to track the MPP reliability. But this method has two major disadvantages:
and extract maximum power out of the PV systems. This paper
acquaints a simulation based comprehensive study for
First, on reaching near the MPP there are unending
comparison of three widely adopted MPPT algorithms: Perturb oscillations in output power causing a reduction in energy
and Observe (P&O), Particle Swarm Optimization (PSO) and yield. Second, this method is unable to cope up with the fast-
Cuckoo Search (CS) using MATLAB Simulink. These MPPT changing irradiance which causes the operating point to shift
algorithms are implemented to regulate the duty cycle of a DC- away from MPP locus causing energy loss [4]. To overcome
DC Boost converter. The performance of the three algorithms is these problems some soft computing based MPPT methods
compared on the basis of their tracking speed, accuracy and are becoming popular. Particle Swarm Optimization (PSO)
tracking efficiency. The CS MPPT algorithm showed the best and Cuckoo Search (CS) are two effective and simple meta-
tracking efficiency among the three under partial shading and heuristic algorithms used for tracking MPP. These MPPT
fast changing conditions of irradiance.
Keywords- Maximum Power Point Tracking, Photovoltaic,
techniques can extract the global maximum power point
Particle Swarm Optimization, Cuckoo Search, Perturb and (GMPP) under partial shading condition (PSC). These
Observe methods do not require the identification of shading pattern
for GMPP location [5].
I. INTRODUCTION
Electricity generation using solar photovoltaic (PV) Rezk et al. [5] studied PSO and CS to compare the
technology has become the most viable option in the recent results with the conventional algorithm of incremental
times. With exhaustion of fossil fuel reserves and rising resistance-based MPP tracker. Performances of the three
concerns about the environmental impact of the use of fossil methods for tracking MPP were compared under partial
fuels in recent times, has led to increase in use of solar PV for shading using MATLAB. PSO and CS showed faster
electricity generation. PV generated electricity is a clean and convergence to MPP than incremental resistance. CS based
noise less source of energy with low maintenance cost. tracker was superior to PSO due to less tracking time
The energy generated by PV modules relies on factors requirement. Mosaad et al. [6] applied CS and compared the
like temperature, solar irradiance and shading condition [1]. results of the MPPT technique with ANN and IC based MPPT
PV modules have non-linear voltage-current (V-I) methods. It was found that CS gave the maximum power
characteristics. The voltage-power (V-P) characteristic curve compared to IC and ANN with varying conditions. Also,
of PV module has a single point where it produces maximum there were no fluctuations in output power on reaching near
power (Pmax). With varying ambient condition MMP also MPP. Koad and Zobaa [7] performed a comparative study
varies. This causes a discrepancy between load and source between the conventional MPPT methods (IC, P&O) and
characteristics, reducing the maximum power delivered to the PSO used in PV systems. These MPPT methods are
load. Maximum power point tracking (MPPT) is used to implemented using Cuk converter in MATLAB to compare
match the characteristics of PV module with load and reduce them on the basis of accuracy,tracking speed, cost and
the power loss [1,2]. The MPPT controller helps in adjusting performance. From simulations it was found that PSO can
the duty cycle of a DC-DC converter, which acts as an track MPP accurately under all conditions. Its tracking
interface between the load and PV modules. efficiency is also high as compared to the other methods. It
In different PV applications to extract the maximum also has fast convergence and easy implementation as
power provided by the PV modules several MPPT techniques compared to IC and P&O. Ahmed and Salam [8] performed
are found in literature. Some commonly used MPPT simulations of CS for tracking MPP in PV system. The results
techniques include Perturb and Observe (P&O), Fractional were correlated against the conventional P&O method. The
Short Circuit Current (FSCC), Incremental Conductance simulation results showed that the CS method tracks the MPP
(IC), etc. Some advanced soft computing based MPPT quickly even under rapidly changing atmospheric conditions
techniques includes Artificial Neural Networks (ANN), with zero steady state oscillations.
Fuzzy logic method and Particle Swarm Optimization (PSO)
methods.

978-1-7281-3455-0/19/$31.00 ©2019 IEEE 1

 represent the number of PV modules connected in series and
parallel respectively, 𝑛𝑠 is the number of PV cells connected
in series in a single string, 𝑅𝑠 and 𝑅𝑝 represents series and
parallel resistances (Ω) of PV module respectively, 𝑎
represents p-n junction ideality factor. The PV module
modelled in this paper is Solarex MSX-60 whose
specifications are compiled in Table 1.

TABLE I. SPECIFICATIONS OF SOLAREX MSX-60

Parameters Abbreviation Values
Fig. 1. PV System with MPPT No. of series connected cells Ns 36
Maximum power Pmp 59.85 W
In the present work, a correlative study of three MPPT Maximum current Imp 3.5 A
techniques viz, PSO, CS and P&O is done at standard test Maximum voltage Vmp 17.1 V
condition (STC) as well as at the changing conditions of Short circuit current Isc 3.8 A
irradiance. The article is organized in five sections. Section I
Open circuit voltage Voc 21.1 V
encloses the introduction and objective. Section II contains
Temperature coefficient of Voc α -0.8 %/oC
the description of the system and its component. Section III
discusses the MPPT techniques used. Interpretation of the Temperature coefficient of Isc β 0.00247 %/oC
simulation results is discussed in Section IV. Conclusion is
of the study is stated in Section V. A series combination of four such PV modules has been
used in this study. The I-V and P-V curve of PV array of four
II. SYSTEM COMPONENTS series connected modules at different solar irradiance (G) and
Schematic of PV system with Maximum Power Point temperature (T) is presented in Fig. 3.
Tracking is illustrated in Fig. 1. For tracking MPP, the
impedance between the load and PV array under varying
atmospheric conditions should be matched by employing a
DC-DC converter. The MPPT controller regulates the duty
cycle of the DC-DC converter by acquiring the real-time
operating parameters required by the governing MPPT
algorithm [9]. The modelling of the components in the above
schematic is discussed the following sub-sections.

A. PV modelling
Equivalent electrical circuit of a PV cell with single
diode is illustrated in Fig. 2. It has a light dependent current
source that produces photocurrent (Iph). A diode is connected
in parallel to the current source which represents the
phenomenon of diffusion current (Id). The contact resistance
between the metal contact and the silicon is represented by
(a)
series resistance (Rs). Loss of power due to manufacturing
defects is represented by the shunt resistance (Rsh)

Fig. 2. Equivalent electrical circuit of the single diode solar cell model
The output current of solar cell is directly proportional
to the solar irradiation falling on it. The series combination of
solar cells forms a PV module whose current is given by (1):
𝑁
𝑉𝑝𝑣+(𝑁 𝑠 )𝑅𝑠 𝐼
𝑝
𝐼 = 𝑁𝑝 𝐼𝑝ℎ − 𝑁𝑝 𝐼𝑠 [𝑒𝑥𝑝 ( ) − 1] −
𝑛𝑠 𝑎𝑣𝑡
𝑁
𝑉𝑝𝑣+(𝑁 𝑠 )𝑅𝑠 𝐼
𝑝
( 𝑁
) (1)
(𝑁 𝑠 )𝑅𝑝 (b)
𝑝

Fig. 3. (a) I-V and P-V characteristics curves at varying G and T=25 oC (b)
where the output current (A) of PV module is designated by
I-V and P-V characteristics curves at varying T and G=1000W/m2
I, 𝑉𝑝𝑣 is output voltage (V) of the PV array, 𝑁𝑠 and 𝑁𝑝

2
B. DC-DC Boost Converter even on reaching the MPP and hence, leading to power losses.
Herein, a boost converter is adopted to transfer maximum It is a major limitation of this method. Also, this method
power from PV module to a load (Fig.4.). The boost converter sometimes is unable to find MPP under frequently changing
helps in stepping up the voltage from supply side to load side. solar irradiance.
The boost converter is turned ON when switch (S) is closed.
ON-state of the converter causes an increase in the inductor
current. On the other hand, opening of switch (S) , shifts the
boost converter into OFF state. In this case, the current flows
through the load (R), diode (D), and capacitor (C).

Fig. 4. Boost converter circuit with PV source and resistive load

TABLE II. VARIOUS PARAMETERS OF DC-DC CONVERTER

Parameters Abbreviation Value
Resistor R 50 Ω
Inductor L 1.15 mH
Fig. 5. Flow diagram showing operation of Perturb and Observe method
Capacitor C1 10 µF
Capacitor C2 470 µF 2) Particle Swarm Optimization (PSO)

The boost converter has input and output voltages, 𝑉𝑖 & It is based upon the behavior of bird flocks. It considers a
𝑉𝑜 respectively related to each other by the following swarm of particles in the n-dimensional space and each of
relation: these particles represent a solution. The particles in the
𝑉𝑜
=
1
(2) multidimensional space keep on adjusting their position to
𝑉𝑖 1−𝑑 find the desired location based on their own previous
where 𝑑 represents duty cycle of the boost converter. experiences and also their neighbors. The best particle in the
The output voltage always exceeds the input voltage as neighborhood ( 𝑃𝑏𝑒𝑠𝑡 𝑖 ) influences the position of each
described by (2) as the duty cycle ranges from 0 to 1. The individual particle. The position of each particle is also
parameters of DC-DC boost converter are summarised in influenced by the best solution found by the whole population
Table 2. of the particles (𝐺𝑏𝑒𝑠𝑡 ) [12]. The particle is brought to the best
C. MPPT Algorithms position by incrementing or decrementing the initial velocity
depending on present value. If the current value of position is
1) Perturb and Observe (P&O) less than best value then there is an increase in the velocity
Its operation involves periodic perturbation (increment or
value and vice versa. The position ( 𝑥𝑖 ) of the particle is
decrement) of terminal voltage or current of the PV array and
adjusted based on [13]:
comparison of the PV output power with the previous
𝑥𝑖𝑘+1 = 𝑥𝑖𝑘 + 𝑣𝑖𝑘+1 (3)
perturbation cycle output power. If operating voltage of the
The velocity (𝑣𝑖 ) in (3) is calculated using [13]:
module changes and power increases, the operating point is
moved to that direction by the control algorithm otherwise, it 𝑣𝑖𝑘+1 = 𝑤𝑣𝑖𝑘 + 𝑐1 𝑟1 {𝑃𝑏𝑒𝑠𝑡 𝑖 − 𝑥𝑖𝑘 } + 𝑐2 𝑟2 {𝐺𝑏𝑒𝑠𝑡 − 𝑥𝑖𝑘 } (4)
moves in the opposite direction. The P&O algorithm follows where 𝑐1 and 𝑐2 are coefficients of acceleration, inertia
similar principle in the next cycle of perturbation [10]. weight is represented by 𝑤 , 𝑃𝑏𝑒𝑠𝑡 𝑖 represents the best
The flow diagram of P&O MPPT technique is shown in personal position of particle 𝑖 , 𝐺𝑏𝑒𝑠𝑡 represents the
Fig. 5. This method has been previously used in [11]. A small neighborhood position of the particle 𝑖 and 𝑟1 , 𝑟2  𝑈 . The
increment in system voltage (V) that causes a change in flow diagram of PSO is shown in Fig. 6. This method has
power (P). A positive change in power causes the operating previously been discussed in [13].
voltage of the system to perturb in the same direction as the 3) Cuckoo Search (CS)
increment. On contrary, if P is negative then the operating It is based on the brood parasitism method (laying eggs in
voltage moves in the direction opposite to the increment. In other birds’ nest) followed by Cuckoo birds for laying eggs.
this method the terminal voltage of PV array is perturbed in Three elementary rules govern CS [14]: (a) each individual
every MPPT cycle that causes the output power to oscillate cuckoo will lay a single egg at a time and place it in any nest

3
which is selected randomly. (b) The finest nest possessing The CS algorithm follows the Lévy distribution, which
superior eggs will be passed on further to the succeeding allows it to obtain local maxima points and, at the same time,
generation. (c) The statistic of convenient nests is pre- shortens the tracking time required to reach global maximum
established and the probability of discovery of the egg laid by power point [15]. The flow diagram of CS [16] is presented
a cuckoo in such pre-determined nests by the host bird is Pa, in Fig. 7.
where Pa∈ [0, 1]. Here, the cuckoo birds represent the
particles assigned to find the solution and the cuckoo bird III. RESULTS AND DISCUSSION
eggs represent the solution of present iteration process.
Comparison of the three MPPT (i.e. P&O, PSO and CS)
methods has been carried out under the following three cases:
(i) Standard Test Condition (ii) Partial shading conditions and
(iii) Fast varying solar radiation. The MATLAB Simulink
model for the system designed is shown in Fig. 8. The MPPT
algorithms are compared on the basis of their tracking speed
and tracking efficiency (Teff); which is calculated using [17]:
𝑡
∫0 𝑃𝑀𝑃𝑃
𝑇𝑒𝑓𝑓 = 𝑡 . 100 (5)
∫0 𝑃𝑃𝑉

A. Under Standard Test Condition (STC)

The array output power at STC is found to be 238.2 W.
At this condition the radiation is 1000 W/m2 and ambient
temperature is 25 oC. The power output comparison of the
three MPPT methods at STC is illustrated in Fig. 9 (a). Here,
it is observed that P&O, PSO and CS MPPT algorithms take
approximately 0.119 s, 1.457 s and 0.646 s, respectively to
track the MPP. The tracking efficiencies calculated for P&O,
PSO and CS are 99.8%, 99.9% and 99.8%, respectively. In
this case the time response of P&O is better than the other
two methods. PSO method takes the maximum time for
tracking MPP with comparable tracking efficiency with the
Fig. 6. Flow diagram of Particle Swarm Optimization method
other two methods.
B. Under Partial Shading Condition (PSC)
In this case the radiation on the four series connected PV
modules are different; 1000, 800, 400 and 600 W/m2
respectively while the temperature was set as 25 oC. In this
case the global maximum power point was found to be 116.2
W. The power output comparison of the three MPPT methods
under partial shading condition is shown in Fig. 9 (b). It was
observed that the time required by the PSO and CS algorithms
to track GMPP (i.e. 116.2 W) is 0.943 s and 0.472 s,
respectively. Contrarily, the P&O algorithm gets trapped at a
local maximum point at around 95.56 W. The tracking
efficiencies of both PSO and CS MPPT algorithms under
partial shading condition is 99.65% while the P&O method
has a low tracking efficiency of 82.23%. The accuracy of
PSO and CS in this case is much higher than the P&O
algorithm.
C. Under Fast Varying Solar Radiation (FVSR)
The fast-changing radiation (Fig.10 (a)) in the form of a
step signal is provided to the PV array. The response of the
three MPPT algorithm in terms of power output to this fast-
changing radiation is presented in Fig. 10 (b). It is found that
from 0-1s of simulation, the P&O algorithm tracks the MPP
fastest (0.077s) with a tracking efficiency of 99.8%. Here, the
PSO and CS track the MPP in 0.55 s and 0.65s with tracking
Fig. 7. Flow diagram showing Cuckoo Search method efficiencies of 99.8%.

4
 Fig. 8. Model of PV array with boost converter in Simulink

(a) (b)
Fig. 9. Comparison of power output from PSO, CS and P&O (a) at STC (b) under PSC

(a) (b)
Fig. 10. (a) Fast changing solar irradiance input, (b) Power output comparison of PSO, CS and P&O under FVSR

and 99.9%. With abrupt change in solar irradiance (at 1s), the and 99.774 %, respectively. In this case as well MPP tracking
P&O tracks the new MPP in 0.05 s with a tracking efficiency by CS has better accuracy and fast tracking.
of 98.42 %. The time taken by PSO and CS for tracking the
MPP is 0.11 s and 0.13 s with tracking efficiency of 99.2 %

5
 IV. CONCLUSION PV system Based Cuckoo Search Algorithm; review
In this study a MATLAB Simulink model is developed and comparison. Energy Procedia. 2019 Apr
to compare three MPPT methods (P&O, PSO and CS). These 1;162:117-26.
MPPT methods are compared in terms of tracking efficiency, [7] Koad RB, Zobaa AF. Comparison between the
tracking time and accuracy. The methods were compared conventional methods and PSO based MPPT
under three conditions: STC, partial shading condition and algorithm for photovoltaic systems.
fast changing irradiance. Under STC, the MPP has been [8] Ahmed J, Salam Z. A soft computing MPPT for PV
tracked in 0.119 s with 99.8% tracking efficiency by P&O system based on Cuckoo Search algorithm. In4th
method. Under partial shading condition, MPP has been International Conference on Power Engineering,
tracked with 99.65% efficiency by PSO and CS methods and Energy and Electrical Drives 2013 May 13 (pp. 558-
hence these methods are found to be better than P&O. Also, 562). IEEE.
CS method takes lesser time (i.e. 0.472 s) compared to PSO [9] Wang Y, Yang Y, Fang G, Zhang B, Wen H, Tang
(0.55 s) to track the MPP in this case. In case of fast changing H, Fu L, Chen X. An advanced maximum power
radiation, MPP has been tracked in 0.13 s with 99.74% point tracking method for photovoltaic systems by
tracking efficiency by CS method. As per the simulation using variable universe fuzzy logic control
results, it was found that MPP is tracked with highest tracking considering temperature variability. Electronics.
efficiency by CS method under the three different conditions. 2018;7(12):355.
[10] Faranda R, Leva S. Energy comparison of MPPT
ACKNOWLEDGEMENT techniques for PV Systems. WSEAS transactions on
The authors are immensely grateful to the Ministry of power systems. 2008 Jun 1;3(6):446-55.
Human Resources and Development (MHRD) and [11] Go, S. I., Ahn, S. J., Choi, J. H., Jung, W. W., Yun,
Department of Hydro and Renewable Energy, Indian Institute S. Y., & Song, I. K. Simulation and analysis of
of Technology (IIT), Roorkee for providing support during existing MPPT control methods in a PV generation
PhD. system. Journal of International Council on
Electrical Engineering. 2011;1(4), 446-451.
REFERENCES [12] Ishaque K, Salam Z, Amjad M, Mekhilef S. An
improved particle swarm optimization (PSO)–based
[1] Verma D, Nema S, Shandilya AM, Dash SK.
MPPT for PV with reduced steady-state oscillation.
Maximum power point tracking (MPPT) techniques:
IEEE transactions on Power Electronics. 2012 Jan
Recapitulation in solar photovoltaic systems.
23;27(8):3627-38.
Renewable and Sustainable Energy Reviews. 2016
Feb 1;54:1018-34. [13] Ishaque K, Salam Z. A review of maximum power
[2] Salam Z, Ahmed J, Merugu BS. The application of point tracking techniques of PV system for uniform
insolation and partial shading condition. Renewable
soft computing methods for MPPT of PV system: A
and Sustainable Energy Reviews. 2013 Mar
technological and status review. Applied Energy.
1;19:475-88.
2013 Jul 1;107:135-48.
[14] Yang XS, Deb S. Multiobjective cuckoo search for
[3] Mohanty P, Bhuvaneswari G, Balasubramanian R,
Dhaliwal NK. MATLAB based modeling to study design optimization. Computers & Operations
the performance of different MPPT techniques used Research. 2013 Jun 1;40(6):1616-24.
[15] Nugraha DA, Lian KL. A Novel MPPT Method
for solar PV system under various operating
Based on Cuckoo Search Algorithm and Golden
conditions. Renewable and Sustainable Energy
Section Search Algorithm for Partially Shaded PV
Reviews. 2014 Oct 1;38:581-93.
System. Canadian Journal of Electrical and
[4] Ahmed J, Salam Z. An improved perturb and
observe (P&O) maximum power point tracking Computer Engineering. 2019 Jul 23;42(3):173-82.
[16] Mohapatra A, Nayak B, Das P, Mohanty KB. A
(MPPT) algorithm for higher efficiency. Applied
review on MPPT techniques of PV system under
Energy. 2015 Jul 15;150:97-108.
partial shading condition. Renewable and
[5] Rezk H, Fathy A, Abdelaziz AY. A comparison of
Sustainable Energy Reviews. 2017 Dec 1;80:854-
different global MPPT techniques based on meta-
heuristic algorithms for photovoltaic system 67.
subjected to partial shading conditions. Renewable [17] Belkaid A, Colak I, Isik O. Photovoltaic maximum
power point tracking under fast varying of solar
and Sustainable Energy Reviews. 2017 Jul
radiation. Applied energy. 2016 Oct 1;179:523-30.
1;74:377-86.
[6] Mosaad MI, abed el-Raouf MO, Al-Ahmar MA,
Banakher FA. Maximum Power Point Tracking of

6
 Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Maximum power point tracking (MPPT) techniques: Recapitulation

in solar photovoltaic systems
Deepak Verma n, Savita Nema, A.M. Shandilya, Soubhagya K. Dash
Department of Electrical Engineering, Maulana Azad National Institute of Technology, Bhopal 462003, Madhya Pradesh, India

art ic l e i nf o a b s t r a c t

Article history: Unfilled gap of prolonged energy demand by conventional energy sources and consent of global warming
Received 5 September 2014 as its vulnerable outcome provides a vent to search compatible option. Augmentation in use of solar
Received in revised form energy reveled through last 3 decades portrays its heterogeneous rewards in the prevailing energy
13 February 2015
scenario. Nevertheless solar PV system arises as viable option in the critical power system era its low
Accepted 23 October 2015
efficiency energy conversion attribute necessitates an efficient power conversion system. The non-
linearity of I–V (current–voltage) characteristic and its alteration for an assorted insolation and tem-
Keywords: perature values may enable the alteration in terminal voltage. This may deviates maximum power point
Maximum power point tracking (MPPT) due to which the available maximum power delivery to load can be differed. Literature of this field
Solar PV systems
reiterated that the uniform insolation and partial shading condition demands undeniable need of
Partial shading
maximum power point tracking. Nonetheless through investigation in this direction furnishes the
Uniform insolation
availability of a bunch of such techniques; each of them posses its own pros and cones. This ubiquitous
trait of available maximum power point tracking (MPPT) techniques unfolds the complexity in its precise
selection. To diminish such complexity this paper offers a state of art of various MPPT technique and their
comprehensive comparative analysis based on 110 standard research articles. The focus of this paper is to
offer a better commencement and to furnish valued information for investigators of this field.
& 2015 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1019
2. Need of maximum power point tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1020
2.1. Impact of insolation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1020
2.2. Impact of temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1020
2.3. Impact of shading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1020
2.4. Operating strategy of MPPT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1021
3. Maximum power point tracking algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
3.1. Constant voltage method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
3.2. Pilot cell method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
3.3. Constant current method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
3.4. Curve fitting method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
3.5. Look up table method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3.6. Perturb and observe (P&O) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3.6.1. Hill climbing method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3.6.2. Beta method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3.6.3. Variable step size incremental resistance (INR) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024
3.6.4. Estimated perturb–perturb (EPP) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.6.5. Three point weight comparison method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.7. Incremental conductance (INC) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025

n
Corresponding author. Tel.: þ 91 9893853816.
E-mail addresses: [email protected] (D. Verma), [email protected] (S. Nema), [email protected] (A.M. Shandilya),
[email protected] (S.K. Dash).

http://dx.doi.org/10.1016/j.rser.2015.10.068
1364-0321/& 2015 Elsevier Ltd. All rights reserved.
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1019

3.8. DC link capacitor droop control or Parasitic capacitance method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025

3.9. dP/dV or dP/dI feedback control method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.10. Load current or load voltage maximization method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.11. Current sweep method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.12. One cycle control (OCC) method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.13. Slide control method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
3.14. System oscillation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.15. Ripple correlation control (RCC) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.16. State space based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.17. Linear current control (LCC)/linearization based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.18. PV output senseless (POS) method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.19. Gradient descent or steepest descent method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.20. Analytic solution based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027
3.21. Variable inductance method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1028
3.22. Temperature based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1028
3.23. Bisect search theorem (BST) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1028
3.24. Fuzzy logic control method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1028
3.25. Artificial neural network (ANN) based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
3.26. Method based on particle swarm optimization (PSO)/ant colony optimization (ACO). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4. Comparison and analysis of MPPT methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.1. Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.1.1. Indirect control methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.1.2. Direct control methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.1.3. Soft computing technique based methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.2. Dependency of PV array (DPVA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4.3. Implementation methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.4. Sensor required . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.5. Stages of energy conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.6. Partial shading enabled (PSE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.7. Grid interaction (GI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.8. Analog or digital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.9. Tracking efficiency (TE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.10. Tracking speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.11. Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.12. Size of the PV array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.13. Product available in market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.14. Comparative analysis of the MPPT methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1032
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1032

1. Introduction energy projects the photovoltaic cell (Solar PV) is on the leading
edge as the promising future energy technology option [8]. The
Electricity demand is growing with highest rate of all the direct conversion of solar radiation to electrical energy by PV cells
energy consumed worldwide. Thus the mankind is facing a mas- has a number of significant advantages. However its proficient
sive challenge of never ending increase in energy demand as a extraction demands accomplishment of some significant chal-
result of overall socio economic growth [1–2]. The declining fossil lenges such as energy fluctuation, huge investment low energy
fuel resources and tremendous rate of its consumption to battle conversion efficiency of module, and energy cost [9–11]. Reducing
the prevailing industrial revolution diverges us on the peak of energy cost of PV system is a big issue since maintenance
consumption of fossil fuel. Incompatibility of conventional sources
requirement is very low and the only real cost savings to be made
to fulfill this bottomless valley of energy requirements, energy
is in efficiency enhancement. Recent literature reveals that
security, and especially the sky-rocketing hike of fossil fuels prices
research efforts target to enhance the power output of the module
gives a work force direction to invent compatible option [3–5].
in terms of MPPT. A photovoltaic (PV) module has non-linear I–V
Despite of these unprovoked concerns, the global warming as an
(current–voltage) characteristics and its P–V (power–voltage)
unavoidable outcome of carbon emissions by the conventional
characteristics shows that there exist only one point (Pmax) where
energy sources proves to be a momentous driver for renewable
energy sources deployment. Ubiquitous accessibility of renewable the module deliver maximum power as shown in Fig. 1. This point
energies like solar and wind offers a striking solution to comply all also varies with the change in insolation and temperature as
these requirements. Continuous efforts of researchers have shown shown in Figs. 2 and 3. Thus the mismatch between load and
an increased efficiency in both the conversion and transport of source characteristics curtails the availability of maximum possi-
these energy sources. Thus they arises as an attractive alternative ble accessible power delivery to load which causes a significant
option to conventional solution [6–7]. It is a trend which is almost power loss. An electronic circuitry used to match the character-
certain to evolve in upcoming power generation. All-pervading istics of module with respect to maximum power to avoid this
and copious availability of solar energy has an outstanding power loss is called maximum power point tracker (MPPT).
potential to make a significant contribution to the world’s energy Approximately 40 various methods are reported in literature
needs. Two ways to extract the solar energy are solar thermal to track the maximum power point, some of them are very much
plants and solar cells i.e. photovoltaic cells. In prevailing renewable close to other method as to their operating principle. This paper
1020 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

10
Isc Cuurent curve
Power curve

Current (A) / Power (W)

6
Pmax
Im
4

0
0 0.1 0.2 0.3 0.4 0.5 Vm 0.6 Voc 0.7
Voltage (Volt)

Fig. 1. Solar cell I–V and P–V characteristic.

10

8 T = 298 K
Current (A) / Power (W)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Voltage (Volt)

Fig. 2. Solar I–V and P–V characteristic with different insolation and for constant temperature.

discusses 31 commonly used methods and compares them on the 2.2. Impact of temperature
basis of 12 parameters.
Temperature is another factor which shows a momentous influ-
ence on the characteristic of solar cell. With increase in temperature
2. Need of maximum power point tracking open circuit voltage increases proportionally but the short circuit
current decreases logarithmically. Fig. 3 reflects this peculiarity of
The concept of MPPT is explain by considering an example of solar cell characteristic for different temperature values.
monocrystalline solar cell Q6LMXP3-G3 made by Q-CELLS. The
simulations are conducted with the cell parameters obtained
from datasheet [12]. Fig. 1 depicts the I–V characteristic and 2.3. Impact of shading
power versus voltage curve of a single solar cell. It indicates that
the solar PV can give maximum power only at a single point. For Along with temperature and insolation partial shading also
extracting the maximum power from the cell the operating vol- shows a major influence on solar cell characteristic. When more
tage or current should be corresponding to the maximum power than one PV module are connected and one of them is partially
point (Pmax) i.e. Vm and Im respectively under a given temperature shaded then the insolation received by both the module are dif-
and insolation [11]. ferent. This mismatch condition is explained by considering an
The characteristics of solar cell are significantly influenced by example of two series connected Q6LMXP3-G3 solar cell with their
terminal voltage V1 and V2, total power P and total voltage V as
 Insolation. shown in Fig. 4. It shows the shading condition with bypass diode
 Temperature.
mitigation strategy. Figs. 5 and 6 show the characteristic of solar
 Partial shading condition [13–17].
PV for this shading condition.
As explained above it is seen that all these time variant and
The impacts of such environmental factors are explained as
environmental dependent factors shows a major contribution in
follows.
the changes of the operating point or maximum power point
2.1. Impact of insolation (MPP) throughout the day. The function of maximum power
point tracking is to shift this changing operating point to the
The changes in characteristic with variation in insolation are point (Pmax) where the module delivers maximum power. The
shown in Fig. 2. The short circuit current (Isc) of the solar cell is the phenomenon of tracking the maximum power point is same as
function of insolation and reduces proportionally with decrease in impedance matching by tape changing transformer in case of AC
insolation. Fig. 2 reflects that the point of maximum power also and in DC a DC–DC converter is used to convert output voltage of
varies with respect to changing insolation. solar cell with respect to Pmax by changing duty ratio.
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1021

10
273 K
303 K
8 333 K

Current (A) / Power (W)

Insolation = 1000 w/m2
6

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Voltage (Volt)

Fig. 3. Solar cell characteristic with different working temperature and constant insolation.

10
00
W
/m
2

Solar Cell 1
500

Bypass
W/m

Diode
2

Solar Cell 2
Bypass
Diode

Fig. 4. Operation of solar PV under partial shading condition.

10
Power Curve
Current Curve
8
Current (A) / Power (W)

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Voltage (Volt)
Fig. 5. I–V and P–V Curve under partial shading condition.

2.4. Operating strategy of MPPT corresponding terminal voltages of the cell are A, B and C and it is
clear from the Fig. 7 and Fig. 2 the power delivered by the solar cell
After analyzing impact of environmental issues such as inso-
with respect to point A, B and C is less than the available power.
lation and temperature as well as shading condition on solar cell
The maximum power point tracking operating strategy evokes
characteristic, the operating strategy of MPPT is explained by
the concept of holding the terminal voltage corresponding to the
considering an example for tracking maximum power with change
in insolation as shown in Fig. 7. It gives the solar cell characte- maximum power point i.e. A’, B’ and C’ instead of operating point
ristic for three different insolation for a linear resistive type load, A, B and C. Thus an electronic circuitry used to drag the operating
leading to different maximum power points which are A’, B’ and C’ point of solar cell to the maximum power point is known as
as shown in Fig. 2. As the load is linear, the operating points and maximum power point tracker.
1022 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

P vs. V1, V2

5
4
3

P
2
1
0
0.6
0.5
0.4 0.6
0.3 0.5
0.3 0.4
V2 0.2 0.2
0.1 0.1 V1
0 0

Fig. 6. Three dimensional curve PV power versus voltages.

10
A'
T = 298 K
8 A
Current (A) / Power (W)

B
B'
6

4 C
C'
2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Voltage (Volt)

Fig. 7. Maximum power point for different insolation.

3. Maximum power point tracking algorithms can be avoided. VREF or voltage corresponding to the MPP can be
calculated by (2):
The noteworthy methods reported in literature are:
V REF ¼ K 3  V oc‘pilot cell ð2Þ

3.1. Constant voltage method where K3 Econstant o1.

Constant voltage method is based on the observation that the 3.3. Constant current method
maximum power point occurs between 72% and 78% of the open
circuit voltage Voc, for the standard atmospheric condition. The Constant current method is based on the same phenomenon of
solar PV module always operates at the constant voltage in this the constant voltage method. In the constant voltage method the
range [18,19]. The duty ratio (δ) of the DC to DC convertor ensures PV array operates at the constant voltage and in this method PV
that the PV voltage is equal to: array operates at the constant current. The maximum power point
V REF ¼ K 1  V oc ð1Þ arrives between 78% and 92% of the short circuit current Isc [21]
thus the sensed parameter is short circuit current.
where K1 ¼0.72–0.78
Fig. 8 shows that after VOC is sampled by a sampler, VREF which IREF ¼ K2  Isc ð3Þ
is calculated by (1) is kept constant during one sampling period by where K2 ¼0.78–0.92
hold circuit, now duty ratio δ is adjusted to make VPV ¼VREF . For
next sample again VOC is sampled and the same procedure is
3.4. Curve fitting method
repeated for each samples. Fig. 9 depicts the flow chart of this
method.
The characteristic of solar PV is non linear, which can be
The method is simple, fast and easy to implement but shows
modeled mathematically by using a curve fitting method. The
limited accuracy, Voc is required to be measured at regular interval
nonlinear characteristic of a solar array can be approximated as:
and used only where lethargic temperature variation is observed.
Based on the same operating principle one more method is P PV ¼ K 4 V 3PV þ K 5 V 2PV þ K 6 V PV þ K 7 ð4Þ
available in the literature named as pilot cell method which is
At maximum power point, dPPV/dVPV ¼ 0
described as below;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 K 5 K 25 3K 4 K 6
3.2. Pilot cell method VM ¼ ð5Þ
3K 4
In this method a pilot cell is used to calculate the open circuit where K 4 ; K 5 ; K 6 ; K 7 ; are constant and can be obtained by
voltage instead of the whole PV array [20]. After simple calculation polyfit command in MATLAB.
array open circuit voltage i.e. Varray can be directly evaluated, Also it has been shown that PPV is a function of array voltage
which reduces the efforts of measuring Voc at regular interval i.e. and temperature. Ref. [22] gives a modified curve fitting method
the problem of disconnection of PV from the load at every sample that predicts the P–V characteristic curve of a PV array with a
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1023

Vsmp VREF VREF

Sampler K1 Hold +
+
Circuit

Voc
- -
Saw tooth
Vpv

Fig. 8. Maximum power point tracker using the constant voltage method.

Table 2
Methodology of hill climbing method.
Start
Perturbation in terminal Change in Next perturbation
voltage power

Positive Positive Positive (increment in duty

ratio ‘δ’)
Positive Negative Negative (decrease in duty
Sense Vpv ratio ‘δ’)
Negative Positive Negative (decrease in duty
ratio ‘δ’)
Yes
Negative Negative Positive (increment in duty
ratio ‘δ’)
Vpv = VREF

No
3.6. Perturb and observe (P&O) method

The perturb and observe (P&O) algorithm is the most com-

No Vpv > VREF Yes
monly used in practice because of its ease of implementation. The
method is basically iterative approach, in which operating point of
solar PV oscillates around the maximum power point.
In Fig. 1 the power versus voltage curve of solar PV shows that,
δ = δ + Δδ δ = δ - Δδ change in power with respect to voltage (dP/dV) is positive,
negative and zero for region before maximum power point, after
maximum power point and at maximum power point, respectively
[24].
Fig. 9. Flow chart of the constant voltage method. This method is applied by perturbing the operating voltage at
regular interval and oscillating around the point dP/dV ¼0 i.e. MPP.
The operation explained in Table 1.
Table 1
The method is easy to implement, shows moderate accuracy,
Methodology of P&O method. operating point oscillate around MPP, the method is slow and not
suitable for fast changing condition, Oscillation can be minimized
Perturbation Change in power Next perturbation by reducing perturbation step size which slow down the MPPT,
Positive Positive Positive
measurement of both voltage and current is required.
Positive Negative Negative Hill climbing, beta method, variable step size INR method,
Negative Positive Negative estimated perturb–perturb and three point weight comparison
Negative Negative Positive method uses the same fundamental principle of P&O with slight
change in operating principle [25–29] and can be sub-classified
fourth-order polynomial for varying temperatures. as below;

P PV ðV PV ; T array Þ ¼ K 8 T array V 4PV þK 9 T array V 3PV þ K 10 T array V 2PV 3.6.1. Hill climbing method
Hill climbing and P&O method are two different methods with
þK 11 T array V PV þ K 12 T array ð6Þ same fundamental principle. P&O involves perturbation in term-
inal voltage to perform MPPT whereas the hill climbing method
involves perturbation in duty ratio (δ) [30–32]. The methodology is
3.5. Look up table method explained in the Table 2 and flow chart given in Fig. 10.

In this method, the measured values of array's voltage and 3.6.2. Beta method
current are compared with previously stored values which har- In this method a coefficient beta (β) is used, which is given as;
monize the operating point of array with respect to the maximum  
Ipv q
power point. The stored database contains different system con- β ¼ ln   Vpv ð7Þ
Vpv kT η
dition for any insolation and temperature, and corresponding
maximum power point for specific solar PV array [23]. where ‘k’ is Boltzmann's constant, ‘η’ is diode quality factor, ‘T’ is
The major disadvantage of this method is the requirement of ambient temperature in Kelvin and ‘q’ is electric charge.
bulk storage memory. Higher accuracy in tracking increases the Eq. (7) indicates that value of β is independent from the inso-
number of operating conditions which requires more storage data. lation but depends on the temperature. In this method the solar
The tracking scheme is specific for array thus the implementation PV operates near to this value β rather than the MPP. The method
is complex, also considering all possible system conditions are fetches the operating point close to the value of beta in few
bothersome to store and archive. iterations thereafter P&O methods with finer steps can be used to
1024 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

Fig. 10. Flowchart of P&O method.

variation in maximum power point is small for fixed temperature

even as the insolation is varied over a wide range. Despite of that
there is an inverse relationship between the value of β and tem-
Start
perature. A range i.e. βmax to βmin is given by [33] for specific PV
system where the algorithm gives appropriate solution. However
Beta method can also be used along with another method such as
incremental conductance as reported in literature.

Sense Vpv, Ipv 3.6.3. Variable step size incremental resistance (INR) method
In case of fixed step size P&O method, if step size is large
method become faster but oscillations occur around MPP is higher
which reduces the system efficiency and similarly smaller step
Calculate β=[ln(Ipv/Vpv)] - (q/(k×T×η))×Vpv size, increases the system efficiency but slows down the tracking
speed. The INR method gives the solution to the problem as:

 dP 
δðkÞ ¼ δðk 1Þ 7N   \2pt\  tj ð8Þ
dV
where N is scaling factor which governs step size in case of the
P&O method (8) becomes
Switch over to Hill Climbing or other methods 
and implement the duty cycle dP 
δðkÞ ¼ δðk 1Þ 7N   \2pt\  tj ð9Þ
dδ
For obtaining scaling factor [34] introduced a simple method
N o δmax =jdV
dP
j
Fig. 11. Flowchart of beta method. where δmax is largest step size, the INR method gives a simple
and effective variable step size angle:
    
track the exact MPP [33]. Fig. 11 shows the flow chart of the dP 
 \2pt\  t  ¼  tan θ; 900 o θ o900 ð10Þ
method.  dI
Since temperature and insolation shows vague dependency, the
value of β is calculated for different insolation and temperature for Sk ¼ ðΔI ref Þmax  sin θk o ðΔI ref Þmax ð11Þ
a particular PV system. β is evaluated with maximum and mini- Around MPP sin θk becomes lesser thus step size Sk becomes
mum value of temperature and insolation and it is observed that smaller.
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1025

Table 3 Fig. 13 shows the schematic diagram of the method. It shows

Methodology of the INC method. that the output of the boost converter is fed to the inverter across
terminals ‘a’ and ‘b’. Now the operating power of the array can be
Before MPP After MPP At MPP
increased by increasing the amplitude of the line current of ac
dP
40 dI
or dV þ VI 4 0 dP
o0 dI
or dV þ VI o 0 dP
¼ 0 or dV
dI
þ VI ¼ 0 system. As long as the output power of the PV is less than the Pmax,
dV dV dV
the power increases with the increase in line current and V*link
remains constant. After reaching the maximum power the V*link
3.6.4. Estimated perturb–perturb (EPP) method starts decreasing thus the equilibrium condition is disturbed
This method is an advanced version of P&O method [35]. EPP which can be controlled by the change in duty ratio of the boost
method uses two operating modes, mode 1 for estimate process converter. Thus in steady state at the constant V*link the maximum
and mode 2 for perturbation. The name “estimated-perturb–per- power point can be achieved [39,40].
turb” gives all information about the principle of this method.
After two perturbations (mode 2 in which determination of next 3.9. dP/dV or dP/dI feedback control method
PV voltage is done) there is one estimation mode in which con-
troller stops tracking MPP by keeping PV voltage constant and Fig. 1 shows that the power kept on increasing till the MPP and
measures only the power variation or voltage variation due to then decreases thereafter that with respect to voltage or current.
environmental changes for the next control period. In the context of the above this method compares two consecutive
An additional estimated mode is there to improve the perfor- powers, but as compared to conventional methods, magnitude of
mance of MPPT significantly, with fast changing insolation. the slope is also considered to determine MPP [41]. Three condi-
tions are there:
3.6.5. Three point weight comparison method Before MPP: p2 4 p1
The principle of this method is same as P&O method but in case At MPP: p2 ¼p1
of P&O, method compares only two operating point and corre- After MPP: p2 op1
sponding power while in this method comparison is done reg- The following formula is used to determine the error in track-
ularly by perturbing the solar PV terminal voltage at three points: ing and duty ratio is adjusted accordingly.
A, B and C. Where A is the current operating point, B is next Z  
dP
operating point after perturbation at A and C is doubly perturbed A ¼ Kc Kp dt ð14Þ
dI
opposite to point B [36].
This method is fast in comparison with P&O also suitable for
fast changing condition. 3.10. Load current or load voltage maximization method

3.7. Incremental conductance (INC) method The method extracts the load parameter i.e. load voltage or
load current to control the MPP instead of input current or voltage,
This method is based on the fact that slop of the PV array Fig. 14 shows the schematic diagram of the method.
power curve is zero at the MPP (Pmax) as shown in Fig. 1, This can MPPT power stage or a matching network has an internal
be expressed as follows: controllable parameter proportional to Vout which controls the
power flow in the network. This matching network may be loss
Power : P ¼ V  I free resistor or a transformer.
dP
dV
¼ I þ V dV
dI The operating principle of this method is based on the single
output parameter extraction, either voltage or current. By
At true MPPT increasing output voltage or current, power output increases until
dP the point of maximum power, thereafter power decreases with
¼0 further increase in voltage or current. In this way the operating
dV
point will converge to the MPP [42,43].
dI
I þV ¼0
dV 3.11. Current sweep method

dI I
¼ ð12Þ To obtain the I–V characteristic of the PV array, this method
dV V
uses sweep waveform for the PV array current which is updated at
where dI/dV: Incremental conductance, a regular time interval. Also for each interval VMPP can then be
I/V: Instantaneous conductance. calculated [44–46].
Eq. (12) indicates that MPP can be found by comparing
instantaneous conductance to the incremental conductance. The 3.12. One cycle control (OCC) method
operation of this method can be divided in three zones as shown
in Table 3. The flow chart of the method is given in Fig. 12 [37]. One cycle control is the nonlinear control technique which is
The method is complex and computationally more demanding based on the integration of a variable (voltage or current), to
as compared to P&O [37,38]. convert the variable value equal to some reference value.
OCC based method is applied to grid connected solar PV system
3.8. DC link capacitor droop control or Parasitic capacitance method as shown in Fig. 15 in which single stage inverter perform the
operation of MPPT [47–49].
This method is based on the power equilibrium at dc link, the
output of the boost converter is kept constant by changing the 3.13. Slide control method
duty ratio ‘δ’ which is given by;
Slide mode control technique is used for nonlinear system, as
V cell
δ ¼ 1 ð13Þ for MPPT the application of this control technique uses two modes
V link
of operation, one is approaching mode and another one is the
1026 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

Start

ΔV = 0

No Yes

ΔI/ΔV = -I/V ? ΔI = 0 ?

No No

ΔI/ΔV > -I/V ? ΔI > 0 ?

Yes
Yes No
Yes Yes Np

Increase Decrease Increase Decrease

Operating Voltage Operating Voltage Operating Voltage Operating Voltage

Return

Fig. 12. Flowchart of the INC method.

Fig. 13. DC link capacitor droop control.

sliding mode, at MPP change in power with respect to voltage or

Iin Iout current is equal to zero.
+ +
MPPT P¼V I
Vin Vout LOAD
- Power Stage -
dP
¼0
dI
Vout Iout
 
δ dh 2 i d
Photovoltaic array
I R ¼ 2I  R þ I 2 R ¼ 0
Control dI dI

This method gives the solution at MPP:

Fig. 14. Schematic diagram of load current or load voltage maximization
d
method. σ ¼ 0 ¼ 2R þ I R ð15Þ
dI
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1027

IL
S1 S3
L
Is + LOAD
+ DC-AC
V1 Vdc Converter
- -
Vs
S4 S2

Photovoltaic
Photovoltaic δ Current
array Control
array Feedback
Fig. 15. Single phase grid connected inverter with solar PV system.
Fig. 16. Control diagram of POS method.

where σ is defined as sliding surface [50–52]. Now duty ratio δ

is updated as given in (16) to achieve the MPP. 3.17. Linear current control (LCC)/linearization based method
(
δ þ Δδ for σ 4 0 The method graphically identifies the intersecting point of two
δupdate ¼ ð16Þ
δ þ Δδ for σ o 0 curves, power curve of PV represented by f ðP; I Þ ¼ 0 and maximum
output power curve represented by dP dI
¼ 0. The maximum output
power curve can be represented by a linear line. To identify the
3.14. System oscillation method
intersecting point of above mentioned two curves, a simple analog
circuit is used. The phenomenon is same as to find the operating
This method is based on the principle of maximum power
point of a transistor amplifier for a specific load and the operating
transfer. It compares the ac component (oscillation due to the
point is intersecting point of transistor I–V characteristic and load
variation of the duty ratio) to the average value of the input vol-
I–V characteristic [56].
tage at the power conversion stage to determine the duty ratio. At
MPP the ratio of oscillation amplitude and average voltage is
constant [53]. In this method only voltage sensor is required and 3.18. PV output senseless (POS) method
methodology is easy to implement.
This method considers only one factor i.e. load current, to find
3.15. Ripple correlation control (RCC) method MPP but it is applicable only for PV system with single phase DC–
AC converter as shown in Fig. 16.
This method takes the advantage of the signal ripple, which is As the duty ratio of DC–AC converter increases, the output
automatically present in power converters. The ripple is inter- current of the converter i.e. load current IL increases but the vol-
preted as a perturbation from which an optimization can be rea- tage decreases thus the relation between duty ratio and voltage
lized. Oscillation in power provided through all pass filter which can be obtained.
makes use of ripples to perform MPPT [54]. Method makes use of The concept of POS control method is that, if duty ratio (δ)
the (17) that at MPP: decreases the output current of power converter or load current
increases i.e. going towards the MPP. After MPP, decrease in duty
dP dV dP dI
 or  ¼0 ð17Þ ratio results decrease in current IL i.e. going away from MPP. Thus
dt dt dt dt
using only one parameter load current, MPP is achieved [57].

3.16. State space based method

3.19. Gradient descent or steepest descent method

The method is based on the mathematical modeling of the PV

Gradient descent is the first order optimization algorithm and it
system with DC–DC converter and can be considers as a dynamic
is used to find a local minimum of a function by tracking steps
system and represented by the state space equation:
proportional to the negative of the gradient. It is also known as
x_ ðt Þ ¼ Axðt Þ þBðt Þuðt Þ þ Dεðt Þ ð18Þ steepest descent method. This method is applied to find nearest
where: x, state variable vector local MPP while the gradient of the function f ðV; P Þ ¼ dVdP
is given
by:
u, switch duty ratio of DC/DC converter
 
ε, disturbance due to load variation  dP V ¼ V k
V k þ 1 ¼ V k þ   ð22Þ
t, independent time variable dV k A
This method gives a control law ‘u’ to obtain MPP
where k A is the step size corrector which decides the next step
u ¼ KT  MT g1 þ nεðt Þ ð19Þ size in the direction of gradient. For MPPT operation it is required
where the vector KT, MT and the parameter n are the controller to find the solution where dVdP
or function become minimum (zero)
parameters that are to be defined during the design process. [58–60].
The vector g1 is the reference signal which satisfy the max-
imum power condition. 3.20. Analytic solution based method
 : T
g 1 ¼ g ðt Þg ðt Þg\  t€\2:5pt ðt Þ ð20Þ This method is based on the mean value theorem which is one
  of the most important method of real analysis which provides the
dV o
g ðt Þ ¼ V o þ io ð21Þ analytic solution of a point which is closer to MPP and inside a
dI o
circle of small radius ε [61].
Vo and io represent the load voltage and current. Line 1 in Fig. 17 is given by:
This method ensures a globally asymptotically stable system.
IL
This method is also applicable to fast changing environmental i¼  v þ IL ð23Þ
V oc
condition [55].
1028 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

Line 2
10
Isc
8

Current (A)
6
Im Line 1
4

0
0 0.1 0.2 0.3 0.4 0.5 Vm 0.6 Voc 0.7
Voltage (Volt)

Fig. 17. Solar cell I–V characteristic with parallel lines.

Variable inductor Membership

grade
LV NB NS ZO PS PB
D M1 1
Cin D1 C1
RL

E
0
-0.04 -0.02 0 0.02 0.04
Vpv Ipv
Photovoltaic array
NB NS ZO PS PB
MPPT δ 1
Controller

Fig. 18. Schematic diagram of variable inductor control of MPPT.

CE
0
where IL is light generated current of the solar cell, and line 2 is -80 -40 0 40 80
parallel to line 1 and tangent to the current curve.

Now ε 4 vpv þipv Rs  ðvMPP þ iMPP Rs Þ

NB NS ZO PS PB
ð24Þ 1

where vpv ; ipv is intersection of line 2 and near to MPP.

δ
3.21. Variable inductance method 0 -0.04 -0.02 0 0.02 0.04

This method introduces variable inductor in place of constant Fig. 19. Membership function for input error change in error and duty ratio.
inductor in DC–DC buck converter along with MPPT controller
[62], shows the experimental results. This method is robust and
reliable with variation in insolation. This method is easy to implement and required simple cir-
Use of variable inductor in place of fixed inductor in DC–DC cuitry. Voltage and temperature of PV array are required to be
converter reduces the overall inductor size by 75%. For high measured.
insolation current would be high, inductance of lower value is
sufficient while in case of lower insolation (lower current) 3.23. Bisect search theorem (BST)
increased inductance required.
The minimum inductance is given by (25): It is a mathematical approach to locate the roots of any function
δ2 ð1 δÞV pv y ¼ f ðxÞ in an interval ½a; b. In context of applying BST in MPPT the
Lmin ¼ ð25Þ function is ΔVΔP
in between the interval ½0; V oc . As is obvious from
2f s I pv
the characteristic of solar cell, it is a function which becomes zero
Fig. 18 shows the schematic diagram of variable inductor con- _
at MPP. Then the root xrepresents the solution [67].
trol, in which any of the conventional method can be apply with
variable inductance which reduces the overall size of the 3.24. Fuzzy logic control method
inductance.
Fuzzy logic is a set of multiple-valued logic, as compared to
3.22. Temperature based method binary set where variable has only two states true or false value.
Fuzzy logic variable have range between zero to one, which
In this method temperature of solar PV is measured. Variation introduces the concept of partial truth, where the variable value
in MPP with respect to the temperature is obtained in similar way may range between complete true and complete false.
of constant voltage method. Eq. (26) decides the reference tem- In the application of fuzzy logic controller for MPPT error
perature with respect to MPP: (E) and change in error (CE) at Kth iteration is:
V MPP ðt Þ ¼ V MPP ðT ref Þ þ T Kvoc ðT  T ref Þ ð26Þ P PV ðkÞ  P PV ðk  1Þ
E ðk Þ ¼ ð27Þ
where VMPP is the MPP voltage, T is the working panel tem- iPV ðkÞ  iPV ðk  1Þ
perature, TKvoc is the temperature coefficient of VMPP, and Tref is the
CEðkÞ ¼ EðkÞ  Eðk 1Þ ð28Þ
standard test conditions temperature [63–66].
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1029

Input Layer Hidden Layer Output Layer

Voc
WI{1,1} WL{2,1} Vmpp(n)

+ +
T(n) b{1} B{2}

2 1
Fig. 20. Feed forward neural network function approximator.

where PPV and iPV are the power and current of the PV array. In 1. Category
case of MPP, E(k) should be zero. 2. Dependency of PV array
For input and output variable Fig. 19 shows the membership 3. Implementation methodology
grades of five fuzzy subsets. The input variable like voltage and 4. Sensor required
current are expressed in terms of labels (NB: negative big, NS: 5. Stages of energy conversion
negative small, ZO: zero, PS: positive small, PB: positive big). Fuzzy 6. Partial shading enabled
controller is designed in such a way that input variable E has to be 7. Grid interaction
always zero, which is the condition of MPP [68–76]. 8. Analog or Digital
9. Tracking efficiency
3.25. Artificial neural network (ANN) based method 10. Tracking speed
11. Cost
ANN is a soft computing technique inspired by central nervous 12. Product available in market.
system (brain) and these computational models are capable of
machine learning and they are represented as the interconnected Some paper also found in literature which provides compara-
neurons (artificial nodes) to form a network similar to biological tive analysis [94–97].
neural network.
The block diagram for application of ANN in MPPT is shown in
4.1. Category
Fig. 20.
Two inputs are there Voc(n), the reference cell’s open circuit
MPPT methods can be classified under Indirect Control (IND),
voltage and time parameter T(n). Training of neurons considers the
connecting weights wI{1,1} and bias b{1}. In training process Direct Control (DIR) and soft computing techniques (SC) on the
connecting weights are modified until best fit is achieved i.e. basis of their control strategy.
reference voltage corresponding to MPP [77–85].
4.1.1. Indirect control methods
Indirect control methods are typically based on mathematical
3.26. Method based on particle swarm optimization (PSO)/ant col-
relationship obtained from the empirical data which may include
ony optimization (ACO)
the parameters and characteristic of the solar PV panel. Maximum
The PSO is a population based search algorithm, which is based power points are basically predicted offline using various algo-
on analysis of the social behavior of birds and school of fishes. The rithms, empirical data or mathematical equations. These methods
PSO approach can be applied to any optimization problem having are not suitable with fast changing environmental condition and
multivariable function with multiple optimal points. partial shading condition, also the MPP given by the method is not
Application of PSO/ACO in MPPT is basically about the tracking true but it is an approximation based on the database. Example:
in case of partial shading condition where more than one max- Constant voltage, constant current, look up table, curve fitting,
imum point are there as shown in Fig. 5. Among all local max- pilot cell methods.
imum there is only one global maximum or MPP. Movement of
PSO agent (swarm) in search space depends on its own previous 4.1.2. Direct control methods
best position and the overall best position for all swarms. For every Direct control methods are search algorithms which locate the
position calculation of power, is done for all agents in this way maximum power points against changing atmospheric conditions
MPP achieved. online. They are generally based on the sampling based control or
The velocity and position for next iteration for ith swarm is modulation based control strategy. Example: Hill Climbing, P&O
given in (29) and (30) respectively: and INC etc.
vki þ 1 ¼ wvki þ c1 r 1 P besti þ c2 r 2 g best ð29Þ
4.1.3. Soft computing technique based methods
Methods based on the soft computing techniques belong to this
Ski þ 1 ¼ ski þ vki þ 1 ð30Þ
category. Example: methods based on the Genetic algorithm,
where vki þ 1 is velocity of ith swarm for iteration k þ 1, w is Artificial Neural Network and particle swarm optimization.
learning factor, c1 ; c2 are position constant and r 1; r 2 are random
numbers (their range is 0–1) [86–93].
4.2. Dependency of PV array (DPVA)

4. Comparison and analysis of MPPT methods This category explains the state of dependency (whether
dependent or independent) of MPPT methods on the type and size
Comparison of MPPT methods is done on the basis of following of PV system. Example: constant voltage method is PV system
12 factors: specific.
1030 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

4.3. Implementation methodology directly related with sensors required, computational procedure
and the circuitry used to implement the same. The methodology
This category explains the degree of complexity of the circuitry used to evaluate the cost of tracking scheme is based on above
required for implementation of the method. Example: constant mentioned parameters. Methods requiring complex circuitry are
voltage method is less complex as compared to the P&O method. more costly. In this paper cost of tracking scheme is separated in
four categories that is inexpensive (INEX), medium (Med.),
4.4. Sensor required expensive (EX) and very expensive (V.EX).
Example:
The control parameter or sensed parameter required to find out
the point of maximum power such as Voltage (V), Current (I),  Constant voltage method and constant current method require
Temperature (T) or may be combination of these three. less computation to reach MPP. Only one sensor is required with
less complex implementation circuitry, hence can be categor-
4.5. Stages of energy conversion ized in inexpensive category.
 P&O and INC require two sensors and more computation to
In solar PV systems energy conversion stages such as DC–DC, obtain MPP as compared to constant current and constant vol-
DC–AC or both, required to control the output of source with tage so these methods are categorized in medium category.
respect to maximum power.  Soft computing techniques require more complex hardware
thus are very expensive.
4.6. Partial shading enabled (PSE)
4.12. Size of the PV array
Under partial shading condition the output of the PV shows
multiple local maxima. This category defines the compatibility of Size of the PV array is another important factor and can be split
the method to find global maximum [98–102]. in three categories as small systems (S) which are less than 1 kW,
medium systems (M) which are in the range of 1 kW to 1 MW and
4.7. Grid interaction (GI) large systems (L) which are above 1 MW.
Example:
Methods can also be classified on the basis of their mode of
connection with the grid such as grid connected or off grid.  Less expensive MPPT techniques are preferred for small size
systems as expensive MPPT will result in increased cost of
4.8. Analog or digital overall system with a little increment in power.
 For medium system P&O and INC is suitable.
Methods can be classified on the basis of the kind of operating  For large systems accurate method is preferred as the power
circuitry required such as analog (A), digital (D), combination increments due to MPPT tracker is significant and partial
of both. shading effect is a major concern.

4.9. Tracking efficiency (TE) 4.13. Product available in market

Tracking efficiency basically defines the tracking accuracy of On the basis of the operating principle of the method, available
the tracker. A MPPT tracker can be efficient upto 100% if it delivers commercial products in the market are comprised in this category.
maximum power which is available. The efficiency of tracker is A survey on available commercial products is done by means of
given as; product datasheet analysis and conversation with the manu-
facturer for evaluation of this category [103–107]. Most of the
P Out manufacturer uses their own MPP methods for example Steca
T:E: ¼ η ¼  100
P MPP Elektronic, Germany uses unique MPP tracking for products Steca
In this paper tracking efficiency is evaluated in terms of three Solarix MPPT 1010 and 2010 [108].
different categories such as good, medium and poor. Table 4 gives the comparative analysis of the MPPT methods.
Example:
4.14. Comparative analysis of the MPPT methods
 Methods based on the soft computing techniques are more
accurate hence their tracking efficiency is good. Table 4 gives the brief comparison of MPPT methods and shows
 P&O method gives medium efficiency. that all methods have their own advantages and disadvantages.
 Constant current method, constant voltage method and pilot Some of the methods show very effective results such as soft
cell method give poor efficiency (as compared to conventional computing technique but the methodologies used were compli-
P&O method). cated. The methods which are simple in implementation such as
constant voltage method, constant current method and pilot cell
4.10. Tracking speed method are less accurate. Perturb and observe method is com-
monly used method because its implementation circuitry is not
Tracking speed of the MPPT tracker defines the speed of the complex but it shows sluggish response where environmental
tracker to achieve the MPP. Example: in case of P&O method conditions changes rapidly. Beta method shows proficient results
increase in perturbation size results increase in Tracking Speed but with fast change in insolation only if variation in temperature is
the accuracy or Tracking efficiency becomes poor. less. Some methods are based on mathematical optimization
algorithms to solve the non linear problem such as state space
4.11. Cost based method, analytic solution based method and steepest des-
cent or gradient descent method, they required more computation
This category defines economy of the MPPT tracker in com- thus the complexity increases and in case of partial shading
parison with the other trackers. The cost of any tracking scheme is tracking efficiency goes down. In partial shading condition soft
Table 4
Comparison of the MPPT methods.

S. No. Method Cate- DPVA Implemen-tation ofme- Sensor required Stages of energy PSE GI Analog or TE Track-ing Cost Size Product available in market
gory thodol-ogy convers-ion digital speed
V I T

1 Method A IND ✔ Simple ✔ DC–DC A,D Poor Fast INEX S

2 Method A(i) IND ✔ Simple ✔ DC–DC A,D Poor Fast INEX S
3 Method B IND ✔ Simple ✔ DC–DC A,D Poor Fast INEX S

D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

4 Method C IND ✔ Simple ✔ ✔ DC–DC D Med. Fast INEX S OutBack Power, USA-FLEXmax80,
FLEXmax60
5 Method D IND ✔ Simple ✔ ✔ DC–DC D Med. Fast INEX S
6 Method E DIR Simple ✔ ✔ Both ✔ A,D Good Slow Med. S/M Genasun, USA GV Boost charge controller
with MPPT
Blue Chip Energy Solution Pvt. Ltd., India-
Solar MPPT Charge controller
7 Method E(i) DIR Simple ✔ ✔ Both ✔ A,D Good Slow Med. S/M
8 Method E(ii) DIR Complex ✔ ✔ Both ✔ ✔ D Good Fast EX M/L
9 Method E(iii) DIR Complex ✔ ✔ Both ✔ D Good Fast EX M
10 Method E(iv) DIR Complex ✔ ✔ Both ✔ A,D Good Med. EX M
11 Method E(v) DIR Complex ✔ ✔ Both ✔ A,D Good Slow EX M
12 Method F DIR Med. ✔ ✔ Both ✔ D Good Med. Med. S/M
13 Method G DIR Complex ✔ ✔ Both ✔ D Good Med. EX M
14 Method H DIR Complex ✔ ✔ DC–DC A,D Good Med. EX M
15 Method I DIR Med. ✔ ✔ DC–AC ✔ A Good Slow Med. S/M
16 Method J DIR Complex ✔ ✔ DC–DC D Med. Med. Med. S/M Blue Sky Energy, USA New-Solar Boost 300i,
2512i-HV & 2512ix-HV
Morningstar Corporation, USA TriStar MPPT
& SunSaver MPPT
17 Method K DIR Med. ✔ DC–AC ✔ A,D Good Fast Med. S/M
18 Method L DIR Med. ✔ ✔ Both ✔ D Good Fast EX M
19 Method M DIR Complex ✔ DC–DC A Good Med. EX M
20 Method N DIR Complex ✔ ✔ DC–DC ✔ A Good Fast EX M
21 Method O DIR ✔ Complex ✔ ✔ DC–DC D Good Med. EX M
22 Method P DIR ✔ Simple ✔ ✔ DC–DC D Good Fast EX M
23 Method Q DIR Simple ✔ DC–DC D Good Med. EX M
24 Method R DIR ✔ Complex ✔ ✔ DC–DC D Good Med. EX M
25 Method S DIR Complex ✔ ✔ DC–DC A,D Good Med. EX M
26 Method T DIR Med. ✔ ✔ Both ✔ A,D Med. Med. EX M
27 Method U SC ✔ Med. ✔ ✔ DC–DC A Good Med. EX M
28 Method V SC Complex ✔ ✔ Both ✔ D Good Med. V. EX L
29 Method W SC Complex ✔ ✔ Both ✔ ✔ D V. Good Fast V. EX L
30 Method X SC Complex ✔ ✔ Both ✔ ✔ D V. Good Fast V. EX L
31 Method Y SC Complex ✔ ✔ Both ✔ ✔ D V. Good Fast V. EX L

1031
1032 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

computing techniques such as ANN based method, GA based [19] Leedy AW, Liping Guo, Aganah KA. A constant voltage MPPT method for a
method and PSO based method give conspicuous tracking effi- solar powered boost converter with DC motor load. In: Proceedings of IEEE
Southeastcon 2012;1–6;18–15.
ciency but overall system becomes expensive. [20] Salameh ZM, Dagher F, Lynch WA. Step-down maximum power point tracker
for photovoltaic systems. Sol Energy 1991;46(5):279–82.
[21] Alghuwainem SM. Matching of a dc motor to a photovoltaic generator using
a step-up converter with a current-locked loop. IEEE Trans Energy Convers
5. Conclusion 1994;9(1):192–8.
[22] Leedy AW, Garcia K E. Approximation of P–V characteristic curves for use in
maximum power point tracking algorithms. In: Proceedings of the system
This review paper presents a summary of 31 MPPT techniques theory (SSST) 45th southeastern symposium. 2013. p. 93–88.
and their classification on the basis of 13 parameters. Availability of [23] Desai HP, Patel HK. Maximum power point algorithm in PV generation: an
bulk options as an MPPT makes its unambiguous selection, a overview. In: Proceedings of the Power Electronics and Drive Systems PEDS
7th International Conference. 2007. p. 630–24.
tougher nut to crack. The comparative analysis provided in this [24] Liu X, Lopes LAC. An improved perturbation and observation maximum
paper may found to be a striking tool which may guide the utility to power point tracking algorithm for PV arrays. In: Proceedings of IEEE 35th
select the most constructive and compatible type of MPPT to fulfill annual conference on power electronics specialists ;3:2005–2010.
[25] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimization of perturb and
the requirements of both operators and consumers. It will avoid the observe maximum power point tracking method. IEEE Trans Power Electron
complicacy in the tiresome job of flawless selection of MPPT which 2005;20:963–73.
[26] Abdelsalam AK, Massoud AM, Ahmed S, Enjeti P. High-performance adaptive
will found to be worthy of complying all the desired attributes of
perturb and observe MPPT technique for photovoltaic-based microgrids.
the system and ecology both. Some MPPT manufacturer names are IEEE Trans Power Electron 2011;26(4):1010–2021.
also provided along with their commercial products. This infor- [27] Elgendy MA, Zahawi B, Atkinson DJ. Assessment of perturb and observe
MPPT algorithm implementation techniques for PV pumping applications.
mation may found to an attractive source to assist the researchers in
IEEE Trans Sustain Energy 2012;3(1):21–33.
context with the prevailing industrial scenario. [28] Aashoor FAO, Robinson FVP. A variable step size perturb and observe algo-
rithm for photovoltaic maximum power point tracking. In Proceedings of the
47th international conference on universities power engineering (UPEC).
2012.1:6; p. 4–7.
References [29] Sera D, Mathe L, Kerekes T, Spataru SV, Teodorescu R. On the perturb-and-
observe and incremental conductance MPPT methods for PV systems. IEEE J
Photovolt 2013;3(3):1070–8.
[1] Chauhan Anurag, Saini RP. A review on Integrated Renewable Energy System [30] Teulings WJA, Marpinard JC, Capel A, O'Sullivan D. A new maximum power
based power generation for stand-alone applications: configurations, storage point tracking system. In: Proceedings of the 24th annual IEEE conference on
options, sizing methodologies and control. Renew Sustain Energy Rev power electronics specialists PESC. 1993. p. 833–38.
2014;38:120–99. [31] Weidong Xiao, Dunford WG. A modified adaptive hill climbing MPPT method
[2] Sebri Maamar, Ben-Salha Ousama. On the causal dynamics between eco- for photovoltaic power systems. In: Proceedings of the IEEE 35th annual
nomic growth, renewable energy consumption, CO2 emissions and trade conference on power electronics specialists. 2004. 3; p. 1957-63.
openness: fresh evidence from BRICS countries. Renew Sustain Energy Rev [32] Koutroulis E, Kalaitzakis K, Voulgaris NC. Development of a microcontroller-
2014;39:14–23. based, photovoltaic maximum power point tracking control system. IEEE
[3] International Energy Agency. Tracking clean energy progress. Paris, France: Trans Power Electron 2001;16(1):46–54.
International Energy Agency; 2014. [33] Jain S, Agarwal V. A new algorithm for rapid tracking of approximate max-
[4] International Energy Agency. World energy investment outlook – special imum power point in photovoltaic systems. IEEE Power Electron Lett 2004;2
report. Paris, France: International Energy Agency; 2014. (1):16–9.
[5] International Energy Agency. World energy outlook. Paris, France: Interna- [34] Mei Qiang, Shan Mingwei, Liu Liying, Guerrero JM. A novel improved vari-
tional Energy Agency; 2013.
able step-size incremental-resistance MPPT method for PV systems. IEEE
[6] Troldborg Mads, Heslop Simon, Hough Rupert L. Assessing the sustainability
Trans Ind Electron 2011;58(6):2427–34.
of renewable energy technologies using multi-criteria analysis: suitability of
[35] Ansari F, Iqbal A, Chatterji S, Afzal A. Control of MPPT for photovoltaic sys-
approach for national-scale assessments and associated uncertainties. Renew
tems using advanced algorithm EPP. In: Proceedings of international con-
Sustain Energy Rev 2014;39:1173–84.
ference on power systems ICPS. 2009.1:6; p. 27-9.
[7] Benson Christopher L, Christopher L Magee. On improvement rates for
[36] Ying-Tung Hsiao, China-Hong Chen. Maximum power tracking for photo-
renewable energy technologies: solar PV, wind turbines, capacitors, and
voltaic power system. In: Proceedings of the 37th IAS annual meeting
batteries. Renew Energy 2014;68:745–51.
industry applications conference. 2002. 2; p.1035–40.
[8] Lesourd Jean-Baptiste. Solar photovoltaic systems: the economics of a
[37] Safari A, Mekhilef S. Simulation and hardware implementation of incre-
renewable energy resource. Environ Model Software 2001;16(2):147–56.
mental conductance MPPT with direct control method using cuk converter.
[9] Dincer Furkan. The analysis on photovoltaic electricity generation status,
IEEE Trans Ind Electron 2011;58(4):1154–61.
potential and policies of the leading countries in solar energy. Renew Sustain
[38] Woyte A, Thong Vu Van, Belmans R, Nijs J. Voltage fluctuations on dis-
Energy Rev 2011;15(1):713–20.
tribution level introduced by photovoltaic systems. IEEE Trans Energy Con-
[10] Parida Bhubaneswari, S Iniyan Ranko Goic. A review of solar photovoltaic
technologies. Renew Sustain Energy Rev 2011;15(3):1625–36. vers 2006;21(1):202–9.
[11] Singh Solanki Chetan. Solar photovoltaics: fundamentals, technologies and [39] Matsui M, Kitano T, De-hong Xu, Zhong-qing Yang. A new maximum pho-
applications. New Delhi: PHI Learning Pvt. Ltd.; 2011. tovoltaic power tracking control scheme based on power equilibrium at DC
[12] Hanwha Q. Cells GMBH, Germany 〈http://www.qcells.com/uploads/tx_ link. In: Proceedings of IEEE Industry Applications Conference, Conference
abdownloads/files/Hanwha_Q_CELLS_Data_sheet_Q6LMXP3-G3_2013-04_ Record of the Thirty-Fourth IAS Annual Meeting. 1999. 2; p. 804–09.
Rev01_EN.pdf〉. [40] Kitano T, Matsui M, De-hong Xu. Power sensor-less MPPT control scheme
[13] Nema Savita, Nema RK, Agnihotri Gayatri. MATLAB/Simulink based study of utilizing power balance at DC link-system design to ensure stability and
photovoltaic cells/modules/array and their experimental verification. Int J response. In: Proceedings of the 27th annual conference of the IEEE Indus-
Energy Environ 2010;1(3):487–500. trial Electronics Society IECON -01. 2001. 2; p. 1309–14.
[14] Nema RK, Nema Savita, Agnihotri Gayatri. Computer simulation based study [41] Bleijs JAM, Gow JA. Fast maximum power point control of current-fed DC-DC
of photovoltaic cells/modules and their experimental verification. Int J converter for photovoltaic arrays. Electron Lett 2001;1:5–6.
Recent Trends Eng 2009;1(3):151–6. [42] Shmilovitz D. On the control of photovoltaic maximum power point tracker
[15] Nema RK, Nema Savita, Agnihotri Gayatri. Design, development and simu- via output parameters. IEE Proc Electr Power Appl 2005;152(2):239–48.
lation of PC-based scheme for characterisation of solar photovoltaic modules. [43] Kislovski AS, Redl R. Maximum-power-tracking using positive feedback. In:
Int J Power Electron 2010;2(3):304–25. Proceedings of the 25th annual IEEE conference on power electronics spe-
[16] Nema RK. Savita Nema, Gayatri Agnihotri. Determination of IV characteristics cialists PESC-1994. 2; p. 1065–68.
of solar photovoltaic module by computer controlled loading. In: Proceed- [44] Noguchi T, Matsumoto H. Maximum-power-point tracking method of pho-
ings of the 15th international photovoltaic science & engineering conference tovoltaic power system using single transducer. In: Proceedings of the 29th
(PVSEC-15); 2005. annual conference of the IEEE Industrial Electronics Society, IECON. 2003. 3;
[17] Kamarzaman Nur, Atharah Chee Wei Tan. A comprehensive review of max- p. 2350–55.
imum power point tracking algorithms for photovoltaic systems. Renew [45] Bodur M, Ermis M. Maximum power point tracking for low power photo-
Sustain Energy Rev 2014;37:585–98. voltaic solar panels. In: Proceedings of the 7th Mediterranean Conference on
[18] Hyeong-Ju Noh, Dong-Yun Lee, Dong-Seok Hyun. An improved MPPT con- Electrotechnical. 1994. 2; p. 761-58.
verter with current compensation method for small scaled PV-applications. [46] Trishan Esram, Chapman Patrick L. Comparison of photovoltaic array max-
IECON 02 [Industrial Electronics Society, IEEE 2002 28th Annual Conference imum power point tracking techniques. IEEE Trans Energy Convers 2007;22
of the]. 2002. 2; p. 1118–19. (2):439–49.
 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034 1033

[47] Chen Yang, Smedley KM. A cost-effective single-stage inverter with max- [74] Patcharaprakiti Nopporn, Premrudeepreechacharn Suttichai, Sriuthaisir-
imum power point tracking. IEEE Trans Power Electron 2004;19(5):1289–94. iwong Yosanai. Maximum power point tracking using adaptive fuzzy logic
[48] Femia N, Granozio D, Petrone G, Spagnuolo G, Vitelli M. Optimized one-cycle control for grid-connected photovoltaic system. Renew Energy 2005;30
control in photovoltaic grid connected applications. IEEE Trans Aerosp (11):1771–88.
Electron Syst 2006;42(3):954–72. [75] Kharb RK, Shimi SL, Chatterji S, Ansari MF. Modeling of solar PV module and
[49] Sreeraj ES, Chatterjee K, Bandyopadhyay S. One-cycle-controlled single-stage maximum power point tracking using ANFIS. Renew Sustain Energy Rev
single-phase voltage-sensorless grid-connected PV system. IEEE Trans Ind 2014;33:602–12.
Electron 2013;60(3):1216–24. [76] Altin Necmi, Saban Ozdemir. Three-phase three-level grid interactive
[50] Zhang Miao, Wu Jie, Zhao Hui. The application of slide technology in PV inverter with fuzzy logic based maximum power point tracking controller.
maximum power point tracking system. In: Proceedings of the fifth world Energy Convers Manag 2013;69:17–26.
congress on intelligent control and automation WCICA. 2004. 6; p. 5594-91. [77] Liu YH, Liu CL, Huang JW, Chen JH. Neural-network-based maximum power
[51] Kim Il-Song, Kim Myung-Bok, Youn Myung-Joong. New maximum power point tracking methods for photovoltaic systems operating under fast
point tracker using sliding-mode observer for estimation of solar array cur- changing environments. Sol Energy 2013;89:53.
rent in the grid-connected photovoltaic system. IEEE Trans Ind Electron [78] Punitha K, Devaraj D, Sakthivel S. Artificial neural network based modified
2006;53(4):1027–35. incremental conductance algorithm for maximum power point tracking in
[52] Aghatehrani R, Kavasseri R. Sensitivity-analysis-based sliding mode control photovoltaic system under partial shading conditions. Energy 2013;62:330–
for voltage regulation in microgrids. IEEE Trans Sustain Energy 2013;4
40.
(1):50–7.
[79] Subiyanto Subiyanto, Mohamed Azah, Hannan MA. Intelligent maximum
[53] Ho BM, Chung HS, Lo WL. Use of system oscillation to locate the MPP of PV
power point tracking for PV system using Hopfield neural network opti-
panels. IEEE Power Electron Lett 2004;2(1):1–5.
mized fuzzy logic controller. Energy Build 2012;51:29–38.
[54] Nema RK, Bhatnagar Pallavee. Maximum power point tracking control
[80] Sahnoun MA, Ugalde HMR, Carmona JC, Gomand J. Maximum power point
techniques: state-of-the-art in photovoltaic applications. Renew Sustain
tracking using p&o control optimized by a neural network approach: a good
Energy Rev 2013;23:224–41.
compromise between accuracy and complexity. Energy Procedia
[55] Solodovnik EV, Liu Shengyi Dougal RA. Power controller design for max-
imum power tracking in solar installations. IEEE Trans Power Electron 2013;42:650–9.
[81] Ocran TA, Cao J, Cao B, Sun X. Artificial neural network maximum power
2004;19(5):1295–304.
[56] Pan Ching-Tsai, Chen Jeng-Yue, Chu Chin-Peng, Huang Yi-Shuo. A fast max- point tracker for solar electric vehicle. Tsinghua Sci Technol 2005;10(2):
imum power point tracker for photovoltaic power systems In: Proceedings of 204–208.
the 25th annual conference of the IEEE Industrial Electronics Society IECON. [82] Whei-Min Lin, Hong Chih-Ming, Chen Chiung-Hsing. Neural-network-based
1993. 1; p. 390-93. MPPT control of a stand-alone hybrid power generation system. IEEE Trans
[57] Lee Seok-Ju, Park Hae-Yong, Kim Gyeong-Hun, Seo Hyo-Ryong, Ali MH, Park Power Electron 2011;26(12):3571–81.
Minwon. et al. The experimental analysis of the grid- connected PV system [83] Takashi Hiyama, Kouzuma Shinichi, Imakubo Tomofumi. Identification of
applied by POS MPPT. In: Proceedings of international conference on elec- optimal operating point of PV modules using neural network for real time
trical machines and systems ICEMS. 2007. p. 1786–91. maximum power tracking control. IEEE Trans Energy Convers 1995;10
[58] Xiao Weidong, Lind MGJ, Dunford WG, Capel A. Real-time identification of (2):360–7.
optimal operating points in photovoltaic power systems. IEEE Trans Ind [84] Karatepe E, Hiyama T. Artificial neural network-polar coordinated fuzzy
Electron 2006;53(4):1017–26. controller based maximum power point tracking control under partially
[59] Xiao Weidong, Dunford WG, Palmer PR, Capel A. Application of centered shaded conditions. Renew Power Gener IET 2009;3(2):239–53.
differentiation and steepest descent to maximum power point tracking. IEEE [85] Mummadi Veerachary, Senjyu Tomonobu, Uezato Katsumi. Neural-network-
Trans Ind Electron 2007;54(5):2539–49. based maximum-power-point tracking of coupled-inductor interleaved-
[60] Pradhan R, Subudhi B. A steepest-descent based maximum power point boost-converter-supplied PV system using fuzzy controller. IEEE Trans Ind
tracking technique for a photovoltaic power system. In: Proceedings of the Electron 2003;50(4):749–58.
2nd international conference on control and embedded systems (ICPCES). [86] Ishaque K, Salam Z, Shamsudin A, Amjad M. A direct control based maximum
2012. 1:6; p. 19–17. power point tracking method for photovoltaic system under partial shading
[61] Rodriguez C, Amaratunga G. Analytic solution to the photovoltaic maximum conditions using particle swarm optimization algorithm. Appl Energy
power point problem. IEEE Trans Circuits Syst I Regul Pap 2007;54(9): 2012;99:414–22.
2054–2060. [87] Anula Khare, Rangnekar Saroj. A review of particle swarm optimization and
[62] Zhang Longlong, Hurley WG, Wö lfle WH. A new approach to achieve max- its applications in solar photovoltaic system. Appl Soft Comput 2013;13
imum power point tracking for PV system with a variable inductor. IEEE (5):2997–3006.
Trans Power Electron 2011;26(4):1031–7. [88] Miyatake M, Toriumi F, Fujii N, Ko H. Maximum power point tracking of
[63] Mutoh N, Matuo T, Okada K, Sakai M. Prediction-data- based maximum- multiple photovoltaic arrays: a PSO approach. IEEE Trans Aerosp Electron
power-point-tracking method for photovoltaic power generation systems In: Syst 2011;47(1):367–80.
Proceedings of the IEEE 33rd Annual Conference on Power Electronics Spe- [89] Lian KL, Jhang JH, Tian IS. A maximum power point tracking method based
cialists. 2002. 3; p. 1489–94. on perturb-and-observe combined with particle swarm optimization. IEEE J
[64] De Brito MAG, Galotto L, Sampaio LP, de Azevedo e Melo G, Canesin CA. Photovolt 2014;4(2):1–8.
Evaluation of the main MPPT techniques for photovoltaic applications. IEEE [90] Sundareswaran Kinattingal, Sankar Peddapati, Palani S. Application of ran-
Trans Ind Electron 2013;60(3):1156–67. dom search method for maximum power point tracking in partially shaded
[65] Park Minwon, Yu In-Keun.A study on the optimal voltage for MPPT obtained
photovoltaic systems. IET Renew Power Gener 2014;8(6):670–8.
by surface temperature of cell. In: Proceedings of the 30th Annual Con-
[91] Kashif Ishaque, Salam Zainal. A deterministic particle swarm optimization
ference of IEEE Industrial Electronics Society IECON. 2004. 3; p. 2040–45.
maximum power point tracker for photovoltaic system under partial shading
[66] Coelho RF, Concer FM, Martins DC. A MPPT approach based on temperature
condition. IEEE Trans Ind Electron 2013;60(8):3195–206.
measurements applied in PV systems. In: Proceedings of the 9th IEEE/IAS
[92] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved Particle Swarm
international conference on industry applications (INDUSCON). 2010. p. 1–6.
Optimization (PSO)–based MPPT for PV with reduced steady-state oscilla-
[67] Wang Peng, Zhu Haipeng, Shen Weixiang, Choo Fook Hoong, Loh Poh Chiang,
tion. IEEE Trans Power Electron 2012;27(8):3627–38.
Tan Kuan Khoon. A novel approach of maximizing energy harvesting in
[93] Liu YH, Huang SC, Huang JW, Liang WC. A particle swarm optimization-
photovoltaic systems based on bisection search theorem. In: Proceedings of
based maximum power point tracking algorithm for PV systems operating
the Twenty-Fifth Annual IEEE Conference and Exposition (APEC) on Applied
Power Electronics. 2010. p. 2143–48. under partially shaded conditions. IEEE Trans Energy Convers 2012;27
[68] Dounis AI, Kofinas P, Alafodimos C, Tseles D. Adaptive fuzzy gain scheduling (4):1027–35.
PID controller for maximum power point tracking of photovoltaic system. [94] Eltawil Mohamed A, Zhao Zhengming. MPPT techniques for photovoltaic
Renew Energy 2013;60:202–14. applications. Renew Sustain Energy Rev 2013;25:793–813.
[69] Othmana Ahmed M, El-arinia Mahdi MM, Ghitasb Ahmed, Fathya Ahmed. [95] Kashif Ishaque, Salam Zainal. A review of maximum power point tracking
Realworld maximum power point tracking simulation of PV system based on techniques of PV system for uniform insolation and partial shading condi-
Fuzzy logic control. NRIAG J Astron Geophys 2012;1(2):186–94. tion. Renew Sustain Energy Rev 2013;19:475–88.
[70] Algazar MM, Al-Monier H, EL-halim HA, Salem MEEK. Maximum power [96] Bidyadhar Subudhi, Pradhan Raseswari. A comparative study on maximum
point tracking using fuzzy logic control. Int J Electr Power Energy Syst power point tracking techniques for photovoltaic power systems. IEEE Trans
2012;39(1):28–221. Sustain Energy 2013;4(1):89–98.
[71] Subiyanto Subiyanto, Mohamed Azah, Hannan M. Intelligent maximum [97] Reza Reisi A, Hassan Moradi M, Jamasb S. Classification and comparison of
power point tracking for PV system using Hopfield neural network opti- maximum power point tracking techniques for photovoltaic system: a
mized fuzzy logic controller. Energy Build 2012;51:29–38. review. Renew Sustain Energy Rev 2013;19:433–43.
[72] Messai A, Mellit A, Guessoum A, Kalogirou SA. Maximum power point [98] Parlak Koray, Şener PV. Array reconfiguration method under partial shading
tracking using a GA optimized fuzzy logic controller and its FPGA imple- conditions. Int J Electr Power Energy Syst 2014;63:713–21.
mentation. Sol Energy 2011;85(2):265–77. [99] Chong BVP, Zhang L. Controller design for integrated PV–converter modules
[73] Bounechba H, Bouzid A, Nabti K, Benalla H. Comparison of perturb & observe under partial shading conditions. Sol Energy 2013;92:123–38.
and Fuzzy logic in maximum power point tracker for PV systems. Energy [100] Lu F, Guo S, Walsh TM, Aberle AG. Improved PV module performance under
Procedia 2014;50:677–84. partial shading conditions. Energy Procedia 2013;33:248–55.
1034 D. Verma et al. / Renewable and Sustainable Energy Reviews 54 (2016) 1018–1034

[101] Dolara A, Lazaroiu GC, Leva S, Manzolini G. Experimental investigation of [105] Blue Chip Energy Solution Pvt. Ltd., E-57 Ground Floor Chattarpur Extn., New
partial shading scenarios on PV (photovoltaic) modules. Energy Delhi-110074. 〈http://www.bluechipenergy.in〉.
2013;55:466–75. [106] Blue Sky Energy, HQ 2598 Fortune Way, Suite K Vista, CA 92081 USA. 〈http://
[102] Murtaza A, Chiaberge M, Spertino F, Boero D, De Giuseppe M. A maximum www.blueskyenergyinc.com〉.
power point tracking technique based on bypass diode mechanism for PV [107] Morningstar Corporation, 8 Pheasant Run, Newtown, PA 18940 USA, 〈http://
arrays under partial shading. Energy Build 2014;73:13–25. www.morningstarcorp.com〉.
[103] OutBack Power, 17825 59th Ave. NE, Suite B Arlington, WA 98223 United [108] Steca Elektronik GmbH, Electronics Services, 〈http://www.steca.com〉.
States, 〈http://www.outbackpower.com〉.
[104] Genasun L.L.C. 1035 Cambridge St. Suite 16B Cambridge, MA 02141 USA,
〈http://genasun.com〉.
 Applied Energy 107 (2013) 135–148

Contents lists available at SciVerse ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

The application of soft computing methods for MPPT of PV system:

A technological and status review
Zainal Salam ⇑, Jubaer Ahmed, Benny S. Merugu
Faculty of Electrical Engineering, Universiti Teknologi Malaysia (UTM), 81310 Johor, Malaysia

h i g h l i g h t s

" We have reviewed the journals on soft computing technique based PV system MPPT.
" Methodologies of all techniques are described.
" Merits and limitations of the techniques are presented.
" All techniques are compared with common benchmarks and future prospects are discussed.

a r t i c l e i n f o a b s t r a c t

Article history: With the availability of powerful and low cost computing power, maximum power point tracking
Received 16 September 2012 (MPPT) that utilizes soft computing (SC) techniques are attracting substantial interests from the PV
Received in revised form 17 January 2013 communities. Due to their flexibility and ability to handle non-linear problems, robust SC-based MPPT
Accepted 2 February 2013
schemes can be developed. Furthermore, the adaptive in nature SC algorithms is suitable in handling
adverse environmental conditions such as partial shading and rapid changes in irradiance. To date,
there are several works on MPPT using SC from which we select approximately 45 published works
Keywords:
that are directly related to MPPT. However, information on these methods are scattered and there
Photovoltaic
MPPT
appears to be an absence for a comprehensive review paper on this topic. This work summarizes
Review the current technology and status of SC MPPT as reported in various literature. It also provides an
Energy evaluation on the performance of various SC methods based on several criteria, namely PV array
Converter dependency, convergence time, ability to handle partial shading conditions, algorithm complexity
Soft computing and hardware/practical implementation. It is envisaged that the information gathered in this paper
will be a valuable one-stop source of information for researchers, as well as providing a direction
for future research in this area.
Ó 2013 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
2. MPPT for PV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.1. Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.2. I–V and P–V characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.3. Partial shading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3. Soft computing MPPT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.1. Artificial Neural Network (ANN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.2. Non-linear predictor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.3. Chaotic search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.4. Fuzzy Logic Control (FLC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.4. Fuzzy Logic Control (FLC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.5. Particle swarm intelligence (PSO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

⇑ Corresponding author. Address: Faculty of Electrical Engineering, Universiti Teknologi Malaysia (UTM), Skudai, 81310 Johor, Malaysia. Tel.: +60 7 5536187; fax: +60 7
5566272.
E-mail addresses: [email protected] (Z. Salam), [email protected] (J. Ahmed).

0306-2619/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.apenergy.2013.02.008
136 Z. Salam et al. / Applied Energy 107 (2013) 135–148

Nomenclature

D average duty cycle PPV PV power

d instantaneous duty cycle SC soft computing
G solar irradiance STC standard temperature and condition
EA evolutionary algorithm VMPP voltage at maximum power point
IMPP current at maximum power point VOC open circuit voltage in PV panel
IPV PV current VPV PV voltage
ISC short circuit current of PV panel T temperature
PMPP power at maximum power point

3.6. Ant colony optimization (ACO). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

3.7. Genetic algorithm (GA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.8. Differential evolution (DE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.9. Bayesian network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.1. PV array dependency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.2. Convergence speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3. Handling partial shading conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.4. Algorithm complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.5. Hardware implementation issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

1. Introduction tion. Other methods include Current Sweep [21], DC-link Capacitor
Drop Control [22], Load Current and Load Voltage Minimization
With the rapid growth in industries and the ever increasing [23], dP/dV or dP/dI Feedback Control [24], Linear Current control
sophistication of modern lifestyles, the world energy supply has [25], State-based MPPT [26], Best Fixed Voltage algorithm [27], Lin-
been subjected to a tremendous strain. These phenomena have ear Reoriented Coordinate [28] and Slide Control method [29,30].
raised concerns over the energy security and environmental sus- These techniques are reported in various in academic journals
tainability. Furthermore, the issues of climate change and the need but it is unclear if they are practically implemented in commercial
to reduce carbon footprints have added to the strong impetus for PV systems.
companies and nations to invest in alternative energy sources, par- With the recent availability of vast and low cost computing
ticularly the renewable energy (RE). As the price of oil is soaring power, MPPT based on soft computing (SC) techniques are attract-
unabated, the positive economic impact of RE is progressively ing considerable interests. The guiding principle of SC is exploiting
being recognized. the tolerance for imprecision, uncertainty, partial truth and
Due to the abundance of the source itself (sun), solar photovol- approximation to achieve tractability, robustness and low cost
taic (PV) is envisaged to an important RE sources of the future. A PV solution. The most important feature of SC is the flexibility of their
system is easy to install, almost maintenance free and environ- algorithms, which allows for the development of robust MPPT
mentally friendly [1–4]. Despite these advantages, the price of PV schemes. This is made possible as SC techniques are fully digital.
energy remains considerably higher than fossil fuel based electric- Furthermore, SC is known to be very effective in handling non-lin-
ity due to the high cost of PV modules. Although enormous amount ear complexities; since MPPT problem primarily centers on its
of work has been carried out to improve the solar cell fabrication non-linear PV curve [31,32], it is natural to solve it using SC tech-
technologies [5–7], it is equally important to boost the power yield niques. In addition, due to the adaptive nature of their algorithms,
of PV system by improving its maximum power point tracking SC is envisaged to be easily adaptable to cater for the adverse envi-
(MPPT) capability. This is because MPPT is the most economical ronmental conditions such as partial shading [33] and rapid
way to improve the overall PV system efficiency. changes in irradiance [34].
Conventional MPPT operates by sensing the current and voltage Despite its potential benefits, information on SC based MPPT are
of the PV array; the power is calculated and accordingly the duty scattered—publications on this topic can be found in energy, com-
cycle of the converter is adjusted to match the maximum power puters, environmental or electrical based journals. Unlike the con-
point (MPP). Despite having the same objectives, the various MPPT ventional MPPT, which are well summarized by several authors
techniques differ markedly in terms of convergence speed, steady [72], there is an absence of a comprehensive review on SC MPPT
state oscillations and cost effectiveness. Among them, Perturb methods. With the growing interest in this topic, it is timely to
and Observe (P&O) [8,9], Incremental Conductance (IC) [10–12] compile all the related works in this area and to present it as a sin-
and Hill Climbing (HC) [13,14] are the most popular. They are gle reference. Hence this paper is written. Its main objectives are to
widely employed in many commercial dc–ac inverters (for grid- summarize the current technology and status of SC MPPT tech-
tied) and dc–dc converters (for battery chargers). Besides, there niques. While doing so, it briefly describes the operation of each
are other simpler methods such as the Fractional Short Circuit Cur- method and highlights its advantages and limitations. Finally, a
rent [15,16], Fractional Open Circuit Voltage [17,18] and Ripple discussion on the performance of various SC methods is made.
Correlation Control [19,20] that are used for low cost applications, Admittedly, it is very difficult to ensure a fair benchmarking
such as street lightings. These MPPTs exhibit limited accuracy but because every researcher tested their algorithm on different PV
they require fewer sensors; thus offering a reliable, low cost solu- system (ratings, technologies, size, etc.) and under different
 Z. Salam et al. / Applied Energy 107 (2013) 135–148 137

environmental conditions. Despite these difficulties, this work at- the values that need to be matched by converter; in most case, only
tempts to summarize the performance based on several logical cri- one variable is selected (usually voltage). Then, the measured
teria. It is envisaged that the information gathered in this paper power value is compared with the present value of MPP. If there
will provide a valuable direction for future research in this area. is a difference between the two, the duty cycle (d) of the converter
is adjusted in an effort to reduce the difference. The control is car-
ried out by a PI or hysteresis controller. In certain cases, the duty
2. MPPT for PV
cycle is determined directly, i.e. without PI controller. Once the
measured equals the reference values, the maximum power from
2.1. Basics
the array is extracted.
The aim of employing MPPT is to ensure that at any environ-
mental condition (particularly solar irradiance and temperature),
maximum power is extracted from the PV modules. This is 2.2. I–V and P–V characteristics
achieved by matching the MPP with its corresponding converter’s
operating voltage and current. The basic block diagram of a typical A PV module can be modeled as a current source that is depen-
PV system with MPPT is shown in Fig. 1. Although a stand-alone dent on the solar irradiance and temperature. The complex rela-
dc–dc system is depicted here, the application can be extended tionship between the temperature and irradiation results in a
to a grid connected system by replacing the dc–dc converter with non-linear current–voltage characteristics. A typical I–V and P–V
an inverter and connecting the latter to the utility grid. curve for the variations of irradiance and temperature is shown
The MPPT works as follows. First, the current and voltage of the in Fig. 2a and b, respectively. As can be observed, the MPP is not
PV array are sensed by a current and voltage sensors, respectively. a fixed point; it fluctuates continuously as the temperature or
These values are fed into an MPPT block that computes the MPP at the irradiance does. Due to this dynamics, the controller needs to
that particular sampling cycle. Once found, the MPPT block delivers track the MPP by updating the duty cycle of the converter at every
the reference values for the current (I) and voltage (V). These are control sample. A quicker response from the controller (to match

PV Array
+
IPV L

CPV DC/DC O
Converer A
D
− VPV

Duty
VPV* /IPV
*
Cycled
MPPT + PI +
− q

Fixed or VPV/IPV
− switching
Adaptive signal

Fig. 1. Basic MPPT system with converter.

Fig. 2. (a) The I–V and P–V curves under varying solar irradiance. (b) The I–V and P–V curve under different temperatures.
138 Z. Salam et al. / Applied Energy 107 (2013) 135–148

Blocking Curve 1
1.2
Diode Shaded I-V curve
with bypass diode
0.9
Curve 2

Ipv (A)
Shaded
PV Module PV
Module 3 Module 3 0.6 Shaded I-V curve
without bypass diode
0.3 Curve 3
Bypass Vpv (V)
0
Diode 0 15 30 45 60
PV PV
Module 2 Module 2 60

Ppv (A)
40

PV PV 20
Module 1 Module 1

Vpv (V)
0
0 15 30 45 60

(a) (b) (c)

Fig. 3. Operation of PV array (a) under uniform irradiance (b) under partial shading (c) the resulting I–V and P–V curve for (a) and (b).

the MPP) will result in better extraction of the PV energy and vice diode. As a result, two stairs current waveform is created on the I–V
versa. curve, while the P–V curve is characterized by multiple maxima
points, as depicted by curve 2 of Fig. 3c. The MPPT needs to ensure
2.3. Partial shading that the tracked maximum point is the true global peak, not one of
the local maxima. If the algorithm is trapped at the local peak, sig-
The MPP tracking becomes more complicated when the entire nificant loss in power incurs.
PV array does not receive uniform irradiance. This condition is
known as partial shading. Typically, it is caused by the clouds that
strike on certain spots of the solar array, while other parts are left 3. Soft computing MPPT
uniformly irradiated [35]. Another source of partial shading-like
characteristics is exhibited by module irregularities; a common To date, there are approximately 45 published work directly re-
example would be the presence of cracks on one or more modules lated to SC MPPT. For simplicity, they are grouped into several cat-
of the PV array. egories, as depicted by the tree diagram in Fig. 4. It is worth noting
Fig. 3a shows a PV array in a typical series–parallel configura- that for SC based MPPT, not all techniques are used as MPPT con-
tion. Commonly, a bypass diode is fitted across the module to en- trollers. Artificial Neural Network (ANN) sometimes utilized to
sure that hot spot will not occur if that module is shaded. In this optimize certain parameters of other MPPT controllers, e.g. Fuzzy
example, three modules are connected in a single string. In a nor- Logic Controller (FLC), P&O and HC. Furthermore, it is common to
mal condition, i.e. when the solar irradiance on the entire PV array combine two or more techniques in a single MPPT controller. For
is uniform, the P–V curve exhibits a unique maximum power point example, ANN can be combined with FLC, or PSO with P&O. In most
(curve 1 of Fig. 3c). However, during partial shading in Fig. 3b, the cases, the performance of the hybrid tends to be superior com-
difference in irradiance between two modules activates the bypass pared to a stand-alone SC method.

Fig. 4. Soft computing technique used for MPPT in PV system.

 Z. Salam et al. / Applied Energy 107 (2013) 135–148 139

3.1. Artificial Neural Network (ANN) Most ANN methods [36–40] employ back-propagation (BP)
training algorithm. More hidden nodes increases its accuracy, but
ANN is an inter connection of artificial neurons (nodes) that mi- at the expense of longer computational time. Hence, an ANN with
mic a biological brain. Primarily, it comprises of an input, hidden high number of nodes may not be able to respond quickly enough
and output layers. A possible structure of ANN, tailored for MPPT to cope with in rapid fluctuations in environment. Additionally,
is shown in Fig. 5. The input variables can be PV array parameters ANN depends on the previous trained data, which remain static
such as irradiance, temperature, wind speed or any combination of for the MPP tracking. In this respect, ANN is unable to comply with
these. The output can be designated as either the voltage, current the dynamic phenomena of partial shading. Moreover, the BP algo-
or duty cycle, depending on the control variables used for the con- rithm remembers the new value but forgets the old, leading to the
verter. In each layer the numbers of nodes varies and is user poor memory re-attainment. Therefore, in general, ANN is unsuit-
defined. able for low cost microprocessors.
The ability of the ANN to track the MPP depends on hidden
layer’s algorithm and how careful and extensively the networks
are trained. Typically, the ANN needs to be trained and tested for 3.2. Non-linear predictor
months (or even years) to ensure that the MPPT responds correctly
to various meteorological conditions. During the training, the neu- Non-linear predictor is an algorithm that performs peak match-
rons are weighted appropriately to match the input–output pat- ing based on certain predictor function. A predictor functions is
tern correlation. It is important to note that once a particular usually selected according to the subject curve that needs to be
ANN is trained and designed for a specific PV module or climate, exploited. For MPPT, the most probable subject curve is the P–V
it may not respond accurately if employed in a different condition. curve itself. Since this curve resembles a parabolic, naturally, the
In certain cases, ANN is not used as the MPP tracker itself; logical predictor function would be a quadratic or parabolic. The
rather it is utilized to identify the optimized parameters of another algorithm first utilizes the inputs of several previous working duty
MPPT controller. For example, ANN is used as optimizers for tradi- cycles and their corresponding power data to define a parabolic
tional MPPT such as P&O or IC methods [37,38]. There are also re- curve based on the predictor function. As slope at MPP and at the
ported works on the application of ANN with other soft computing extreme point of the parabola is zero, the objective of the process
techniques like GA, FLC or DE. In general, these combinations result is to match these two points.
in improved performances [40,41,68]. A summary of the ANN re- As depicted in Fig. 6, D1, D2 and D3 are the input duty cycle
lated researches for MPPT are shown in Table 1. points to generate a downward parabola using corresponding P1,
P2 and P3. Then the duty cycle Dm1 is found where the parabola it
at its peak. The power Pm1 at Dm1 is then calculated. As Pm1 and
the MPP point of the PV curve (Pmpp) do not match, a new working
curve is generated using P1, P2, P3 and Pm1. Afterwards in the sec-
Layer h Layer j Layer k ond iteration, new point Pm2 is found, which also does not match
1 with the MPP. However, it is obvious from the graph that for every
renewal of the predictor function, the curve is heading towards the
MPP. Eventually the peak of the parabola will match the MPP.
The main advantage of prediction is that it works very effi-
1 1 ciently under fast changing irradiance or temperature, as the algo-
Solar I MP
2 rithm is less complex and easy to implement. However, under
Insolation (W/m 2 )
partial shedding condition such method is not applicable because
the predictor function does not constitute multiple peaks. Once
2 2 the extreme point of the predictor curve matches with local max-
Temperature V MP
( o C) ima it will stop searching for MPP and get trapped. Consequently,
j there is no reported work on the application of predictor for partial
j = 20 shading condition.
Presently in literature, there are two papers on the application
Fig. 5. A typical ANN structure for MPPT [39]. of non-linear predictor function for MPPT. Authors in [42] utilized

Table 1
Summary of ANN related work for MPPT.

Authors Ref. Control Converter type/Application Remarks

variable
Veerachary [36] Voltage Buck-boost converter for ANN is used as the MPPT controller. The MPP is identified using gradient descent algorithm
and standalone applications training. Work is extended for permanent magnet series motor
Yadaiah
Alabedin [37] Duty cycle Buck converter for ANN is used as an optimizer for P&O MPPT controller. Improved performance in dealing with the
et al. standalone applications fluctuations in the array power
Jinbang et al. [38] Duty Cycle Boost converter for ANN is used as an optimizer for IC/P&O MPPT controller. Faster than the IC and exhibits smaller
standalone applications steady-state error than the P&O algorithm
Islam and [39] Voltage Boost converter for ANN is used as the MPPT controller. Utilizes only 20 nodes in the hidden layer, which reduces its
Kabir and standalone applications complexity and increases the execution time
Current
Jie and Ziran [40] Voltage Buck-boost converter for ANN used as the MPPT controller. The method uses a 2 level ANN, which has higher speed and
standalone applications accuracy compared to single level ANN
Veerachary [41] Voltage Boost Converter for ANN is used as an optimizer for the feed forward FLC MPPT. The ANN is trained using the BP
et al. standalone applications algorithm to estimation the reference voltage on-line. Tracking performance is improved. It also
avoids the tuning of PI controller parameters
140 Z. Salam et al. / Applied Energy 107 (2013) 135–148

previous work has shown that it takes longer time to converge to

MPP than the conventional HC method [44]. The dual-carrier cha-
otic search improves the efficiency, precision and system robust-
ness. Furthermore, it makes the control mode greatly simplified
[44]. Using dual carrier search for MPPT, two logistic mapping is
used to generate bifurcation diagram that are given below, i.e.

xnþ1 ¼ kxn ð1  xn Þ ð1Þ

and

ynþ1 ¼ l sinðpyn Þ ð2Þ

where n = 1, 2, . . . , N. The two optimization variables are defined as

xri ¼ a þ xn ðb  aÞ ð3Þ

Fig. 6. Parabolic predictor for MPPT [43]. and

1
the quadratic as the predictor function. A second-order Lagrange yri ¼ a þ ðyn þ 2Þðb  aÞ ð4Þ
4
interpolating polynomial is proposed to improve its convergence.
In another work [43], the MPP is predicted using the parabolic con- where a and b defines the range of the initial searching zone. These
vex function, similar to the one shown in Fig. 8. The results show two variables are used to determine several initial points on a sub-
good dynamic efficiency under variable atmospheric condition. It ject function with a large search zone shown in Fig. 7.
has higher convergence speed and accuracy than the traditional When the output power displays a low–high–low phenomenon
P&O method. as denoted by points PðY 02 Þ; PðX 02 Þ and PðY 03 Þ in Fig. 7a, the two
low points PðY 02 Þ and PðY 03 Þ are selected as the end points to narrow
the searching zone. Through consecutive iteration, the search zone
3.3. Chaotic search progressively gets smaller and stops when its value is less than a
set threshold. The main difference between chaotic search and
Chaos is mathematically defined as ‘‘randomness’’ that results other search algorithm is that the search is initiated randomly in
from the sensitivity of a system to a small difference in the initial a selected range. The search from both ends of a PV curve forces
conditions. This feature can be exploited to search for the opti- a faster to convergence by rapid narrowing of the search range.
mized point of a dynamic system such as MPPT. For a chaotic Furthermore, the end-to-end zonal search guarantees the tracking
search to take place, an optimization variable and fitness function of multiple maxima points; hence its suitability for partial shading
is needed. In the case of MPPT, the PV voltage and power are se- problem. In [44], the dual carrier chaotic search is shown to be
lected as the optimization variable and the fitness functions, more efficient than the HC and the variable step HC methods.
respectively. The optimization can be done using single or multiple The reported convergence time is 60 ms with an error between
carriers. The single carrier search is not a feasible option because theoretical and experimental maximum power is less than 0.03%.

Fig. 7. Chaotic search for MPPT [44].

 Z. Salam et al. / Applied Energy 107 (2013) 135–148 141

3.4. Fuzzy Logic Control (FLC)

FLC provides a systematic approach to create automatic control

algorithm by exploiting linguistic variables, based on experts’
knowledge. In contrast to the binary logic, fuzzy variables may as-
sume a value between 0 and 1. Such controllers are advantageous
when working with imprecise inputs as it does not require an
accurate mathematical model. The latter is a significant advantage
because the uncertainties such as un-modeled physical quantities,
non-linearity and unpredictable changes in operating point can be Fig. 9. Membership function of a fuzzy set.
excellently dealt with [55]. Furthermore, FLC is known to be very
efficient in handling non-linearity problems.
increase the MPPT efficiency and tracking ability. Some of the
The general FLC structure is shown in Fig. 8. It consists of three
important research on FLC based MPPT are summarized in Table 3.
processing stages, namely fuzzification, rules inferences and
One of the major disadvantages of (pure) FLC is its inability to
defuzzification. In addition, it has a rule table in which the de-
handle partial shading. In order for it to do so, the rule table needs
signed rules are stored. The process in which the FLC performs
to change dynamically. However, this is not possible because once
the calculation is called rules inference. In a FLC-based MPPT, the
the rule table is set, it would be very difficult to change it as the
inputs are usually an error E and a change in error DE. Since dP/
controller is in tracking mode. Hence, there is no reported work
dV vanishes at the MPP, both inputs can be calculated as follows:
on partial shading using standalone FLC. To overcome this draw-
PðkÞ  Pðk  1Þ PðkÞ  Pðk  1Þ back, the FLC is coupled with other controller, such as the Hopfield
EðKÞ ¼ or EðKÞ ¼ ð5Þ Neural Network (HNN) [57] or the traditional P&O controller [58].
VðkÞ  Vðk  1Þ IðkÞ  Iðk  1Þ
Adaptive FLC [54] also can handle partial shading by re-tuning its
and membership functions automatically.

DE ¼ EðkÞ  Eðk  1Þ ð6Þ

3.5. Particle swarm intelligence (PSO)
where P and V represent the power and voltage in P–V curve,
respectively. Once E and DE are computed, they are converted to Swarm intelligence is the collective behavior of decentralized,
the linguistic variables based on a membership function, similar self-organized systems, inspired by the natural or biological behav-
to the one in Fig. 9. Here five fuzzy variables are used: Negative ior. Particle swarm optimization (PSO) is an optimization method
Big (NB), Negative Small (NS), Zero (ZE), Positive Small (PS), and Po- that iteratively trying to improve a candidate solution (known as
sitive Big (PB). Such variable assignment is user dependent and are particle) with regard to a given measure of quality. These particles
typically designed based on the user’s experience for a specific con- move around in the search-space according to simple mathemati-
trol problem. The membership function is sometimes made less cal formulae, exploiting their position and velocity. Each particle’s
symmetric to give more importance to specific fuzzy levels as de- is influenced by its own best position and the best known positions
scribed in [49,51]. In the defuzzification process, the FLC output is in the search-space, guided by the neighborhoods best position.
defined linguistically in terms of voltage (DV), current (DI) or duty The general idea of particles movement in PSO is illustrated by
ratio (DD) of the power converter. It can be looked up in a rule table, Fig. 10.
such as Table 2. The linguistic variables assigned to the output for Based on Fig. 10, the individual particles position is defined by
the different combinations of E and DE are based on the power con-
xkþ1
i ¼ xki þ v kþ1
i ð8Þ
verter as well as the knowledge of the user. As an example, in Ta-
ble 2, when operating point is far to the left from the MPP means where vi represent the velocity component is calculated by
E is PB; if DE is ZE, duty ratio may need to be increased. Conse-
quently, DD will be PB. v kþ1
i ¼ wv ki þ c1 r 1 ðPbesti  xki Þ þ c2 r 2 ðGbest  xki Þ ð9Þ
Next, the linguistic output signal is converted to numerical val-
In (9), w is the inertia weight; c1 and c2 are the acceleration con-
ues by a process known as defuzzification. This can be achieved
stants, while Pbest and Gbest are the personal and global best posi-
using several methods; one example is the ‘‘centre of gravity’’,
tions, respectively. To start the optimization process, a solution
which utilize the following formula:
vector of duty cycles initialized and the algorithm transmits duty
Pn
lðDj Þ  ðDj Þ cycles to the power converter. These duty cycles (represented by
j¼1
D¼ Pn ð7Þ xi in (9)) serve as the initial particles in the first iteration. All parti-
j¼1 lðDj Þ
cles are heading towards their local best position Pbest. Among these
where j is the number of sampled duty cycle. In most work, FLC is particles, one of them is the global best Gbest. It gives the best fitness
used to design MPPT by manipulating different types of PV inputs. value. After calculating the velocity, which serves as a perturbation
Combination of FLC with other SC algorithms is also done to to the duty cycle, a new position of the duty cycle is found. Through
successive iteration all particles move towards global best position.

Rule table Table 2

Example of FLC rule table.
Crisp input Crisp output
E DE
NB NS ZE PS PB

Fuzzification Rules NB NB NB NB NS ZE
Inferences Defuzzification
NS NB NS NS ZE PS
ZE NB NS ZE PS PB
PS NB ZE PS PB PB
Fig. 8. Basic fuzzy logic structure. PB ZE PS PB PB PB
142 Z. Salam et al. / Applied Energy 107 (2013) 135–148

Table 3
Summary of FLC related work for MPPT.

Authors Ref. Control Converter type and application Remarks

variable
Mahmoud et al. [45] Voltage Cuk converter for stand-alone Converge in all solar irradiance with good efficiency and robustness. This method
application does not need any modification in membership functions and rules while testing with
different resistive loads
Veerachary et al. [46] Voltage Interleaved dual boost converter A feed-forward MPPT for dual boost converter. The reference voltage for the feed-
for stand-alone application forward loop, is obtained by an off-line trained ANN. Advantages: input and output
ripple cancellation and reduced electromagnetic interference
Khaehintung et al. [47] Voltage Boost converter for stand-alone FLC membership functions are made less symmetric to give more significance to the
application specific linguistic variables
Chung-Yuen et al. [48] Duty Boost converter for stand-alone FLC is operated in two modes: ‘‘coarse and fine’’. Has higher efficiency than traditional
cycle application hill climbing method
Simoes et al. [49] Duty Boost converter for stand-alone PV array power variation and duty ratio are used as inputs for the FLC. It does not
cycle application need any parameter information (no sensors needed), thus leads to fast convergence
Masoum et al. [50] Duty Buck converter for stand-alone Three inputs FLC MPPT: array current, power and duty cycle of converter. Very robust
cycle application. under variable temperature and solar irradiance
Chian-Song and Fuzzy [51] Duty Buck converter for stand-alone FLC-MPPT based on A Takagi–Sugeno (T–S) observer for state feedback to achieve
cycle application asymptotic control. Unlike traditional MPPT, the proposed controller directly drives
the system to MPP without searching the maximum power point and measuring
irradiance
Kottas et al. [52] Duty Boost converter for stand-alone A Fuzzy Cognitive Networks (FCN) with the voltage, current, temperature and solar
cycle application irradiance used as the nodes. The weights of the nodes are trained using a wide range
of different climatic condition to ensure it can used with various PV system. It exhibits
excellent tracking speed but at the expense of an additional switch and a sensor
Alajmi et al. [53] Duty Boost converter for micro-grid Application of FLC to improve the performance of the conventional HC method. The
cycle application. FLC is developed by translating the HC algorithm into 16 fuzzy rules to ensure better
tracking speed and efficient convergence
Patcharapraand Premrudeepreechacharn [54]
Duty cycle Boost converter for grid connected An adaptive FLC is proposed to facilitate the constant tuning of the membership
application functions and the rule based table in order to achieve optimum performance
Wu et al. [55] Duty Inverter for grid connected The scaling factor of both fuzzy inputs and output are automatically tuned to achieve
cycle application the better dynamic performance of MPPT
Purnama et al. [56] Duty Boost converter for DC–DC FLC is optimized by Hopfield Neural Network which is proven applicable in partial
cycle application shading. Convergence time is less than P&O and typical FLC controller
Subiyanto et al. [57] Voltage Boost converter for DC–DC Fuzzy P&O MPPT (FMPPT) is developed which is supported by offline tracking
application function to avoid local maxima. It shows better performance than traditional P&O
method
Syafaruddin et al. [59] Duty Buck-Boost converter for DC–DC A novel method is proposed which is a combination of ANN and polar coordinated
Cycle application FLC. ANN is offline trained under several conditions including partial shading. In FLC
linguistic variables are presented in angle and radius like polar coordinate. Such
fusion makes the controller capable of handling partial shading and converges at MPP
very quickly

As the particles approach the MPP, they get closer to the Gbest posi- global position in sequential iteration. Such feature is very useful
tion. Correspondingly, the Pbest factor and Gbest factor in velocity to track MPP in partial shading condition [60,61,63,65]. Further-
term moves towards zero. Eventually a zero velocity is achieved more, PSO is normally used in conjunction with other MPPT meth-
and the duty cycle remains almost unchanged. Under this condition, ods to improve the tracking efficiency. A summary of various PSO
the PV system reaches at MPP. work are shown in Table 4.
Interestingly, the PSO algorithm will not be trapped in a local
maximum because some particles would always look the better

3.6. Ant colony optimization (ACO)

The ant colony optimization (ACO) is another swarm optimiza-

tion technique based on the behavior of ants seeking a path be-
tween their colony and a source of food. The original idea has
since diversified to solve a wider class of numerical problems
[66]. Usually, ACO is applied for problems that exhibit frequent dy-
namic changes. The main advantage of ACO is the ability of the
algorithm to run continuously while adapting to changes in real
time. In the case of MPPT, the PV power serves as the target func-
tion and the duty cycle is the control variable. In the beginning,
each ant is generated in different starting regions randomly. Every
region has its own attraction strength to all ants. Due to the differ-
ence in attraction strength, the ants make transitions from lower to
higher strength. Depending on the region strength, the transition
probability of ants is calculated. In the next iteration, ants move to-
wards a stronger strength position. Then again the attraction
strength of every region is calculated. By successive iteration ants
Fig. 10. Particles movement in PSO. move towards the optimized point, i.e. the MPP.
 Z. Salam et al. / Applied Energy 107 (2013) 135–148 143

Table 4
Summary of PSO related work for MPPT.

Authors Ref. Technique Remarks

Ishaque [60] Improved MPPT with PSO PSO is applied for MPPT specially in partial shading condition and showed that tracking
et al. efficiency is 99.5%
Ishaque [61] Deterministic PSO (DPSO) with direct duty cycle The random element of the velocity is replaced by constants. Faster tracking speed, zero
et al. control oscillations at MPP. Able to locate MPP for any environmental variations including partial
shading and large irradiance fluctuations
Kuo et al. [62] Orthogonal PSO (OPSO) + Petri net (PN) The OPSO has higher searching efficiency. Local minima problem is avoided. Tracking efficiency
is better when compared to conventional P&O
Miyatake [63] PSO based MPPT control of several PV arrays with Simple and reduced cost for data acquisition system. Able to cater for partial shading
et al. one pair of current and voltage sensors conditions. Response is almost independent partial shading pattern and search space
dimensions
Soon [64] PSO with Inverse Barrier constraint model Able to determine the unknown parameters of PV module dynamically, thus eliminates the
et al. need to assume the ideality factor. It incorporates the effects of temperature for determining
the un-known model parameters. Accurate results have been obtained despite changes in
temperature
Fu and [65] Complex PSO (CPSO) CPSO integrates the global search ability of the PSO and the local search ability of the Complex
Tong method. The algorithm effectively solves problems of local search efficiency and premature
convergence in the PSO

Considering an N-dimensional search space where M number of ations of chromosomes. This new generation is evaluated through
colonies has N number of ants each, the transition probabilities of fitness function and is assigned a new fitness value. After consecu-
the ants are defined as follows: tive iteration chromosome with the highest fitness value is chosen
! as the optimized parameter for MPP.
saj gbij j GA is usually used to optimize other algorithms like ANN or FL.
Pij ¼ Pm a b when gi > 0 elsewhere Pij ¼ 0 ð10Þ
s
j¼1 j ij g In [67] FLC is optimized using GA where GA chooses optimally and
simultaneously both membership functions and control rules for
where sj is the attraction strength in region j, gij refers to the differ- the FLC. GA-FLC-based MPPT is better than the ones obtained with
ence between target function searched from the region i to j. The classical P&O controller, since the response time in the transitional
variables a and b are the influence factors. Attraction strength can state is shortened and the fluctuations in the steady state are con-
be calculated using siderably reduced. In another work [68], GA is used to optimize
X
m ANN. GA based offline trained ANN provides the reference voltage
sj ðt þ 1Þ ¼ qsj ðtÞ þ Dskj and Dskj ¼ Q=Lkj ð11Þ corresponding to the maximum power for any environmental
k¼1 changes.
where L is the variation of target function, q is a constant [0, 1] and
Q is the pheromone density released from ant colonies. At particular 3.8. Differential evolution (DE)
duty cycles, the power is measured and compared to (11). Based on
the difference between the compared values, the attraction strength DE is a stochastic, optimizer based evolutionary algorithm. It
on ants increase and consequently the ants change the region to- optimizes a problem by maintaining and creating a population of
wards the probable optimum zone. candidate solutions according to its simple formulae, and then
Despite its potential, ACO has not been directly used as the keeping whichever candidate solution that exhibits the best fit-
MPPT controller. Rather, it is utilized in conjunction with other ness. As with other evolutionary techniques, DE relies on initial
SC techniques as an optimization tool. In [66], ACO is proven be random population generation, which is then, improved using
highly efficient in optimizing the adaptive parameters of FLC selection, mutation, and crossover. It has three main advantages
through reducing its steady state error and settling time. By opti- (1) able to locate the true and accurate global optimum regardless
mizing these parameters, the efficiency of the FLC-based MPPT is of the initial parameter values (2) fast convergence (3) utilizing
improved. The optimized FLC controller is capable of tracking the few control parameters so that it will be easy to use [70].
MPP under rapidly changing environmental condition. In terms of optimization procedure, DE is similar to GA. How-
ever, unlike GA, which relies on crossover, DE primarily utilizes
the mutation operation (i.e. difference vector) as a search and
3.7. Genetic algorithm (GA)
selection mechanism to direct the search toward the prospective
regions in the search space. Fig. 11 depicts the basic concept of DE.
GA is categorized under the evolutionary algorithm. Specifi-
In the MPPT problem, the fitness function is the PV equation,
cally, it is a problem-solving techniques based on principles of bio-
while the duty cycle, Xi(j) is defined as the population. The latter
logical evolution. In the process, some inputs are assigned as
is randomly initialized within the initial parameter bounds, i.e.
chromosomes, which are recombined or mutated and then tested
to fulfill a predefined fitness function. Since the objective of the X i ðjÞ ¼ X iL þ randð0; 1Þ  ½X iH  X iL  ð12Þ
evolution is to create a better species than its predecessor, GA finds
the best solution by a random combination of different genes. In each generation, the individuals of the current population be-
The initial set of chromosome is defined as the searching come target vectors. For each target vector, the mutation operation
parameters of the optimization problem. In case of MPPT, such produces a mutant vector, by adding the weighted difference be-
parameters can be either voltage or duty cycle. Chromosome can tween two randomly chosen vectors to a third vector, as presented
be defined in a real or binary coded numbers, while the PV equa- below:
tion is considered the fitness function. It is important to decide X i;Gþ1 ¼ X r1:GX þ FðX r2;G  X r3;G Þ ð13Þ
the length of the chromosome because a larger population requires
less time to converge but such generation increases processing Then the crossover operation generates a new vector, called trial
time. Next, the algorithm utilizes the crossover and mutation oper- vector Ui,G, by mixing the parameters of the mutant vector Vi,G with
ation to change the DNA of the chromosomes, creating new gener- those of the target vector Xi,G, Every trial vectors then applied in the
144 Z. Salam et al. / Applied Energy 107 (2013) 135–148

Fig. 11. Basic flow of DE algorithm [73].

fitness function. If the trial vector obtains a better fitness value than mated by the PSO method. The remaining 10 nodes refer to the 10
the target vector, then the trial vector replaces the target vector in points, estimated by the IC method. Finally, the symbol S denotes
the next generation. The process is repeated through generations the estimated location after fusion process. Nodes that matches
until the stopping condition is reached, usually a good fitness value. from both sides are given a state value 1, otherwise are given 0.
In literature DE is used both for dynamic PV modeling [69] and For all selected nodes an previously offline trained output vector
designing MPPT [70]. In [69] the PV module parameters are deter- is assigned. To decide which output value is best for the input
mined dynamically when controller is in operation and method is nodes, conditional probability is applied. With successive iteration,
justified using three different modules from different manufac- the best suited value is given as the final output, i.e. MPP. Despite
turer. Authors in [70] proposed a novel method where DE can track its effectiveness, the complexity of the method is the major con-
MPP faster than P&O under both constant and sudden variations in straint for implementing it using a low-cost microprocessor.
partial shading conditions. The algorithm is less complex than In [71] the fusion of two type of PSO algorithms and Incremen-
other SC-based MPPT, such as ANN and FLC. tal Conductance is done using Bayesian network. This novel meth-
od converges to the MPP with great speed and efficiency under
both normal and partial shading condition. However, the solution
3.9. Bayesian network
is highly complex and needs powerful sensors for accuracy.
A Bayesian network is a probabilistic graphical modeling meth-
od which performs mapping based on conditional probability of 4. Discussion
random variables. In multidimensional MPPT problems based on
the combination of two or more SC algorithms, this can be an The works using different SC MPPT techniques described above
efficient method. The structure of a Bayesian network for PSO-IC are carried out in non-standardized situations. Furthermore, every
combination is shown in Fig. 12. The network has N = 20 nodes, researcher tested their algorithm on different PV system (ratings,
represented by [A1, A2, . . . , A20]. They are divided into two sectors, technologies, size, etc.) and under different environmental condi-
left and right observation nodes, respectively. Ten nodes tions. There is very little uniformity among the various works
[A1, A2, . . . , A10] correspond to 10 points of the I–V curve to be esti- and in that sense, it is difficult to benchmark their capabilities
 Z. Salam et al. / Applied Energy 107 (2013) 135–148 145

[37] simulate two MPPT controllers: (1) a stand-alone FLC and

S (2) an ANN with P&O. The stand-alone FLC tracks the MPP within
27 ms, with its steady-state MPP fluctuation kept below 2%. With
the combination of ANN and P&O, the tracking speed is reduced
to 2 ms, while the fluctuation is shown to be nil. In [38] the ANN
is used to improve the IC; it tracks the MPP within 40 ms, with
and additional response time 10 ms if the environment changes
rapidly. Furthermore, an ANN scheme with GA optimization [68]
A1 A10 A11 A20 shows steady state fluctuation of only 1.45%. In [67] the GA opti-
mized FLC presents a tracking speed 715 ms with fluctuations
Left observation Right observation
nodes nodes 0.02%. The Hopfield Neural Network (HPNN) [57] (a modified ver-
sion of ANN), coupled with FLC can track MPP in 10–40 ms with
Fig. 12. Bayesian network for MPPT information fusion. very low fluctuation, i.e. 0.002%. In another work on standalone
FLC [47], the MPP is tracked in 150 ms. Furthermore, authors in
[53] proposed an FLC integrated with HC, which improves the
based on a single performance index. Despite these shortcomings, tracking time to 70 ms while an adaptive fuzzy in [54] tracks the
this section attempts to summarize the performance of SC MPPT MPP in 3 ms. The latter is one of the fastest response achievable
techniques based on several logical criteria, which is thought to by any SC technique to date. It has to be noted that, however, all
be relevant. These are PV array dependency, convergence time, the above-mentioned methods, none are implemented in hard-
the ability to handle partial shading conditions, algorithm com- ware, and hence their practical feasibility could not be ascertained.
plexity and hardware/practical implementation. Only in [41], ANN based FLC is verified by both simulation and
hardware. The scheme converges within 1 s.
Simulation work for intelligent predictor is carried out in [43];
4.1. PV array dependency
it successfully tracks the MPP in 5–28 ms, depending on the level
of irradiance. The MPP fluctuation kept well below 4%. When
Ideally, an MPPT should not be array dependent; the algorithm
implemented in hardware, the tracking time increased to 130 ms,
ought to perform equally well for all cell technologies. From Ta-
but the fluctuation is decreased to 2.2%. Chaotic search [44] shows
ble 5, it can be observed that majority of the SC algorithms are
a good performance by tracking MPP in 0.06–1.47 s, while fluctua-
independent of PV modules, with the exception of ANN and FLC
tion rate is reduced to almost zero. Swarm intelligence offers MPPT
MPPT. In the case of ANN, once the controller is trained, its nodes
with great accuracy. However, it requires longer time than other
are weighted based on the characteristics of a particular module.
techniques; for example, DPSO [61] takes 2.7 s while OPSO [62]
Consequently, it may not yield accurate results if applied to a dif-
and standalone PSO [63] require 2 s to track MPP. The MPP fluctu-
ferent module technology or products from other manufacturers.
ation in every case is below 0.001%. On the other hand, ACO with
To cope with the variations, the ANN need to be re-trained to the
FLC [66] requires 7 s to converge with very high precision. Finally,
new conditions and re-installed.
the work on DE [70] suggests that the MPP is tracked in less than
For FLC, the designs of the rule table and membership functions
300 ms with very low fluctuation.
largely depend on the individual designer’s understanding of the
Furthermore, SC techniques prove to be highly efficient in con-
PV system behavior and how it interprets the response of the MPPT
verging to the specified MPP under difficult conditions, particularly
[72]. In that sense, it can be considered as array dependence. Any
for partial shading [62] and rapid fluctuations of irradiance [57]. It
alteration in the arrays/modules or environmental conditions re-
has also been shown that the combination of two methods (SC
quires a redesign of the FLC rule table and membership functions.
with SC or SC with conventional MPPT) exhibit improved perfor-
One approach to mitigate this shortcoming is to apply the adaptive
mance compared to standalone techniques. Furthermore, utilizing
FLC, as proposed by [54]. In this technique, the membership func-
optimization algorithms (such as ANN) to optimize certain param-
tions of the FLC are altered dynamically in accordance to the mod-
eters for a particular MPPT appear to improve the tracking time
ule technology or the changes in atmospheric conditions.
and precision significantly.
In general, SC tends to be slower than the conventional meth-
4.2. Convergence speed ods. This is to be expected because the SC algorithm is more com-
plex. In many cases, especially the one requires the use random
It is difficult to rank the convergence speed of various SC meth- numbers, substantially higher amount of iterations are required
ods in a general way due to several reasons: (1) in the literature, for the results to converge towards the desirable fitness values.
MPPT schemes are implemented on different system size, module However, as the computing power and speed of processors in-
technologies (crystalline silicon, thin film, etc.), manufacturers and creases, the convergence time of SC MPPT is expected to improve
array configurations, (2) the computer specifications (RAM, clock accordingly.
speed, etc.) used to simulate the algorithm are different, (3) the
PV systems are installed in different atmospheric and geographical 4.3. Handling partial shading conditions
location, (4) the programming efficiency, i.e. the codes, may or may
not be optimized, and (5) the hardware (prototype) used to imple- Due to its inherent structure, the prediction technique cannot
ment the algorithm is different to each other. Besides that, each re- cater for partial shading conditions. It is extremely difficult to find
searcher defines his own preferable parameters and condition to a suitable predictor function that resembles the partial shading
test the algorithm. Based on these factors, a fair benchmarking curve. Even if such function exists, it would impossible to consis-
for the computational speed is not feasible. Notwithstanding these tently match the varying global peak to the peak of the predictor
difficulties, a summary of the convergence for various MPPT tech- function. For ANN, handling partial shading is also impractical, un-
niques is presented in Table 5. This has to be viewed in context to less the shading is predictable; for example fixed spots shadows
the unique implementation conditions as described below. from building structures or trees. In such cases, the ANN can be
In fairness, a trade-off between the response time and operating trained to adapt for such conditions. However, with regard to
point fluctuations around MPP need to be considered. Authors in environmental uncertainties, i.e. shading due to clouds, training
146 Z. Salam et al. / Applied Energy 107 (2013) 135–148

Table 5
Comparison of soft computing techniques.

Methods PV array Convergence Partial shading Hardware Algorithm

dependency implementation complexity
Speed Published Possible? Published
work work
Fuzzy logic Yes Fast [47,53,54] Standalone – No Coupled – [52–54] Easy Moderate
Yes
ANN Yes Fast [37,38,41] Standalone – No Coupled – [52,53,57] Difficult Moderate
Yes
DE No Fast/ [70] Yes [70] Easy Moderate
Varies
GA No Fast [67,68] No Easy Difficult
PSO No Fast [61,62] Yes [61,62] Moderate Moderate
ACO No Fast [66] Yes [66] Moderate Moderate
Chaotic search No Fast [44] Yes [44] Difficult Very difficult
Bayesian network No Very fast [71] Yes Difficult Difficult
Non-linear No Fast [43] No [43] Easy Easy
predictor

the ANN is not possible due to the random nature of their 4.5. Hardware implementation issues
occurrences.
For the standalone FLC, tracking the varying global peak is not a At present, most of the SC MPPT techniques described in lit-
straightforward task. This is because its membership function and erature are demonstrated using simulation. Despite this fact,
control variables are static, while partial shading incidences can be there are several authors include experimental validation of their
highly dynamic. One attempt to improve the dynamism of the work. For ANN, the major implementation issue is the training
membership function is by introducing an adaptive capability in that requires high cost microcontroller, as noted in [41]. For
the FLC [54]. Other efforts include combining FLC with other MPPT FLC, it is possible to use much simpler and less expensive con-
techniques, such as ANN [37], cognitive network (CN) [52], HC [53] trollers; in [47] and [48] the low cost RISC microcontroller,
and HFNN [57]; these solutions appear to be workable in partial namely the PIC 16F872 and PIC16C74 are used implement the
shading conditions. FLC algorithm, respectively. In [51] the Infineon TriCore
The search based SC techniques are naturally suitable to handle TC1796, which is also low cost processor, is used to code the
partial shading. This is primarily due to their ability to scan the en- Fuzzy and HC algorithms. The work for OPSO [62] utilized
tire P–V curve and subsequently discriminate between the global the16-bit PIC microcontroller.
and local peaks. The challenge, however, is to optimize the search On the other hand, several authors, for example [51,61] have
time so that the MPPT dynamic response can be improved. The implemented the FLC using the more expensive Digital Signal
chaotic search is proven highly efficient under partial shading, as Processor Controller (DSP). Similar is the work carried out for
exhibited by [44]. Equally, swarm intelligence methods, for exam- DPSO [60]. The objective of using such high performance chip
ple DPSO [61], OPSO [62] are also effective. There are claims that is to achieve better accuracy and faster computational time. In
DE [70] is being applied for partial shading conditions, but the most cases, the DSP is embedded in a user-friendly development
work described by this reference does not provide sufficient proof platform, such as DSpace [60,61]. The programming and debug-
of its performance. Nevertheless, the potential of DE to handle par- ging facilities of DSpace provide an ideal tool for rapid prototyp-
tial shading should not be ruled-out. ing of the MPPT algorithm into the DSP hardware. Once the
MPPT codes are correctly programmed, an optimized DSP system
4.4. Algorithm complexity with only the necessary peripherals can be designed for mass
production. For improved precisions, several authors imple-
Generally, SC MPPTs are more complex than the conventional mented the MPPT on the Field Programmable Gate Array (FPGA)
methods. Even among the SC techniques, their complexities vary chip. Such work is carried out by [67] for the GA optimized FLC
greatly. For example, ANN is highly complicated because the mul- and in [43] for the intelligent predictor algorithm. As chaotic
ti-layered network that need to be trained in order to correctly re- search [44] is complex in algorithm it is implemented in RT-
spond to different irradiance, temperature and other PV inputs. If LAB [44] platform.
the training is sufficient and correctly done, the accuracy of the
MPPT can be very high. On the other hand, a standalone FLC algo-
rithm is considerably simpler but its tracking performance is much 5. Conclusion
inferior to a well-trained ANN. To improve its performance, FLC is
coupled with other techniques like HC [52], HFNN [57], CN [51] or This work gathers and summarizes the current technology and
GA [67]. However, these combinations result in longer algorithm status of SC MPPT, which are scattered in various literatures. It also
and hence the increase in complexities. provides an evaluation on their performance based on several
Based on the algorithm structure, the chaotic search [44] is important criteria. Although it is very difficult to benchmark the
found most complex, while the predictor technique [43] is the sim- merit of each method based on a generalized performance index,
plest one among all SC techniques. Others methods exhibit com- a detailed discussion on the PV array dependency, convergence,
plexities somewhere in between the two. However, there is no partial shading conditions, algorithm complexity and practical
exact scale to rank the complexities these techniques. It is worth implementation are given. With the rapid expansion in computing
noting that swarm intelligence methods, like DPSO [61], OPSO capability, the interest in SC MPPT appears to be growing. It is
[62] have a simple structure and are more adaptable to hardware hoped that this paper will be a valuable source of information for
implementation. researchers, power converter designers and PV practitioners.
 Z. Salam et al. / Applied Energy 107 (2013) 135–148 147

Acknowledgment [27] De Carvalho PCM, Pontes RST, Oliviera Jr DS, Riffel DB, De Oliviera RGV,
Mesquita SB. Control method of a photovoltaic powered reverse osmosis plant
without batteries based on maximum power point tracking. In: Proc IEEE PES
The authors would like to thank Universiti Teknologi Malaysia transmiss distrib conf & expo. Latin, America; 2004. p. 137–42.
and the Ministry of Higher Education Malaysia for providing the [28] Rtiz-Rivera EI, Peng F. A novel method to estimate the maximum power for a
photovoltaic inverter system. In: Proc 35th annu IEEE power electron spec
facilities and financial support (Research University Grant No.
conf; 2004. p. 2065–9.
2423.00G40) to conduct this research. [29] Zhang M, Wu J, Zhao H. The application of slide technology in PV maximum
power point tracking system. In: Proc 5th world congr intell contr automat;
2004. P. 5591–4.
[30] Kim IS. Sliding mode controller for the single-phase grid-connected
References photovoltaic system. Appl Energy 2006;83:1101–15.
[31] Ioulia T, Purvins PA. Mathematical and graphical approach for maximum
power point modeling. Appl Energy 2012;91:59–66.
[1] Cacciato M, Consoli A, Attanasio R, Gennaro F. Soft-switching converter with
[32] Amrouche B, Guessoum A, Belhamel M. A simple behavioural model for solar
HF transformer for grid-connected photovoltaic systems. IEEE Trans Ind
module electric characteristics based on the first order system step response
Electron 2010;57(5):1678–86.
for MPPT study and comparison. Appl Energy 2012;91:395–404.
[2] Ishaque K, Salm Z, Taheri H, Syafaruddin. Modeling and simulation of
[33] Ishaque K, Salam Z, Taheri H. A comprehensive MATLAB Simulink PV system
photovoltaic (PV) system during partial shading based on a two-diode
simulator with partial shading capability based on two-diode mode. Sol
model. Simul Modell Practice Theor 2011;19:1613–26.
Energy 2011;85:2217–27.
[3] Bruendlinger R, Bletterie B, Milde M, Oldenkamp H. Maximum power point
[34] Kobayashi K. Matsuo H, Sekine Y. A novel optimum operating point tracker of
tracking performance under partially shaded PV array conditions. In: Proc. 21st
the solar cell power supply system. In: 35th Power electron spec conf, vol. 3;
EUPVSEC; 2006. p. 2157–60.
2004. p. 2147–51.
[4] Esram T, Kimball JW, Krein PT, Chapman PL, Midya P. Dynamic maximum
[35] Di Piazza MC, Vitale G. Photovoltaic field emulation including dynamic and
power point tracking of photovoltaic arrays using ripple correlation control.
partial shadow conditions. Appl Energy 2012;87:814–23.
IEEE Trans Power Electron 2006;21:1282–91.
[36] Veerachary M, Yadaiah N. ANN based peak power tracking for PV supplied DC
[5] Parida B, Iniyan S, Goic R. A review of solar photovoltaic technologies. Renew
motors. Sol Energy 2000;69:343–50.
Sust Energy Rev 2011;15:1625–36.
[37] Alabedin AMZ, El-Saadany EF, Salama MMA. Maximum power point tracking
[6] Han X, Wang Y, Zhu L. Electrical and thermal performance of silicon
for photovoltaic systems using fuzzy logic and artificial neural networks. IEEE
concentrator solar cells immersed in dielectric liquids. Appl Energy
Power Energy Soc Gen Meet 2011:1–9.
2011;88(12):4481–9.
[38] Jinbang X, Shen A, Yang C, Rao W, Yang X. ANN based on IncCond algorithm for
[7] Krebs FC. Fabrication and processing of polymer solar cells: a review of
MPP tracker. In: Bio-inspired computing: theories and applications (BIC-TA),
printing and coating techniques. Sol Energy Mater Sol Cells
2011. Sixth international conference on; 2011. p. 129–34.
2009;93(4):394–412.
[39] Islam MA, Kabir MA. Neural network based maximum power point tracking of
[8] Jainand S, Agarwal V. A new algorithm for rapid tracking of approximate
photovoltaic arrays. In: TENCON IEEE region 10 conf; 2011. p. 79–82.
maximum power point in photovoltaic systems. IEEE Power Electron Lett
[40] Jie L, Ziran C. Research on the MPPT algorithms of photovoltaic system based
2004;2(1):16–9.
on PV neural network. In: Chinese control and decision conference; 2011.
[9] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimization of perturb and
[41] Veerachary M, Senjyu T, Uezato K. Neural-network-based maximum-
observe maximum power point tracking method. IEEE Trans Power Electron
power-point tracking of coupled-inductor interleaved-boost-converter-
2005;20(4):963–73.
supplied PV system using fuzzy controller. IEEE Trans Ind Electron
[10] Kuo YC, Liang TJ, Chen JF. Novel maximum-power-point-tracking controller for
2003;50(4):749–58.
photovoltaic energy conversion system. IEEE Trans Ind Electron
[42] Sheng PF. Min CRA new algorithm to photovoltaic power point tracking
2001;48(3):594–601.
problems with quadratic maximization. IEEE Trans Energy Convers
[11] Kobayashi K, Takano I, Sawada Y. A study on a two stage maximum power
2010;25:262–4.
point tracking control of a photovoltaic system under partially shaded
[43] Sheng PF, Min CR, Hong KS, Sheng LT. Performance evaluation of parabolic
irradiance conditions. IEEE Power Eng Soc Gen Meet 2003:2612–7.
prediction to maximum power point tracking for PV array. IEEE Trans Sust
[12] Lin CH, Huang CH, Du YC, Chen JL. Maximum photovoltaic power tracking for
Energy 2011;2:60–8.
the PV array using the fractional-order incremental conductance method. Appl
[44] Zhou L, Chen Y, Guo K, Jia F. New approach for MPPT control of photovoltaic
Energy 2011;88:4840–7.
system with mutative-scale dual-carrier chaotic search. IEEE Trans Power
[13] Koutroulis E, Kalaitzakis K, Voulgaris NC. Development of a microcontroller-
Electron 2011;26:1038–48.
based photovoltaic maximum power point tracking control system. IEEE Trans
[45] Mahmoud AMA, Mashaly HM, Kandil SA, El-Khashab, Nashed HMN. Fuzzy
Power Electron 2001;16(21):46–54.
logic implementation for photovoltaic maximum power tracking. In: Indust
[14] Xiao W, Dunford WG. A modified adaptive hill climbing MPPT method for
Electron Soc, IECON 26th annu conf IEEE, vol. 1; 2000. p. 735–40.
photovoltaic power systems. In: Proc 35th annu IEEE power electron spec
[46] Veerachary M, Senjyu T, Uezato K. Feed forward maximum power point
conf; 2004. p. 1957–63.
tracking of PV systems using fuzzy logic controller. IEEE Trans Aerospase
[15] Masoum MAS, Dehbonei H, Fuchs EF. Theoretical and experimental analyses of
Electron Syst 2002;38(3):969–81.
photovoltaic systems with voltage and current-based maximum power-point
[47] Khaehintung N, Pramotung K, Tuvirat1 B, Sirisuk P. RISC-microcontroller built-
tracking. IEEE Trans Energy Convers 2002;17(4):514–22.
in fuzzy logic controller of maximum power point tracking for solar-powered
[16] Noguchi T, Togashi S, Nakamoto R. Short-current pulse based adaptive
light-flasher applications. In: Indust Electron Soc, IECON 30th annu conf IEEE,
maximum power point tracking for photovoltaic power generation system.
vol. 3; 2004. p. 2673–8.
In: IEEE Int Symp Indust Electron; 2000. p. 157–62.
[48] Chung-Yuen W, Duk-Heon K, Chen KS. A new maximum power point tracker of
[17] Hart GW, Branz HM, Cox CH. Experimental tests of open-loop maximum-
photovoltaic arrays using fuzzy controller. In: Power electron specialists 25th
power-point tracking techniques. Sol Cells 1984;13:185–95.
annu IEEE conf, vol. 1; 1994. p. 396–403.
[18] Noh HJ, Lee DY, Hyun DS. An improved MPPT converter with current
[49] Simoes MG, Franceschetti NN, Friedhofer M. A fuzzy logic based photovoltaic
compensation method for small scaled PV-applications. In: Proc 28th annu
peak power tracking control. In: Proce of Indust Electron IEEE Int Symp, vol. 1;
conf Indust Electron Soc; 2002. p. 1113–8.
1998. p. 300–5.
[19] Midya P, Krein PT, Turnbull RJ, Reppa R, Kimball J. Dynamic maximum power
[50] Masoum MAS, Sarvi M. Design, simulation and implementation of a fuzzy-
point tracker for photovoltaic applications. In: Proc 27th annu IEEE power
based maximum power point tracker under variable irradiance and
electron spec conf; 1996. p. 1710–6.
temperature conditions. Iran J Sci Technol 2005;29:27–32.
[20] Arcidiacono V, Corsi S, Lambri L. Maximum power point tracker for
[51] Chian-Song C. T–S fuzzy maximum power point tracking control
photovoltaic power plants. In: Proc IEEE photovoltaic spec conf; 1982. p.
of solar power generation systems. IEEE Trans Energy Convers
507–12.
2010;25:1123–32.
[21] Bodur M, Ermis M. Maximum power point tracking for low power photovoltaic
[52] Kottas TL, Boutalis YS, Karlis AD. New maximum power point tracker for PV
solar panels. In: Proc 7th Mediterranean electrotechnical conf; 1994. p. 758–
arrays using fuzzy controller in close cooperation with fuzzy cognitive
61.
networks. IEEE Trans Energy Convers 2006;21:793–803.
[22] Kitano T, Matsui M, Xu DH. Power sensor-less MPPT control scheme utilizing
[53] Alajmi BN, Ahmed KH, Finney SJ, Williams BW. Fuzzy-logic-control approach
power balance at DC link-system design to ensure stability and response. In:
of a modified Hill–Climbing method for maximum power point in microgrid
Proc 27th annual conf IEEE Indust Electron Soc; 2001. p. 1309–14.
standalone photovoltaic system. IEEE Trans Power Electron 2011;26:1022–30.
[23] Shmilovitz D. On the control of photovoltaic maximum power point tracker via
[54] Patcharaprakiti N, Premrudeepreechacharn S. Maximum power point tracking
output parameters. IEE Proc Electr Power Appl 2005:239–48.
using adaptive fuzzy logic control for grid-connected photovoltaic system.
[24] Sugimoto H, Dong H. A new scheme for maximum photovoltaic power
Power Eng Soc Winter Meet IEEE 2002;1:372–7.
tracking control. In: Proc IEEE power conv conf; 1997. p. 691–6.
[55] Wu TF, Yang CH, Chen YK, Liu R. Photovoltaic inverter systems with self-tuning
[25] Pan CT, Chen JY, Chu CP, Huang YS. A fast maximum power point tracker for
fuzzy control based on an experimental planning method. In: 34th indust app
photovoltaic power systems. In: Proc 25th annu conf IEEE ind electron soc;
annu meet conf IEEE, vol. 3; 1999. p. 1887–94.
1999. p. 390–3.
[56] Purnama I, Lo YK, Chiu HJ. A fuzzy control maximum power point tracking
[26] Solodovnik EV, Liu S, Dougal RA. Power controller design for maximum power
photovoltaic system. In: Fuzzy systems (FUZZ) IEEE int conf; 2011. p. 2432–9.
tracking in solar installations. IEEE Trans Power Electron 2004;19(5):1295–304.
148 Z. Salam et al. / Applied Energy 107 (2013) 135–148

[57] Subiyanto, Mohamed A, Hopfield Shareef H. Neural network optimized fuzzy [66] Besheer AH, Adly M. Ant colony system based PI maximum power point
logic controller for maximum power point tracking in a photovoltaic system. tracking for standalone photovoltaic system. In: Indust tech IEEE international
Int J Photoenergy 2012:6–13. conference; 2012. p. 693–8.
[58] Chin CS, Neelakantan P, Yoong HP, Yang SS, Teo KTK. Maximum power point [67] Messai A, Mellit A, Guessoum A, Kalogirou SA. Maximum power point tracking
tracking for PV array under partially shaded condition. In: 3rd int conf on using a GA optimized fuzzy logic controller and its FPGA implementation. Sol
comput intell, comm sys and net; 2011. p. 72–7. Energy 2011;85:265–77.
[59] Syafaruddin, Karatepe E, Hiyama T. Artificial neural network-polar coordinated [68] Ramaprabha R, Gothandaraman V, Kanimozhi K, Divya R, Mathur BL.
fuzzy controller based maximum power point tracking control under partially Maximum power point tracking using GA-optimized artificial neural
shaded conditions. IET Renew Power Gener 2009;3(2):239–53. network for Solar PV system. In: 1st Int conf on elec energy sys IEEE; 2011.
[60] Ishaque K, Salam Z, Shamsudin A, Amjad M. A direct control based maximum p. 264–8.
power point tracking method for photovoltaic system under partial shading [69] Ishaque K, Salam Z. An improved modeling method to determine the model
conditions using particle swarm optimization algorithm. Appl Energy parameters of photovoltaic (PV) modules using differential evolution (DE). Sol
2012;99:414–22. Energy 2011;85:2349–59.
[61] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved particle swarm [70] Taheri H, Salam Z, Ishaque K, Syafaruddin G. A novel maximum power point
optimization (PSO)-based MPPT for PV with reduced steady-state oscillation. tracking control of photovoltaic system under partial and rapidly fluctuating
IEEE Trans Power Electron 2012;27:3627–38. shadow conditions using differential evolution. IEEE Symp Ind Electron Appl
[62] Kuo JL, Chao KL, Lee LS. Dual mechatronic MPPT controllers with PN and OPSO 2010:82–7.
control algorithms for the rotatable solar panel in PHEV system. IEEE Trans Ind [71] Keyrouz F, Georges S. Efficient multidimensional maximum power point
Electron 2010;57:678–89. tracking using Bayesian fusion. In: Elec power and energy conv syst 2nd int
[63] Miyatake M, Toriumi F, Endo T, Fujii N. Maximum power point tracking of conf; 2011. p. 1–5.
multiple photovoltaic arrays: a PSO approach. IEEE Trans Aerospace Electron [72] Esram T, Chapman PL. Comparison of photovoltaic array maximum power
Sys 2011;47:367–80. point tracking techniques. IEEE Trans Energy Convers 2007;22(2):439–49.
[64] Soon JJ, Low KS. Photovoltaic model identification using particle swarm [73] Da Costa WT, Fardin JF, Simonetti DSL, Neto VBM. Identification of
optimization with inverse barrier constraint. IEEE Trans Power Electron photovoltaic model parameters by differential evolution. In: Indust tech IEEE
2012;27:3975–83. conf on Dig Obj Ident; 2010. p. 931–6.
[65] Fu Q, Tong N. A complex-method-based PSO algorithm for the maximum
power point tracking in photovoltaic system. In: 2nd Int conf on InfoTech and
com sci; 2010. p. 134–7.
 Renewable and Sustainable Energy Reviews 38 (2014) 581–593

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

MATLAB based modeling to study the performance of different

MPPT techniques used for solar PV system under various
operating conditions
Parimita Mohanty a,n, G. Bhuvaneswari b, R. Balasubramanian b, Navdeep Kaur Dhaliwal b
a
The Energy and Resources Institute (TERI), New Delhi, India
b
Indian Institute of Technology, Delhi, New Delhi, India

art ic l e i nf o a b s t r a c t

Article history: Although solar photovoltaic technology is one of the matured technologies, its initial high cost and low
Received 27 March 2013 efficiency have not made it fully attractive as an alternative option for electricity users. Hence it is
Received in revised form very critical to utilize the maximum available solar power of the array and to operate the PV array at
11 May 2014
its highest energy conversion output. For this, the solar PV generating system has to operate at the
Accepted 12 June 2014
maximum power output point. Since the maximum power point varies with radiation and temperature,
Available online 15 July 2014
it is difficult to maintain optimum power operation at all radiation levels. Over the years, many MPPT
Keywords: techniques have been advocated, developed and implemented. These methods vary in several aspects
Maximum power point tracking (MPPT) such as complexity, required number of sensors, convergence speed, cost, range of effectiveness, ease of
Solar photovoltaic
hardware implementation etc. Although different methods have been developed by different research
Perturb and observe
groups, very little literature is available, where different MPPT techniques/methods are compared in
Incremental conductance
Performance assessment terms of energy capture, conversion efficiency, response time and reliability. This paper compares the
PV converter performance of different MPPT methods that are currently used in a solar PV system and also advocates a
new MPPT technique which offers better performance than the existing ones. The methodology adopted
for analysis is as follows: Initially, a MATLAB based solar PV array model is first developed and validated; then,
different MPPT techniques are employed on this PV array under varying temperature and insolation conditions
to study the effectiveness of the particular MPPT technique under consideration.
& 2014 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
2. Overall system configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
3. MPPT control algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
3.1. MPPT control with constant voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
3.2. MPPT control with perturb and observe method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
3.3. MPPT control with variable step perturb and observe method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
3.4. MPPT control with incremental conductance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
3.5. MPPT control with fractional short circuit current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585
4. Modeling and simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585
4.1. Modeling and simulation of solar PV array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585
4.2. Modeling and simulation of 1.6 kWp solar PV array with MPPT for battery charging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
5. Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
5.1. Variation of power profile of 1.6 kWp SPV system with different MPPT techniques at fixed solar insolation (G) ¼1000 W/m2 and
ambient temperature (T) ¼300 K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
5.1.1. PV system with constant voltage MPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
5.1.2. PV system with incremental conductance MPPT technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
5.1.3. PV system with perturb and observe MPPT technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587

n
Corresponding author.
E-mail addresses: [email protected] (P. Mohanty), [email protected] (G. Bhuvaneswari), [email protected] (R. Balasubramanian).

http://dx.doi.org/10.1016/j.rser.2014.06.001
1364-0321/& 2014 Elsevier Ltd. All rights reserved.
582 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

5.1.4. PV system with variable step size perturb and observe MPPT technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
5.2. Comparative assessment of different MPPT techniques under uniform G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
5.3. Variation of power profile of 1.6 kWp SPV system with different MPPT techniques at varying G and constant T ¼ 300 K . . . . . . . . . . . 588
5.3.1. PV system with constant voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
5.3.2. PV system perturb and observe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
5.3.3. Incremental conductance with varying G and fixed T ¼300 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
5.4. Comparative assessment of different MPPT techniques under varying G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
6. Proposed model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
6.1. Comparison of the proposed MPPT technique with others . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593

1. Introduction point (MPP) for both hill climbing and P&O methods. They have
developed variable perturbation size that gets smaller as the
Solar photovoltaic (SPV) systems are considered as one of the operating point approaches maximum power point (MPP). Hsiao
most reliable and matured technologies amongst various renew- and Chen [6] have presented a three-point weight comparison
able energy sources. However, high installation cost and low P&O method to address the issue of failure of P&O method under
commercial efficiency (12–15%) of the SPV system make it an rapidly changing atmospheric conditions. Hussain [7] developed a
unattractive choice for those electricity users who look for an new MPPT algorithm based on the fact that the maximum peak
alternative despite the government subsidies. Although SPV sys- operating point (MPOP) of a PV generator can be tracked accu-
tem prices have decreased considerably during the last years due rately by comparing the incremental and instantaneous conduc-
to new developments in the thin-film technology and manufactur- tance of the PV array. The work was carried out both in simulation
ing process [1], electricity from PV arrays is still expensive as and experimental prototype, with results showing that the devel-
compared to the existing fossil fuel generated electricity from the oped incremental conductance (IntCond) algorithm successfully
utility grid. Hence, it is important that the SPV array is used to its tracks the MPP, even in rapidly changing atmospheric conditions;
maximum potential. In order to achieve that, the SPV system has it also results in higher efficiency than the other MPPT algorithms
to be operated at its maximum power point, which will be tracked in terms of total PV energy transferred to the load. Patcharaprakiti
by different Maximum power point tracking (MPPT) techniques. et al. [8] presented an adaptive fuzzy logic controller that con-
Standalone PV generation systems are attractive for remote stantly tunes the membership functions and the rule based tables
areas. However, the efficiency of the PV system depends on several so that the optimum performance in tracking the true MPP is
climatic factors like the solar radiation, ambient temperature and achieved.
the state of the solar panels such as their age, cleanliness etc. Since Several research papers propose different techniques for achieving
the maximum power point varies with radiation and temperature, MPPT, however, a comprehensive comparison of these techniques,
it is difficult to maintain optimum matching at all radiation levels. in terms of their cost effectiveness, conversion efficiency, range of
Therefore, many research papers have focused on increasing the temperatures and insolation levels under which they can operate and
efficiency of the overall solar PV system by ensuring maximum reliability is not available in the literature. In some papers [4–6], a
power capture by adjusting the operating point of the DC–DC comparison of two MPPT methods is carried out but the comparison is
converter. A DC–DC converter acts as the interface between the made at the solar module level rather than for the entire solar PV
load and the SPV module. By changing the duty cycle, the load system. Since I–V and P–V characteristic of the single module do not
impedance as seen by the source is varied and matched such that depict multiple peaks which are commonly shown in the I–V and P–V
maximum power is harnessed from the PV source by appropriately characteristics of the solar PV array under non-uniform insolation,
maintaining the voltage–current relationship. Over the years, many solar PV array is a far more practical representation of the actual solar
MPPT techniques have been advocated, developed and implemen- PV system, attention should be given to study and compare different
ted. These methods vary in several aspects such as complexity, MPPT methods used with a solar PV array. This paper envisages
required number of sensors, convergence speed, cost, range of comparing the performances of some of the existing MPPT techniques
effectiveness, ease of hardware implementation etc [2]. There are in a solar PV system in terms of energy captured, conversion efficiency,
around 20 MPPT techniques that have been developed in the last reliability etc. Further, it also advocates a new MPPT methodology
two decades. Some of these methods/techniques are [3]: which yields a better performance than the existing techniques.
The paper is structured as follows. The overall configuration of
 Hill climbing, the SPV system is described in Section 2. The description of
 Perturb and observe (P&O), different MPPT techniques and algorithms under study is pre-
 Incremental conductance (IncCond), sented in Section 3. Modeling and simulation of entire SPV system
 Fractional open circuit voltage (Voc), is presented in Section 4. Performance assessment of different
 Fractional short circuit current (Isc), MPPT techniques under similar operating conditions is presented
 Ripple correlation control (RCC), in Section 5. A novel MPPT technique is advocated in Section 6
 Current sweep, along with its performance characteristics. Finally conclusion is
 Fuzzy logic and neural network control, presented in Section 7
 Load current or load voltage maximization,
 DC link capacitor droop control etc.

Over the years, several researchers have studied the character- 2. Overall system configuration
istics of PV modules and various MPPT techniques have been
developed [4–15]. Xiao and Dunford [5] have worked on P&O When a PV array is directly connected to a load or for charging
method to reduce the system oscillation near the maximum power a battery, the operating point of the system is at the intersection of
 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 583

Battery
DC/DC converter
48V, 300 Ah

1.6 kWp Solar PV array MPPT control

Fig. 1. Block diagram of the proposed PV system with MPP Tracker.

Table 1
Different MPPT techniques.

MPPT technique PV array True Analog or Periodic Convergence Implementation Sensed

dependent MPPT digital tuning speed complexity parameter

Constant voltage Yes No Both Yes Medium Low Voltage

Incremental No Yes Digital No Varies Medium Voltage, Current
conductance
Perturb and observe No Yes Both No Varies Low Voltage
Fractional Isc Yes No Both Yes Medium Medium Current

the I–V curve of the SPV array and the load line. In general, this PV DC/DC CONVERTER LOAD
operating point is not at the MPP of the SPV array. Thus in a SPV ARRAY
battery charging system, the PV array is oversized to charge the
Array Voltage
battery in a typical day with average insolation. This leads to a
huge capital investment. In order to overcome this disadvantage, a
switching mode power converter called Maximum Power Point ERROR AMPLIFIER COMPARATOR
Tracker is used to maintain the operating point of the SPV array
at MPP. The block diagram of the proposed solar PV system for
charging a lead acid battery with MPP Tracker is shown in Fig. 1.
The proposed system consists of a 1.6 kWp solar PV array Voltage Reference
which is connected to a buck converter to charge a 48 V, 300 Ah
Fig. 2. Control topology for the constant voltage based MPPT.
battery. According to the MPPT technique used, a Vref correspond-
ing to VMPP is created which regulates the duty cycle of the
converter and forces the solar array to work at the MPP. If a that the load impedance visualized by the solar PV array will make
proper MPPT algorithm is chosen and implemented, then the it operate at the MPP for a given temperature and insolation. The
MPPT can locate and track the MPP of the PV array continuously. following sections describe some of the MPPT algorithms.
Although the current requirement for charging the battery
depends upon the charging profile and varies with state of charge 3.1. MPPT control with constant voltage
of the battery, for a battery charging option, it might not be
possible to run the module always at the MPPT mode. However, The basis for the constant voltage (CV) algorithm lies in the fact
since the purpose of this paper is to assess the performance of that the ratio of the voltage of the array at MPP Vmpp to its open-
different MPPT techniques under similar operating conditions, a circuit voltage Voc is approximately a constant and generally
larger size battery whose current requirement (at C/10 rate) is around 0.76 time of the Voc of the solar module in any given solar
much larger as compared to the current provided by solar module insolation [9]. In other word, Vmpp/Voc–K o1(in this research work,
at maximum power is considered. There are several MPPT tech- it is considered as 0.76).
niques/algorithms that exist and Table 1 shows different MPPT The block diagram representation for the MPPT control with
techniques that were considered in this research work with their constant voltage is given in Fig. 2
varied features [4]. The constant voltage algorithm can be implemented using the
The purpose of this work is to compare the performances of flowchart given below in Fig. 3
various MPPT techniques in a solar PV system working under Constant voltage control can be easily implemented with analog
similar operating conditions, i.e. solar insolation and temperature. hardware. Here, the reference voltage is obtained by multiplying
The following sections describe the control design for different 0.76 with the Voc of the solar module for any given solar insolation.
MPPT algorithms.
3.2. MPPT control with perturb and observe method

3. MPPT control algorithms Perturb and observe (P&O) method is also known as hill
climbing method. The P&O algorithms operate by periodically
MPPT algorithms work in such a way as to modify the duty perturbing (i.e. incrementing or decrementing) the array terminal
ratio of the DC–DC converter at the output of the solar array such voltage or current and comparing the PV output power with that
584 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

of the previous output power. In P&O method, a small voltage the MPP. Therefore,
perturbation in a certain working voltage of PV array is made dP
simultaneously observing the change in direction of output power. If 4 0; V ref ¼ V ref þ dV ð1Þ
dV
If the output power increases, then the perturbation of the voltage
is made in the same direction otherwise perturbation against the dP
If o 0; V ref ¼ V ref  dV ð2Þ
original direction should be made. Due to the fixed perturbation dV
step, oscillations occur near the MPP, which reduces the power and accordingly the perturbation is given.
generation efficiency and affects stability of the entire system
adversely. Also, if the step change in the perturbation is large, then 3.3. MPPT control with variable step perturb and observe method
the response time to achieve the maximum power is fast, however
oscillations near MPP will also be large. On the other hand, if the In P&O, if the PV array operating voltage changes and power
step change is small, oscillations will be minute at the MPP but the increases (dP/dVPV40), the control system moves the PV array
tracking time will be very long. Therefore, selecting the appro- operating point in the same direction further; otherwise the
priate step size is the key for P&O method to achieve the desired operating point is moved in the opposite direction. A common
effect. P&O involves perturbation in the PV array voltage to problem in P&O algorithms is that the array terminal voltage is
achieve the MPP. When connected to a DC–DC converter, perturb- perturbed every MPPT cycle; therefore when the MPP is reached,
ing the duty ratio of power converter perturbs the PV array current the output power oscillates around the maximum, resulting in
and consequently perturbs the PV array voltage. It is known that power loss in the PV system. This is especially true in constant or
incrementing the voltage increases the power when operating on slowly-varying atmospheric conditions. Furthermore, P&O meth-
the left of the MPP and decreases the power when on the right of ods can fail under rapidly changing atmospheric conditions. In
order to circumvent this problem, P&O method with variable step
size is advocated in several papers; in this modified P&O method,
a large perturbation step is adopted when the present operating
Isolate Array point is far from the MPP whereas a small step size is adopted in
from MPPT the close vicinity of MPP. Fig. 4 shows the flow chart for P&O based
MPPT technique.

3.4. MPPT control with incremental conductance

Record Voc
The P&O method occasionally fails to track the fast changing
MPP point in the correct direction during fast changing atmospheric
Calculate Vmpp from % of Voc
condition. For this reason, incremental conductance method is used.
Fig. 5 shows the control topology for incremental conductance
based MPPT.
Vary array voltage till MPP The incremental conductance algorithm is derived by differ-
entiating the PV array power with respect to voltage and setting
the result equal to zero. This is shown in Eq. (3)
Wait X second
dP dðVIÞ dI
¼ ¼ 1þV ¼ 0 at the MPP ð3Þ
Fig. 3. Flow diagram of constant voltage algorithm. dV dV dV

Input: V(t),I(t)

P(t)= V(t) *I(t)

dP= P(t)-P(t-1)

Yes
dP>0

No
Yes
dV> 0 dV > 0
Yes
No
No
V ref= V ref - K V ref= V ref + K V ref= V ref - K V ref= V ref + K

Return

Fig. 4. Flow chart for perturb and observe based MPPT.

 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 585

Rearranging Eq. (3) gives 3.5. MPPT control with fractional short circuit current
I dI
 ¼ at MPP ð4Þ In fractional short circuit current, IMPP ¼k Isc. This ‘k’ has to be
V dV
determined from the solar array in use; generally ‘k’ lies between
 
dI I dP 0.78 and 0.82. But measuring Isc during operation is difficult. An
4 ; 40 ð5Þ additional switch usually has to be added to the power converter
dV V dV
to periodically short the PV array so that Isc can be measured
and using a current sensor. This increases the number of components
 
dI I dP and cost. Further, MPP is never perfectly matched when this
o ; o0
dV V dV algorithm is adopted. Fig. 7 shows the topology of the fractional
current
Therefore,
 
dI I
If 4 ; V ref ¼ V ref þ dV ð6Þ
dV V 4. Modeling and simulation
 
dI I 4.1. Modeling and simulation of solar PV array
If o ; V ref ¼ V ref  dV ð7Þ
dV V
A mathematical model is developed for the 1.6 kWp SPV array.
The MPP can thus be tracked by comparing the instantaneous
Fig. 8 depicts the equivalent circuit of a single solar cell using 2 diodes.
conductance (I/V) to the incremental conductance (dI/dV) as shown in
The equation that represents I–V characteristics of solar cell is
the flow diagram below in Fig. 6
derived from a more accurate two exponential diode model, where

PV DC/DC CONVERTER LOAD PV

DC/DC CONVERTER LOAD
ARRAY ARRAY

dI/dV+ I/V PWM GENERATOR PV current sensing PWM GENERATOR

-
-
PV power
calculation & + PI CONTROL
PI CONTROL I ref adjust
+ I ref

Fig. 5. Control topology for incremental conductance based MPPT. Fig. 7. Topology for factional current.

Input: V(t),I(t)

dI/dV = I(t)-I(t-dt)

dI/dV=V(t)-V(t-dt)

Yes
dV = 0

No

Yes Yes
dI/dV= -I/V dI = 0

No No

Yes dI> 0 Yes

dI/dV> -I/V

No No

Increment Vref Decrement Vref Decrement Vref Increment Vref

I (t-dt) =I (t)

V (t-dt) =V(t)

Return

Fig. 6. Flow diagram of incremental conductance.

586 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

the output current of the solar cell can be derived as 4 numbers of 12 V, 150 Ah batteries are connected in series and
qðV þ IRsÞ=n kT qðV þ IRsÞ=n kT there are two such strings connected in parallel to make a 48 V,
I ¼ I L I 0 ½exp 1  1–I 01 ½exp 2  1Þ  ðV þ IRs Þ=Rp
300 Ah battery bank. Various MPPT algorithms such as constant
ð8Þ voltage, Incremental conductance, perturb and observe and vari-
where IL is light-induced current of the solar cell; I0 is saturation able step size perturb and observe have been implemented by
current of the first diode; I01 is saturation current of the second writing the corresponding algorithms in the form of m-files in
diode; k is Boltzmann constant; T is device operating temperature MATLAB environment. The basic simulated model of 1.6 kWp solar
parameter value; q is elementary charge on an electron; n1 is PV with MPPT for battery charging is shown in Fig. 10
ideality factor (diode emission coefficient) of the first diode; n1 is Based on the type of MPPT techniques used, the specific m-file
ideality factor (diode emission coefficient) of the first diode; n2 corresponding to the specific MPPT technique is called to simulate the
is ideality factor (diode emission coefficient) of the second diode; MPPT blocks. The MPPT algorithm generates a reference voltage (Vref)
V is voltage across the solar cell electrical ports. which corresponds to voltage at MPP. A PI controller is used to process
A 1.6 kWp solar PV array model is developed in simulink by the difference between the output voltage of the solar PV array and
using the mathematical model where eight number of solar PV Vref to yield the duty ratio that would make the error zero. The output
module, each with rated power of 50 Wp, is connected in series to of the buck converters is regulated by modulating the duty ratio in
form the solar PV sub array and there are 4 such sub arrays that order to get a regulated voltage of 60 V which is sufficient to charge a
are connected in parallel to make the 1.6 kWp solar PV array. The 48 V battery.
PV array has maximum power of 1.602 W at VMPP of 89.9 V. The
basic model of SPV array is shown in Fig. 9.

4.2. Modeling and simulation of 1.6 kWp solar PV array with MPPT 5. Results and discussions
for battery charging
5.1. Variation of power profile of 1.6 kWp SPV system with different
1.6 kWp SPV array is connected with the 48 V, 300 A h battery MPPT techniques at fixed solar insolation (G) ¼ 1000 W/m2 and
bank through buck converter. The existing battery model in the ambient temperature (T) ¼300 K
simulink is used to create the 48 V, 300 Ah battery bank where
5.1.1. PV system with constant voltage MPPT techniques
1.6 kWp SPV system is simulated for fixed value of G and T
using Constant voltage MPPT algorithm. The value of G is 1000 W/m2
and T is 300 K. Constant voltage MPPT algorithms fix the reference
voltage at approximately 0.76 of open circuit voltage (Voc). This might
be away from the exact MPP point but is very stable when the
variation in insolation is small. Fig. 11 shows the power, voltage and
current characteristics. It is observed that voltage ripple is very small
Fig. 8. Circuit diagram for solar cell. and energy delivered in to load is appreciable in this case.

Fig. 9. Basic model of SPV array.

 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 587

5.1.2. PV system with incremental conductance MPPT technique 1000 W/m2 and T is 300 K. This algorithm continuously perturbs
1.6 kWp system is simulated for fixed value of G and T using the voltage by 0.01 V in every 5 ms. and accordingly allows the
incremental conductance MPPT algorithm. The value of G is reference voltage to change at every step in order to track the MPP.
1000 W/m2 and T is 300 K. Incremental conductance is very Since the step function is quite fast, a large oscillations in voltage,
accurate with very little voltage ripple at stable point and large current and power are observed, and hence more power loss
energy delivered. As seen from Fig. 12 power ripple is just 0.8 W. occurs in perturb and observe MPPT technique. As seen from
Fig. 13, power ripple is 60 W and energy delivered is much less
than constant voltage or incremental conductance. However
5.1.3. PV system with perturb and observe MPPT technique through simulation, it is found that the oscillation around MPP
1.6 kWp system is simulated for fixed value of G and T using can be reduced from 60 W to 10 W if the voltage is perturbed to
perturb and observe (P&O) MPPT algorithm. The value of G is 0.0001 V in every 5 ms.

Fig. 10. Simulated model of 1.6 kWp solar PV with MPPT for battery charging.

18.06
Cu rre n t (A)

18.04

18.02

18
0.49 0.491 0.492 0.493 0
Vo lta g e (V)

88.8

88.6
0.49 0.491 0.492 0.493 0

1601.5
Po wer (W )

1601

1600.5
0.49 0.491 0.492 0.493 0

Fig. 11. Power, voltage and current for constant voltage MPPT with fixed G ¼1000 W/m2 and T ¼ 300 K.
588 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

17.6

Current (A)
17.55

17.5
0.49 0.491 0.492 0.493 0.4

91.4

Voltage (V)
91.2

91
0.49 0.491 0.492 0.493 0.4

1602

Power (W)
1601

1600
0.49 0.491 0.492 0.493 0.4

Fig. 12. Power, voltage and current for incremental conductance with fixed G ¼ 1000 W/m2 and T ¼ 300 K.

19

Current (A)
18
17
16
0 0.05 0.1 0.15 0.2

95
Voltage (V)

90

85
0 0.05 0.1 0.15 0.2

1600
Power (W)

1550

0 0.05 0.1 0.15 0.2

Fig. 13. Power, voltage and current for perturb and observe with fixed G ¼1000 W/m2 and T ¼300 K.

5.1.4. PV system with variable step size perturb and observe observed that in uniform insolation, the energy generated by P&O
MPPT technique is the least as compared to other three MPPT techniques. It is
Here initially the voltage step change is 0.01 V for every 5 ms primarily due to the large oscillation of power at around MPP
and then the perturbation size is reduced to 0.0001 V near the (refer to Fig. 13).
maximum power point i.e. till the power reaches710% of the At the same time it is also noticed that although the energy
Maximum power point (in this case 720 W of 1600 Wp). It can be generated by using variable step size P&O and Inc Conductance are
observed that oscillations in power are negligible (as shown in higher than the other two methods; the response time in these
Fig. 14); however, the response time to achieve the MPP is quite case are very large, particularly in case of variable step size P&O.
large as compared to that of the normal P&O method. Thus a trade-off between energy generated and response time
needs to be arrived at to get the most effective MPPT algorithm.
5.2. Comparative assessment of different MPPT techniques under
uniform G

The P–V curve for the solar PV array is plotted for all the four 5.3. Variation of power profile of 1.6 kWp SPV system with different
MPPT techniques under uniform solar insolation of 1000 W/m2 MPPT techniques at varying G and constant T¼ 300 K
and temperature of 300 K. Fig. 15 shows the P–V plots of all the
four MPPT techniques. 5.3.1. PV system with constant voltage
Table 2 shows that comparative performance assessment of For constant voltage G is stepped down to 200 W/m2 from
different MPPT techniques when connected to the same 1.6 kW 1000 W/m2 at 0.1 s, depicting cloudy weather and again it comes
solar PV array and operated under uniform G of 1000 W/m2 and to 900 W/m2 at 0.12 s depicting clearing of the cloudy region.
temperature of 300 K. The performance is assessed in terms of Fig. 16 shows the current, voltage and power profile of PV system
energy generated by PV array in half an hour; the oscillation in with constant voltage MPPT algorithm and when operated under
power at MPP, the response times to achieve the MPP. It can be varying G and constant T.
 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 589

Current (A)
17.85

17.8
0.49 0.491 0.492 0.493

Voltage (V)
89.8

89.6

0.49 0.491 0.492 0.493

1602.84

Power(W)
1602.82

1602.8
0.492 0.493 0.494

Fig. 14. Power, voltage and current for variable step size perturb and observe with fixed G ¼ 1000 W/m2 and T ¼ 300 K.

P&O
Const Voltage

Power

Incr Cond
Variable step size
P&O

Voltage
Fig. 15. P–V plots of 1.6 kWp PV array for all four MPPT techniques.

5.3.2. PV system perturb and observe system becomes unstable and the operating point deviates away. But
For perturb and observe technique, with varying insolation (same such situation is not observed in simulation and it still oscillates
as mentioned in the previous section), it is seen practically that the around the maximum power point as shown in Fig. 17.
590 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

5.3.3. Incremental conductance with varying G and fixed T ¼300 K weather and again it comes to 900 W/m2 at 0.12 s depicting
Fig. 18 shows that the current, voltage and power profiles do clearing of the cloudy region) and temperature of 300 K. Fig. 19
not vary much even in varying insolation condition and deliver the shows the P–V plots of all the four MPPT techniques.
maximum power. From Table 3, it is observed that under varying conditions,
the total energy extracted half an hour is maximum for variable
step size P&O MPPT. To obtain a very good performance, MPP
5.4. Comparative assessment of different MPPT techniques variable step size perturb and observe is used which gives very
under varying G good performance. In this the step size of the perturbation is
varied. Initially it is kept large and when it approaches the MPP
The P–V curve for the solar PV array is plotted for all the four point it is reduced. Only disadvantage is that computation time
MPPT techniques under varying solar insolation (G is stepped increases.
down to 200 W/m2 from 1000 W/m2 at 0.1 s, depicting cloudy

6. Proposed model
Table 2
Comparative assessment of different MPPT techniques at fixed solar insolation of As discussed earlier, it is observed that the constant voltage
1000 W/m2 and temperature of 300 K. MPPT technique works quite well under uniform insolation and
particularly when the solar insolation is low. The response time to
Algorithms Energy (W h) Oscillation Response time (s)
around MPP (W) achieve the MPP is also low. On the other hand, variable step size
P&O works very well in both uniform and varying insolation
Perturb and observe 749.5 60 0.087 although the response time is relatively high. In order to get an
Inc Conductance 768.7 0.8 0.1 overall better performance with lesser response time, a Two-
Variable step size P 775 0.015 0.21
and O
Model MPPT Control algorithm is being proposed here (Fig. 20).
Constant voltage 755 0.5 0.075 It combines Constant voltage and the variable step P&O method. If
the irradiation is lower than 30% of the nominal insolation level

16.24
Current (A)
16.22

16.2
0.49 0.491 0.492 0.493

88.6
Voltage (V)

88.5

88.4

88.3
0.49 0.491 0.492 0.493

1436
Power (W)

1435.5

1435
0.49 0.491 0.492 0.493

Fig. 16. Power, voltage and current for constant voltage with fixed T ¼ 300 K and varying G.

17
Current (A)

16

15
0 0.05 0.1 0.15 0.2 0.25

95
Voltage (V)

90

85
0 0.05 0.1 0.15 0.2 0.25

1440
Power (W)

1430
1420
1410
0 0.05 0.1 0.15 0.2 0.25
Time

Fig. 17. Power, voltage and current for perturb and observe with fixed T ¼300 K and varying G.
 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 591

16.26

Current (A)
16.24

16.22
0.49 0.491 0.492 0.493

88.6

Voltage (V)
88.5

88.4

88.3
0.49 0.491 0.492 0.493

1436

Power (W)
1435.5

1435
0.49 0.491 0.492 0.493

Fig. 18. Power, voltage and current for incremental conductance with fixed T ¼ 300 K and varying G.

Power (W) Power (W)

Const voltage
P&O

Voltage (V)
Voltage (V)

Power (W) Power (W)

Incremental cond Variable step size

Voltage (V) Voltage (V)

Fig. 19. P–V plots for solar PV array for different MPPT techniques.
592 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593

the CV method is used, otherwise, variable step P&O method is respectively. Although the total energy extracted from the new
adopted. Since the short circuit current (Isc) of the PV array is method is slightly less as compared variable step size P&O (less by
proportional to the insolation, in the model which simulates this 1%), but the response time has drastically reduced from 0.21 s to
proposed MPPT technique, Isc is taken as the indicator of solar 0.054 s.
insolation.
Fig. 21 shows the power, voltage and current profile of the
newly designed two model MPPT techniques.

6.1. Comparison of the proposed MPPT technique with others 7. Conclusion

The performance of newer two model modified P&O technique The performance assessment of different MPPT techniques
is compared with other MPPT techniques and their comparative used in a 1.6 kW solar PV system for charging the battery is
performances under varying insolations (G is stepped down to critically assessed under uniform and varying solar insolation. A
200 W/m2 from 1000 W/m2 at 0.1 s, depicting cloudy weather and new two model modified P&O MPPT technique is also developed
again it comes to 900 W/m2 at 0.12 s depicting clearing of the and compared with other MPPT techniques. From the simulation
cloudy region) are given in Table 4. results, it is observed that under uniform insolation, the constant
It is observed that under varying conditions, the total energy voltage and P&O MPPT techniques work quite well where as their
extracted from the two model modified P&O in half an hour is 2% performance degrades as compared to other MPPT techniques
and 5% more than that of the P&O and constant voltage method, under varying solar insolation. Again under both uniform and
varying solar insolation, although the energy generated by using
variable step size P&O and Inc Conductance is quite high as
compared to other MPPT techniques, the response time in these
Table 3
Comparative assessment of different MPPT techniques at varying solar insolation case are very large, particularly in case of variable step size P&O.
and temperature of 300 K. Thus a trade-off between energy generated and response time
needs to be developed to get the most effective MPPT algorithm.
Algorithm Energy (W h) Oscillation Response
The newly simulated two model modified P&O technique works
around time (s)
MPP (W)
quite well in both uniform and varying solar insolation condition
and is working effectively as a trade-off between energy generated
Perturb and observe 700 60 0.087 and response time.
Inc Conductance 720 0.9 0.07
Variable step size P and O 725 0.015 0.21
Constant voltage 680 0.5 0.05
Table 4
Comparative assessment of newer developed MPPT technique with other different
MPPT techniques at varying solar insolation and temperature of 300 K.

Algorithms Energy (W h) Oscillation Response

V PV
around time (s)
MPP (W)

I sc V ref
MPPT control Perturb and observe 700 60 0.087
Inc Conductance 720 0.9 0.07
I PV Two model Modified P&O Variable step size P and O 725 0.015 0.21
Constant voltage 680 0.5 0.05
Two model modified P&O 715 0.9 0.054
Fig. 20. Proposed MPPT model.

16.26
Current (A)

16.24

16.22
0.49 0.491 0.492 0.493

88.6
Voltage (V)

88.5

88.4

88.3
0.49 0.491 0.492 0.493

1436
Power (W)

1435.5

1435
0.49 0.491 0.492 0.493

Fig. 21. Power, voltage and current profile of PV system with two model MPPT technique.
 P. Mohanty et al. / Renewable and Sustainable Energy Reviews 38 (2014) 581–593 593

References [9] Hohm DP, Ropp ME. Comparative study of maximum power point tracking
algorithms. Prog Photovoltaics Res Appl 2003;11:47–62.
[10] Noguchi T, Togashi S, Nakamoto R. Short-current pulse-based maximum-
[1] Faranda R, Sonia L. Energy comparison of MPPT techniques for PV systems.
power-point tracking method for multiple photovoltaic-and-converter mod-
WSEAS transactions on power systems 2008;6(3):446-55.
ule system. IEEE Trans Ind Electron 2002;49(1):217–23.
[2] Schaefer J. Review of photovoltaic power plant performance and economics.
[11] Walker G. Evaluating MPPT converter topologies using MATLAB PV model. J
IEEE Trans Energy Convers 1990;5:232–8.
Electr Electron Eng 2005;21:49–56.
[3] Leva S, Zaninelli D, Contino R. Integrated renewable sources for supplying remote
[12] Femia Nicola. Optimization of perturb and observe maximum power point
power systems. In: WSEAS transactions on power systems 2007;2:41–8.
[4] Esram T, Chapman PL. Comparison of photovoltaic array maximum power tracking method. IEEE Trans Power Electron 2005;20:4.
point tracking techniques. IEEE Trans Energy Convers 2007;22(2):439–49. [13] DP Hohm, ME. Ropp. Comparative study of maximum power point tracking
[5] Xiao W, Ozog N, Dunford WG. Topology study of photovoltaic interface algorithms using an experimental, programmable. In: IEEE Maximum power
for maximum power point tracking. IEEE Trans Ind Electron 2007;54(3): point tracking test bed; 1699–1702, 2000.
1696–704. [14] Nafeh AA, Fahmy FH, Mahgoub OA, El-Zahab EM. Developed algorithm
[6] Hsiao, Y-T, Chen, C-H. Maximum power point tracking for photovoltaic power of maximum power tracking for stand-alone photovoltaic system. Energy
system. In: 37th IAS annual meeting; 2002. p. 1035–40. Sources 1998;20:45–53.
[7] Hussain KH, Muta I. Maxium photovoltaic power tracking: an algorithm for [15] Jaboori MG, Saied MM, Hanafy. AA. A contribution to the simulation and
rapidly changing atmospheric conditions. In: Proceeding of industrial elec- design optimization of photovoltaic systems. IEEE Trans Energy Convers
trical engineering, generation transmission and distribution 1995:142;59–64. 1991;6:401–6.
[8] Patcharaprakiti N, Premrudeepreechacharn S. Maximum power point tracking
using adaptive fuzzy logic control for grid connected photovoltaic system. IEEE
Power Eng 2002:372–7.
 Applied Energy 150 (2015) 97–108

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

An improved perturb and observe (P&O) maximum power point tracking

(MPPT) algorithm for higher efficiency
Jubaer Ahmed, Zainal Salam ⇑
Centre of Electrical Energy Systems, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 Johor Bahru, Malaysia
Institute of Future Energy, Universiti Teknologi Malaysia, 81310 Johor Bahru, Malaysia

h i g h l i g h t s

 A improved P&O based MPPT method for PV system is developed.

 The method enhances the steady state performance of the conventional P&O.
 Proposed method minimizes the probability of losing the tracking direction.
 The energy yield is increased by 2% on average.
 No complex computation is required and hardware implementation is very simple.

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes a method to improve the efficiency of the P&O maximum power point tracker
Received 19 August 2014 (MPPT) by reducing the steady state oscillation and eliminating the possibility of the algorithm to lose
Received in revised form 31 March 2015 its tracking direction. A dynamic perturbation step-size is employed to reduce the oscillation, while
Accepted 5 April 2015
boundary conditions are introduced to prevent it from diverging away from the MPP. To prove its effec-
tiveness, the proposed P&O is compared with the conventional and adaptive P&O using the Ropp, sinu-
soidal and ramp irradiance tests. In addition, the performances are evaluated based on a one-day
Keywords:
(10 h) irradiance and temperature profile. The algorithm is implemented on a buck-boost converter
Solar energy
Photovoltaic
and benchmarked by the standard MPPT efficiency (gMPPT) calculation. It was found that, for all the tests,
P&O the gMPPT of the proposed P&O scheme is increased by approximately two percentage points. Besides, the
MPPT proposed algorithm does not require any extra hardware components; only several lines of additional
P–V curve software codes are to be embedded into the conventional P&O MPPT control program.
MPPT Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction own constraint, primarily due to the physical limitations of silicon

itself. Recently, new and more exotic semiconductor materials for
Solar photovoltaic (PV) is envisaged to be one of the key sources PV cells are also being experimented [3,4], but the present cost
of the future energy mix. The cost of modules has declined sharply of these technologies appears to be prohibitive. Consequently, they
over the last decade and nowadays PV power plants with MW are mostly used in high-end products, for example for space
capacity are becoming a norm in many countries [1,2]. Despite this application.
encouraging trend, the per-watt price of PV energy is still con- The more economical way to enhance the performance of the
siderably expensive compared to its fossil fuel counterpart. The PV installation is to consider the balance of system (BOS) compo-
reason is clear; the capital investment of PV project (in terms of nents, particularly the power converter (inverter or charger). One
dollar per watt) is significantly higher than the latter. There are possible area of improvement is to increase the efficiency of the
extensive research and development efforts to increase the PV maximum power point tracking (MPPT) algorithm [5–7]. This is
energy throughput—mostly they are focused on improving the per- because the MPPT comprises only of software codes that can be
formance of the solar cell efficiency. However, this approach has its embedded within the power converter firmware, i.e. without an
additional cost. The MPPT automatically locates the maximum
⇑ Corresponding author. Tel.: +60 07 5536187; fax: +60 07 5566272. power point (MPP)—that is the desired operating voltage (VMPP)
E-mail addresses: [email protected] (J. Ahmed), [email protected]. or current (IMPP) in order to achieve the maximum output
my (Z. Salam). power (PMPP). Furthermore, the tracking must be dynamic, i.e. the

http://dx.doi.org/10.1016/j.apenergy.2015.04.006
0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.
98 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

operating point has to be continuously adjusted in response to the V new ¼ V old þ DV  slope ðif P > P old Þ
ð1Þ
measured irradiance and temperature conditions. Thus locating V new ¼ V old  DV  slope ðif P < P old Þ
the MPP is a complicated process because the environmental con-
dition is constantly changing, while at the same time, the P–V In Eq. (1), the slope indicates the direction of the perturbation,
curve itself exhibits non-linear characteristics. i.e. to the right (climbing) or left (descending). The flowchart to
Among the numerous MPPT techniques described in literature implement P&O is given in Fig. 1(a). Clearly, the size of the
[8,9], the perturb and observe (P&O) [10,11] is the most popular. perturbation, DV is crucial; if DV is large, the convergence is fast,
It is widely installed in commercial PV inverter using low cost but it results in large fluctuation in P, and vice versa. Whatever
microprocessors. Despite the simplicity and reliability of the algo- the case, algorithm will caused to operating point to continuously
rithm, it inherently has two main drawbacks. First, when the track- oscillate around the MPP, as depicted in Fig. 1(b). Obviously, the
ing reaches the vicinity of the MPP, the operating point is forced to loss is more if the perturbation size is large. The oscillation is
go back and forth around MPP, resulting in unending oscillation in highly undesirable as it results is significant energy loss. That is
the output power. As a result, the energy yield is reduced and why in the adaptive P&O is introduced [15,20].
hence a drop in the efficiency. Secondly, P&O is prone to lose its
2.2. Divergence from the tracking locus
tracking direction, i.e. it could not cope with the change in the
irradiance and causes the operating point to diverge away from
Apart from the steady state loss, P&O is prone to diverge from
the MPP locus. This divergence also results in energy loss.
its tracking locus, i.e. moving away from the MPP. This divergence
This work is carried out to increase the efficiency the conven-
usually happens when a gradual change of irradiance takes place
tional P&O by simultaneously addressing both abovementioned
[9]. As illustrated in Fig. 2, assume that in the beginning, P&O
problems. Structure-wise, of proposed algorithm is similar to the
tracks the MPP at point A. Thus, the operating point oscillates
conventional MPP, but with two additional features: (1) it incorpo-
around the MPP, i.e. moving back and forth between A, B and B0 .
rates a dynamic built-in checking mechanism that guarantees the
Let’s consider a case whereby while moving towards A, the irradi-
consistent detection of the oscillation and varies the perturbation
ance starts to increase gradually. The operating point will reach at
size accordingly and (2) it establishes voltage boundaries to ensure
C instead of B. In this situation, the algorithm assumes the
that the operating point will not deviate from its tracking locus. It
perturbation is in the direction of increasing power; thus it keeps
is acknowledged that there exist previous works to resolve the
on providing a perturbation in the same direction. As a result,
power oscillation by means of the adaptive P&O [10–18].
the operating point is moves away from the MPP, i.e. following
However, as shall be shown in this paper, the adaptive mechanism
the path A-C-D-E, as shown in Fig. 2(a). Consequently, the operat-
has a tendency to provoke the MPPT to go in the wrong direction
ing point diverges from the MPP and will keep on doing so if the
and eventually causes the operating point to diverge from the MPP.
irradiance steadily increases. However it is necessary to note that,
Another point to consider is that, for majority of the cases
the loss of tracking direction only occurs when the irradiance
reported in literature, the MPPT algorithms are tested against sim-
increases gradually. When irradiance falls (both slow and faster
plistic irradiance profiles. In actuality, these tests do not reflect the
rate), the same phenomena does not happen.
conditions that occur in the real environment. Subsequently, the
improvements in performance (as claimed) cannot be truly verified. 2.3. The adaptive P&O
Since there is a lack of benchmarking, this paper proposes a com-
prehensive evaluation of the proposed method, in comparison to The adaptive P&O algorithms can be based on the voltage
the conventional and adaptive P&O. Three types of test, namely [10,18,20,21] or duty cycle [13,22,23]. The adaptive mechanism
the Ropp [19], sinusoidal and ramp are used for this purpose. intelligently varies the perturbation size to minimize steady state
These tests are very stringent as it demands the MPPT to track the oscillation. When the operating point is far from the MPP, the per-
irradiance profiles of various shapes, gradient and duration turbation is made larger. This is to ensure that the tracking speed is
very accurately. Furthermore, to understand how the proposed maximized. As the operating point gets closer the MPP, the per-
P&O performs under the real environmental condition, it is sub- turbation size is dynamically reduced until it becomes very small.
jected to a one-day of continuous irradiance and the temperature For example, in [21], the algorithm has an initial perturbation of
variation. 10% of Voc but at steady state, its value is reduced to 0.5%.
The paper is organized as follows: after this introduction, a brief There are various ways to vary the perturbation size. In [16], the
overview is given on the limitations of the conventional and adap- following relationship is used:
tive P&O. These are well-known problems—thus for brevity, only a
DP n
cursory information is given. In the next section, a detailed descrip- DV nþ1 ¼ M ð2Þ
tion on the proposed P&O is given. To prove the concept, a PV sim- DV n
ulation based on two-diode model is explained. The type and where DP and DV are the changes in power and voltage, respec-
specifications of module are also provided. A buck-boost experi- tively, while M is a constant that requires tuning. In another
mental set-up to verify the effectiveness of the algorithm is approach, [18] the perturbation size is determined as the logarith-
described next. The three performance tests, as well as the daily mic function as follows.
profile as described above are carried out in the following section.  
DP n
The results are complemented by a comprehensive discussion. DV nþ1 ¼ Mlog10 ð3Þ
DV n
The adaptive process is somewhat improved due to the loga-
2. Problems with the conventional P&O rithmic nature of the equation.

2.1. Steady state oscillation near the MPP 3. The proposed P&O

In the conventional P&O, the power (P) is computed using the 3.1. Outline of the scheme
measured values of the voltage (V) and current (I) of the PV array.
The algorithm provides a perturbation (DV) in V, based on the The proposed scheme is designed to improve the performance
change of P by the following rule: of the P&O by removing the steady state oscillation and preventing
 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108 99

increased. However, under this circumstance, there are possibili-

ties that the divergence from the MPP locus can occur. To prevent
this from happening, a boundary condition is imposed to restrain
the V⁄MPP within the range of the MPP locus. Ultimately, the change
in irradiance stops; the oscillation is detected and the perturbation
size will be reduced to the minimum again. On the other hand, if
DP/P > Tr2, the irradiance is considered to change rapidly from
one level to the next (step). In such case, it is sufficient to increase
the perturbation size to ensure that the MPP is tracked at a new
level of irradiance.

3.2. Setting of the initial points

The initial voltage (V⁄MPP) is set to 65% of Voc, where Voc is the
open circuit voltage of the PV array. The slope is determined in a
similar way to conventional P&O, i.e. using the signs of DP and
DV. The ‘‘value’’ of the slope is the sign multiplication of these
two quantities and normalized to unity, as shown in Table 1.
The detection of oscillation is done by the recording three con-
secutive slope values. When voltage is increasing, the sign of the
slope is positive, and vice versa. Thus, during the increase or
decrease of the voltage, the three consecutive sign of slope will
be either positive or negative. Thus the absolute value of the sum-
mation of the slope value will be 3. After reaching at the MPP when
oscillation starts, operating point will move two times in one direc-
tion and then move to the opposite direction. As a result under
continuous oscillation three consecutive sign of slope will never
be same. It will be a combination of positive and negative values.
Thus, the absolute summation of the slope value will be less than
3. Thus the following inequality can be derived:
X 
3 MPPT not converged to steady state
if slope ¼ ð4Þ
< 3 MPPT converged to steady state

Fig. 1. (a) The basic flowchart of the conventional P&O, (b) the occurrence of
oscillation of P&O around the steady state MPP.

the divergence of the algorithm from the MPP locus. The flowchart
is presented by Fig. 3. It starts by climbing the P–V curve until it
reaches the vicinity of MPP. Similar to the other P&O, the operating
point (V⁄MPP) oscillate around MPP. However, the oscillation is
detected by a special mechanism (that shall be described later)
and the perturbation size is reduced until it reaches a certain mini-
mum value. A very small power tolerance is allowed to handle tiny
flickers as a result of circuit noise and small deviation of the
irradiance.
Due to the small oscillation, there will be always a difference in
power (DP) at every consecutive sample. The value of DP/P (the
normalized power change) is kept below a certain threshold limit
(Tr1) because the perturbation size has reached its minimum value.
If DP/P > Tr1, it implies that the irradiance is changing (either
ascending or descending). In the real environment, the irradiance
varies in three possible ways: (1) slow-gradual (0.1–5 W/m2/s),
(2) fastgradual (>5 W/m2/s) and (3) rapid (step) change. These
possibilities are illustrated in Fig. 4. For the gradual changes (both
slow and fast), the power increases or decreases in tandem with
the change in the irradiance. On the other hand, rapid change con-
stitutes a step change in the irradiance. To determine whether the
change in irradiance is gradual or rapid, another threshold value
(Tr2) is introduced. If Tr1 < DP/P < Tr2, the change in the irradiance Fig. 2. Losing the tracking direction by P&O (a) towards left and (b) towards right
is considered as gradual. In this case, the perturbation size is side of MPP.
100 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

Fig. 3. General outline of the proposed P&O algorithm.

because when the irradiance decreases, the position of the VMPP

shifts slightly to the left. The shifting is observed to be approxi-
mately 5% of the Voc. This is demonstrated by Fig. 5. Thus by apply-
ing the 5% margin, the MPPT is forced to remain near the MPP
during gradual irradiance change—thus avoiding the loss of track-
ing condition.
It is reported in [25], under normal condition, the maximum
irradiance change can be 0.027 KW/m2/s. Thus the maximum grad-
ual occurrence of DP/P can be 0.027. To make the algorithm more
robust, Tr1 and Tr2 are chosen to be 0.001 (lower than 0.027) and
Fig. 4. Different types of irradiance change. 0.05 (higher than 0.027), respectively. Thus, any irradiance change
in between 1.0 W/m2/s to 50 W/m2/s is considered as the gradual
The initial perturbation size is set to / = 2% of Voc [24]. As the change. A value higher than 50 W/m2/s is detected as the step
operating point converges near to MPP, the oscillation is detected change. In case of the step changes, the perturbation size is
by Eq. (4). Then the size of the perturbation is reduced by 0.5% of increased to 2% of Voc. However, the voltage boundaries are not
Voc for every iteration step. The perturbation size is continuously imposed because MPP divergence is not relevant for the step
reduced until it reaches 0.5% of Voc. The tiny oscillation will result change in irradiance.
almost zero power loss and at the same time immunize the algo-
rithm from very small changes of the irradiance and the noise in 4. Implementation
the circuits.
4.1. PV modeling
3.3. Change of irradiance and setting the boundary condition
To simulate the PV system, the multiple string two-diode model
For the proposed P&O scheme, as operating point converges to [26], as shown in Fig. 6 is utilized. Variable NS is the number of
MPP, the perturbation size becomes very small; thus it is able to modules per string, while NP indicates the number of string in
handle slow-gradual change in the irradiance easily. However, the array. If N is NS/NP, the PV current drawn from the system
when irradiance changes at a faster rate, it may lose the tracking can be written as
direction. To avoid this, a flag is introduced. Initially, the flag is
I ¼ IPV Np  ID1
low, but once the oscillation is detected, it is toggled to high.      
This flag remains high until DP/P > Tr1. Then the second threshold V þ NIRs V þ NIRs
¼ Id1 Np exp  Id2 Np exp
value, i.e. Tr2, is checked. If Tr1 < DP/P < T2, it implies that the a1 V T1 Ns a2 V T2 Ns
change of power is small; thus a gradual change in irradiance is V þ NIRs
occurring. In this case, the perturbation size is restored to 2% of  ð5Þ
NRp
Voc. Additionally, since the gradual change occurs; the voltage
boundary condition is imposed. The value of the boundaries values where I and V are the PV current and voltage, respectively; Rs is the
are estimated as follows. The initial voltage boundaries are set, series, while Rp is the parallel resistance; VT1 = VT2 is the thermal
i.e. 0 to Voc. When the flag is high, the boundaries are changed to voltage of the diodes. The light generated current (IPV) is given by
(V⁄MPP-5% of Voc) and (V⁄MPP+ 5% of Voc). These values are selected G
IPV ¼ ðIPV STC þ K I ðT  T STC ÞÞ ð6Þ
Table 1 GSTC
Determination of the slope ‘‘value’’.
Note that IPV_STC is measured in the standard test condition
Sign of DP Sign of DV Slope value (STC),1 Variable KI is the short circuit current coefficient, which is
Positive Positive +1 usually provided by the manufacturer. The diode saturation current
Positive Negative 1 is given by
Negative Positive 1
Negative Negative +1
1
At STC, temperature T = 298 K and irradiance G = 1000 W/m2.
 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108 101

Table 2
The specifications of the PV module.

Parameters Variable Value Unit

Short circuit current ISC 3.8 A
Open circuit voltage VOC 21.1 V
Current at Pmax IMPP 3.5 A
Voltage at Pmax VMPP 17.1 V
Maximum power PMPP 59.85 W
VOC coef. of temperature KV 0.08 V/°C
ISC coef. of temperature KI 3e3 A/°C
No. of modules in series (per string) NS 10 –
No. of parallel strings NP 2 –

Fig. 5. The position change of Vmpp depending of the irradiance level.

ISC STC þ K I ðT  T STC Þ

Id1 ¼ Id2 ¼ ð7Þ
expððV OC STC þ K V ðT  T STC ÞÞ=V T Þ  1

In Eq. (5), ISC_STC and VOC_STC are the short circuit current and the
open circuit voltage in STC, respectively. Variable KV is the open cir-
cuit voltage coefficients. The specifications for the PV module used
in this paper are given in Table 2. Using these parameters, the
corresponding P–V and I–V curves for the variation in irradiance
and temperature are plotted in Figs. 7(a) and (b), respectively.
Each curve presents its unique maximum power point (MPP)
location.

4.2. Experimental set-up

Fig. 8 presents the MATLAB simulation model for the overall PV

system. The P&O algorithm is realized using a buck-boost con-
verter. This converter is selected to its suitability for MPPT prob- Fig. 7. The I–V and P–V curves for two modules in parallel with specifications given
in Table 2. (a) under varying solar irradiance. (b) under varying temperature.
lem, as suggested by various authors, for example [27]. The
buck-boost is designed to operate in the continuous inductor cur-
rent mode, with the specifications shown in Table 3. voltage, i.e. V⁄MPP. The best values of the PI controller are tuned by
The PV model described in Section 4.1 is labeled as ‘‘PV Array’’ trial and error method and are shown in Table 3.
block in Fig. 8. Its inputs are the G and T. These are profiles of
irradiance and temperature that will be imported using look-up
4.3. Performance benchmarking
tables. The outputs are the PV voltage (VPV) and current (IPV).
They measured using voltage and current sensor. Multiplying these
The effectiveness of the proposed P&O algorithm is measured
two quantities gives the present value of power. This value is fed to
by its MPPT efficiency. The instantaneous MPPT efficiency is which
the MPPT block, which in turn will return the value of Vout (V⁄MPP).
is computed by the following formula:
An error voltage (Verror) is obtained by subtracting Vref (VPV) from
the Vout, which is then fed to a proportional–integral (PI) controller. P MPP ðtÞ
The output of the PI controller is compared to a sawtooth wave-
gMPPT ¼  100 ð8Þ
P MPP ðtÞ
form to produce the duty cycle (D) for the converter. The converter
is forced to operate using this value of D, which results desired The average MPPT efficiency is given as
R
P MPP ðtÞdt
gMPPTðav gÞ ¼ R  100 ð9Þ
PMPP ðtÞdt
In Eqs. (8) and (9), PMPP⁄ is the maximum theoretical power that
can be achieved, which is the target of the algorithm. It is com-
puted using the PV model. On the other hand, PMPP is the actual
power that is extracted using the MPPT algorithm. It depends on
how close the algorithm tracks the irradiance at any instant of
time. To calculate PMPP, the measured values of VPV and IPV are
required.
The performance of the conventional, adaptive and proposed
P&O is benchmarked using three tests, namely (1) Ropp (2) sinu-
soidal and (3) ramp irradiance. In addition, to establish how these
algorithms perform under real environmental condition, they are
subjected to a one-day irradiance and temperature profile. For
Fig. 6. The two-diode model of PV cell that is used throughout this paper. the adaptive P&O, the logarithmic scheme [18] is used.
102 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

Fig. 8. The buck boost circuit used to implement the MPPT.

Table 3 falls to the 90% region. For the proposed P&O, the steady state oscil-
Specifications for the buck-boost converter. lation is extremely small. Furthermore, during the change of irradi-
Component/parameter Label Value ance—unlike the conventional and adaptive P&O, it is able to follow
Switching frequency f 100 kHz the power locus very closely. This is the results of the effective con-
Buck-boost inductor L 1 mH trol of the boundary conditions, as explained in Section 3.2.
Buck-boost capacitor C2 470 lF However, when the steady state condition is detected, it very
Input capacitor C1 470 lF quickly reduces the perturbation size to avoid large oscillations.
Proportional constant (P) P 0.08
The computed average MPPT efficiency, i.e. gMPPT(ave) over the total
Integral constant (I) I 15
Ropp profile cycle for the conventional, adaptive and proposed ver-
sion is 97.52%, 98.84% and 99.4%, respectively.
5. Results and discussions
5.2. Sine irradiance
5.1. The Ropp test
In the Ropp test, although the irradiance profile is highly
The Ropp test was originally proposed by [19]—using an irradi- dynamic, the rise or fall of the slopes is linear. This is not necessary
ance profile shown in the Fig. 9(a). The test determines if the MPPT true in the real environmental conditions. To represent for the non-
algorithm is able to cope with the gradual as well as sudden step linear slopes, another test named as the sinusoidal irradiance is
changes of the irradiance. In addition, it also includes the steady- developed. Its objective is to determine whether the MPPT algo-
state condition. The profile starts at 200 W/m2. During the initial rithm is able to track an irradiance profile with varying irradiance
ascend (60–140 s), the irradiance increases gradually from 200 rate at every sampling time. Fig. 11(a) depicts the profile for the
to 1000 W/m2. Then four step changes are made: (1000–200), sinusoidal test. It begins at 800 W/m2. Similar to the Ropp test, a
(200–600), (600–200) and (200–1000) W/m2. Finally irradiance 5 s dwelling time provided. Then irradiance starts to vary between
descends gradually from 1000 to 200 W/m2. 600 W/m2 and 1000 W/m2, according to the change in the sinusoi-
The tracking performances of conventional and proposed P&O dal curve.
based MPPT can be determined from the output power plot shown The tracking performance of the conventional and the proposed
in Fig. 9(b). For the conventional P&O, the perturbation size is fixed P&O is illustrated in Fig. 11(b). As can be observed in the enlarged
at 2% of Voc. As can be observed, the proposed P&O tracks the irradi- image, as the irradiance increases, the conventional P&O repeat-
ance profile almost perfectly. The amplitude of oscillation is very edly loses the tracking direction. As a result, its efficiency drops
low and it never loses its tracking locus. On the other hand, the for brief periods, as depicted by the Fig. 12(b). In the case of adap-
conventional P&O exhibits large oscillation at steady state as tive P&O, an interesting situation occurs during the positive rise of
shown by the inset (zoomed portion) of the figure. Furthermore, the sine wave, i.e. between 8 and 12 s. As can be seen in Fig. 11(c),
during the gradual ascending of the irradiance, the conventional the algorithm is not able to maintain the tracking, particularly
P&O lost its tracking direction numerous times. However, during when irradiance gradually increased at a higher rate. The reason
the descends of irradiance, no loss of tracking is observed. This is is that, when the perturbation size becoming smaller (as it reaches
consistent with the explanation given in Section 2.2. The compar- the steady state), it could not cope with the larger change in the
ison between the adaptive and proposed P&O is illustrated in irradiance. In such condition, what the algorithm needs is the con-
Fig. 9(c). It can be seen, the adaptive P&O still loses the tracking trary, i.e. larger perturbation size. As a result of the insufficient per-
direction at several points; an example is shown by the enlarged turbation size, the voltage drops and it loses the track of the MPP.
image. Furthermore, during steady state, it keeps on oscillating, On the other hand, the proposed P&O tracks the power locus quite
even though the perturbation size is very small. As a result, the with near perfection. The only departure from the reference occurs
tracked power is slightly below the actual MPP. when the operating point reaches the peaks and valleys of the
Fig. 10(a), (b) and (c) shows the MPPT efficiency (gMPPT) as com- sinusoidal. This happens due to the fact that when the irradiance
puted by Eq. (8). Due to the oscillation, the gMPPT of the conventional is descending, the MPP voltage shifts slightly to the left. Since
P&O varies in the range of 95–99.5%. However, during the loss of the sinusoidal waveform is continuously changing, the operating
tracking direction, the gMPPT drops to as low as 75%. In the case point is not given an opportunity to settle at a new MPP for a while,
of adaptive P&O, the efficiency is maintained around 99% for most unlike the Ropp test. As a result, the efficiency drops slightly
of the time. However during the tracking divergence, the efficiency around these points.
 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108 103

gradients from 5.0 W/m2/s to 50 W/m2/s. Thus, it covers a compre-

hensive range of irradiance change profile, i.e. from a slow to very
fast (almost step change). At the peak and valley of each ramp, a
small dwell time (0.5 s) is given to allow for the operating point
to settle at a particular MPP, before ascending or descending again.
The full irradiance profile of the test is shown in Fig. 13(a).
The tracking performance of the conventional and proposed
P&O is presented in Fig. 13(b). During the slow irradiance change,
i.e. ramp No. 1, the proposed P&O tracks the power locus almost
perfectly. This is illustrated by the inset, i.e. image 1. On the other
hand, as expected, the conventional P&O exhibits steady state
oscillation. There is also an occasion whereby it diverges from its
intended power locus. As the slope of the ramp is increased, it loses
the tracking direction more frequently. These are depicted in the
images 2, 3 and 4, respectively. In image 4, (50 W/m2/s) the algo-
rithm almost fail to reach the peak value of the ramp. For the adap-
tive P&O, it tracks the initial ramps quite successfully. As the
gradient of the ramp increases, it loses the tracking direction
occasionally. However, the occurrence is not as frequent as the
conventional P&O. Nevertheless, the reduction in the power yield
is significant. Furthermore, it can be seen from the enlarged images
in Fig. 13(c) that it failed to reach the peaks of the ramps with gra-
dients 20, 30 and 50 W/m2/s.
The steady state oscillation and tracking divergence is reflected
in the efficiency. Fig. 14(a) shows the efficiency profile for the con-
ventional P&O. As can be observed, the efficiency drops signifi-
cantly during tracking divergence. Additionally, the drop is also

Fig. 9. (a) Irradiance profile for the Ropp test (b) Comparison of the tracking
performance of the conventional P&O and proposed P&O. (c) Tracking performance
of the adaptive P&O and proposed P&O.

Due to the repeated loss of the tracking direction, the gMPPT(ave)

for the conventional P&O dropped down to 98.3%. In several cases,
the instantaneous efficiency goes as low as 95%. In the case of
adaptive P&O, the large drop during the gradual change in irradi-
ance causes gMPPT(ave) to be in the range of 98%. On the other hand
the proposed P&O exhibits excellent tracking over the whole test
curve. Its instantaneous efficiency is continuously above 99.5%
under all conditions, while the gMPPT(ave is approximately 99.8%.
The efficiency profiles for all cases are shown in Fig. 12(a)–(c).

5.3. The ramp irradiance

The objective of the ramp test is to quantify the ability of the

MPPT to track ascending and descending irradiance triangle with Fig. 10. Efficiency profile of (a) conventional P&O (b) adaptive P&O (f) proposed
various slopes values. A total of seven ramps are utilized, with P&O.
104 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

Fig. 11. (a) Irradiance profile for the sinusoidal irradiance test. (b) Comparison of tracking performance between the conventional P&O and proposed P&O (c) Comparison of
tracking performance between the adaptive P&O and proposed P&O.

Fig. 12. Efficiency profile of the (a) conventional P&O (b) adaptive P&O (c) proposed P&O.
 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108 105

Fig. 13. (a) Ramp irradiance profile (b) Tracking performance of the conventional P&O and proposed P&O. (c) Tracking performance of the adaptive P&O and proposed P&O.

contributed by the steady state oscillation. Its average efficiency, the MPP more frequently due to the higher rate of change in the
(gMPPT(avg)) is approximately 96.5%. However, as can be observed, irradiance (10 W/m2/s). During the steady-state condition, (image
for certain parts of the profile, the efficiency dips well below 3), the conventional P&O maintains its tracking direction.
90%. The efficiency profile for the adaptive P&O is shown in However, the power oscillates continuously due to the fixed per-
Fig. 14(b). The gMPPT(avg) is 97.5%. The efficiency plot for the pro- turbation size. Image 4 represents the condition in which the
posed P&O is shown in Fig. 14(c). As can be seen, the gMPPT is irradiance descends gradually. During this interval, as explained,
almost uniform throughout the irradiance profile. It recorded an both the conventional and proposed P&O do not diverge from the
gMPPT(avg) of almost 99.5%. MPP. However, the steady sate oscillation of the conventional
P&O still persists. In Fig. 15(c) the adaptive and proposed P&O is
5.4. One-day irradiance and temperature profile compared. For clarity, four sections of the profile are zoomed. In
these four images, it can be seen that the adaptive P&O is prone
The MPPT techniques are subjected to a daily (ten-hour) irradi- to lose the tracking due to the small perturbation sizes. On the con-
ance and temperature profile as shown in Fig. 15 (a). This is a plot- trary the proposed P&O detects the steady state almost perfectly
ted using a meteorological data of a typical tropical day that and reduces the oscillation to the minimum level.
includes the presence of intermittent clouds and occasional The efficiency of the conventional, the adaptive and the pro-
showers. The irradiance and temperature increase almost linearly posed P&O is presented in Figs. 16(a), (b) and (c), respectively.
since morning until midday. During the afternoon, irradiance The measured gMPPT(avg) for the conventional P&O is approximately
remains almost constant with the sporadic occurrence of clouds 97.5%. In certain cases the gMPPT drops below 85% due of the loss of
and rain. As it approaches evening, the irradiance and temperature tracking direction. For the adaptive P&O gMPPT(avg) is 98.76%.
gradually falls. Though it drops near to 95% occasionally, most of the time it is well
The output power harvested by the conventional and proposed above 97%. On the other hand, gMPPT(avg) for the proposed P&O is
P&O is illustrated in Fig. 15(b). Since the traces for both are almost calculated at 99.2%. At most times, its efficiency is maintained well
indistinguishable, four enlarged snapshots (images) are presented. above 99%, except for the lower irradiance region, where it dips to
Image 1 focus on the responses for the sudden increase in the near 97%. On the overall, the proposed P&O increases the gMPPT(avg)
irradiance increases (5 W/m2/s). As can be seen, the conventional by 1.7%.
P&O loses the tracking continuously. On the other hand, the pro- To conclude all tests carried, a summary of the performance are
posed P&O track the power locus almost perfectly. In image 2, a tabulated on Table 4. It can be deduced that, in general, the pro-
similar case is presented; the conventional P&O diverges from posed P&O offers an increases in the average efficiency of
106 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

Fig. 14. Efficiency profile of (a) conventional P&O (b) adaptive P&O (c) proposed P&O.

Fig. 15. (a) Daily irradiance profile (b) Tracking performance of the conventional P&O and proposed P&O. (c) Tracking performance of the adaptive P&O and proposed P&O.
 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108 107

Fig. 16. Efficiency of (a) the conventional P&O (b) the adaptive P&O (c) the proposed P&O.

Table 4
Summary of MPPT efficiency (gMPPT) for all tests.

Test type Max efficiency (%) Min efficiency (%) Avg. efficiency (%)
Conventional Adaptive Proposed Conventional Adaptive Proposed Conventional Adaptive Proposed
Ropp 99.65 99.80 99.88 72.65 91.50 97.30 97.52 98.84 99.45
Sinusoidal 99.65 99.70 99.93 95.42 85.70 99.58 98.30 98.00 99.82
Ramp 99.65 99.60 99.93 80.03 85.50 99.42 96.45 97.5 99.55
One day profile 99.65 99.70 99.82 86.49 90.50 96.55 97.51 98.76 99.20

approximately two percentage points. For a PV system, this Acknowledgments

improvement is very significant due to the fact that an installed
system has a life-time of approximately twenty years. Based on The authors would like to thank Universiti Teknologi Malaysia
this observation, there is a strong justification to consider its appli- and the Ministry of Higher Education, Malaysia for providing the
cation over the conventional and adaptive P&O. Besides, the facilities and financial support (Research University Grant No.
proposed algorithm does not require additional hardware; only a Q.J130000.2509.06H78) to conduct this research.
handful line of software codes need to be embedded into the
MPPT control program.
References

6. Conclusion [1] Rehman S, Bader MA, Al-Moallem SA. Cost of solar energy generated using PV
panels. Renew Sustain Energy Rev 2007;11:1843–57.
[2] Wei H, Liu J, Yang B. Cost-benefit comparison between Domestic Solar Water
A scheme to improve the efficiency of the P&O MPPT has been Heater (DSHW) and Building Integrated Photovoltaic (BIPV) systems for
described. It reduces the steady state oscillation and eliminates households in urban China. Appl Energy 2014;126:47–55.
the probability of divergence form the MPP locus. The proposed [3] Krebs FC. Fabrication and processing of polymer solar cells: a review of
printing and coating techniques. Sol Energy Mater Sol Cells 2009;93:394–412.
method is evaluated against the conventional and adaptive P&O [4] Zhang W, Zhu R, Liu B, Ramakrishna S. High-performance hybrid solar cells
based MPPT by using the Ropp, sinusoidal and ramp irradiance employing metal-free organic dye modified TiO2 as photoelectrode. Appl
tests. In addition, these are subjected to a one-day (10 h) irradiance Energy 2012;90:305–8.
[5] Ishaque K, Salam Z. A review of maximum power point tracking techniques of
and temperature profile. In all the cases, the proposed method out- PV system for uniform insolation and partial shading condition. Renew Sustain
performs the conventional and the adaptive P&O. It was found that, Energy Rev 2013;19:475–88.
for all the tests, the gMPPT of the proposed P&O increases by [6] Ahmed J, Salam Z. A Maximum Power Point Tracking (MPPT) for PV system
using Cuckoo Search with partial shading capability. Appl Energy
approximately two percentage points. This improvement is impor-
2014;119:118–30.
tant since a PV installation has a life-time of approximately twenty [7] Ramli MZ, Salam Z. A simple energy recovery scheme to harvest the energy
years. Since there proposed version maintains the similar algo- from shaded photovoltaic modules during partial shading. Power Electron,
rithm structure with the conventional, the former can be easily IEEE Trans 2014;29:6458–71.
[8] Salam Z, Ahmed J, Merugu BS. The application of soft computing methods for
implementable with a low-cost microcontroller, similar to the con- MPPT of PV system: a technological and status review. Appl Energy
ventional one. 2013;107:135–48.
108 J. Ahmed, Z. Salam / Applied Energy 150 (2015) 97–108

[9] Esram T, Chapman PL. Comparison of photovoltaic array maximum power [19] Ropp M, Cale J, Mills-Price M, Scharf M, Hummel S. A test protocol to enable
point tracking techniques. Energy Conversion, IEEE Trans 2007;22:439–49. comparative evaluation of maximum power point trackers under both static
[10] Piegari L, Rizzo R. Adaptive perturb and observe algorithm for photovoltaic and dynamic irradiance. Photovoltaic Specialists Conference (PVSC), 2011 37th
maximum power point tracking. Renew Power Generation, IET IEEE: IEEE; 2011. p. 003734-7.
2010;4:317–28. [20] Sera D, Teodorescu R, Hantschel J, Knoll M. Optimized Maximum Power Point
[11] Mamarelis E, Petrone G, Spagnuolo G. A two-steps algorithm improving the Tracker for fast changing environmental conditions. Industrial Electronics,
P&O steady state MPPT efficiency. Appl Energy 2014;113:414–21. 2008 ISIE 2008 IEEE International Symposium on 2008. p. 2401–7.
[12] Alonso R, Ibaez P, Martinez V, Roman E, Sanz A. An innovative perturb, observe [21] Zhang L, Al-Amoudi A, Bai Y. Real-time maximum power point tracking for
and check algorithm for partially shaded PV systems. Power Electronics and grid-connected photovoltaic systems, 2000.
Applications, 2009 EPE ‘09 13th European Conference on 2009. p. 1–8. [22] Femia N, Petrone G, Spagnuolo G, Vitelli M. A technique for improving P&O
[13] Fermia N, Granozio D, Petrone G, Vitelli M. Predictive & adaptive MPPT perturb MPPT performances of double-stage grid-connected photovoltaic systems. Ind
and observe method. Aerospace Electron Syst, IEEE Trans 2007;43:934–50. Electron, IEEE Trans 2009;56:4473–82.
[14] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimization of perturb and [23] Xiao W, Dunford WG. A modified adaptive hill climbing MPPT method for
observe maximum power point tracking method. Power Electron, IEEE Trans photovoltaic power systems. Power Electronics Specialists Conference, 2004
2005;20:963–73. PESC 04 2004 IEEE 35th Annual: Ieee, 2004. p. 1957–63.
[15] Abdelsalam AK, Massoud AM, Ahmed S, Enjeti P. High-performance adaptive [24] Bennett T, Zilouchian A, Messenger R. A proposed maximum power point
perturb and observe MPPT technique for photovoltaic-based microgrids. tracking algorithm based on a new testing standard. Sol Energy 2013;89:23–41.
Power Electron, IEEE Trans 2011;26:1010–21. [25] Bletterie B, Bruendlinger R, Spielauer S. Quantifying dynamic MPPT
[16] Yang Y, Zhao FP. Adaptive perturb and observe MPPT technique for grid- performance under realistic conditions first test results–The way forward.
connected photovoltaic inverters. Procedia Eng 2011;23:468–73. In: 21st European photovoltaic solar energy conference, Dresden, 2006.
[17] Pradhan R, Subudhi B. Design and real-time implementation of a new auto- [26] Ishaque K, Salam Z, Taheri H. Simple, fast and accurate two-diode model for
tuned adaptive MPPT control for a photovoltaic system. Int J Electr Power photovoltaic modules. Sol Energy Mater Sol Cells 2011;95:586–94.
Energy Syst 2015;64:792–803. [27] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved particle swarm
[18] Zhang F, Thanapalan K, Procter A, Carr S, Maddy J. Adaptive hybrid maximum optimization (PSO)–based MPPT for PV with reduced steady-state oscillation.
power point tracking method for a photovoltaic system. Energy Conversion, Power Electron, IEEE Trans 2012;27:3627–38.
IEEE Trans 2013;28:353–60.
 Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A comparison of different global MPPT techniques based on meta-heuristic MARK

algorithms for photovoltaic system subjected to partial shading conditions
⁎
Hegazy Rezka,b, , Ahmed Fathyc, Almoataz Y. Abdelazizd
a
College of Engineering at Wadi Addawaser, Prince Sattam Bin Abdulaziz University, KSA
b
Electrical Engineering Department, Faculty of Engineering, Minia University, Egypt
c
Electrical Power and Machine Department, Faculty of Engineering, Zagazig University, Egypt
d
Electrical Power & Machines Department, Faculty of Engineering, Ain Shams University, Egypt

A R T I C L E I N F O A BS T RAC T

Keywords: The characteristics of photovoltaic array under partial shading comprises multiple local MPPs and one global.
Particle swarm optimization The classical maximum power point tracking (MPPT) algorithms can’t reach to global MPP. Accordingly, this
Cuckoo Search work aims to study the behavior performance of two optimization techniques. They have been developed for
Global MPPT extracting the global MPP from the partially shaded PVPS. The two studied techniques include Particle Swarm
Partial shading
Optimization (PSO) and Cuckoo Search (CS). A comprehensive assessment of the two techniques has been
carried out against a conventional algorithm of INR−based tracker. The tracking performances of PSO and CS
based trackers are evaluated for different partial shading patterns based on MATLAB software. Results confirm
that PSO and CS based trackers guarantee the convergence to the global MPP. Furthermore, they have the best
performance in comparison with the conventional one. Additionally; the obtained results show that the
CS−based tracker has superiority compared with PSO. The tracking time in case of CS−tracker is reduced
compared to PSO in all the studied cases.

1. Introduction and neural network [3,6]. These algorithms are suitable for MPP
tracking under uniform irradiance, where there is only single MPP that
The needs of energy have increased in the last decades. exists in the power against voltage or current curves. Conversely, such
Furthermore, predict end for fossil fuel has emphasized on seek for algorithms aren't convenient to track the global peak under PSC [4].
alternative ones which are renewable energy resources (RES). For solving this problem, global MPPT based on meta-heuristic
Photovoltaic power system (PVPS) has been considered an attractive optimization algorithms have been introduced [6–20]. Such algorithms
solution among RES. It is environmental friendliness, long-term and contain; Genetic algorithm, Cuckoo Search (CS), Particle Swarm
maintenance-free [1]. PVPS comprises several photovoltaic panels Optimization (PSO), Ant Colony Optimization (ACO) and Teaching–
which are connected in series and in parallel for reaching desired Learning–Based Optimization (TLBO). PSO and CS are simple and
rating. Therefore, probability of occurring partial shading condition effective approaches. They are effectively utilized to solve different
(PSC) is high. This condition occurs in case of PVPS receives a non- engineering optimization issues that have many peaks [3]. They
uniform irradiance [2]. Perfectly, PV system should established in demonstrate numerous merits more than other techniques utilized
shadow free region. However, it is generally installed in urbanized for extracting global MPP under PSC. Such merits include; identifica-
regions and PV panels may be subjected into PSC produced by near tion of shadow patterns is not required, and the algorithm can be used
constructing. Under PSC, the power against voltage (P–V) or against in single-stage arrangement. Furthermore, they are characterized by
current (P–I) curves of PVPS contains multiple local peaks (points B fast and simplicity in design [20].
and C in Fig. 1) and one global peak (point A in Fig. 1) which is Jubaer and Salam [2,21] proposed MPPT based on Cuckoo Search
required to be tracked. This phenomenon decreases the efficiency of (CS) optimization technique. In the first work [21], three initial values
classical MPPT algorithms [3,4]. In order to enhance PVPS conversion of PV voltage have selected over span of PV voltage. Such voltage values
efficiency under uniform irradiance, many classical MPPT algorithms according to Lévy flight were directed into optimal. CS technique has
are developed [5]. Such algorithms are; hill climbing, incremental been modeled via MATLAB. After that the comparison between CS and
conductance, incremental resistance, fuzzy logic, perturb & observe, P & O is done without taking the condition of partial shading into

⁎
Corresponding author at: Electrical Engineering Department, Faculty of Engineering, Minia University, Egypt.
E-mail addresses: [email protected] (H. Rezk), [email protected] (A. Fathy).

http://dx.doi.org/10.1016/j.rser.2017.02.051
Received 16 August 2016; Received in revised form 27 December 2016; Accepted 10 February 2017
Available online 24 February 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 1. Power against voltage, power against current and current against voltage curves of PV system under uniform irradiance and PSC.

consideration. The obtained results showed that CS technique reaches sion system has been tracked and achieved. Kofinas et al. [25] used
to MPP faster than P & O. On the second work [2], CS technique has single neuron in artificial neural network (ANN) for presenting the
been compared against PSO with focusing on the gradual, step and MPPT control, the performance of the buck converter has been
rapid changes of irradiance and temperature conditions. Such condi- improved via applying a directional control scheme containing one
tions occur under uniform irradiance. Only single irradiance pattern adaptive neuron and learning algorithm of Big Bang-Big Cruch (BB-
was taken into consideration for checking the capability of CS to catch BC) algorithm. The initial parameters of the ANN have been adapted
global MPP under PSC. Accordingly, this paper tries to present a via off-line learning approach of the (BB-BC) to avoid local optima. In
comprehensive comparison for CS against PSO with global MPPT for Ref. [26]; meta-heuristic optimization algorithms used in simulating of
PV system under several different irradiance patterns. For evaluating MPPT for shaded PV array have been reviewed. Additionally, the
the response of CS technique with various cases and guarantee the condition of operation, the advantages, the limitations of each one have
reliability of CS in tracking the global, the position of global peak is been given. The reviewed methods focused on P & O algorithm, PSO,
changed varying PSC shapes. ant colony, artificial neural network, Fuzzy-logic hill climbing algo-
Finally, a comparison of the CS and PSO techniques is done against rithm, cuckoo search, differential evolution, flashing firefly, extreme
conventional one, namely incremental resistance (INR). Four partial seeking control approach and irradiance measurement. The authors
shading patterns are taken into consideration during the analysis; the concluded that, all algorithms have similar idea in catching the global
tracking performances of studied MPPT techniques are evaluated for maximum power point trackers which are using sorted threshold
each pattern using MATLAB simulation. equations to check the occurrence of partial shading condition.
Dileep et al. [27] reviewed different approaches based on improved P
2. Literature review & O technique for designing MPPT incorporated in shaded PV system.
In each approach; the dependency degree, the tracking speed, the
A summary of previously proposed global MPPT based different complexity, the used parameters, the cost, the type of converter, the
optimization algorithms is shown in Table 1. Dileep et al. [22] advantage, the limitation and application have been shown. The
introduced a comprehensive study for soft computing techniques for authors summarized that the artificial intelligence (AI) approaches
simulating the MPPT; in addition to the applications, advantages and are superior to classical ones especially for tracking the MPP in shaded
drawbacks of each reviewed approach have been reported. The authors PV system. Additionally; a controller ship of FPGA facilitated the
classified the reported approaches regarding to the control strategy to implementation of AI. Authors in [28] presented a global MPPT based
direct and indirect while the classification with respect to number of on TLBO. Simulation results proved that TLBO is accurately tracked
variables are one and two variables techniques. Classification in terms global MPP for different studied cases. Authors in [29] used mine blast
of circuitry is analogue and digital algorithms; finally regarding to algorithm (MBA) and TLBO in order to capture the global MPP for
applications the reported methods are expensive and inexpensive ones. partially shaded PV system. Various cases of PSC are taken into
The soft computing algorithms studied in that paper are artificial considerations. Such cases include; uniform distribution, changing
neural network (ANN), Fuzzy logic (FL), Chaos theory, probability, from 0% to 375% with steps of 25%, and non-uniform patterns with
non-linear predictor, evolutionary algorithms (GA and DE) and swarm various positions of global MPP. Ramli et al. [30] presented a review of
intelligence algorithms (ACO and PSO). In Ref. [23], a modification in MPPT methods used for normal and PSC operations of PV system. The
PSO has been presented for simulating the maximum power point reported techniques are artificial intelligence, hybrid algorithm and
tracker incorporated with shaded PV array. The modification process others. The authors mentioned that, the researchers focused on the
has been performed in determining the parameters of PSO for partial shading condition as it is required to enhance the output power
acceleration purpose by using the control method of exponential-form extracted from the PVPS under PSC. The reported methods in that
parameter. The control method used exponential increase or decrease work are Hopfield neural network with FLC, ANFIS and biological
in the cognition, social learning factors and the inertia weight of each swarm chasing technique for PVPS operated under normal irradiance.
particle. PIC microcontroller has been programmed with conventional Additionally, other methods were reviewed like testing standard based
and modified algorithm and the curves are compared. Fathabadi [24] MPPT algorithm, a fast and energy efficient MPPT circuit, sliding mode
presented a universal MPPT for a hybrid system containing PV, wind control strategy, adaptive control design for MPPT, variable scaling
and fuel cell. The presented MPPT didn’t need to any sensor as it uses factor based MPPT control technique, rapid MPPT algorithm of PV
only a unified algorithm and controller for implementation, the input systems, discrete time ripple correlation control (RCC), geometrical
signals are the PV module, the fuel cell stack and wind turbine voltages prediction, time-domain array-reconfiguration, classical root-finding
and currents. The highest output power from the wind energy conver- method, modified differential evolution, sliding mode control strategy,

378
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

cuckoo search, modified FL controller, Tabu search, Takagi-Sugeno

Exponential increase or decrease in the cognition, social learning factors

ANN have been adapted via off-line learning approach of the (BB-BC)
(TS) fuzzy, PSO and modified PSO.

Studying CS performance under gradual, step and rapid changes in

Extra coefficients are added to conventional PSO for accelerating

Increasing speed of MPPT via improvement of PSO using variable

the tracking speed [8]. Firefly optimization algorithm is designed for
extracting global MPP for PVPS under PSC [11]. Integration between
PSO and P & O is developed to track the global MPP [12]. GA has been
proposed for determining number of neurons in multilayer ANN [13].
A modified DE algorithm is proposed for dealing with rapid changed of

Evaluating the performance of TLBO algorithm

Evaluating the performance of MBO algorithm

and the inertia weight of each particle in PSO
Design of MPPT with no need of any sensors

irradiance and tracking global MPP under PSC [15]. Renaudineau et al.
Proposing using Cuckoo Search technique

[31] presented PSO algorithm for designing MPPT inserted in PVPS for
controlling the dc/dc converter. In Ref. [32]; grid-connected PVPS
under weather fluctuation operating conditions has been modeled
based on different PV parameters. Shivashankar et al. [33] reviewed
irradiance and temperature

the methods associated with the connecting PV to the grid in addition

sampling time strategy to the smoothing methods of PV power fluctuation and the controlling
approaches for storing the energy in battery. Methods of extracting the
parameters of solar cell based on mathematical techniques used single
Contribution

diode and double diode models have been reviewed in Ref. [34].

3. Partially shaded PVPS

PVPS contains many photovoltaic panels which are interconnected

in series and parallel to reach the desired capacity. Under PSC, the
Matlab Simulation only

shaded panel consumes the generated power by others and dissipates

PIC microcontrollers

heat [35,36]. PSC limits the current of series connected panels by

Matlab Simulation

Matlab Simulation

Matlab Simulation
microcontroller

current of shaded one. In order to overcome this dilemma, bypass

MC68HC11E9

diodes have been parallel connected with each panel. In normal

Controller

operation, such diode is reversed and do not has any effects [6].
Conversely under PSC, it is forward biased and carry current instead of
DSP

DSP

shaded panel. Fig. 1 illustrates power against voltage, the power against
current and the current against voltage curves of PVPS, respectively,
DC-DC converter

under uniform irradiance of 1 kW/m2 and two different cases of PSC.

This condition is given by four PV modules connected in series with
buck-boost

buck-boost
converter

converter

solar irradiances of 1 kW/m2, 1 kW/m2, 0.8 kW/m2, and 0.8 kW/m2

Boost

Boost

Boost

Boost
Boost
Buck

respectively strikes PV modules for the first PSC case. The solar
irradiances are 1 kW/m2, 1 kW/m2, 0.8 kW/m2 and 0.6 kW/m2 re-
Control parameter

spectively in the second PSC case. From Fig. 1, one must note that,
under PSC P–V and P–I graphs contain a number of peak points, while
the I–V curve contains many steps. The number of peak points is the
Duty cycle

Duty cycle

Duty cycle

Duty cycle
Duty cycle
Voltage

Voltage
current

same as the number of different irradiance levels incident on panels.

Referring to Fig. 1 (P–V curve); the points B and C are called the local
MPPs while point A is called global MPP. For PVPS; it is important to
Grid connection

track global MPP.

Off grid

4. PSO based MPPT technique

Yes

NO
No

No

No

No

No
Comparison of different global MPPT approaches from previously works.

PSO technique includes a number of particles, and each one

proposes a solution by exchange the data obtained in their particular
Big Bang–Big Crunch with

search process to find the best solution. Such particles transport

Improved particle swarm

Teaching–learning-based

Mine blast optimization

Cuckoo Search & PSO

through search-space based on a simple mathematical relation [4].

The particles position and velocity are updated based on (1) and (2). In
unified algorithm
MPPT technique

this work, the particle swarm position and the fitness are taken as the
Cuckoo Search

Improved PSO
optimization

optimization

duty cycle and PV system output power respectively. The particle

position di is updated by the following relations [20,37,38];
ANN

dik +1 = dik + vi k +1 (1)

2015

2016

2015

2013

2014

2016

2016
2015
Year

Where;.
vi denotes step size at iteration k+1
Jubaer and Salam [2]

Mirhassani et al. [20]

vik +1 = wvik + c1r1(Pbest − dik ) + c2r2(Gbest − dik ) (2)

Jubaer and Salam
Chao et al. [23]

Fathabadi [24]

Where;.
Kofinas [25]

Fathy [29]
Rezk [28]
Authors

w denotes inertial weight;

[21]

c1 and c2 denote the acceleration coefficients;

Table 1

r1 and r2 random values from [0,1];

1

7
8

379
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 3. PV module output power and its derivative under changing current density.

tracker is that PV power derivative w.r.t its current is zero at the MPP
(Fig. 3, dark cyan line). The derivative of PV power w.r.t the current can
be written as (4) whereas the error signal can be estimated by (5).

dP d (V × I ) dV
= =V+I =0
dI dI dI (4)

dV V V (i ) − V (i − 1) V (i )
+ = + =0
dI I I (i ) − I (i − 1) I (i )

Also,

V (i ) − V (i − 1) V (i )
e= +
I (i ) − I (i − 1) I (i ) (5)

The value of step size (SS) is considered a main factor in fixed SS

based techniques. With large SS, the tracking speed to catch MPP is
high, but the oscillation around the MPP at steady state is high. On
contrary, with small SS, the tracking speed is decreased with the
minimization of the oscillation around MPP. Therefore, the trade-off
between tracking speed and accuracy is considered high important
issue. In general, MPPT techniques which use the fixed-step size
approach must be take the trade-off dilemma into consideration. In
Fig. 2. Flowchart for searching mechanism by PSO-based tracker.
sum, such techniques do not have a good accuracy due to oscillation
occurs around the MPP in steady state and have the problem that is
Pbest denotes best position of particle i;
failure to MPP with environment changing. Accordingly, for mitigation
Gbest denotes best position in entire population.
such dilemma several variable SS-based MPPTs are developed. The
core idea of variable SS-based MPPT is based on starting the tracking
Fig. 2 shows the flowchart for searching mechanism by PSOebased
process with large SS in order to quickly catch MPP. Next, the SS is
tracker. At beginning, number of duty cycles has been randomly
reduced in case of the operating point is close to MPP in order to
nominated. Then, each one is applied to PVPS. The system current
mitigate the oscillation at steady state. In the current work, VSS−based
and voltage are measured in order to estimate PVPS power. Such power
MPPT includes a discrete integrator with gain k and fed by the error.
represents the fitness function of particle i. Next, the comparison
The integrator gain k is employed to adapt the error to be in a suitable
between new fitness value and power corresponding to pbest stored in
range before integration stage [39–41]. Fig. 4 illustrates the Matlab/
the history is done. In case of the new estimated power is more than the
Simulink model for INR−based tracker.
old one, it is selected as the best fitness value. After evaluating all
particles, old velocity and position for each can be updated based on (1)
and (2). When the stopping criterion is achieved, PSO−based tracker
stopped and gives optimal duty cycle which corresponds to global
power.

5. Incremental resistance based MPPT algorithm

One of the well-known algorithms developed to track MPP of PVPS

is the incremental resistance (INR). In order to explain the procedures
followed in such technique consider Fig. 3 which illustrates power
against current for PV panel under uniform irradiance. There is a single
point located at power curve. That point represents MPP at which PV
module produces its maximum power. The core idea of INR−based Fig. 4. Implementation of INR‒based tracker using Matlab/Simulink.

380
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

6. Cuckoo Search based global MPPT

Cuckoo Search (CS) has been considered one of the best optimiza-
tion algorithms recently proposed [42]. The core idea of CS can by
summarized by brood parasitism. It lays its eggs inside nests of other
birds. Brood parasitism can be classified into interspecific, cooperation
and nest takeover [43]. When host bird detects cuckoo egg in her nest,
it may break the egg or leave off the nest and create a new one inside
another location. For increasing possibility to have a new cuckoo,
cuckoos imitate the colors, devolve the capacity of cuckoo to mimic the
call of host bird to gain access to more feeding opportunity. The model
explained by [42] has been using the following rules;

• Only one egg generated by cuckoo every time;

• Only best nests having good quality eggs pass to next iteration;
• Host nests have been considered fixed.

Lévy flights which are used for generating new eggs have been
considered a robust feature of CS. It is a random walk where SS is
derived from Lévy distribution based on a power law.

L évy(β ) = l −β (6)
where;
l denotes the length of flight whereas 1 < β < 3.
The new egg is represented by the coordinates reached at the end of
the flight. For controlling flight size a coefficient α has been provided.
On the other hand, the fraction of eggs to be discarded (pa) has been
considered an important parameter which should be taken into
consideration [42]. The main target is reaching the optimum boost
converter duty cycle related to global MPP under different PSC
patterns. Accordingly, boost converter duty cycle has been used as a
variable to be optimized. PV power has been used as the fitness
function. Based on Lévy flight, the duty cycle can be updated by the
following [44]:
⎛ u ⎞ k
di k +1 = di k + α ⊕ L évy(β ) ≈ di k + kL évy × ⎜ 1/ β ⎟(dbest − di k )
⎝v ⎠ (7)
Where β=1.5, kLévy denotes the Lévy multiplying coefficient, u and v
can be estimated based on normal distribution curves.

u ≈ N (0, σu2 ) and v ≈ N (0, σv2 ) (8)

Fig. 5. Flowchart for searching mechanism by CS-based tracker.
Where the variables σu and σv are defined as follows:

⎛ ⎞1/ β extensive MATLAB simulation studies under different PSC patterns

⎜ Γ (1 + β ) × sin(π × β /2) ⎟ have been performed. LA361K51S PV module is chosen to be used in
σu = ⎜ ⎟ and σv = 1
⎜ Γ ( 1 + β ) × β × 2 (β −1) ⎟ the simulation model. The PV system illustrated by Fig. 6 consists of
⎝ 2
2 ⎠ (9) four series-connected modules and DC/DC boost converter. The input
Fig. 5 shows the flowchart for searching mechanism by CS−based inductance is 1 mH whereas output capacitor value equal to 47 μF and
tracker. At beginning, number of duty cycles is randomly nominated. 80 Ω resistive load. The operating scenario of the system shown in
Then, each one is applied to PVPS. The system current and voltage are Fig. 6 was based on sensing shaded PVPS voltage and current which are
measured in order to estimate PVPS power. Such power represents the fed to MPPT block to search duty cycle at which global maximum
fitness value. The duty cycle related to best fitness function has been power can be derived.
selected as current best nest (dbest). Next, Lévy flight is applied based Three MPPT techniques have been used as a controller feeding the
on Eq. (7) in order generate new nests. New finesses' values are tested converter with suitable duty cycle for studying and comparing dynamic
through the PVPS. Afterwards, the worst nest is randomly destroyed response of PVPS during PSC.
with a probability Pa, this process simulates the behavior of the host Such techniques are compared in terms of tracking speed, steady
bird discovering the cuckoo's eggs and destroying them. The new nest state performance and tracking efficiency for each pattern of irra-
is replaced by the destroyed one via Lévy flight, then the PV power is diances under study. Three different PSC patterns are employed for
measured and the current best nest is selected. When the stopping evaluating the tracking performance. For the first case of the partial
criterion is achieved, CS−based tracker stopped and gives optimal duty shading patterns, the solar irradiances stroked each PV module are
cycle which corresponds to global power. 1.0 kW/m2, 0.9 kW/m2, 0.8 kW/m2, and 0.7 kW/m2 respectively. The
P−I and P−V graphs are illustrated Fig. 7(a) and (b). Under this case,
7. Analysis and discussion of the results there are four peaks and the global MPP of 157 W is located at the first
one on P−I curve. Fig. 7(c) illustrates PVPS output power for studied
For analyzing the tracking performances of PSO and CS−based algorithms. Based on such figure, it can be seen that, the three
trackers and compare them with the conventional one, INR technique, algorithms successfully reached to global MPP with high tracking

381
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 6. Implementation of a PVPS using Matlab/Simulink.

efficiency. The detailed simulation results (power, voltage, current and until stopping criterion is achieved. CS, PSO and INR converge to the
duty cycle of boost converter) of PVPS with different MPPT techniques global MPP with tracking time 0.24 s, 0.4 s and 0.4 s respectively.
under the first shading pattern are shown in Fig. 8. As illustrated in Therefore, it can be concluded that using CS in determining global MPP
Fig. 8, the PSO and CS−based trackers begin initialization by setting decreases the tracking time by 40% compared with PSO and INR.
boost converter duty cycles to be 0, 0.3, 0.5 and 0.9 respectively in The solar irradiances are 1.0 kW/m2, 0.8 kW/m2, 0.6 kW/m2, and
order to ensure that the search space covers the whole P–V curve. After 0.4 kW/m2 in second studied PSC pattern. The P–I and P–V graphs are
that, the values of duty cycle are modified. Such process keeps on going illustrated in Fig. 7(a) and (b). For this condition, global MPP of 100 W

Fig. 7. (a) P−I curve (b) P−V curve (c) PVPS output power for PSO, INR and CS techniques under different three partial shading patterns.

382
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 8. Detailed simulation results for PV system under first shading pattern of 1.0 kW/m2, 0.9 kW/m2, 0.8 kW/m2, and 0.7 kW/m2; (a) PSO−based tracker and (b) INR−based tracker
(c) CS−based tracker.

Fig. 9. Detailed simulation results for PV system under second shading pattern of 1.0 kW/m2, 0.8 kW/m2, 0.6 kW/m2, and 0.4 kW/m2; (a) PSO−based tracker and (b) INR−based
tracker (c) CS−based tracker.

383
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 10. Detailed simulation results for PV system under third shading pattern of 1.0 kW/m2, 0.8 kW/m2, 0.4 kW/m2, and 0.2 kW/m2; (a) PSO−based tracker and (b) INR−based
tracker (c) CS−based tracker.

was found at second one on P–I curve. From Fig. 7(c), it can be noted tracking time and number of iterations for all different studied shading
that, CS and PSO bypass the first and second local MPPs (P–V curve) patterns. The updating process of the particle swarms in PSO and host
and extract the global one with 0.24 s and 0.36 s tracking time nests positions in CS during searching process are shown in Fig. 11.
respectively. Therefore, one can derive that the usage of CS decreases For PSO all particles of the agents converge at the optimal value after
the tracking time by 33.3% relative to PSO under this case. On the 10 iterations while CS needs only 6 iterations. A comparison between
other hand; the conventional INR–based tracker failed in reaching the different studied MPPT techniques is listed in Table 2. From this table,
global MPP and caught the first local MPP of 90.4 W (P–I curve) due to the superiority of CS –based tracker over the other two techniques can
its disability to differentiate between the local and global MPP. Under be clearly observed.
this condition the tracking efficiency decreased by 9.6% compared with
CS and PSO. The detailed simulation results for PV system with
different MPPT techniques under the second shading pattern are 8. Conclusion
shown in Fig. 9.
For the third PSC pattern; the solar irradiances are 1.0 kW/m2, In this paper, the performance of two global MPPT based on meta-
0.8 kW/m2, 0.4 kW/m2, and 0.2 kW/m2. The P–I and P–V graphs are heuristic optimization algorithms is evaluated. Such algorithms include;
displayed in Fig. 7(a) and (b). Under this case, the global MPP of PSO and Cuckoo Search (CS) techniques. The presented algorithms aim to
82.3 W is the located at the third point on P–I curve. From Fig. 7(c), improve the efficiency of partially shaded PVPS. A comprehensive
one can derive that, CS and PSO ignore the first local MPP (P–V curve) assessment of the two techniques is carried out against incremental
and catch the global MPP with very high tracking efficiency. The resistance (INR) based tracker. The suggested PV system comprises
detailed results for the PVPS with studied algorithms under the such MPPT system, boost converter and PV array modeled and is implemented
case are illustrated in Fig. 10. Based on such figure, the convergence in MATLAB/Simulink software. For three different partial shading
time for CS was 0.24 s and equals 0.4 s in case of PSO. This means that patterns, the tracking efficiency and the overall performance of studied
usage of CS decreases the tracking time by 40% compared to PSO. Also MPPT techniques were evaluated. The results confirm that CS and PSO
it can be noted that, INR still failed in catching the global MPP. INR based trackers have high accuracy and stability in extracting the global
extracts the local MPP of 66 W and the tracking efficiency is decreased MPP in all the tested cases no matter where the global MPP locates
by 19.8% compared with CS and PSO. The performance of the compared with the conventional one. Finally; it is concluded that CS is
CS−based tracker is far better than the PSO based tracker in terms of better than PSO. The tracking time decreased by more than 30% in all
different studied partial shading patterns.

384
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

Fig. 11. Update of particles and host nests positions for each iteration.

Table 2
Comparison between different studied MPPT techniques.

Shading pattern Technique Power, W Tracking speed Global power Tracking efficiency %

Milliseconds Iterations

Fist case PSO 156.88 40 10 157 W 99.92

INR 156.65 40 – 99.14
CS 156.90 24 6 99.94

Second case PSO 99.90 36 9 100 W 99.90

INR 90.40 18 – 90.40
CS 99.90 24 6 99.90

Third case PSO 82.3 40 10 82.4 W 99.87

INR 66.00 24 – 80.01
CS 82.3 24 6 99.87

References for photovoltaic systems under the influence of partial shading, In: Proceedings of
the IECON 2013-39th Annual Conference of the IEEE, 10-13 Nov. 2013, pp. 1490–
1495.
[1] Liu C. et al., A PSO-based MPPT Algorithm for Photovoltaic Systems Subject to [8] Phimmasone V, Kondo Y, Kamejima T, Miyatake M. Evaluation of extracted energy
Inhomogeneous Insolation, In: Proceedings of the 6th International Conference on from PV with PSO–based MPPT against various types of solar irradiation changes,
Soft Computing and Intelligent Systems (SCIS) and 13th International Symposium In: Proceedings of the IEEE International Conference on Electrical Machines and
on Advanced Intelligent Systems (ISIS), pp. 721–726; 2012. Systems, Korea, pp. 487–492; 2010.
[2] Ahmed J, Salam Z. A Maximum Power Point Tracking (MPPT) for PV system using [9] Jiang L, Maskell D, Patra J. A novel ant colony optimization-based maximum power
Cuckoo Search with partial shading capability. Appl Energy 2014;119:118–30. point tracking for photovoltaic systems under partially shaded conditions. Energy
[3] Liu Y, et al. A particle swarm optimization3based maximum power point tracking Build 2013;58:227–36.
algorithm for PV systems operating under partially shaded conditions. IEEE Trans [10] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved particle swarm
Energy Convers 2012;27(4):1027–10.35. optimization (PSO)–based MPPT for PV with reduced steady-state oscillation.
[4] Sarvi M, Ahmadi S, Abdi S. A PSO-based maximum power point tracking for IEEE Trans Power Electron 2012;27(8):3627–38.
photovoltaic systems under environmental and partially shaded conditions. Prog [11] Sundareswaran K, Peddapati S, Palani S. MPPT of PV systems under partial shaded
Photovolt: Res Appl 2015;23(2):201–14. conditions through a colony of flashing fireflies. IEEE Trans Energy Convers
[5] Ma J, Man K, Ting T, Zhang N, Lei C, Wong N. Low–Cost Global MPPT Scheme for 2014;29(2):463–72.
Photovoltaic Systems under Partially Shaded Conditions, In: Proceedings of the [12] Sundareswaran K, kumar V, Palani S. Application of a combined particle swarm
IEEE International Symposium on Circuits and Systems (ISCAS), pp. 245–248; optimization and perturb and observe method for MPPT in PV systems under
2013. partial shading conditions. Renew Energy 2015;75:308–17.
[6] Tey KS, Mekhilef S, Yang H, Chuang M. A differential evolution based MPPT [13] Kulaksız AA, Akkaya R. A genetic algorithm optimized ANN–based MPPT algo-
method for photovoltaic modules under partial shading conditions. Int J Photo rithm for a stand-alone PV system with induction motor drive”. Sol Energy
2014;2014:1–10. 2012;86(9):2366–75.
[7] Daraban S, Petreus D, Morel C, A novel global MPPT based on genetic algorithms [14] Paraskevadaki EV, Papathanassiou SA. Evaluation of MPP voltage and power of

385
H. Rezk et al. Renewable and Sustainable Energy Reviews 74 (2017) 377–386

mc-Si PV modules in partial shading conditions. IEEE Trans Energy Convers [29] Fathy A, Rezk H. A novel methodology for simulating maximum power point
2011;26(3):923–32. trackers using mine blast optimization and teaching learning based optimization
[15] Tajuddin MFN, Arif MS, Ayob SM, Salam Z. Perturbative methods for maximum algorithms for partially shaded photovoltaic system. J Renew Sustain Energy
power point tracking (MPPT) of photovoltaic (PV) systems: a review. Int J Energy 2016;8:023503. http://dx.doi.org/10.1063/1.4944971.
Res 2015. [30] Ramli MAM, Twaha S, Ishaque K, Al-Turki YA. A review on maximum power point
[16] Besheer AH, Adly M. Ant colony system based PI maximum power point tracking tracking for photovoltaic systems with and without shading conditions. Renew
for stand-alone photovoltaic system, In: Proceedings of the IEEE International Sustain Energy Rev 2017;67:144–59.
Conference on Industrial Technology (ICIT), pp. 693–698; 2012. [31] Renaudineau H, Donatantonio F, Fontchastagner J, Petrone G, Spagnuolo G,
[17] M. Lei, S. Yaojie, L. Yandan, B. Zhifeng, T. Liquin and S. Jieqiong, " A high Martin JP, Pierfederici S. A PSO-based global MPPT technique for distributed PV
performance MPPT control method", Materials for Renewable Energy & power generation. IEEE Trans Ind Electron 2013;62(2):1047–58.
Environment (ICMREE), In: Proceedings of the 2011 International Conference , [32] Humada AM, Hojabria M, Sulaiman MHB, Hamada HM, Ahmed MN. Photovoltaic
Vol. 1, 2011, pp. 195–199. grid-connected modeling and characterization based on experimental results. PLoS
[18] Kazmi S, Goto H, Ichinokura O, Guo H. An improved and very efficient MPPT One 2016;11(4):1–13.
controller for PV systems subjected to rapidly varying atmospheric conditions and [33] Shivashankar S, Mekhilef S, Mokhlis H, Karimi M. Mitigating methods of power
partial shading , In: Proceedings of the Power Engineering Conference (AUPEC fluctuation of photovoltaic (PV) sources-A review. Renew Sustain Energy Rev
2009), pp. 1–6; 2009. 2016;59:1170–84.
[19] Kobayashi K, Takano I, Sawada Y. A study of a two stage maximum power point [34] Humada AM, Hojabri M, Mekhilef S, Hamada HM. Solar cell parameters extraction
tracking control of a photovoltaic system under partially shaded insolation based on single and double-diode models: a review. Renew Sustain Energy Rev
condition. Sol Energy Mater Sol Cells 2006;90:2975–88. 2016;56:494–509.
[20] Mirhassaniet SM, Golroodbari SZ, Golroodbari SM, Mekhilef S. An improved [35] Chen K, Tian S, Cheng Y, Bai L. An improved MPPT controller for photovoltaic
particle swarm optimization based maximum power point tracking strategy with system under partial shading condition. IEEE Trans Sustain Energy
variable sampling time. Electr Power Energy Syst 2015;64:761–70. 2014;5(3):978–85.
[21] Ahmed J, Salam Z. A soft computing MPPT for PV system based on Cuckoo Search [36] Rezk H, Eltamaly AM. A comprehensive comparison of different MPPT techniques
algorithm, In: Proceedings of the 4th International Conference on Power for photovoltaic systems. Sol Energy 2015;112:1–11.
Engineering, Energy and Electrical Drives, Istanbul, Turkey, 13-17 May, pp. 558– [37] Boutasseta N. PSO-PI based control of photovoltaic arrays. Int J Comput Appl
562; 2013. 2012;48(17):36–40.
[22] Dileep G, Singh SN. Application of soft computing techniques for maximum power [38] Ishaque K, Salam Z. A deterministic particle swarm optimization maximum power
point tracking of SPV system. Sol Energy 2017;141:182–202. point tracker for photovoltaic system under partial shading condition. IEEE Trans
[23] Chao KH, Lin YS, Lai UD. Improved particle swarm optimization for maximum Ind Electron 2013;60(8):3195–205.
power point tracking in photovoltaic module arrays. Appl Energy [39] Mei Q, Shan M, Liu L, Guerrero JM. A novel improved variable step-size
2015;158:609–18. incremental-resistance mppt method for PV systems. IEEE Trans Ind Electron
[24] Fathabadi H. Novel highly accurate universal maximum power point tracker for 2011;58(6):2427–34.
maximum power extraction from hybrid fuel cell/photovoltaic/wind power gen- [40] Ahmed EM, Shoyama M. Stability study of variable step size incremental
eration systems. Energy 2016;116:402–16. conductance/impedance MPPT for PV systems, In: Proceedings of the IEEE 8th
[25] Kofinas P, Dounis AI, Papadakis G, Assimakopoulos MN. An Intelligent MPPT International Conference on Power Electronics and ECCE Asia (ICPE & ECCE), pp.
controller based on direct neural control for partially shaded PV system. Energy 386–392; 2011.
Build 2015;90:51–64. [41] Ahmed EM, Shoyama M. Novel stability analysis of variable step size incremental
[26] Ahmed J, Salam Z. A critical evaluation on maximum power point tracking resistance INR MPPT for PV systems, In: Proceedings of the IEEE 37th
methods for partial shading in PV systems. Renew Sustain Energy Rev International Conference on Industrial Electronics Society, pp. 3894–3899; 2011.
2015;47:933–53. [42] Yang X-S, Deb S. Cuckoo search via Lévy flights, In: Proceedings of the World
[27] Dileep G, Singh SN. Maximum power point tracking of solar photovoltaic system Congress on Nature & Biologically Inspired Computing (NaBIC 2009), India, pp.
using modified perturbation and observation method. Renew Sustain Energy Rev 210–214; 2009.
2015;50:109–29. [43] Payne RB, Sorenson MD, Kiltz K. The Cuckoos. Oxford University Press; 2005.
[28] Rezk H, Fathy A. Simulation of global MPPT based on teaching–learning-based [44] Yang X-S, Deb S. Multiobjective Cuckoo Search for design optimization. Comput
optimization technique for partially shaded PV system. Electr Eng 2016. http:// Oper Res 2013;40(6):1616–24.
dx.doi.org/10.1007/s00202-016-0449-3, (Oct).

386
 ScienceDirect
Energy Procedia 00 (2019) 000–000
Energy Procedia 00 (2019) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
Availableonline
Available onlineatatwww.sciencedirect.com
www.sciencedirect.com

ScienceDirect
ScienceDirect
Energy
EnergyProcedia
Procedia162 (2019) 000–000
00 (2017) 117–126
www.elsevier.com/locate/procedia
Special Issue on Emerging and Renewable Energy: Generation and Automation
Special Issue on Emerging and Renewable Energy: Generation and Automation
Maximum Power Point Tracking of PV system Based Cuckoo
Maximum Power Point Tracking of PV system Based Cuckoo
Search
The Algorithm;
15th International review
Symposium and comparison
on District Heating and Cooling
Search Algorithm; review and comparison
Assessing the feasibility of using the heat demand-outdoor
temperature function for a long-term district heat demand forecast
Mohamed I. Mosaad* a M. Osama abed el-Raouf b Mahmoud A. Al-Ahmar c Fahd A.
a b c
Mohamed I. Mosaad*
a,b,c
M. Osama
a
abed el-Raouf
Banakher
a
a
b
Mahmoud A. Al-Ahmar c
Fahd
c
A.
I. Andrić *, A. Pina , P. Ferrão , J. Fournier
Yanbu Industrial College, KSA, Banakher
a a ., B. Lacarrière
[email protected], [email protected], O. Le Corre
a Building physics
b a and Industrial
Yanbu environmental research
College, KSA, institute. Housing and Building
[email protected], National research, Cairo, Egypt.
[email protected]
IN+ Centerb for Innovation, Technologyc and Policy Research - Instituto Superior Técnico, Av. Rovisco Pais 1, 1049-001 Lisbon, Portugal
Faculty of engineering, Benha university, Shoubra, Egypt. research, Cairo, Egypt.
Building physics
b and environmental research institute. Housing and Building National
Veolia Recherche
c
Faculty&ofInnovation, 291Benha
engineering, Avenue DreyfousShoubra,
university, Daniel, 78520
Egypt.Limay, France
c
Département Systèmes Énergétiques et Environnement - IMT Atlantique, 4 rue Alfred Kastler, 44300 Nantes, France

Abstract
Abstract
Abstract
The article presents a review about Maximum Power Point Tracking (MPPT) of PV system based cuckoo search algorithm. Cuckoo
District
The
search heating
article
(CS) presents networks
provides a review are
about
several commonly
Maximum
advantages addressed
Power
such thein process
as Point the literature
Tracking of (MPPT)as one
tuning of the
of PV most
system
parameters effective
is based
few cuckoo
with solutions
search
high for decreasing
algorithm.
efficiency Cuckoo
beside the
fast
greenhouse
search (CS) gas
convergence. emissions
provides
Cuckoo searchfrom
severaluses the building
advantages
a random sector.
such
walk as These
the
according systems
process requireinparameters
of tuning
to le’vy flight high investments
searching is fewwhich
process. with
MPPT arebyreturned
high through
efficiency
using cuckoo the fast
beside
search heat
is
sales. Due
convergence.
compared to to thetwo
Cuckoo
other changed
search
methods,climate
uses neural conditions
a randomnetwork and building
walkmethod
according renovation
to le’vy
which needs flight inpolicies,
training for dataheat
searching demand
process.
and MPPTin the
the incremental future
by using could search
cuckoo
conductance decrease,
method.is
prolonging
compared
DC-DC to the
converter investment
other istwo
utilized return
methods, period.
with neural
direct network
duty cyclemethod
controlwhich
of PWM needs training
based for data and
PID controller. ThethePID
incremental
controllerconductance
parameters aremethod.
tuned
The main
DC-DC
using scope
converter
particle of
swarmis this paper
utilized is to
with
optimization assess
direct
(PSO) dutythecycle
and feasibility
controlof
compared ofusing
with PWM thebased
classical heat demand
PID
methods. – outdoor
controller.
The Thetemperature
results PID controller
show that CS function for heat
parameters
can track aredemand
MPP tuned
under
forecast.
using Theswarm
particle
different operatingdistrict of Alvalade,
optimization
conditions with (PSO) located
lower and incompared
power Lisboncompared
losses (Portugal),
with classical wasother
to the usedtwo
methods. asThe
a case study.
results
methods. show The
thatdistrict
CS canistrack
consisted
MPP of 665
under
buildingsoperating
different that varyconditions
in both construction
with lower power periodlosses
and typology.
compared Three weather
to the other twoscenarios
methods. (low, medium, high) and three district
renovation© scenarios
Copyright 2019 Elsevier wereLtd.
developed
All rights (shallow,
reserved. intermediate, deep). To estimate the error, obtained heat demand values were
©compared
2019 Thewith Authors.
resultsPublished
from a by Elsevier
dynamic heat Ltd
demand
Copyrightand
Selection © 2019 Elsevierunder
peer-review Ltd. All of the model,
rights reserved.
responsibility scientific previously
committee developed and validated
of the Special Issue onby the authors.
Emerging and Renewable
Selection
The resultsandshowed
peer-review underonly
thatAutomation.
when responsibility
weather change of theisscientific
considered, committee of the
the margin 6th International
ofSpecial
error could Conference
be Emerging
acceptable foron Emerging
some and
applications
Selection
Energy:
Renewable and peer-review
Generation
Energy: and under
Generation responsibility
and Automation, of the scientific
ICEREGA committee
2018. of the Issue on and Renewable
(the error in annual demand
Energy: Generation and Automation. was lower than 20% for all weather scenarios considered). However, after introducing renovation
scenarios,PV;
Keywords: thecuckoo
error search;
value increased
MPPT; le’vy upflight;
to 59.5%
PSO; (depending
DC-DC converter;on thePID weather and renovation scenarios combination considered).
controller.
The valuePV;
Keywords: ofcuckoo
slope search;
coefficient
MPPT; increased onPSO;
le’vy flight; average
DC-DC within the range
converter; of 3.8% up to 8% per decade, that corresponds to the
PID controller.
decrease in the number of heating hours of 22-139h during the heating season (depending on the combination of weather and
renovation scenarios
Corresponding author: Yanbuconsidered).
Industrial On the (YIC),
College other hand,
Alnahdah,function
Yanbuintercept increased
Al Sinaiyah, for 7.8-12.7%
Yanbu 46452, per decade (depending on the
Saudi Arabia.
E-mail addresses:
coupled
Corresponding [email protected],
scenarios).
author: The values
Yanbu Industrial [email protected]
suggested couldAlnahdah,
(YIC), be used Yanbuto(M.I. Mosaad).
modify
Al the function
Sinaiyah, parameters
Yanbu 46452, for the scenarios considered, and
Saudi Arabia.
E-mail addresses:
improve [email protected],
the accuracy [email protected] (M.I. Mosaad).
of heat demand estimations.

© 2017 The Authors. Published by Elsevier Ltd.

Peer-review
1876-6102 under ©
Copyright responsibility
2019 Elsevierof theAllScientific
Ltd. Committee of The 15th International Symposium on District Heating and
rights reserved.
Cooling.and
Selection
1876-6102 peer-review
Copyright under
© 2019 responsibility
Elsevier Ltd. All of the scientific
rights reserved. committee of the Special Issue on Emerging and Renewable Energy: Generation
and Automation.
Selection and peer-review under responsibility of the scientific committee of the Special Issue on Emerging and Renewable Energy: Generation
and Automation.
Keywords: Heat demand; Forecast; Climate change

1876-6102 © 2017 The Authors. Published by Elsevier Ltd.

Peer-review under responsibility of the Scientific Committee of The 15th International Symposium on District Heating and Cooling.
1876-6102 © 2019 The Authors. Published by Elsevier Ltd.
Selection and peer-review under responsibility of the scientific committee of the 6th International Conference on Emerging and
Renewable Energy: Generation and Automation, ICEREGA 2018.
10.1016/j.egypro.2019.04.013
118 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126
2 M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000

1. Introduction

Due to environmental problems and energy crises, there is an increase in the world's attention in the field of
clean and renewable energy generation. Solar energy generation is the most promising renewable energy technologies
[1]. In the last few decades, PV systems became common in grid-connected applications and had an important role in
power generation in the new century [2]. PV system converts sun light to electricity directly when photons of the sun
light hit on the PV array. The performance of solar PV system is powerfully relaying on the operating conditions like
sun’s geometric location, the ambient temperature and the irradiation levels of sun. PV system must be operated at
maximum power to get more efficiency, so that the maximum power point (MPPT) tracking is utilized. In PV cell,
P/V and P/I characteristics are non-linear relations, so there are several search methods are popularly used [2-4].
However, there are some problems when using these methods if there are rapid changes in irradiance and temperature.
Moreover they cause huge power losses and inability to handle partial shading conditions. These problems can be
handled by using artificial intelligence MPPT methods as artificial neural networks [5], Fuzzy logic controller [6],
Genetic Algorithm [7], Differential Evaluation [8] and particle swarm optimization [9].
Recently many researches start on a new topology called cuckoo search optimization (CS), [10]. This method
has a several advantages over conventional methods and over others methods [10].
In this paper, CS for MPPT of PV system is applied and compared with two methods, namely artificial neural
network (ANN) and incremental conductance methods (IC). To ensure the capability of the proposed search method,
CS, different operating conditions are considered and tested. PID controller tuned by PSO to control the duty cycle of
the DC-DC converter to guarantee MPP is applied.

2. Problem Formulation
The proposed system, Fig. 1, comprises PV module feeding DC-DC converter and its duty ratio is controlled using
PID controller to maintain maximum power operation of the PV. CS is used for MPPT with different ecological
conditions represented by solar irradiation and cell temperature to ensure the capability of that method to track MPP
of the module. The reference voltage calculated from MPPT algorithm tracked by CS is generated at each temperature
and irradiation level, Vref. The difference between Vref and the measured output voltage from PV, Vpv, is calculated to
produce the error voltage, Verror that is used to stimulate the PID controller. The output of PID controller is compared
to saw-tooth waveform to produce the duty cycle for DC-DC converter. PID control parameters are tuned using PSO
to obtain better performance [11].

Fig.1. Overall system of PV with MPPT using CS.

 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126 119
M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000 3

2.1. PV Modeling
The commonly used equivalent circuit of PV solar cell is illustrated in Fig. 2. The model composed of a light
dependent current source in parallel with an equivalent diode structure [12]. The output of the current source is directly
proportional to the light falling on the cell. The solar cell fails to maintain a fixed current as the load resistance
increases. The output current reaches to zero when the load resistance becomes very large. The PV model included
temperature dependence of photo current IL, the saturation current of the diode I0 and RS & Rp series and parallel
resistances respectively as depicted in Fig.2.
In this paper, the solarex MSX60, 60 W panel will be used [13]. The specifications of the solarex MSX60, 60 W PV
panel at steady state conditions (STC) of 250c and 1000 W/m2 irradiation are shown in table1, [14].
The characteristics of the MSX60, 60W PV cell are plotted in Figs. 3 and 4.

Fig.2. Solar cell equivalent circuit

Table 1: Specifications of Solarex MSX60, 60 W PV panel.

Open circuit voltage VOC 21.0 V
Short circuit current Isc 3.74 A
Voltage at max. power Vm 17.1 V
Current at max. power Im 3.5 A
Maximum power Pm 59.9 W

(a) (b)

Fig. 3. PV characteristics at STC

a- P-V characteristics b- I-V characteristics.

2.2. DC-DC converter

The purpose of the DC-DC converter is to transfer the maximum power from Module to the load. There are some
types of DC-DC converters such as buck, boost and buck-Boost. In this paper, the boost converter is used. In the boost
converter, the output voltage is higher than the input voltage. The topology of the boost power stage is depicted in
Fig.4, [15].

Fig.4 boost converter schematic

120 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126
4 M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000

The principle operation of the boost converter is:

a) In on-state, the switch S is closed resulting in increased the inductor current.
b) In off-state, the switch is opened and the path to the conductor is through the diode D, load R and C.
The relation between the output and input voltages (VO and Vi) respectively, is
𝑉𝑉0 1
𝑉𝑉𝑉𝑉
= 1−𝐷𝐷 (1)
where D is the duty cycle.
As the duty, cycle ranges from zero to one, and then the output voltage is always higher than the input voltage as in
equation (1).
The parameters of DC-DC boost converter used in this paper are shown in table 2.

Table 2: parameters of DC-DC converter

Parameter value
Resistor R 50 Ω
Inductor L 120 µH
Capacitor C 330 µF
Capacitor C1 330 µF

3. Cuckoo Search

Cuckoos are fascinating birds, due to their beautiful sounds, and their aggressive reproduction strategy. There are
some types of cuckoo birds like ani and guira. These types lay their eggs in communal nests and these may remove
others’ eggs to increase the hatching probability of their own eggs [16]. Brood parasitism is the behaviour of some
cuckoo birds, Tapera are intelligent birds that imitates the host birds in shape and colour, that may led to
increase reproduction probability. It is surprising and magnificence to look at the timing of the process of laying eggs
for Tapera. Firstly, cuckoos female select a group of host species with similar nest sites and egg characteristics to their
own, then choosing the best from these nests. Host birds could be fooled and accept foreign eggs but if these eggs
discovered, they are dumped outside the nest or the nest completely destroyed and go to new area to build a new nest.
Usually there are three kinds of brood parasitism namely intraspecific, cooperative and nest takeover [17].

3.1 Le’vy flight

Cuckoo birds start in looking for the best nest, but this step has an important role in cuckoo’s reproduction method.
The process of searching the nest is the same as the process of searching food, the walks and directions are chosen
and modelled on certain mathematical functions, le’vy flight is one of the most common models.
A recent study by Reynolds and Frey shows that the fruit flies or Drosophila melanogaster, explore their landscape
using a series of straight flight paths punctuated by sudden 90 0 turn leading to a le’vy flight style [18]. This behavior
is used in making optimization for different problems [19]. le’vy flight is a random walk in which the step lengths
have a probability distribution and steps are defined in terms of the step lengths. In CS, the step length is drawn from
le’vy distribution according to power law [20], i.e.
Le’vy ∼ u =t−λ where (1 < λ ≤ 3), (2)
Here the steps essentially form a random walk process with a power-law step-length distribution with a
heavy tail [21]. Le’vy flight in two-dimensional plan is shown in Fig.5.

Fig.5. Le’vy Flight

 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126 121
M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000 5

3.2 Cuckoo Search Algorithm

There are three idealized rules used for CS have been used by Yang and Deb. [21]. These rules are:
1. Each cuckoo lays one egg at a time and places it in a randomly chosen nest.
2. The best nest with high quality of eggs will carry over to the next generation.
3. The number of available nests is fixed and the egg laid by a cuckoo is discovered by the host bird with a probability
of Pa, where Pa ∈ [0, 1].

If the host birds are discovered the cuckoo’s eggs, the host bird can abandon its nest or destroy cuckoos’ eggs or in
this case a new nest will be generated with probability Pa. In simple form, the last assumption can be approximated
by the fraction Pa and the n nests are replaced by new nests (with new random solutions). In a simple form the
following simple representation is used, each egg represents a solution and a cuckoo egg represents a new solution.
In this this work, a simple approach is used where each nest has only a single egg. CS can be summarized as pseudo
code as shown in Fig.6.

Begin
Objective function F(x), x=(x1,………xn)t.
Generate initial population of size n.
Host nests Xi (i=1, 2,… n).
While (t< max generation) or (stop criterion)
Get a cuckoo randomly by le’vy flight.
Evaluate its quality / fitness value, Fi.
Choose a nest among n (say, j) randomly.
If (Fi >Fj)
Replace j by the new solution;
End
A fraction (pa) of worse nests is abandoned;
In addition, new ones are built;
Keep the best solution.
(Or nests with quality solutions);
Rank the solutions are find the current best
End while.
Print the results
End

Fig.6.pseudo code of cuckoo search (CS).

When generating a new solutions X(t+1) for a cuckoo i a levy flight is performed
Xi(t+1) =Xi(t) + α ⊕ le’vy flight (λ) (3)
Where α > 0, is the step size related to the scales of the optimization problem, in most cases it is assumed that α=1,
[22]. Equation (3) consists of two terms, the first term is the current location and the second term is the transition
probability. The product ⊕ means an entry wise multiplication, which is similar to those used in PSO. The random
step length is drawn from equation (2).
For MPPT Algorithm there is a similarity between CS and Hill climbing in some large scale randomization, but
there are some differences such as
1. CS depends on the population as in both PSO and GA, but it uses some sort of elitism and/or selection similar to
that used in harmony search.
2. The possibility for any large step in addition to the step length has weighty tail that led to the process of
randomization is more efficient.
3. CS has less tuned parameters compared to GA or PSO, so that CS can thus be extended to the type of Meta-
population Algorithm.

4. MPPT Algorithm

In this work, MPPT using cuckoo search is proposed. In cuckoo search algorithm proposed, there is a random initial
solution of the operating voltage of PV panel, Va, is equal to [0, 25, 0].Va according to the CS procedure represents
122 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126
6 M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000

the nest. The fitness value of power is calculated as in (4)

Pa=Va×Ia (4)
Then find the best current, choose a random nest and generate a new solution by random walk as
Vi (t+1) = Vi (t) + α ⊕ L´evy (λ), (5)
A simplified scheme of a random walk is used which is less efficient than levy flights. In addition to the step size
should be a vector for problems with different scales, with step size=0.05, [23].
After new solutions are generated, values of the fitness are tested again and best current is selected. Iterations are
performed until all nests reach at the maximum power point. A MATALB code is developed to search MPP using CS
of for PV 60W module.
The PV voltages and currents are measured using voltage and current sensors, respectively. The reference voltage
calculated from MPPT algorithms tracked by CS is generated at each temperature and irradiation level, V ref. The
difference between Vref and the measured output voltage from PV, Vpv, is calculated to produce the error voltage, Verror.
This error is used as an input signal to the PID controller, the output of the PID controller is compared to saw tooth
waveform to produce the duty cycle for DC-DC converter. The PID controller is preferable over the P or PI controller
[24]. Trial and error method is used for tuning the parameters of PID controller. This method is not guaranteed to
calculate the controller coefficients so that artificial intelligent are used for tuning PID control parameters [25]. There
are several methods used for tuning PID control parameters [26-29]. In this work, PSO as one of AI techniques will
be used [30-34].

5. Simulation Results

5.1 CS MPPT

The simulation of the overall system depicted in Fig.1 is carried out using Matlab/Simulink. The calculated MPP of
PV module at STC conditions is 60.4728 W. The Cuckoo search parameters are set as the following: number of
iterations is Ngen=100, number of nests, which represent different solutions, are n=25 and the probability (discovery
rate of alien eggs/solutions), is pa=0.25.
The current, voltage and power of PV module using CS for MPPT are shown in Fig. 7.

(a) (b)
(a)

(c)

Fig.7. PV charatcteritcis with MPPT using CS at STC

a- Current b- Voltage C-power.
 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126 123
M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000 7

5.2 MPPT Algorithms

To evaluate the performance of the proposed CS MPPT, the conventional incremental conductance method is also
used for the same simulated conditions. This method used as a benchmark, we compare the results of cuckoo search
not only with this method but also with a neural network MPPT at the same conditions.
The performance of the algorithms (incremental conductance, neural network and CS) are compared at STC, i.e. T=25
C and G=1000 W/m2.
From Fig.11, the CS succeeds in tracking the power more than neural network and incremental conductance as shown
in table 3.
At steady state the IC reaches the MPP at time 0.1 s. For the time from t=0 to t= 0.1s, the simulation output power
from PV has overshoot, after the time t = 0.1 s the waveforms vibrate up and down regularly.
ANN algorithm reaches the MPP at t= 0.021 s. for the time form t=0 to t=0.021s, the output power waveforms has
overshoot, after this time the waveform oscillations are very smaller than IC and the tracked power is more accurate.
CS reaches the MPP faster than NN and IC. From starting, the system almost stable and operating nearly at the MPP.
CS keeps tracking the exact MPP, which is considered an extra feature that leads to higher MPPT efficiency tracked
power which equal 60.4728 watt. As incremental conductance keeps on swinging around MPP as in Fig. 8, this causes
a losing energy. In case of neural energy, loss is smaller, but in CS, power loss near zero at steady state.
Table 3: power tracked by (CS, ANN and IC)
Algorithm Tracked power
Cuckoo Search (CS) 60.4728 W
Neural Network (NN) 60.19 W
Incremental Conductance (INC) 58.89 W

Fig.8. MPPT using CS, IC and ANN.

At steady state the IC reaches the MPP at time 0.1 s. For the time from t=0 to t= 0.1s, the simulation output power
from PV has overshoot, after the time t = 0.1 s the waveforms vibrate up and down regularly.
ANN algorithm reaches the MPP at t= 0.021 s. for the time form t=0 to t=0.021s, the output power waveforms has
overshoot, after this time the waveform oscillations are very smaller than IC and the tracked power is more accurate.
CS reaches the MPP faster than NN and IC. From starting, the system almost stable and operating nearly at the MPP.
CS keeps tracking the exact MPP, which is considered an extra feature that leads to higher MPPT efficiency tracked
power which equal 60.4728 watt. As incremental conductance keeps on swinging around MPP as in Fig. 8, this causes
a losing energy. In case of neural energy, loss is smaller, but in CS, power loss near zero at steady state.

5.3 DC-DC converter with MPPT and PID controller

The tracked maximum power and the corresponding duty cycle of the DC-DC converter used are shown in Figs.9 a
and b respectively. There are several methods used for tuning the PID control parameters [35]. In this work the try
and error method is used, but this method is not guaranteed to obtain accurate and optimized values. So PSO as one
of the artificial intelligence optimization techniques is used. The objective is to achieve minimal settling time in the
step response of the process. By tuning KP, Ki and Kd values in the transfer function of the controller given by
𝑈𝑈(𝑠𝑠) 𝐾𝐾𝐾𝐾
G(s) = = 𝐾𝐾𝑝𝑝 + + 𝐾𝐾𝑑𝑑 𝑆𝑆 (6)
𝐸𝐸(𝑠𝑠) 𝑠𝑠
124 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126
8 M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000

where 𝐾𝐾𝑝𝑝 , 𝐾𝐾𝑖𝑖 𝑎𝑎𝑎𝑎𝑎𝑎 𝐾𝐾𝑑𝑑 are the proportional, integral and differential gains respectively.
𝐸𝐸(𝑠𝑠): laplace transform of the input to the controller.
𝑈𝑈(𝑠𝑠): laplace transform of the output to the controller.
Particle swarm optimization technique is used, in which a three dimensional solution spaces are generated according
to the possible range of KP, Ki and Kd in Ziegler Nichols method. The initial position of the ith particle of the swarm is
𝑋𝑋𝑖𝑖 𝑖𝑖𝑖𝑖 = (𝑋𝑋𝑋𝑋1, 𝑋𝑋𝑋𝑋2, 𝑋𝑋𝑋𝑋3)𝑡𝑡 (7)
When the block is implemented in MATLAB/Simulink, the best values of Kd, Ki and KP are 1.9688, 0.0504 and 0.630
respectively.

(a)
(b)

Fig.9. MPPT using PID controller

a- Duty cycle of DC-DC converter b- Power tracked of PV module.

5.4 Step changes in irradiance with constant temperature

Irradiance and temperature increases in atypical sunny day, then decreases towards the evening. A set of irradiance
steps are imposed on PV module as shown in table 4.

Table 4: MPPT with step changes

Time(s) 0 – 0.2 0.2 – 0.4 0.4 – 0.6
G(w/m2) 1000 800 1000
P at MPPT(w) CS 60.47 48.24 60.47
P at MPPT(w) NN 60.19 48.00 60.19
P at MPPT(w) IC 58.89 46.67 58.89

Step response of cuckoo search, neural and incremental conductance is illustrated in Fig 10. From this Figure, CS
continuously sticks to the MPP. The faster response for CS over the neural and incremental conductance is due to the
larger step size according to the simplified le’vy flight used in code. It is shown from these results that CS has the
ability to track MPP under step changes. Neural also has this ability with lower efficiency than cuckoo. IC method
fall to track system under these changes so cuckoo is the best.

Fig. 10. MPPT using CS, IC, ANN with irradiance step change
 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126 125
M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000 9

5.5 Rapid Atmospheric change

The system is tested under rapid atmospheric changes; the irradiance and temperature are varied with a time 0.2 s .The
test patterns are shown in table 5 and also plotted in Figs. 11 a and b.

Table5: test conditions for rapid changes

Time Temperature Irradiance w/m2 MPPT (W)
0 to 0.2 s 25 1000 60.4728
0.2 to 0.4 s 15 600 38.8
0.4 to 0.6 s 20 800 49.95

(b)
(a)

Fig.11. Test pattern for step change

a- Irradiance b- Temperature

It is shown in Fig.12 that CS can track the MPP of PV array with fast changes of temperature and irradiance. It is
observed that cuckoo search sticks to MPP without any further fluctuations.

Fig.12. Rapid tracking of CS

6. Conclusions

This paper presents MPPT algorithm employing Cuckoo search algorithm, depending on the cuckoo natural behaviour
and simplified form of levy flight distribution. Results show that cuckoo is capable of tracking the MPP efficiently. It
is confirmed by comparing results with other methods such namely, incremental conductance and artificial neural
network. Cuckoo search gave maximum power greater than incremental conductance and artificial network. Moreover
in cuckoo search method, no further fluctuations.
126 Mohamed I. Mosaad et al. / Energy Procedia 162 (2019) 117–126
10 M. Osama abed el-Raouf et al./ Energy Procedia 00 (2019) 000–000

References

[1] Mukund R. Patel, “Wind and Solar Power Systems”, Ed CRC Press, New York, 1999.
[2] W J Praiselina and J Belwin Edward, “Voltage profile improvement of Solar PV Grid – Connected Inverter with Micro Grid Operation using
PI Controller”, Energy Procedia, Vol.117, June 2017, pp.104-111.
[3] Osisioma Ezinwanne, Fu Zhongwen, Li Zhijun “Energy Performance and Cost Comparison of MPPT Techniques for Photovoltaics and other
Applications” Energy Procedia, Vol. 107, February 2017, P. 297-303.
[4] Fu Wang, Yingming Zhu, Jinyue Yan,” Performance of solar PV micro-grid systems: A comparison study” Energy Procedia,
Vol. 145, July 2018, P. 570-575.
[5] Loubna Bouselham, Mohammed Hajji, Bekkay Hajji, Hicham Bouali “A New MPPT-based ANN for Photovoltaic System under Partial
Shading Conditions Energy Procedia, vol. 111, March 2017, P. 924-933.
[6] C. Chian-Song, “T-S Fuzzy maximum power point tracking control of solar power generation systems,” IEEE Energy Conv. Trans.vol.25,pp
1123-1132, 2010.
[7] R. Ramaprabha, V. Gothandaraman, K. Kanimozhi, R. Divya and B.L. Mathur, “ Maximum power point tracking using GA-Optimized artifitial
neural network for Solar PV system,” 1st Int. Conf. on IEEE Elec. Energy Sys. p. 264-268, 2011.
[8] H. Taheri, Z. Salam,K. Ishaque and Syafaruddin, “A novel maximum power point tracking control of photovoltaic system under partial and
rapidly fluctuating shadow conditions using Differential Evolution,” IEEE Indust. Electron. & App. Symp. p. 82-87, 2010.
[9] M. Osama abed el-Raouf, Mohamed I. Mosaad, Mahmoud A. Al-Ahmar and Fahmy M. El Bendary “MPPT of Hybrid solar-wind-grid power
generation system” Int. J. Industrial Electronics and Drives-Inderscience Publishers, Vol. 2, No. 4, 2015, p. 234-241.
[10] Yang X. S. and Deb S, Engineering Optimization by cuckoo search, Int. J. Math. Modelling& Numerical Optimization, 1 330-343 2010.
[11] S. Easter Selvan, Sethu Subramanian，S. Theban Solomon，Novel Technique for PID Tuning by Particle Swarm Optimization, in Proc.7th
Annul. Swarm Users/Researchers Conf. (Swarm Fest 2003), Notre Dame, IN, 2003.
[12] T. Markvart, Solar Electricity, John Wiley & Sons, 1994.
[13] M. Osama abed el-Raouf, Mohamed I. Mosaad, Adel Mallawany, Mahmoud A. Al-Ahmar and Fahmy M. El Bendary “MPPT of PV-Wind-
Fuel cell of off-grid Hybrid System for a New Community” Twentieth International Middle East Power Systems Conference (MEPCON),
480 – 487, 2018.
[14] Solarex data sheets, www.solarex.com.
[15] UL1741, Inverter, Converter, and Controllers for Use in Independent Power System.
[16] Payne R. B., Sorenson M. D., and klitz K., THE Cuckoos, OXFORD University PRESS, (2005).
[17] Schlesinger M. F., Search research, Nature, 443, p.281-282, 2006.
[18] A. M. Reynolds and M. A. Frye, “Free -flight odor tracking in Drosophila is consistent with an optimal intermittent scale-free search.” PloS
one. 2, e.354, 2007.
[19] R. Rajabioun, “Cuckoo Optimization Algorithm,” Applied Soft Computing. Vol. 11, pp. 5508-5518, 2011.
[20] A.M.Z. Alabedin, E. F. El-Saadany and M. M. A.Salama, “Maximum power point tracking for photovoltaic systems using fuzzy logic and
artificial neural networks,” IEEE Power and Energy Soc. Gen.Meet. pp. 1-9, 2011.
[21] X.-S. Yang, S. Deb, “Cuckoo search via L´evy flights”, in: Proc. of World Congress on Nature & Biologically Inspired Computing (NaBIC
2009), December 2009, India. IEEE Publications, USA, pp. 210-214 (2009).
[22] Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Comput oper Res 2011.
[23] X.S. Yang Nature-inspired metaheuristic algorithms, luniver, (2008).
[24] Viagundamoorthi M, Ramesh Experimental investigation of chaos in input regulated solar PV powered cuk converter. Int J compute APPL
2012; 34:11-6.
[25] S.Easter Selvan Sethu Subramanian, S.Theban Solomon, Novel Technique for PID Tuning by Particle Swarm Optimization, in Proc7th Annul.
Swarm Users/Researchers Conf. (Swarm Fest 2003), Notre Dame, IN, 2003.
[26] Mohamed I. Mosaad, Nagy I. Elkalashy and Mohamed G. Ashmawy, “Integrating Adaptive Control of Renewable Distributed Switched
Reluctance Generation and Feeder Protection Coordination “ Electric Power System Research journal, Elsevier, Vol. 154, January 2018, P.
452–462.
[27] Mohamed I. Mosaad “Model reference Adaptive Control of STATCOM for Grid-Integration of Wind Energy Systems” IET Electric Power
Applications Journal, Vol. 12, Issue 5, May 2018, pp. 605 – 613.
[28] Hang C.C., J.K. Astrom, W.K. Ho;”Refinements of Ziegler Nichols Tuning formula”, IEE Proceedings, 138(2), 111(1991).
[29] Fawzan Salam, Mohamed I. Mosaad “A Comparison Between MPC and Optimal Pid Controllers: Case Studies” IET conference, Michael
Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India,pp. 59-65.
[30] Gaing, Z.L. (2004). A particle swarm optimization approach for optimum design of PID controller in AVR system. IEEE Transaction on
Energy Conversion, Vol.19 (2), pp.384-391.
[31] Zhao, J., Li, T. and Qian, J. (2005). Application of particle swarm optimization algorithm on robust PID controller tuning. Advances in
Natural Computation: Book Chapter. Springer Berlin / Heidelberg, pp. 948-957.
[32] Pratik Shantaram Gavhane, Smriti Krishnamurthy, Ridhima Dixit, J. Prasanth Ram, N. Rajasekar, “EL-PSO based MPPT for Solar PV under
Partial Shaded Condition” Energy Procedia, Vol. 117, June 2017, P. 1047-1053.
[33] M.M.El Metwally, A.A.El Emary, F.M.El Bendary and M.I.Mosaad “Optimal Power Flow Using Evolutionary programming Techniques”
International Middle East Power System Conference MEPCON 2008, pp260-264.
[34 M.I. Mosaad, M.M. ElMetwally, A.A. ElEmary, F.M. Bendary, On line optimal power flow using evolutionary programming techniques,
Thammasat International Journal of Science and Technology 15 (1) (2010) 20-28.
[35] Mohamed I. Mosaad and H.S. Ramadan “Power Quality Enhancement of Grid Connected Fuel Cell Using Evolutionary Computing
Techniques”, International Journal of Hydrogen, Volume 43, Issue 25, 21 June 2018, Pages 11568-11582.
View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by Brunel University Research Archive

Comparison Between the Conventional Methods

and PSO Based MPPT Algorithm for
Photovoltaic Systems.
Ramdan B. A. Koad, Ahmed. F. Zobaa

 development that has occurred in the semiconductor field,

Abstract—Since the output characteristics of photovoltaic (PV) which has made it possible to increase the energy output to
system depends on the ambient temperature, solar radiation and load meet the required load power. Thus PV systems can replace
impedance, its maximum power point (MPP) is not constant. Under conventional energy sources in the future as a result of its
each condition PV module has a point at which it can produce its
MPP. Therefore, a maximum power point tracking (MPPT) method is
inexhaustible source. Furthermore, it is clean without pollution
needed to uphold the PV panel operating at its MPP. This paper and has no moving parts which reduce the cost of maintenance
presents comparative study between the conventional MPPT methods [2]. Since the output characteristics of the PV system depends
used in (PV) system: Perturb and Observe (P&O), Incremental on the ambient temperature, solar radiation and load
Conductance (IncCond), and Particle Swarm Optimization (PSO) impedance, it is important to operate the PV panel at its
algorithm for (MPPT) of (PV) system. To evaluate the study, the maximum power point (MPP) which as explained is not
proposed PSO MPPT is implemented on a DC-DC cuk converter and
has been compared with P&O and INcond methods in terms of their
constant as it varies with the weather conditions [3].
tracking speed, accuracy and performance by using the Matlab tool Several MPPT methods have been developed in relation to
Simulink. The simulation result shows that the proposed algorithm is PV systems in order to reach the MPP. These range from
simple, and is superior to the P&O and IncCond methods. using simple methods to more complex analysis depending on
the weather conditions and the application [7]-[9]. The main
Keywords— Incremental Conductance (IncCond) Method, aim of the MPPT is to extract maximum output power from
Perturb and Observe (P&O) Method, Photovoltaic Systems (PV) and
the PV module under different sunlight radiation and
Practical Swarm Optimization Algorithm (PSO).
temperatures. In this survey, the Perturb and Observe (P&O),
I. INTRODUCTION the Incremental Conductance (IncCond) and Practical Swarm
Optimization (PSO) MPPT algorithms are presented and
R enewable energy sources have many advantages over
conventional energy sources, as they are green, do not
emit carbon dioxide and in many cases are sustainable.
compared under different atmospheric conditions.

II. THE TERMINAL CHARACTERISTICS OF PV CELLS

Consequently, there is much interest in using renewable
energy as a solution of the problems created by burning finite Fig. 1 shows the equivalent-circuit diagram of PV cell that
fossil fuels. However, the use of renewable energy still has a consist of a source current (Iph), a diode (D), and series and
number of limitations, as most renewable energy sources parallel resistances (Rs, Rp).
depend on the weather conditions, such as wind in wind power
generation, rain in hydropower and clear skies in photovoltaic
(PV) systems. Furthermore, the cost of renewable energy
sources is higher than the conventional energy sources when
generating large volumes of energy. Therefore, the main issue
in renewable energy research is to reduce the cost and increase
the efficiency of production. In recent years, wind power and
PV energy have been the two main areas of the research and
development. However, much work needs to be done in this
field in order to make renewables as efficient and reliable as Fig. 1 The equivalent-circuit of PV cell model [4].
possible [1]-[2]. Nevertheless, the PV system has become an
important source for generating electricity due in part to the The output current-voltage (I-V) characteristics can be
calculated by using the following equation:
.
Ramdan B A Koad., is with Brunel University, Uxbridge, UB8 3PH, / 1 . (1)
Middlesex, United |Kingdom. He is currently pursuing the Ph.D. degree in
renewable energy sources. (email: [email protected] ).
A. F. Zobaa, he was a Senior Lecturer in renewable energy with the Where, I , V - the PV cell output current and voltage
University of Exeter, Cornwall, U.K. Currently, he is a Senior Lecturer in respectively, I is the photon generated current, and Io is the
power systems with Brunel University, Uxbridge, U.K. ph
diode reverse saturation currents. A is the ideality factors, T is
the cell temperature in Kelvin, k is the Boltzmann’s constant
-23
(K=1.380 x 10 J/K) and q is the Electronic charge
=1.6×10−19C). The selected PV module for this study is BP
Solar BP SX 150S PV module, and it is able to generate an
output power of 60 watt. Its electrical specifications are shown
in (see Table I).
TABLE I
ELECTRICAL SPECIFICATIONS OF THE SIMULATED PV MODULE
Maximum Power (Pmax) 60 W

Voltage @ Pmax (Vmp) 17.1 V

Current @ Pmax (Imp) 3.5 A

Open-circuit voltage (Voc) 21.1 V

Fig. 3 I-V Characteristics with different irradiance values and a
Short-circuit current (Isc) 3.8 A constant temperature (25°C).
Temperature coefficient of Short- From the above figures, it is clear that the PV module has a
.(0.065±0.015)%/°C
circuit current (Isc) non-linear characteristic which differs according to solar
Temperature coefficient of Open- radiation, temperature and the load condition. Each curve has
–(80±10)mV/°C
circuit voltage (Voc)
a different point at which the module can produce its
Temperature coefficient of power –(0.5±0.05)%/°C maximum power. Hence, in order to overcome this problem, a
MPPT controller is required. The major job in an MPPT
III. THE PV MODULE PERFORMANCE system is to choose and design a high efficiency DC-DC
converter that can operate the PV module at its MPP.
By using (1), and the electrical specifications of the PV
module, the current-voltage (I-V) characteristics at different IV. CUK CONVERTER
environmental conditions, temperature and irradiance are
Fig. 5 illustrates the Cùk converter circuit diagram which
displayed in Fig. 2and Fig. 3.
was designed according to the electric specification is shown
.
in TABLE II.

Fig. 4 Basic electrical circuit of DC-DC cùk converter [4].

The voltage transfer function can be written as following;

2
1

TABLE II
THE ELECTRIC SPECIFICATION OF CUK CONVERTER
Specification
Input Voltage (Vs) 12-18V
Fig. 2 I-V Characteristics with variable temperatures and constant
irradiance (1KW/m2). Input Current (Is) 0-5A(<5% ripple)
Output Voltage (Vout) 40V(<5% ripple)
Output Current (Io) 0-5A(<5% ripple)
Maximum Output Power (Pmax) 60W
Switching Frequency (f) 10KHz
Duty Cycle (D) 0.6≤D≤1
 V. TECHNIQUES OF MAXIMUM POWER POINT dI
TRACKING 9
dV V
A. Perturbation and Observation algorithm (P&O)
These equations show that the PV module operates at its
This technique is based on the relationship between the PV MPP when the IncCond dI/dV is equal to its direct
module output power and the its output voltage, and the MPP conductance -I/V. while if the PV module IncCond dI/dV is
is obtained by adjusting the switching mood of the converter greater than the its conductance -I/V, then the controller would
(duty ratio) until dp/dv is equal to zero, The drawback of P&O increase the PV module voltage by adjusting the duty ratio of
methods are that it produces osculation around the MPP in the a DC –DC converter, otherwise, the perturbation would be in
steady state. Reference [5] shows that a continuous oscillation the opposite direction or to increase the duty ratio of the
in P&O methods in the steady state causes a reduction in the converter in order to reduce the voltage and shift the operating
PV module output power. In addition it cannot operate the point back to the MPP [8]-[9]. Figure 8 shows the flowchart
module at its maximum output power in rapidly changing of of the IncCond algorithm.
weather conditions, the flowchart of the P&O algorithm is
depicted in Fig. 8[6]-[7]-[8].
Start

Measure Vk,Ik

P=Ik*Vk
dP=Pk-P(k-1)

Yes No Yes
Yes
dI/dV =0 dV= 0 dI=0
No No
Increase Module Decrease Decrease Increase Module
Voltage Module Voltage Module Voltage Voltage

Update History
Fig. 6. Flowchart of the IncCond algorithm
Fig. 5 Flowchart of P&O algorithm.
C. Over View of the Particle Swarm Optimization Algorithm
B. Incremental Conductance (IncCond) Algorithm Particle swarm optimization (PSO) is an intelligence
The Incremental Conductance (IncCond) algorithm was optimization theory was developed by Eberhart and Kennedy
developed to overcome the drawback of the P&O method in 1995. The principle of this algorithm was inspired from the
under rapidly changing weather conditions. The relationship foraging behavior of birds and fish schooling, and the two
between the voltage and power can be expressed as follows; scholars were applied this phenomenon to overcome the
dP problems associated with search and optimization. In this
0 3 algorithm, several cooperative birds are used, and each bird,
dV
dP referred to as a particle, each particle flying in the space has its
0 4
dV own fitness value that mapped by an objective function and
dP velocity which uses to decide the direction and distance of
0 5
dV their movement. Each particle exchanges information
obtained in its respective search process. The typical process
The P-V characteristic slope (dP/dV) can be calculated using of optimization the particles are shown in fig.1 [10]-[11]-[12].
the PV module output voltage and its output current as
follows:
dP dV dV dI dV
6
dV dV dV

Hence, the PV module operating point at its maximum output

power can be calculated based on (16) as follows;
dI I
7
dV V
dI I
8
dV V
Fig. 7.Movement of a PSO particle.
 The movement of particles impact by two variables; the Pbest Step1. Parameter Selection
that used to store the best position of each particle as an For the proposed MPPT algorithm, the duty cycle of the
individual best position, and the Gbest that found by comparing converter was defined as the particle position, and the
individual positions of the particle swarm and store it as best generated output power was chosen to be the fitness value
position of the swarm. The particle swarm uses this process to evaluation function, the position and initial velocity of each
move towards the best position and continuously it revise its particle was randomly initialized in a uniform distribution
direction and velocity, by this way, each particle quickly over the search space.
converge to an optimal or close to a global optimum. The Step2. Fitness Evaluation
standard PSO method can be defined by the following The fitness value of particle i, is computed after the
equations; controller sent the duty cycle command which represents the
position of particle i.
1 . . 10 Step3. (Update Individual and Global Best Data): Update
the fitness values, individual best positions (Pbest ) and global
1 1 11 best fitness values (gbest) of each particle by comparing the
i=1, 2…, N new calculated fitness values against the previous ones and
where xi and vi are the velocity and position of particle i, ; k replacing the pbest, and gbest corresponding to their positions
represents the iteration number; w is the inertia weight; r1 ,r2 as necessary.
are random variables and their values are uniformly Step4. (Update Velocity and Position of Each Particle):
distributed between [0,1]; c1, c2 represents the cognitive and After evaluating all particles, update the velocities and
social coefficient respectively. pbest,i is the individual best positions of each particle in the swarm by using the PSO
position of particle i, and gbest is the swarm best position of all formulas (1) and (2).
the particles. If the condition (14) of initialization was Step5. (Convergence Determination): The converge
satisfied, the method updated like (13) criterion are either locating to optimal solution or reaching the
pbesti = xik (13) maximum number of iterations. If the convergence criterion is
f(xik) > f(pbesti ) (14) met, the process would terminate; otherwise, rerun Steps 2
through 5.
Where f represents the objective function that should be Step6. (Reinitialization); in standard PSO method the
maximized. converge criterion are either locating to optimal solution or
The basic operating principle of this method can be explained reaching the maximum number of iterations. However in PV
as follows; systems the fitness value is not constant as it changes with the
Step 1 (PSO Initialization): Particles are usually initialized weather condition and load. Therefore, the PSO must be
randomly following a uniform distribution over the search reinitialized and search again for a new to search the new MPP
space, or are initialized on grid nodes that cover the search when the PV module output changed .for this application the
space with equidistant points. Initial velocities are taken proposed PSO algorithm is reinitialized whenever the
randomly. following functions are satisfied;
Step 2 (Fitness Evaluation): Evaluate the fitness value of
each particle. Fitness evaluation is conducted by supplying the 1
candidate solution to the objective function. 15
Step 3 (Update Individual and Global Best Data):Individual
and global best fitness values (pbest,i and gbest) and positions VI. DESIGN AND SIMULATION OF MPPT ALGORITHM
are updated by comparing the newly calculated fitness values
Fig.7 illustrates the Simulink module of the proposed
against the previous ones, and replacing the pbest,i and gbest
system that was simulated in Matlab, in which the terminal
as well as their corresponding positions as necessary.
voltages of the PV module was controlled by the DC-DC cuk
Step 4 (Update Velocity and Position of Each Particle): The
converter and its output was coupled to the load. The switch of
velocity and position of each particle in the swarm is updated
the converter was controlled by different MPPT algorithms
using (10) and (11).
and their tracking efficiency were analyzed and compared
Step 5 (Convergence Determination): Check the
under various conditions.
convergence criterion. If the convergence criterion is met, the
Firstly the proposed MPPT system was simulated with the
process can be terminated; otherwise, the iteration number will
MATLAB model at (1000w/m2, 25°C), and then simulated at
increase by 1 and go to step 2.
rapidly atmospheric conditions. The performance of each
D.Application of PSO to MPPT MPPT technique was evaluated when the steady state
This section describes the implementation of PSO method in condition is reached.
solving the problem involved to MPPT controller in PV
system. The flowchart of the proposed PSO-based MPPT
algorithm is illustrated in Fig.2, and the main blocks of the
proposed algorithm can be described as following;
 Perturb and Observe (P&O) Methods

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

P V M od ule O u tp ut P ow e r (W )
Incremental Conductance (INcond) Method

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Particle Swarm Optimization (PSO) algorithm

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)

Fig. 10 11The PV module output power (w) simulated with the

MATLAB model at 1000kw/m2, 25°C.

From Fig. 10 it can be seen that the theoretical maximum

power value is 60 W, and it is 60.7 W, by the PSO MPPT
control algorithm. In addition, the optimization time of PSO is
0.059 s and has a very fast convergence speed which proves
the accuracy of the proposed PSO-MPPT. While the tracking
efficiency of the P&O method was the lowest 59.7 W
comparing to PSO and IncCond. methods. The IncCond
method tracking efficiency was higher than P&O method
59.89 W, as a result of its independent to the solar radiation
level. Thus, this algorithm is usually used at high and fast
radiance variations.
20 Perturb and Observe (P&O) Method
P V m o d u le O u tp u t P o w e r ( W )

10
Fig. 8 PSO method flowchart 0
Incremental Conductance (INcond) Method
20
10
0
Particle Swarm Optimization (PSO) algorithm
20
10
0
0 0.5 1 1.5
Time (s)
Fig. 9 Simulink module of the MPPT system.
Fig. 12 The PV module output power (w) simulated with the
Fig 10 shows the simulation result for the response of the MATLAB model at 200kw/m2, 25°C.
three methods in the first stage at (1000kw/m2, 25°C). While From the result in condition 2 Fig. 11 under low solar
fig.10 shows the output power of the PV module under low radiation (G=200w/m2, T=25°C), the P&O efficiency is lower
solar radiation (G=200w/m2, T=25°C). than PSO and IncCond methods and it was not able to
converge to the MPP. In that time, the IncCond method was
able to open the PV module near to the MPP, and its
efficiency is higher than P&O. However, PSO algorithm was
the best control method throughout the simulation results, as it
was quickly and successfully tracked the MPP of the module
in both conditions.
 Perturb and Observe (P&O) Method [6] Y. Kuo, T. Liang and J. Chen, "Novel maximum-power-point-tracking
80 1000w/m2 controller for photovoltaic energy conversion system," Industrial
60 800w/m2 Electronics, IEEE Transactions on, vol. 48, pp. 594-601, 2001.
600w/m2 [7] T. Esram and P. L. Chapman, "Comparison of Photovoltaic Array
PV Module Output Power (W)

500w/m2
30
200w/m2 Maximum Power Point Tracking Techniques," Energy Conversion,
0
IEEE Transactions on, vol. 22, pp. 439-449, 2007.
0 5 10 15 [8] W. Xiao, "Improved control of photovoltaic interfaces," 2008.
80 Incremental Conductance (IncCond) Method [9] R. Faranda and S. Leva, "Energy comparison of MPPT techniques for
60 PV Systems," WSEAS Transactions on Power Systems, vol. 3, pp. 446-
455, 2008.
30 [10] W. Xiao and W. O. Dunford, "A modified adaptive hill climbing MPPT
method for photovoltaic power systems," in Proc. 35th Annu. IEEE
0
0 5 10 15 Power Electron.Spec.Conf., 2004,pp.1957-1963.
80 Particle Swarm Optimization (PSO) algorithm [11] K.H. Hussein, I. Muta, T. Hoshino, and M.Osakada, “Maximum
60
photovoltaic power tracking: an algorithm for rapidly changing
atmospheric conditions,”IEEE, vol.142, pp.59,Jan1995.
30 [12] K. Ishaque, Z. Salam, M. Amjad and S. Mekhilef, "An improved particle
swarm optimization (PSO)–Based MPPT for PV with reduced steady-
0
0 5 10 15 state oscillation," Power Electronics, IEEE Transactions on, vol. 27, pp.
Time 3627-3638, 2012

Fig. 13 The PV module output Power (w) simulated with the

MATLAB model at rapidly changing solar radiation, 25°C.

Fig. 12 shows output power of the PV module under rapidly

changing atmospheric conditions. The results highlight that
the tracking efficiencies of the systems with PSO in all
conditions was higher than 99.8%. The P&O method had large
amount of power losses and causes an oscillation around the
MPP, while the tracking efficiencies of the systems with the
IncCond method was better than the P&O performances.
However, the IncCond method is more complicated than the
P&O method as it requires a fast controller speed and more
sensor devices that leading to increase the system cost.

VII. CONCLUSIN
This study presents the simulation of MPPT of PV system
using P&O, IncCond and PSO techniques in terms of their
tracking efficiency, convergence speed, cost and performance.
According to the simulation results, the PSO method was able
to track the MPP correctly in all conditions, and it has
advantages over other techniques such as a very high tracking
efficiency, simple structure, easy implementation, and has a
very fast convergence speed to the desired solution .however,
the choice of its parameters has some impacts on the
optimization performance.

REFERENCES
[1] E. Koutroulis, K. Kalaitzakis and N. C. Voulgaris, "Development of a
microcontroller-based, photovoltaic maximum power point tracking
control system," Power Electronics, IEEE Transactions on, vol. 16, pp.
46-54, 2001.
[2] A. N. A. Ali, M. H. Saied, M. Z. Mostafa and T. M. Abdel-Moneim, "A
survey of maximum PPT techniques of PV systems," in Energytech,
2012 IEEE, 2012, pp. 1-17.
[3] Y. Liu, Advanced Control of Photovoltaic Converters, 2009.
[4] H. Abidi, A. B. Ben Abdelghani and D. Montesinos-Miracle, "MPPT
algorithm and photovoltaic array emulator using DC/DC converters," in
Electrotechnical Conference (MELECON), 2012 16th IEEE
Mediterranean, 2012, pp. 567-572.
[5] A. Durgadevi, S. Arulselvi and S. Natarajan, "Study and implementation
of maximum power point tracking (MPPT) algorithm for photovoltaic
systems," in Electrical Energy Systems (ICEES), 2011 1st International
Conference on, 2011, pp. 240-245.
4th International Conference on Power Engineering, Energy and Electrical Drives Istanbul, Turkey, 13-17 May 2013

A Soft Computing MPPT for PV System Based on

Cuckoo Search Algorithm

Jubaer Ahmed, Zainal Salam

Faculty of Electrical Engineering
Universiti Teknologi Malaysia
81310 Johor Bahru, Malaysia

Email: [email protected], [email protected]

Abstract— This paper presents a novel approach to determine the To overcome this problem MPPT based on soft computing
maximum power point (MPP) in Photovoltaic (PV) System using (SC) techniques, for example Artificial Neural Network [5],
the Cuckoo Search (CS) algorithm. In CS, three samples of Fuzzy logic Controller [6], Genetic algorithm [7], Differential
voltage are generated randomly over the span of the PV voltage. Evolution [8] and Particle Swarm Optimization [9] are
Based on the Le’vy distribution, the voltage samples are directed attracting considerable interests. Recently, there are several
towards the best solution and on the basis of power comparison, works on a SC method known as “Cuckoo Search (CS)”.
the best position is found. The algorithm is simulated using Compared to other SC techniques, CS is proven (in other
MATLAB and compared to the conventional P&O method. It applications) to be more robust, has better convergence and
exhibits very fast convergence with zero steady state oscillation. exhibits higher efficiency. Despite these advantages, the use of
In addition, it tracks the MPP perfectly when PV system is
CS for MPPT has not been reported anywhere in the literature.
subjected to rapid changes of atmospheric condition.
Hence, this paper proposes the application of CS for MPPT.
Keywords- MPPT, Cuckoo Search, Le’vy Flight, Convergence
speed, Rapid tracking II. PV MODELING
To investigate the performance of a PV system, first, a
I. INTRODUCTION model for the PV module needs to be developed. A simplified
Solar power is considered one of the most important cell model is presented in Fig. 1.
renewable energy (RE) sources of the future. Besides the
abundance of the sun, photovoltaic (PV) system is easy to
install, almost maintenance free and environmentally friendly.
One of the most economical ways of increasing efficiency of
PV system is to ensure it is operated at the maximum power
point (MPP). This can be achieved by employing a MPP
tracker (MPPT). Since the relation between the power and
voltage/current in PV cell is highly non-linear, MPPT
algorithms needs to track the MPP with the following
constraints: (1) convergence speed, (2) steady state condition, Figure 1. Simplified PV panel Modeling
(3) adaptability with changing atmosphere as irradiance or
temperature and (4) compatibility with abnormal condition The Current from the PV cell can be presented in (1)
such as partial shading [1]. V + IR
I =I − I −( s)
(1)
Conventional MPPT operates by sensing the current and PV d R
p
voltage of the PV array; the power is calculated and
accordingly the duty cycle of the converter is adjusted to match Where, light generated current is given in (2)
the MPP. Among them, Perturb and Observe (P&O) [2], I pv = (I + K I (T − T )) G G (2)
PV _ STC STC
Incremental Conductance (IC) [3] and Hill Climbing (HC) [4] STC
are the most popular. One of the major problems of these Note that IPV_STC is the light generated current in standard test
methods is their incompatibility with the rapid change in the condition (STC), i.e. temperature T=298 K and irradiance
irradiance and temperature. Besides, P&O and HC oscillate G=1000 w/m2. KI is the short circuit current coefficient, which
around MPP that cause huge loss in power during operation. is usually provided by the manufacturer. The diode current can
None of these techniques are capable of handling partial be written as in (3)
shading condition. V + IRS
I d 1 = I o exp ( ( ) − 1) (3)
VT

978-1-4673-6392-1/13/$31.00 ©2013 IEEE

POWERENG 2013 558
4th International Conference on Power Engineering, Energy and Electrical Drives Istanbul, Turkey, 13-17 May 2013

Where diode saturation current is expressed in (4) adopted in meta-heuristic search algorithm for optimization
T 3 qE g 1 1 problem [13].
(
I o = I o _ STC STC )
exp [ (− )] (4) In CS, nest searching steps of cuckoo is determined by
T k TSTC T
Le’vy Flight. Mathematically, a Le’vy flight is a random walk
In (4), Io_STC is the saturation current under STC, q is the where step sizes are extracted from levy distribution according
charge of electron, Eg is the band gap energy of the fabrication to a power law in (5), i.e.
material and k is the Boltzmann constant. Based on this −λ
equation the P-V curves for the PV cell at STC are shown in y=t (5)
Fig. 2. Where, 1 <Ȝ <3. Thus (5) has an infinite variance. Fig.3 is
showing an example of le’vy flight in a two dimensional plane.

Figure 3. A Le’vy Flight

C. Cuckoo Search Algorithm

Figure 2. P-V and I-V curve at STC Yang and Deb [10] have used three idealized rules for CS
based on cuckoo’s brood parasitic behavior: (1) each cuckoo
lays one egg at a time and places it in a randomly chosen nest,
III. CUCKOO SEARCH
(2) the best nest with high quality of eggs will carry over to
A. Cuckoo Behaviour the next generation and (3) the number of available nests is
fixed and the egg laid by a cuckoo is discovered by the host
Cuckoo Search (CS) was first proposed by [Yang & bird with a probability of Pa, where 0 < Pa <1. If the cuckoo’s
Deb][10]. The algorithm emulates the aggressive reproduction
eggs are discovered, host bird can abandon its nest or destroy
strategy of cuckoo birds. It is observed that several species of
cuckoos’ eggs. Either way a new nest will be generated with a
cuckoos perform brood parasitism [10], i.e. by laying their eggs
in other birds’ (host birds) nests. Usually three types of brood probability of Pa of fixed number of nest. Based on these three
parasitism are seen (1) intraspecific, (2) cooperative and (3) rules, the CS algorithm can be summarized in a pseudo code
nest takeover. Some cuckoo species as ‘Tapera’ are intelligent below:
enough to mimic the shape and color of host bird that increases
the reproduction probability. It is also presented in [11] that Begin
cuckoos lay their eggs at some specific time so that their eggs T
Objective function f ( x ) , x = ( x1 , x2 .....xn ) ;
hatch some time earlier than the host bird’s own. After early
hatching, cuckoos destroy some host bird’s eggs to increase the Generate initial population of n host nest, xi (i = 1, 2,......., n) ;
chance of their chicks getting more food. It is also a common While(t< Max Generation or stop criterion)
phenomenon that host birds discover the cuckoo’s eggs and Get a cuckoo randomly by levy flight
destroy these. Sometimes they abandon their nest completely Evaluate its quality/fitness, f i
and go elsewhere to build a new nest.
Choose a nest among n (say, j) randomly
B. Le’vy Flight If ( f i > f j )
Searching for a suitable host bird’s nest is an important Replace j by the new solution
part of cuckoo’s reproduction method. Normally, the search End
for nest is similar for a search for food, which takes place A fraction(Pa) of worst nest occurs;
randomly or in a quasi-random form. In general, while Worst nests are abandoned and new nests are built;
searching for food, animals choose directions or trajectories Keep the best solutions;
that can be modeled on certain mathematical functions. One of Rank the solutions and find the current bests;
most common model is the Le’vy flight. A recent study by End while
Reynolds and Frye [12] shows that that fruit flies or Post process of results and visualization;
Drosophila melanogaster, explore their landscape using a End
series of straight flight paths punctuated by a sudden 90
degree turn, leading to a Le’vy flight style. Such behavior is

POWERENG 2013 559

4th International Conference on Power Engineering, Energy and Electrical Drives Istanbul, Turkey, 13-17 May 2013

When generating a new solution x

( t +1)
for a cuckoo, a le’vy while keeping the better nests. New nest are generated in place
flight is performed as in (6), i.e. of worst nest. Values of fitness are tested again and current
(t +1) best is selected. Iteration continues until all nests reach at the
xi = xit + α ⊕ le′vy ( λ ) (6) maximum power point.
B. Simulation
Where, Į > 0 is the step size which is related to the constraint
In the simulation, the direct duty cycle control method
of the optimization problem. Based on the constraints it is
MPPT [15] is utilized. The CS MPPT algorithm is tested on a
important to tune the value of Į to get desired step size. The
DC-DC buck-boost converter. The circuit is shown in Fig.4.
product ⊕ means entry wise multiplication. The value of le’vy Two sensors are used to measure the voltage and current of the
(Ȝ) is found from le’vy distribution given in (7), i.e. PV array. Using the CS algorithm, the reference value of the
duty cycle is computed. This value is directly fed to the input
−λ (7) in the PWM generator. This PWM signal adjusts the duty
Le′vy ≈ u = t , (1 < λ < 3)
cycle of the MOSFET. The common parameters for the tested
PV system are given in Table I.
In the context of MPPT algorithm, the structure of CS
algorithm in (6) is in the same form of the hill climbing (HC)
method. This similarity, of course, does not include the step
size from le’vy flight. However, there are several significant
features [14] that allow for CS to be much more robust than
HC: (1) CS is a population based algorithm like GA and PSO
but it exhibits elitism in selection procedure like harmony
search, (2) in CS, the randomization is much more efficient as
the steps sometimes get bigger because of levy flight which
provide faster convergence and (3) the number of parameters
for tuning in CS is two; this is less than GA and PSO where
the tuning parameters are three and above . Figure 4. Simulation Circuit

TABLE I. LIST OF PV PANEL PARAMETERS

IV. MPPT DESIGN USING CS Panel Parameters Value
Short Circuit current 7.5 amp
A. MPPT Algorithm Open Circuit voltage 20.7 v
Maximum current 6.6 amp
Considering fixed number of nest (n) initially population is Maximum voltage 16.7 v
generated as voltage, Vi = V1 , V2 ,......,Vn and initial step size α 0 . Temperature coefficient of current 0.00033 amp/C
Temperature coefficient of voltage -0.084 v/C
Voltages are applied in the PV panel and the power is found as Number of cells in series inside panel 36
the fitness value. The maximum value among the fitness is Number of panels in series 20
considered as the current best. Then le’vy flight is performed Number of panels in parallel 1
and new nests are generated as in (8)
(t +1) t To evaluate the performance of the proposed CS MPPT, the
Vi = Vi + α ⊕ le ′vy ( λ ) . (8)
standard P&O algorithm is also developed for the same
Where, α = α 0 ( x − xi ) . A simplified scheme of le’vy simulated conditions. The P&O is used as a benchmark
best
because it is the most commonly used MPPT in the PV
distribution is presented in [x] as (9)
industry.
u
s = α 0 ( xbest − xi ) ⊕ Le′vy (λ ) ≈ 0.01( )( xbest − xi )
§v 1 · (9)
¨ β ¸
© ¹ V. RESULTS
where ȕ=1.5, u=random(1, n)× and v=random(1, n). The
A. At STC
variable is defined in (10)
1 The performance of the algorithms (P&O and CS) are
§ · β compared at STC, i.e. T=25 C and G=1000 w/m2. The
¨ ¸
¨ Γ (1+ β ).sin(π .β / 2) ¸
(10) simulation is carried out for 1 sec. The results are shown in
φ= ¨
β −1 ¸
¸
Fig. 5 up to 200 ms. From the figure, it can be observed that
¨
¨¨ Γ ((1+ β ).β .2 2 ) ¸¸ MPP tracking time for CS, i.e. the time taken to track the MPP
© 2 ¹
from the initial condition, is 22 ms. On the other hand, the
Where, Γ denotes the integral gamma function. New fitness P&O tracks at 45 ms. Furthermore, at steady state, the P&O
values are tested through PV panel. Afterwards, probability of oscillates around MPP, while the CS keeps tracking the exact
discovered worst nests is applied and such nests are destroyed MPP point. This is an added advantage as the power losses

POWERENG 2013 560

4th International Conference on Power Engineering, Energy and Electrical Drives Istanbul, Turkey, 13-17 May 2013

associated to power oscillation for CS is almost nil at steady

state, leading to a higher MPPT efficiency. In the case of
voltage and current waveforms, shown in Figs. 6 and 7
respectively, the P&O keeps on oscillating in steady state but
CS sticks on Vmpp and Impp. However, it can be observed that
initially CS exhibits much higher power fluctuations as the
algorithm starts to check for random samples in different
region (in transient). Nevertheless, since the samples get
closer to MPP in very quick succession, the power loss
associated to this fluctuation is not a big issue.

Figure 7. Combined Current Graph of P&O and CS

B. Rapid Atmospheric Change

The CS is tested to verify its performance under rapid
atmosphere changes. In the simulation, the irradiance and
temperature are changed abruptly at every one second. The
test patterns for the atmospheric conditions are shown in Table
II.

TABLE II. TEST CONDITION FOR RAPID CHANGES

Figure 5. Combined Power Graph of P&O and CS Temperature( C) Irradiance(w/m2) MPP(W)
25 1000 2205
As the P&O keeps on oscillating around MPP presented in 35 1200 2504
Fig. 5, it looses energy at an average rate of 3 mJ/sample. As 20 800 1769
45 1400 2675
20 panels are in series, energy loss per sample per panel is 15 600 1295
0.15 mJ. Since the simulation is designed as 1 ms per sample,
which makes energy loss 0.15 J/sec/panel (0.136% of the
panel capacity), which is really huge when PV panel is
operating for long period of time. In case of CS power loss
tends to zero at steady state.

Figure 6. Combined Voltage Graph of P&O and CS Figure 8. Rapid Tracking of CS

It is clear from Fig. 8 that CS is quite capable of tracking MPP

under drastic changes in environmental conditions. However,
the initial convergence times are not constant; they vary from
20 to 100 ms. This is to be expected due to the random nature
of CS algorithm. Furthermore, it can be observed that after

POWERENG 2013 561

4th International Conference on Power Engineering, Energy and Electrical Drives Istanbul, Turkey, 13-17 May 2013

steady state is reached, CS sticks to the MPP without any [5] A. M. Z. Alabedin, E. F. El-Saadany and M. M. A. Salama, “Maximum
power point tracking for photovoltaic systems using fuzzy logic and
further fluctuation. artificial neural networks,” IEEE Power and Energy Soc. Gen. Meet. pp.
1-9, 2011.
VI. CONCLULSION
[6] C. Chian-Song, “T-S Fuzzy maximum power point tracking control of
In this paper a novel MPPT algorithm is proposed based on solar power generation systems,” IEEE Energy Conv.Trans. vol. 25,
cuckoo search algorithm. Following the cuckoos natural pp.1123-1132, 2010.
behavior and le’vy flight distribution, MPP is tracked very [7] R. Ramaprabha, V. Gothandaraman, K. Kanimozhi, R. Divya and B.L.
Mathur, “ Maximum power point tracking using GA-optimized artificial
efficiently. It is confirmed from the simulation result that this neural network for Solar PV system,” 1st Int. Conf. on IEE Elec. Energy
algorithm converges faster than P&O. Besides, it exhibits zero Sys. pp. 264-268, 2011.
oscillation at steady state, thus saves a large amount of power. [8] H. Taheri, Z. Salam,K. Ishaque and Syafaruddin, “A novel maximum
Additionally this algorithm can track MPP successfully when power point tracking control of photovoltaic system under partial and
rapidly fluctuating shadow conditions using Differential Evolution,”
atmospheric condition changes very rapidly. IEEE Indust. Electron. & App. Symp. pp. 82-87, 2010.
[9] K. Ishaque, Z. Salam, M. Amjad and S. Mekhilef, “An improved particle
swarm optimization (PSO)-based MPPT for PV with reduced steady-
ACKNOWLEDGMENT state oscillation,” IEEE Power Electron. Trans. vol. 27, pp. 3627-3638,
2012.
The authors would like to thank Universiti Teknologi
[10] X. S. Yang and S. Deb, “Cuckoo search via levy flights,” World
Malaysia and the Ministry of Higher Education Malaysia for Congress on Nature & Biologically Inspired Computing. pp. 210-214,
providing the facilities and ¿nancial support (Research 2009.
University Grant No.2423.00G40 to conduct this research. [11] Y. Teuschl, B. Taborsky and M. Taborsky, “How do cuckoos find their
hosts? The role of habitat imprinting,” Animal Behaviour. vol. 56, pp.
1425–1433, 1998.
REFERENCES [12] A. M. Reynolds and M. A. Frye, “Free-flight odor tracking in
[1] K. Ishaque , Z. Salam and H. Taheri, “A comprehensive MATLAB Drosophila is consistent with an optimal intermittent scale-free search,”
simulink PV system simulator with partial shading capability based on PLoS One. 2, e. 354, 2007.
two-diode mode,” Solar Energy, vol.85, pp. 2217-2227, 2011. [13] R. Rajabioun, “Cuckoo Optimization Algorithm,” Applied Soft
[2] N. Femia, G. Petrone, G. Spagnuolo and M. Vitelli, “Optimization of Computing. vol. 11, pp. 5508–5518, 2011.
perturb and observe maximum power point tracking method,” IEEE [14] P. Civicioglu and E. Besdok, “A conceptual comparison of the Cuckoo-
Power Electron. Trans. vol. 20(4), pp. 963–973, 2005. search, particle swarm optimization, differential evolution and artificial
[3] Y. C. Kuo, T. J. Liang and J. F. Chen, “Novel maximum-power-point- bee colony algorithms,” Artificial Intelligence Review, July 2011.
tracking controller for photovoltaic energy conversion system,” IEEE [15] K. Ishaque, Z. Salam, A. Shamsudin and M. Amjad, “ A direct control
Indust. Electron. Trans. vol. 48(3), pp. 594–601, 2001. based maximum power point tracking method for photovholtaic system
[4] W. Xiao and W.G. Dunford, “A modified adaptive hill climbing MPPT under partial shading conditions using particle swarm optimization
algotithm,” Applied Energy. vol. 99, pp. 412-422, 2012.
method for photovoltaic power systems,” Proc. 35th Annu. IEEE Power
Electron. Spec. Conf. pp. 1957–1963, 2004.

POWERENG 2013 562

 electronics
Article
An Advanced Maximum Power Point Tracking
Method for Photovoltaic Systems by Using Variable
Universe Fuzzy Logic Control Considering
Temperature Variability
Yiwang Wang 1,2, *, Yong Yang 3 , Gang Fang 4 , Bo Zhang 2 , Huiqing Wen 5 , Houjun Tang 1 ,
Li Fu 6 and Xiaogao Chen 7
1 Department of Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China;
[email protected]
2 School of Electronic and Information Engineering, Suzhou Vocational University, Suzhou 215104, China;
[email protected]
3 School of Rail Transportation, Soochow University, Suzhou 215131, China; [email protected]
4 Jiangsu GOODWE Power Supply Technology Co., Ltd., Suzhou 215163, China; [email protected]
5 Department of Electrical and Electronic Engineering, Xi’an Jiaotong-Liverpool University, Suzhou 215123,
China; [email protected]
6 Department of New Chemical Materials Engineering, Shandong Polytechnic College, Jining 272067, China;
[email protected]
7 Wuxi Solartale PV Technology Co., Ltd., Wuxi 214174, China; [email protected]
* Correspondence: [email protected]; Tel.: +86-512-6658-6691




Received: 27 October 2018; Accepted: 21 November 2018; Published: 26 November 2018


Abstract: In this study maximum power point tracking (MPPT) is applied to the photovoltaic (PV)
system to harvest the maximum power output. The output power of the PV effect changes according
to external solar irradiation and ambient temperature conditions. In the existing MPPT strategies,
most of them only take variations in radiation level into account, rarely considering the impact of
temperature changes. However, the temperature coefficients (TC) play an important role in the PV
system, especially in applications where ambient temperature changes are relatively large. In this
paper, an MPPT method is presented for a PV system that considers the temperature change by using
variable universe fuzzy logic control (VUFLC). By considering the ambient temperature change in PV
modules, the proposed control method can regulate the contraction and expansion factor of VUFLC,
which eliminates the influence of temperature variability and improves the performance of MPPT,
therefore achieving fast and accurate tracking control. The proposed method was evaluated for a
PV module under different ambient conditions and its control performance is compared with other
MPPT strategies by simulation and experimental results.

Keywords: maximum power point tracking (MPPT); photovoltaic (PV) system; variable universe
fuzzy logic control (VUFLC); temperature variability

1. Introduction
With the development of photovoltaic (PV) technologies, an increasing number of PV power
generation systems have been presented for large-scale applications. The PV module is one of the
key components of PV power generation systems; its performance and efficiency directly affect the
high-efficiency operation of the entire system. However, the power energy generated from PV modules
relies highly on environmental factors such as solar insolation and the ambient temperature [1,2].

Electronics 2018, 7, 355; doi:10.3390/electronics7120355 www.mdpi.com/journal/electronics

Electronics 2018, 7, 355 2 of 18

Therefore, in order to harvest the maximum power output and improve the efficiency of the entire PV
system, many advanced MPPT control methods have been implemented in PV systems [3–5].
Many MPPT control algorithms have been proposed and developed in recent years [6], such as the
classic methods, including open-circuit voltage (OCV)/short-circuit current (SCC) [7,8], incremental
conductance (INC) [9], perturbation-and-observation (P&O) [10] and other hybrid strategies [3]. Due to
the non-linear problems of PV cells, some soft computing techniques have been applied to the MPPT
of PV systems, such as the artificial neural networks method (ANN) [11] and fuzzy logic control
(FLC) [12–14].
Most of the MPPT approaches only take the variability in radiation level into account, while rarely
considering the effects of temperature. Some new MPPT methods based on temperature measurements
were discussed in Reference [15]. Many control algorithms use temperature as a feedback parameter
to realize MPPT. For example, studies in References [16–18] proposed an MPPT-temperature algorithm
where the PV module temperature was used to determine the maximum power point voltage to
track the maximum power point (MPP). In Reference [19], a sun tracking system that included the
temperature effect was presented, and an optimum system design was achieved. Compared to other
approaches under the same control algorithm, the MPPTs based on temperature measurement directly
consider the temperature variations leading to MPP changes, which can obtain a faster tracking speed,
especially in engineering applications where temperature changes are relatively large.
Furthermore, to improve the tracking accuracy, some artificial intelligence techniques have been
employed for the MPPT implementation [20–22]. Fuzzy logic control (FLC) is a relatively popular and
mature artificial intelligence algorithm and has been applied to track the MPP in PV systems [23,24].
In Reference [25], results indicated that FLC had the best performance when compared to some MPPT
techniques with INC, P&O, and others in both dynamic response and steady-state under most of the
normal operating range. The variable universe fuzzy logic control (VUFLC) can adaptively change the
input and output universes to improve the control effect and obtain higher control accuracy [26–29].
However, how to choose a variable universal scalable or contraction-expansion factor is a
challenging issue practically, according to the nonlinear characteristic of PV systems [29]. Hence,
this works to exploit the PV modules’ real-time temperature variable as the constraints of variable
universal factor selection, then a new VUFLC-temperature MPPT algorithm was designed to obtain
efficient tracking performance in the external working environment (environmental condition)
variations. The proposed VUFLC-temperature MPPT method selects the variable universal factor
according to the dynamic change of temperature by combining the modules’ temperature coefficients
(TC) characteristic, which can accelerate the MPPT and improve the tracking accuracy when compared
to conventional MPPT strategies. The proposed VUFLC-temperature MPPT method was validated by
simulation and experimental results.
This paper is organized as follows: Section 2 describes the characteristics of the PV system.
Section 3 presents the proposed VUFLC-temperature MPPT algorithm and its development techniques.
Section 4 presents the simulation results. Section 5 provides the results of the experimental tests.
Finally, the paper is concluded in Section 6.

2. Model and Characteristics of a PV System

PV cells and modules are the key components of PV systems, which absorb photons of light
and release electron charges, and can directly convert solar energy into electricity [30]. The electric
energy generated from the PV effect is highly dependent on environmental factors [1]. The PV cell is a
nonlinear device and can be represented as a current source model [31]. The common and popular PV
models are single diode and double diode [4,31,32]. Here, we use the single diode model to describe
the characteristics of PV cells.
Electronics 2018,
Electronics 2018, 7,
7, 355
x FOR PEER REVIEW 33 of
of 18
18

I pv = I ph − I d − I R (1)
The single diode equivalent circuit is shown in Figuresh1, where D is a parallel diode, Rsh is the
shunt resistance and Rs is the series resistance. The output mathematical equations were obtained in
q (V pv + Rs I pv )  V + Rs I pv
References [4,30,33,34]. I pv = I ph − I 0  exp( ) − 1 − pv (2)

Electronics 2018, 7, x FOR PEER REVIEW  I pv = I phAKT − Id − IRsh  Rsh 3 of 18 (1)
I q(V=pvI + −RIs I pv ) VpvI+ Rs I pv
 
where ph is the light-generated I0 exp( Iofpv the
I pv = I ph − current d − I)
elementary
ph Rsh−PV 1 − cell, d is the current of the (1) parallel
(2)
AKT Rsh
I sh R
diode, I Ris
where is the
the shunt current
light-generated of resistance
current (V pv sh
+ ,Rand
of the qelementary I 0 cell,
s I pv ) PV I +isreverse
isVthe
pv Rsthe
I pv current
saturation current diode,
of the parallel of the
ph I pv = I ph − I 0  exp( ) − 1 − d (2)
I pv of resistance Rsh, and I0 isAKT
IRsh is the shunt current Rsh
the reverse saturation current the diode. Vpv and
q isofthe
diode. V pv and are the output voltage and current, respectively. electron
−19 C), Acharge
I ph voltage and current, respectively. q is the electronI dcharge (1.602 × 10
I pv are the output is the
where
× 10 C),
(1.602 ideality
−19 is the
A is andlight-generated
the diode current
ideality of the elementary
factor,constant
and K is(1.38 PV cell,
−
the Boltzmann23 isconstant (1.38 × parallel
the current of the 10−23 J/K).
diode factor, K is the Boltzmann × 10 J/K).
I R
diode, Rsh is the shunt current of resistance sh , and I 0 is the reverse saturation current of the
I pv
diode. V pv and are the output voltage and current, respectively. q is the electron charge
(1.602 × 10−19 C), A is the diode ideality factor, and K is the Boltzmann constant (1.38 × 10−23 J/K).

Figure
Figure 1. Single diode
1. Single diode model
model and
and equivalent
equivalent circuit
circuit of
of the
the photovoltaic
photovoltaic (PV)
(PV) cell.
cell.

According
According to toEquations
Equations
Figure (1)(1)
1. Single andand
diode (2), theand
(2),
model electrical
the electrical
equivalent characteristics
characteristics
circuit of theofPV
of the photovoltaic cell
the
(PV) PV arecell
cell. generally drawn
are generally
as a current
drawn versus voltage
as a current versus (Ipv-Vpv) curve and
voltage (Ipv-Vpv) a power
curve and aversus
powervoltage
versus(Ppv-Vpv) curve under
voltage (Ppv-Vpv) curve different
under
environmental According to Equations
conditions [29]. (1) and (2), the electrical characteristics of the PV cell are generally
different environmental conditions [29].
drawn as a current versus voltage (Ipv-Vpv) curve and a power versus voltage (Ppv-Vpv) curve under
Figure
Figure 22 shows
shows the
thecharacteristics
characteristics ofofthethePVPVcell inin
cell different
different environmental
environmental conditions.
conditions. As As
cancan
be
different environmental conditions [29].
seen from
be seen from the relation between
the 2relation
Figure shows the between the PV output
the PV
characteristics parameters
output
of the PV cell and
parameters the environment
and the environment
in different environmental variables,
conditions.variables,it
As can is highly
it is
nonlinear and from
be seen
highly nonlinear dependent
and on the
thedependent
relation solar
between
on the
the radiation
PV output
solar level
radiation and
leveltemperature
parameters and temperature
and changes
the environment on the
variables,
changes onPV iscell
itthe PV[35].
cell
The output power
highly energy
nonlinear and of the
dependentPV system
on the is
solaraffected
radiation by radiation
level and and temperature.
temperature
[35]. The output power energy of the PV system is affected by radiation and temperature. Figure 2a changes onFigure
the PV 2a
cell presents
the [35].under
curves The output power energy of the
different PVthe system is affected by radiation
ISCcurrent and temperature. Figure
the 2a
presents the curves under radiation, i.e.,
different radiation, current
i.e., the increases ISC quasi-linearly with
increases quasi-linearly radiation
with the
while presents
the voltage the curves
VOCvoltage under
increases different radiation, i.e., the current ISC increases quasi-linearly with the
slightly, and the maximum electric power Pmax changes asmax
the radiation
radiation while the V OC increases slightly, and the maximum electric
radiation while the voltage VOC increases slightly, and the maximum electric power Pmax changes as power P changes as
changes. Figure
the radiation 2b gives
changes. the
Figure relationship
2b 2bgives of I_V and P_V at different temperatures, where I slightly
the radiation changes. Figure givesthe the relationship
relationship ofofI_V I_VandandP_VP_V at different
at different temperatures,
temperatures, SC
increases
where IwhereandISC VOC
SC slightly
strongly
increases
slightly decreases
increasesandand VOC with
VOCstrongly temperature
strongly decreases
decreases with changes.
with The maximum
temperature
temperature changes.
changes. Theelectric
Thepower
maximum maximum also
significantly
electric electric
power decreases
power alsowith a temperature
significantly
also significantly decreaseswith
decreases rise,
withas illustrated rise,
aa temperature
temperature inrise,
Figure 3 [30].in Figure
as illustrated
as illustrated 3 [30]. 3 [30].
in Figure

(a) (b)

Figure 2. Characteristics of the PV cell at different environmental conditions: (a) Different solar
radiation level effect; (a)
(b) different temperature variability effect. (b)
Electronics 2018, 7, x FOR PEER REVIEW 4 of 18

Figure
Electronics 2018,2.7, Characteristics
355 of the PV cell at different environmental conditions: (a) Different solar4 of 18
radiation level effect; (b) different temperature variability effect.

Figure 3. Power variations with temperature

Figure 3. temperature changes
changes under
under different
different radiation
radiation levels.
levels.

In
In recent
recentyears, many
years, researchers
many have have
researchers conducted MPP control
conducted MPP studies
controlunder
studiesradiation
underconditions
radiation
and proposed
conditions andmany various
proposed manyMPPT
variousalgorithms in the literature
MPPT algorithms [1–30]. This
in the literature paper
[1–30]. Thisfocuses on the
paper focuses
impact of temperature on MPPT and proposed corresponding control strategies
on the impact of temperature on MPPT and proposed corresponding control strategies and and implementation
methods. The short-circuit
implementation methods. current ISC and open-circuit
The short-circuit current ISCvoltage VOC at the reference
and open-circuit voltage Vnominal operating
OC at the reference
cell temperature
nominal operating(NOCT) TNOCT can be
cell temperature calculated
(NOCT) TNOCTatcana given temperature
be calculated TC with
at a given some temperature
temperature TC with
variation
some temperature variation [28,30], respectively. The PV output parameters consideringeffect
[28,30], respectively. The PV output parameters considering temperature can be
temperature
obtained
effect canasbefollows:
obtained as follows:
ISC = ISC− NOCT ·[1 + αSC ·( Tc − TNOCT )] (3)
I SC = I SC − NOCT ⋅ [1 + α SC ⋅ (Tc − TNOCT )] (3)
VOC = VOC− NOCT ·[1 + βOC ·( Tc − TNOCT )] (4)
VOC==PV
Pmax NOCT ⋅·[[1
OC−−NOCT
max 1++ βγOC
max⋅·(
(TTcc −−TTNOCT
NOCT)])] (4)
(5)

where the TC is the operating temperature of the PV cell and TNOCT is the temperature at the nominal
ambient environment. ISC− NOCT Pmax = Pmax − NOCT
, VOC ⋅ [1 + γ max ⋅ (Tcare− Tthe
NOCT )] (5)
− NOCT , and Pmax − NOCT short-circuit current, open-circuit
voltage, and maximum power at the reference NOCT, respectively. αSC , βOC , and γmax are the
T
and NOCT is
temperature is the operating
where the TC coefficients (TC)temperature
of ISC , VOCof the PV
, and Pmaxcell
, respectively. the temperature
According at the nominal
to Equations (3)–(5),
I
the short-circuit current, open-circuit V P
ambient environment. SC − NOCT , OCvoltage − NOCT , and
and maximum
max − NOCT power of PV cells will be affected by
are the short-circuit current, open-
temperature changes. Therefore, the impact of temperature variability can be considered when
circuit voltage, and maximum power atthrough
the reference NOCT, respectively.
α SC , βOC , and γ max are
designing the MPPT control strategy the direct or indirect measurement of the operating
temperature of the PV cells. I V P
the temperature coefficients (TC) of SC , OC , and max , respectively. According to Equations (3)–
(5),MPPT
3. the short-circuit current,
Control System andopen-circuit voltage Method
Proposed Control and maximum power of PV cells will be affected by
temperature changes. Therefore, the impact of temperature variability can be considered when
designing
3.1. the MPPT
MPPT Control control strategy through the direct or indirect measurement of the operating
System
temperature of the PV cells.
Power electronic converters are commonly applied in a PV system to achieve different MPPT
control methods, where the converters act as the interface between the PV source and different loads.
3. MPPT Control System and Proposed Control Method
In order to efficiently track the MPP, the converter needs to adjust the duty cycle under varying
operating
3.1. MPPT atmospheric
Control System conditions [1,4,25]. The MPPT controller acquires the real-time operating
parameters depending on the control algorithm, then outputs the corresponding control signal to
Power
control electronic
the DC/DC converters
converter. Thearemost
commonly
common applied
solar PV in aMPPT
PV system
systemtoconsists
achieveofdifferent MPPT
a PV module,
control methods,
DC/DC where the
boost converter, converters
MPPT act as
controller theainterface
and load, as between
shown inthe PV source
Figure 4. Theand
PVdifferent loads.
cells generate
In order to efficiently track the MPP, the converter needs to adjust the duty
power energy and its output is connected to the DC/DC converter. The converter is controlled by the cycle under varying
operating
MPPT atmospheric
controller conditions
where different [1,4,25].
control The MPPT
algorithms can becontroller acquires the real-time operating
carried out.
parameters depending on the control algorithm, then outputs the corresponding control signal to
 Electronics 2018, 7, x FOR PEER REVIEW 5 of 18
Electronics 2018, 7, x FOR PEER REVIEW 5 of 18
control the DC/DC converter. The most common solar PV MPPT system consists of a PV module,
control
DC/DCthe DC/DC
boost converter.
converter, MPPT The most common
controller solar as
and a load, PVshown
MPPTinsystem
Figureconsists of acells
4. The PV PV module,
generate
DC/DC boost converter, MPPT controller and a load, as shown in Figure 4. The PV cells
power energy and its output is connected to the DC/DC converter. The converter is controlled by the generate
power
MPPT energy2018, 7, and
controller
Electronics its output
355 where is connected
different to the DC/DC
control algorithms canconverter.
be carriedThe
out.converter is controlled by5 ofthe
18
MPPT controller where different control algorithms can be carried out.

Figure 4. The block diagram of a general PV maximum power point tracking (MPPT) system.
Figure 4. The block diagram of a general PV maximum power point tracking (MPPT) system.
3.2. Variable Universe Fuzzy Logic Control (VUFLC)
3.2. Variable Universe
3.2. Variable Universe Fuzzy
Fuzzy Logic
Logic Control
Control (VUFLC)
(VUFLC)
Due to the nonlinear characteristics of the PV system, intelligent MPPT control algorithms in PV
Due
Due to to thenonlinear
nonlinearcharacteristics
characteristicsofofthe the PV system, intelligent MPPT control algorithms in
systems arethe very promising and some havePVbeen system, intelligent
successfully MPPT
employed control algorithms
for maximum in PV
power
PV systems
systems are
are[4,6].
veryvery promising
promising and
and some some have been successfully employed for maximum power
extraction Fuzzy logic control (FLC)have
is onebeen of thesuccessfully
most prevalent employed for control
intelligent maximum power
techniques,
extraction
extraction [4,6]. Fuzzy logic control (FLC) is one of the most prevalent intelligent control techniques,
which has[4,6]. Fuzzy logic
advantages like acontrol (FLC) is
fast response oneless
time, of the most prevalent
fluctuation and high intelligent control techniques,
control accuracy. Therefore,
which
which has
has advantages
advantages like aa fast
like fast response
response time,
time, less
less fluctuation
fluctuation and and high
high control
control accuracy.
accuracy. Therefore,
Therefore,
it is effective in controlling nonlinear systems [4].
it
it is
is effective
effective in
in controlling
controlling nonlinear
nonlinear systems [4].
However, the conventional FLCsystems
with fixed[4]. fuzzy control rules will not perform well when
However,
However, the
the conventional
conventional FLC with
FLC with fixed
fixed fuzzy
fuzzy in control
control rules
rules will
will not perform well when
working with large uncertainties or unknown variations the systems [28], not
and perform
the control well when
precision
working
working with
with large
large uncertainties
uncertainties or
or unknown
unknown variations
variations in
in the
the systems
systems [28],
[28], and
and the
the control
control precision
precision
is commonly not high. Hence, adaptive fuzzy logic controllers (AFLC) have been proposed to solve
is commonly
isthis
commonly not
not high.
high. Hence, adaptive
Hence, adaptive fuzzy
fuzzy logic
logic controllers
controllers (AFLC)
(AFLC) have have been
been proposed
proposed to to solve
solve
issue [26–28]. The VUFLC is one of the AFLC that has been applied to various control engineering
this
this issue
issue [26–28].
[26–28]. The
The VUFLC
VUFLC is
is one
one of
of the
the AFLC
AFLC that
that has
has been
been applied
applied to
to various
various control
control engineering
engineering
projects such as specialty vehicle control [27], analog circuit implementation [28] and liquid lever
projects
projects such
suchThe as
as specialty
specialty vehicle
vehicle control
control [27],
[27], analog
analog circuit
circuitandimplementation
implementation [28] and
and liquid lever
system [36]. VUFLC combined with the characteristics advantages of [28] liquid
the variable lever
universe
system
system [36]. The VUFLC combined with the characteristics and advantages of the variable universe
control[36]. The VUFLC
is introduced intocombined
the MPPT with the characteristics
control, which can improve and advantages
the controlofspeed
the variable universe
and precision of
control
control is
is introduced
introduced into
into the
the MPPT
MPPT control,
control, which
which can improve
can improve thethe
control
controlspeed
speed andandprecision
precisionof the
of
the PV system.
PV
the system.
PV system.
The VUFLC was proposed in Reference [26] and its discourse universes of the input and output
The
The VUFLC
VUFLC was
was proposed
proposed in
in Reference
Reference [26] and
and its
[26]control discourse
itsconditions universes
discourse instead
universes of
of the
the input
input and output
output
variables can be adjusted according to changed of adjusting theand
fuzzy rules,
variables
variables can be adjusted
can be adjusted according
according to changed
to changed control conditions instead of adjusting the fuzzy
thus illustrating more control accuracy and control
flexibilityconditions instead of adjusting
than conventional the fuzzy
FLCs [27,28]. rules,1
Figure
rules,
thus thus illustrating
illustrating more
more diagram control
control of accuracy
accuracy and flexibility
and flexibility than conventional
than conventional FLCs
FLCs [27,28]. Figure 11
illustrates the process the variable universe, where Figure 5b shows an[27,28].
originalFigure
universe
illustrates
illustrates the process diagram of the variable universe, where Figure 5b shows an original universe
with five the fuzzyprocess diagram
partitions of the sets
as fuzzy variable
of NB universe,
(negative where big),Figure 5b showssmall),
NS (negative an original universe
ZE (zero), PS
with
with five fuzzy partitions as as
fuzzy setssets
of NB (negative big), NS (negative small), ZE (zero), PS(zero),
(positive
(positive small) and PB (positive big) with a piecewise linear membership function [28]. FigurePS
five fuzzy partitions fuzzy of NB (negative big), NS (negative small), ZE 5a
small)
(positive and PB (positive
small) and PBofbig) with abig)
(positive piecewise
withFigure linear membership
a piecewise linearthe function [28]. Figure[28].
membership 5a shows
Figurethe
shows the contracting the universe and 5c presents expanding function
of the universe. 5a
contracting of the universe
shows the contracting of theand Figureand
universe 5c presents
Figure 5c the expanding
presents of the universe.
the expanding of the universe.

μ μ μ
μ μ μ

−α ( x ' ) E α ( x' )E −α ( x ) E α ( x) E
−α ( x ' ) E α ( x )E
' −α ( x ) E α ( x) E
(a) (b) (c)
(a) (b) (c)
Figure Illustration
Figure 5. Illustration of variable
of variable universe:
universe: (a) Contracted
(a) Contracted universe; universe;
(b) original(b) original
universe; universe;
(c) expanded
Figure
(c) 5. Illustration
expanded
universe. of variable universe: (a) Contracted universe; (b) original universe; (c) expanded
universe.
universe.
In order to simplify the analysis, common two-input and single-output systems were taken as
an example. Let the universes of input (x1 , x2 ) and y output variables be Xi = [−Ei , + Ei ] (i = 1, 2) and
 In order to simplify the analysis, common two-input and single-output systems were taken as
x ,x
an example. Let the universes of input ( 1 2 ) and
y
output variables be Xi = [−Ei, + Ei] (i = 1, 2) and
Y = [−U, +U], respectively. According to Figure 5, the input and output universes Xi and Y can be
Electronics 2018, 7, 355 6 of 18
adjusted with the change of variables
x1
and
y , respectively. Their relationship is as follows:

X i ( xi ) = [to−αFigure
Y = [−U, +U], respectively. According i ( xi ) E5, +αi input
i , the ], i =output
( xi ) Eiand 1, 2 universes Xi and Y can(6)be
adjusted with the change of variables x1 and y, respectively. Their relationship is as follows:
Y ( y ) = [ − β ( y )U , + β ( y )U ] (7)
Xi ( xi ) = [−αi ( xi ) Ei , +αi ( xi ) Ei ], i = 1, 2 (6)
α (x )
where i i is the input universe contraction factor and β ( y ) is the output universe contraction
Y (y) = [− β(y)U, + β(y)U ] (7)
factor. With a contraction factor change, the input variable and output variable will change to better
adapt α
where toi ( xdifferent control
i ) is the input conditions
universe and achieve
contraction more
factor and β(yprecise
) is the control objectives.
output universe Therefore,
contraction the
factor.
design
With and selection
a contraction of contraction
factor change, the factors
inputisvariable
also important
and outputand variable
the details about
will contraction
change to betterfactors
adapt
will be introduced
to different controlbased on theand
conditions proposed
achieveMPPT
moresystem
preciseincontrol
the nextobjectives.
section. Therefore, the design
and selection of contraction factors is also important and the A B contraction factors will be
details about
The fuzzy rule is essential for a VUFLC system, let xi and y be regarded as linguistic
introduced based on the proposed MPPT system in the next section.
The fuzzy
variables of inputrule xisi essential for ay,VUFLC
and output system,
respectively. The A xi and
letfuzzy By be regarded
IF-THEN as linguistic
control rule variables
[26] is formed as
of input xi and output y, respectively. The fuzzy IF-THEN control rule [26] is formed as follows:
follows:

IF x1 isIF Ax1x1 isand x 2 isx2 is

A x1 and Ax2A,xTHEN y is B
2 , THEN y is By y
(8)
(8)

Unlike
Unlike the
the conventional
conventional FLC,
FLC, the
the universes
universes of VUFLC can
of VUFLC can correspondingly
correspondingly adjust
adjust along with
along with
changes
changes to the input variables. The membership of input and output variables use the triangle. For aa
to the input variables. The membership of input and output variables use the triangle. For
complete
complete fuzzy
fuzzy controller
controller design,
design, aa defuzzification
defuzzification process
process is
is needed,
needed, that
that is,
is, the VUFLC output
the VUFLC output is
is
converted from a linguistic variable to a numerical variable. There are many defuzzification
converted from a linguistic variable to a numerical variable. There are many defuzzification methods, methods,
and
and in
in this
this design, the center-of-gravity
design, the center-of-gravity (COG)
(COG) defuzzification
defuzzification method
method was
was employed
employed [25,28,37].
[25,28,37].

3.3. Proposed Control Method

Figure 6 shows the block diagram of the PV MPPT control system. The ordinary PV module was
used for the
the PV
PV power
powergeneration
generationmodel,
model,the
theDC/DC
DC/DCwas
wasconverted
convertedand
andthe
the output
output was
was connected
connected to
to a load.
a load.

Block diagram of the PV MPPT control system.

Figure 6. Block

As previously
As previously described,
described, the
the proposed
proposed design
design steps
steps of
of the
the VUFLC
VUFLC applied
applied to
to PV
PV MPPT
MPPT system
system
are presented in the remainder of this subsection.
are presented in the remainder of this subsection.
In terms of the input and output variables, the proposed VUFLC has two input variables and one
output variable. The two VUFLC input variables are the error E(k) and the difference in error CE(k),
which are calculated as follows [38].

p ( k ) − p ( k − 1)
E(k) = (9)
v ( k ) − v ( k − 1)
Electronics 2018, 7, 355 7 of 18

CE(k) = E(k) − E(k − 1) (10)

where k refers to the iteration number; p(k) is the instantaneous output power of PV; and v(k) is the
instantaneous output voltage of PV corresponding to kth sample.
In terms of the fuzzy control rules, according to the actual operation of the PV power generation
project and the previous MPPT control engineering experience, the linguistic expressions and the
initial universes of the inputs and output variables are given in Table 1.

Table 1. The input and output variables and initial universe.

Type Universe
Parameter Linguistic
I/O Min Max
Power error/Volt error (x1 ) Input NB NS ZE PS PB −40 +40
Error change (x2 ) Input NB NS ZE PS PB −80 +80
Duty change (y) Output NB NM NS ZE PS PM PB −0.09 +0.09

The input and output variables have five and seven linguistic expressions, respectively. The input
variables have five linguistic expressions as follows: NB (negative big), NS (negative small), ZE (zero),
PS (positive small) and PB (positive big), and the output variables have seven linguistic expressions
as follows: NB (negative big), NM (negative medium), NS (negative small), ZE (zero), PS (positive
small), PM (positive medium), and PB (positive big), which adds two linguistic expressions. The initial
universe of the input variables x1 and x2 are normalized to the range (−40, +40) and (−80, +80),
respectively. The initial universe of the output variable y is normalized to the range (−0.09, +0.09).
All of the membership functions of the input and output variables use a triangular form, which is easy
to calculate and specifies the entire fuzzy partition of these variables.
According to Equations (9) and (10), the sign of the input variable x1 shows if the operating point
is located on the left or right side when compared to the actual MPP position P–V curve, while x2
expresses the moving direction of this operation point [38,39]. The output control variable y can be
obtained under the fuzzy control rules. Based on Equation (8), the fuzzy control rules are given in
Table 2, which determine the VUFLC output control signal.

Table 2. The input and output variables and initial universe.

CE(x2 ) U(y) E(x1 ) NB NS ZE PS PB

NB NB NS PS PM PB
NS NS PS PM PM PB
ZE NM NS ZE PS PM
PS NS ZE PS PM PB
PB NB NM NS PS PB

In terms of universe control factor design, the input and output variables can be adaptively
adjusted by the contraction factor in Equations (6) and (7), and the variable universe process can
be described from Figure 5. The conventional contraction factors are presented and discussed in
Reference [26], which achieved the contraction of the universe when the input variable was small.
In order to speed up the response time and improve the control accuracy, an improved universe control
factor for the input variable was proposed and designed, considering the influence of temperature
characteristics. The control factors of the new input variable are defined as follows:

α0 ( xi ) = 1 − λ1 exp[−k1 (γmax ·∆TC )2 ], 0 < λ1 < 1, k1 > 0, i = 1, 2; |∆Tc | ≤ θset (11)

where 0 < α0 ( xi ) < 1, and is related to the maximum power temperature coefficient γmax and the
temperature change value ∆TC = T. The γmax can be found in the manufacturer’s datasheet.
Electronics 2018, 7, 355 8 of 18

In Equation (8), the value of γmax ·∆TC represents the power change with ambient temperature.
When ∆TC increases, the α0 ( xi ) value decreases, the universe of input variables x1 and x2 are expanded,
the output power change becomes larger, and VUFCL will achieve fast MPP tracking. However,
if |∆Tc | ≥ θset (θset is the maximum threshold for the temperature change setting), then α0 ( xi ) will
become 1, and the universe variables take the maximum value. On the other hand, when ∆TC drops,
the α0 ( xi ) value rises, the universe of input variables x1 and x2 are contracted, and the VUFLC will
limit the oscillations and improve the MPP control tracking accuracy. Similarly, the modified output
variable control factor can be expressed as:

β(y) = 1 − λ2 exp[−k2 ( βOC ·∆TC )2 ], 0 < λ1 < 1, k1 > 0; |∆Tc | ≤ θset (12)

where β(y) is related to the βOC ·∆TC , because the output control signal is ∆D, which can adjust and
change the voltage ratio of the converter. When the control factor β(y) takes the voltage temperature
effect into account, the control compensation can be achieved, and the control accuracy is improved.
Different value selection of parameters λ1 , λ2 , k1 , and k2 will have a certain impact on the range
of variable universes and affect the convergence of the proposed control method. Considering the
correlation between the variables and actual engineering application, the related parameters were
chosen as λ1 = 0.4, λ2 = 0.6, and k1 = k2 = 104 . Hence, α0 ( xi ) ranges from 0.4 to 1.0, and the range of
E(k) and CE(k) are limited to the limits (−40/−36, +36/+40) and (−80/−48, +48/+80), respectively.
Similarly, the range of U (output universe) is tuned to the limits (−0.09/−0.054, +0.054/+0.09).
Furthermore, the θset was selected as 2 ◦ C. Therefore, the proposed VUFLC-temperature can change
the input and output universes with control factor variation. According to the selected parameters,
the input and output universes control factors α0 ( xi ) and β(y) are calculated by follows:
)
α0 ( xi ) = 1 − 0.4 exp[−104 ·(γmax ·∆TC )2 ]
when|∆TC | ≤ 2 ◦ C
β(y) = 1 − 0.6 exp[−104 ·( βOC ·∆TC )2 ] (13)
else α0 ( xi ) = β(y) = 1, |∆TC | > 2 ◦ C

Figure 7 shows the curves of the variable universe control factors.

Electronics 2018, 7, x FOR PEER REVIEW 9 of 18

Figure 7.
Figure Curves of
7. Curves of variable
variable universe
universe control
control factors.
factors.

The TC of Pmax , γmax , VOC and βOC were chosen as −0.40%/◦ C and −0.30%/◦ C, which can
The
γ , VOC and βOC were chosen as −0.40%/°C and −0.30%/°C, which can adjust
of Pmax, ofmax
adjust theTC
universes the input and output variables to achieve the MPPT of a PV system by using
the universes of the input
control factors. Therefore, andtheoutput variables to achieve
VUFLC-temperature the MPPTthe
can improve of acontrol
PV system by using
accuracy and control
reduce
factors. Therefore, the
power fluctuations VUFLC-temperature
in the PV MPPT. can improve the control accuracy and reduce power
fluctuations
In regardin the PVimplementation
to the MPPT. of VUFLC for MPP, the proposed VUFLC-temperature MPPT
In regard to the implementation
control algorithm was implemented as offollows.
VUFLC First,
for MPP, the proposed
the controller VUFLC-temperature
detects the output Vpv andMPPT
I pv of
the PV module and computes the E(k) and CE(k), then it measures the PV module temperatureV TI (k)
control algorithm was implemented as follows. First, the controller detects the output pv and C pv
and evaluates ∆T (or calculates by related evaluation method of the test standard). According to
of the PV moduleCand computes the E(k) and CE(k), then it measures the PV module temperature
the different ∆TC values, the VUFLC-temperature controller selects different universe control factors
TC(k) and evaluates △TC (or calculates by related evaluation method of the test standard). According
to the different △TC values, the VUFLC-temperature controller selects different universe control
factors based on Equation (13). Finally, the updated duty cycle control signal is output to control the
power converter and achieve MPPT tracking. The detailed and complete control implementation flow
chart is shown in Figure 8.
fluctuations in the PV MPPT.
In regard to the implementation of VUFLC for MPP, the proposed VUFLC-temperature MPPT
V I
control algorithm was implemented as follows. First, the controller detects the output pv and pv
of the PV module and computes the E(k) and CE(k), then it measures the PV module temperature
Electronics 2018, 7, 355 9 of 18
TC(k) and evaluates △TC (or calculates by related evaluation method of the test standard). According
to the different △TC values, the VUFLC-temperature controller selects different universe control
based on
factors Equation
based (13). Finally,
on Equation the updated
(13). Finally, duty cycle
the updated dutycontrol signal is
cycle control output
signal to control
is output the power
to control the
converter and achieve MPPT tracking. The detailed and complete control implementation
power converter and achieve MPPT tracking. The detailed and complete control implementation flow flow chart is
shownisin
chart Figure
shown in8.Figure 8.

Start

Control initialization

Input θ set value

Detect Vpv and Ipv of PV module,then

compute E(k) and CE(k)

Measure the PV module

temperature Tc(k),and evaluateΔTC
Parameters
acquisition and
processing Yes No
If Δ TC ≤ θ set ?

α' (xi ) =1− 0.4exp[−104 ⋅ (γ max ⋅ΔTC )2 ] α ' ( xi ) = 1

β ( y ) = 1 − 0.6exp[−104 ⋅ ( βOC ⋅ ΔTC )2 ] β ( y) = 1

Rules base
Fuzzifization Fuzzy inference

VUFLC-temp
MPPT Defuzzifization
Algorithm

Update duty cycle

control signal

Return

Flowchart
Figure 8.8. Flowchart of proposed
of the the proposed variable
variable universe
universe fuzzy
fuzzy logic logic temperature
control control temperature
(VUFLC-
(VUFLC-temperature) MPPT control algorithm
temperature) MPPT control algorithm procedure. procedure.

4. Simulation Results
The MPPT control system based on VUFLC-temperature algorithm was simulated and developed
to test and confirm the proposed method, as shown in Figure 6. A boost circuit was selected to be
the converter to achieve the MPP by adjusting the control signal through the VUFLC-temperature
controller. The proposed control algorithm and MPPT system were simulated in MATLAB/Simulink
(version 9.1, the MathWorks, Inc., Natick, MA, USA). The simulation model consisted of radiation and
temperature input units, a PV module, a converter, a load and the proposed controller. The output of
the PV module was connected to the boost converter, then the controller adjusted the duty cycle of the
converter control signal to achieve maximum power control. The proposed VUFLC-temperature MPPT
PV system is shown in Figure 9. The control logic was implemented through software programming.
The PV module, boost converter and load were built with the related components in SimPowerSystems.
output of the PV module was connected to the boost converter, then the controller adjusted the duty
cycle of the converter control signal to achieve maximum power control. The proposed VUFLC-
temperature MPPT PV system is shown in Figure 9. The control logic was implemented through
software programming. The PV module, boost converter and load were built with the related
Electronics 2018, 7, 355 10 of 18
components in SimPowerSystems.

9. Simulation model of the VUFLC-temperature PV MPPT system.

FigureFigure
9. Simulation model of the VUFLC-temperature PV MPPT system.
The simulation parameters of the Crystalline Silicon PV module are listed in Table 3.
TheThesimulation
simulationparameters of the of
input was composed Crystalline Silicon
five listed PV PVconnected
modules module inare listed
series andin
theTable
total 3. The
simulation input
input peakwas composed
power of five listed PV modules connected in series and the total input peak
was 1650 W.
power was 1650 W.
Table 3. Simulation PV module parameters.

Electrical(STC)
Table 3. Simulation PV moduleTemperature Characteristics
parameters.
Specification Data Specification Data
Electrical(STC)
Maximum Power (Pmax ) 330 W Temperature
Temperature Characteristics
Coefficient of Pmax −0.41%/◦ C
Specification Data Specification
Temperature Coefficient of VOC −0.38%/◦ Data
C
Optimum Operating Voltage (Vmp ) 37.5 V
Maximum Power Temperature Coefficient of
Temperature Coefficient of ISC 0.05%/ ◦ C
330W −0.41%/°C
(Pmax) Open Circuit Voltage (V ) Pmax
Nominal Operating Cell Temperature 45±2 ◦ C
oc 46.2 V
Operational Temperature −40~+85 ◦C
Optimum Operating Temperature Coefficient of VOC −0.38%/°C
37.5V
Voltage (Vmp) Temperature Coefficient of ISC 0.05%/°C
Figures 2 and 3 demonstrate that the temperature will affect the PV output power and that the MPP
Nominal Operating Cell
Open Circuit Voltage
of the PV module also shifts with temperature changes [30]. At maximum and minimum operating 45±2°C
temperatures 46.2V Temperature
per day, the output power can vary by about 20%. The simulation model system (Figure 9)
(Voc)
can simulate the temperature and irradianceOperational
fluctuation of Temperature −40~+85°C
the input PV array, which indicates the
adaptability and superiority of the system under different operation conditions. In order to better
demonstrate
Figures 2b andthe proposed control
3 demonstrate strategy,
that its control effects
the temperature were
will compared
affect the PV with conventional
output powerMPPTand that the
FLC (fuzzy logic control) and INC (incremental conductance) under the same conditions.
MPP of the PV module also shifts with temperature changes [30]. At maximum and minimum
Figure 10 shows that the simulation results of the MPPT control tracking process under the
operating temperatures per day, the output power can vary by about 20%. The simulation model
solar radiation intensity remained at 500 W/m2 constantly and the temperature changed slowly;
system (Figure 9) can rose
the temperature simulate
from athe temperature
minimum of 0 ◦ C toand irradiance
a maximum of 72fluctuation of theback
◦ C, then dropped input PV array,
to the
which indicates the adaptability
lowest temperature value as and
seen superiority
in Figure 10a,of the system
where undervoltage
the PV output different operation
and power conditions.
changed
slowly. In Figure 10b, using the proposed method, the output and power could respond
In order to better demonstrate the proposed control strategy, its control effects were compared with quickly with
precise tracking. However, in Figure 10c with FLC and Figure 10d with INC, respectively, the output
conventional MPPT FLC (fuzzy logic control) and INC (incremental conductance) under the same
power had relatively large fluctuations at 0.43 s and 0.56 s, which could not accurately adapt to the
conditions.
temperature changes.
Figure 10 shows that the simulation results of the MPPT control tracking process under the solar
radiation intensity remained at 500 W/m2 constantly and the temperature changed slowly; the
temperature rose from a minimum of 0 °C to a maximum of 72 °C, then dropped back to the lowest
temperature value as seen in Figure 10a, where the PV output voltage and power changed slowly. In
Figure 10b, using the proposed method, the output and power could respond quickly with precise
tracking. However, in Figure 10c with FLC and Figure 10d with INC, respectively, the output power
had relatively
Electronics large fluctuations at 0.43 s and 0.56 s, which could not accurately adapt to
2018, 7, 355 the
11 of 18
temperature changes.

(a) (b)

(c) (d)

MPPTtracking
Figure 10. MPPT trackingsimulation
simulationresults
results
ofof different
different control
control methods
methods in variable
in variable temperature:
temperature: (a)
(a) Change temperature and keep the solar radiation constant; (b) VUFLC; (c) fuzzy logic control
Change temperature and keep the solar radiation constant; (b) VUFLC; (c) fuzzy logic control (FLC); (FLC);
(d) incremental conductance (INC).

In Figure
In Figure11, 11,the
theradiation
radiation andandtemperature
temperature simultaneously
simultaneously changed, with random
changed, changeschanges
with random during
the 1.0 sthe
during period.
1.0 s Before
period.0.3 s, the0.3
Before irradiance and temperature
s, the irradiance almost rose
and temperature synchronously,
almost then kept atthen
rose synchronously, the
maximum of 1000 W/m 2 and 60 ◦ C at 0.37 s. At 0.5 s, the radiation declined to 500 W/m2 , and at 0.78 s
kept at the maximum of 1000 W/m and 60 °C at 0.37 s. At 0.5 s, the radiation declined to 500 W/m2,
2

reached
and at 0.781000 W/m2 1000
s reached again. W/mAt 20.715
again. s, At
the0.715
temperature changed to
s, the temperature 25 ◦ C and
changed at°C
to 25 0.92
and s back to s60
at 0.92
◦ C.
back
In 60
to Figure
°C. In11b, with VUFLC,
Figure 11b, withwhen VUFLC, the radiation
when theand temperature
radiation varied, the power
and temperature varied,curve had smooth
the power curve
tracking, no power loss and the power ripple maximum power
had smooth tracking, no power loss and the power ripple maximum power point oscillations point oscillations were eliminated.
were
In Figure 11c,d,
eliminated. with FLC
In Figure and with
11c,d, INC, the FLCpower
and INC,trackingtheerror
powerandtracking
fluctuation were
error andcomparatively
fluctuation large,
were
when the temperature
comparatively changed
large, when thethe power curves
temperature overshot
changed theand loss were
power curves occurred.
overshot and loss were
Figure
occurred. 12 shows the simulation results of the MPPT tracking process under the temperature
step Figure
change.12Before ◦
shows0.4 thes,simulation
the temperature results was
of the 20MPPTC. Attracking
0.4 s, the temperature
process under thestep increased from
temperature step
20 ◦ C to 50 ◦ C. Additionally, before 0.2 s, the initial radiation was 0 W/m2 , and it quickly increased
change. Before 0.4 s, the temperature was 20 °C. At 0.4 s, the temperature step increased from 20 °C
to 50
1000 2 at 0.25 s, then at 0.4 s began to drop to 250 W/m2 at 0.45 s. As shown in Figure 12b,
to °C.W/m
Additionally, before 0.2 s, the initial radiation was 0 W/m2, and it quickly increased to 1000
the VUFLC has a quick
W/m at 0.25 s, then at 0.4
2 MPPT
s beganresponse.
to drop However,
to 250 W/m Figure 12c with
2 at 0.45 s. AsFLC
shownandinFigure
Figure 12d with
12b, theINC
VUFLChad
a large overshoot and dynamic error; at 0.4 s, the power tracking
has a quick MPPT response. However, Figure 12c with FLC and Figure 12d with INC had a large with FLC and INC both had power
fluctuationand
overshoot loss.dynamic error; at 0.4 s, the power tracking with FLC and INC both had power
The simulation
fluctuation loss. results for the proposed VUFLC-temperature based MPPT control method are
presented and compared
The simulation resultsto for
thethe
conventional FLC with temperature
proposed VUFLC-temperature change.
based MPPT The variable
control universe
method are
control factors of the VUFLC controller will dynamically adjust according
presented and compared to the conventional FLC with temperature change. The variable universe to the change in atmosphere;
when
controlthefactors
moduleoftemperature
the VUFLC rises or drops,will
controller the input and output
dynamically universe
adjust controlto
according factors are chosen
the change in
with differentwhen
atmosphere; values thetomodule
speed up the MPPTrises
temperature and or control
drops,convergence
the input and andoutput
all theuniverse
power tracking
control
waveforms with VUFLC
factors are chosen are smooth
with different valueswith toless
speedlossup andtheno overshooting,
MPPT and control which has a faster
convergence andtracking
all the
speed and more precise control effect than the other methods.
power tracking waveforms with VUFLC are smooth with less loss and no overshooting, which has It can be seen from the results thata
VUFLC had a significant
faster tracking speed and impact on thecontrol
more precise MPP tracking
effect thancontrol, where
the other a relatively
methods. It cansmall universe
be seen from can
the
improve
results theVUFLC
that control hadaccuracy and reduce
a significant impact theon oscillation
the MPPattracking
the MPP. The comparisons
control, between
where a relatively the
small
simulation results and existing others are briefly summarized in Table 4.
Electronics 2018, 7, x FOR PEER REVIEW 12 of 18
Electronics 2018, 7, x FOR PEER REVIEW 12 of 18
universe can 7,improve
Electronics 2018, 355 the control accuracy and reduce the oscillation at the MPP. The comparisons
12 of 18
universe can improve the control accuracy and reduce the oscillation at the MPP. The comparisons
between the simulation results and existing others are briefly summarized in Table 4.
between the simulation results and existing others are briefly summarized in Table 4.

(a) (b)
(a) (b)

(c) (d)
(c) (d)
Figure 11. MPPT tracking
tracking simulation
simulation results
results of different control methods in variable radiation and
Figure 11. MPPT
temperature: (a) tracking
Changing simulation
solar resultsand
radiation of different control
temperature methods in variable
simultaneously; radiation and
temperature: (a) Changing solar radiation and temperature simultaneously; (b) (b) VUFLC;
VUFLC; (c) FLC;
(c) FLC; (d)
temperature:
(d) INC. (a) Changing solar radiation and temperature simultaneously; (b) VUFLC; (c) FLC; (d)
INC.
INC.

(a) (b)
(a) (b)

(c) (d)
(c) (d)
Figure 12. MPPT tracking simulation results of different control methods in variable radiation and
Figure 12.
Figure MPPT tracking
12. MPPT tracking simulation
simulationresults
resultsofofdifferent control
different methods
control in variable
methods radiation
in variable and and
radiation step
step temperature: (a) Changing solar radiation and step temperature simultaneously; (b) VUFLC; (c)
temperature: (a) Changing solar radiation and step temperature simultaneously; (b) VUFLC;
step temperature: (a) Changing solar radiation and step temperature simultaneously; (b) VUFLC; (c) (c) FLC;
FLC; (d) INC.
(d) INC.
FLC; (d) INC.
According tothethe
According toto simulation
simulation and comparative
and and
comparative analysis
analysis results,
results, the the proposed VUFLC-
proposed
According the simulation comparative analysis results, the VUFLC-temperature
proposed VUFLC-
temperature
MPPT method MPPT method
had method had
a betterhad a better
control control performance,
performance, especially under especially under conditions
conditions of
temperature MPPT a better control performance, especially underofconditions
temperatureof
temperature change.
change. Thechange. The VUFLC-temperature
VUFLC-temperature method couldmethod could
also obtain also obtain
a fastobtain a
trackingfast tracking
speed, speed,
small small
oscillation,
temperature The VUFLC-temperature method could also a fast tracking speed, small
oscillation,
and improvedandaccuracy
improved accuracy
during the during
step the step temperature
temperature change. change.
oscillation, and improved accuracy during the step temperature change.
Electronics 2018, 7, x FOR PEER REVIEW 13 of 18
Electronics 2018, 7, 355 13 of 18
Table 4. Performance comparison of different MPPT control methods.

Table 4. Performance comparison ofMPPT Methods

different MPPT control methods.
Items Proposed
P&O INC ANN
MPPT Methods FLC
VUFLC
Items Proposed
Dynamic
PoorP&O INC
Medium ANN
High FLC
Medium VUFLCHigh
response
Dynamic
Transient
response BadPoor Medium
Bad High
Good Medium
Good High
Good
fluction
Transient
Steady Bad Bad Good Good Good
fluction Large Moderate Zero Small Zero
oscillation
Steady
Static error HighLarge High
Moderate Low
Zero Low
Small ZeroLow
oscillation
Control
Static error LowHigh Accurate
High Accurate
Low Accurate
Low Excellect
Low
accurcy
Control accurcy
Tracking Low Accurate Accurate Accurate Excellect
Tracking speed
SlowSlow
Slow
Slow
Moderate
Moderate
Fast
Fast
Very fast
Very fast
speed
OverallOverall
Medium
Medium Medium
Medium High
High High
High HighHigh
efficiency
efficiency
System System
Simple Simple Medium Medium Medium
complexity Simple Simple Medium Medium Medium
complexity
Temperature
Temperature
characteristics PoorPoor Poor
Poor Good
Good Good
Good Excellect
Excellect
characteristics

5. Experimental Validation
In order to further verify the analysis and simulation results, the proposed VUFLC-temperature
MPPT control
control algorithm
algorithmwas
wasexperimentally
experimentallyvalidated onon
validated a PV system
a PV systemprototype. A photograph
prototype. of the
A photograph of
experimental prototype hardware is shown in Figure 13.
the experimental prototype hardware is shown in Figure 13.

Figure 13. Experimental setup of the PV MPPT system.

A PV module analog
analog programmable
programmable DC DC power
power supply
supply 6215H-600S
6215H-600S (CHROMA
(CHROMA ATE ATE (SUZHOU)
(SUZHOU)
CO., LTD.,
LTD.)Suzhou, China) was
was employed employed
as the input foras the
the input
test; itforwas
the also
test; used
it wastoalso used todifferent
emulate emulate working
different
working environments and temperature changes. The main control chip was a
environments and temperature changes. The main control chip was a DSP (Digital signal processor) DSP (Digital signal
processor) TMS320F28035
TMS320F28035 which waswhich was employed
employed for implementing
for implementing the proposedthe proposed control algorithm.
control algorithm. A boost
A boost converter was employed to achieve the power conversion and MPPT.
converter was employed to achieve the power conversion and MPPT. A grid-connected inverter was A grid-connected
inverter was
connected to connected
the outputtoofthe
theoutput of the boost converter.
boost converter.
The experimental
The experimental waveforms
waveforms underunder different
different operating
operating conditions
conditions were
were captured
captured using
using the
the
Chroma dedicated photovoltaic power generation monitoring software (F/W
Chroma dedicated photovoltaic power generation monitoring software (F/W Version: Chroma ATE Version: Chroma ATE
61250H-600S,00368,01.10). The
61250H-600S,00368,01.10). Theexperimental
experimental results for the
results for temperature changechange
the temperature and radiation constant
and radiation
operationoperation
are shownare
in shown
Figure 14. The radiation 2 ◦
constant in Figure 14. Thewas 1000 W/m
radiation andW/m
was 1000 temperature was 52 C. was
2 and temperature The MPPT
52 °C.
Electronics 2018, 7, 355 14 of 18
Electronics 2018, 7, x FOR PEER REVIEW 14 of 18

The
P–V MPPT
curve isP–V curveasisashown
shown brightlyascolored
a brightly colored
thick line inthick line14a.
Figure in Figure 14a.was
The MPP The1636.50
MPP wasW, 1636.50
then as
W,
the then
temperature rose to 60 ◦ C
as the temperature rose to 60
as the °C as
light thecurve
color light color
shows,curve
whenshows, when downloading
downloading this
this I–V input fileI–V
to
input file to the 6215H-600S, the experimental waveform becomes the P–V curves in
the 6215H-600S, the experimental waveform becomes the P–V curves in Figure 14b, and the MPP was Figure 14b, and
the MPPW.
1582.30 was 1582.30 W.

(a) (b)

Figure 14.
14.Experimental
Experimental P–VP–V tracking
tracking curvecurve
resultsresults of the irradiation
of the constant constant irradiation
and change and change
temperature.
2
(a) 1000 W/m(a) , 52 ◦ C W/m 2
(b) 1000 2 ◦ 2
temperature. 1000 2 , 52W/m
°C (b), 1000
60 C.W/m , 60 °C.
2

The maximum
The maximum power
power reduced
reduced can
can be
be calculated
calculated as
as follows:
follows:

Δ∆P max==γ γ
Pmax max ⋅Δ
max T ⋅· PPmax
·∆T max (14)
(14)

According
According to tothe
theset
setexperimental
experimental temperature
temperature coefficient, γ max
coefficient, γmaxwas
was −0.41%/a◦nd
−0.41%/°C thePset
the set
C and max
max

Pmax maximum
maximum power power
was was
16501650
W ofW6215H-600S
of 6215H-600S output.
output. Because
Because ΔT∆Twas
the the
◦ C, the theoretical
was88 °C, the
calculation
calculation was was 54.12
54.12 W.
W. Additionally,
Additionally, thethe experimental
experimental test
test value was 54.20
value was 54.20 W,
W, which
which could rapidly
could rapidly
maintain maximum power
maintain maximum power tracking. tracking.
Figure 15
Figure 15shows
shows thethe
experimental
experimentalresults of theofMPPT
results the tracking process under
MPPT tracking process different
undermaximum
different
powers
maximum (thepowers
maximum (thepower was set
maximum at 1500
power W),set
was irradiations andirradiations
at 1500 W), temperatures.
andWhen the temperature
temperatures. When
varied from 35 ◦ C to 45 ◦ C and the radiation synchronization increased from 500 to 1000 W/m2 ,
the temperature varied from 35 °C to 45 °C and the radiation synchronization increased from 500 to
the MPP
1000 W/malso changed
22, the MPP alsofrom 798.7 tofrom
changed 1498.5 W, and
798.7 the control
to 1498.5 W, and output poweroutput
the control fast tracked
powerthe new
fast MPP
tracked
quickly, as shown in Figure 15b.
the new MPP quickly, as shown in Figure 15b.

(a) (b)
Experimental P–V tracking
Figure 15. Experimental tracking curve
curve results
results of
of the
the change
change irradiation
irradiation and
andchange
changetemperature.
temperature.
W/m222, ,35
(a) 500 W/m ◦ C (b) 1000 W/m
35°C 2 2 , 45 ◦ C.
(b) 1000 W/m , 45 °C.
2

To verify the effectiveness and advantages of the proposed VUFLC-temp MPPT algorithm,
comparative experiments of different MPPT control methods were conducted in the same
experimental system. Figure 16 shows the experimental results of the three MPPT algorithms.
Electronics 2018, 7, 355 15 of 18

To verify the effectiveness and advantages of the proposed VUFLC-temp MPPT algorithm,
comparative experiments of different MPPT control methods were conducted in the same experimental
system. Electronics
Figure 2018, 7, the
16 shows x FOR PEER REVIEWresults of the three MPPT algorithms.
experimental 15 of 18
Electronics 2018, 7, x FOR PEER REVIEW 15 of 18

Electronics 2018, 7, x FOR PEER REVIEW 15 of 18

(a) (a) (b) (b)

(a) (b)

(c) (c) (d) (d)

Figure 16.
Figure Figure
16.Under 16. Under step
Undertemperature
temperaturetemperature
step change:
change: step
(a) (a) change: (a)results
Experimental
Experimental Experimental
results results
of MPPT
of MPPT of MPPT
control;
control; using using
(b) control;
(b)
VUFLC, using (b)
VUFLC,
(c) (c)VUFLC,
FLC and FLC and(c)
(d) INC. (d)FLC and (d) INC.
INC.

(c) W/m2.21000 (d)

As shownAs
As shown in shown16a,
in Figure in Figure 16a, the irradiation
the irradiation intensity was intensity was 1000
constant
a constant a constant
1000 W/m W/m
. Before
Before 2. Before
5.0
5.0 s,s,the
theset5.0 s, the set
set
temperature temperature
was 47 ◦ C.was
At 47
5.0 °C.
s, At
the 5.0
temperatures, the temperature
steps from steps
47 to from
52 ◦ 47
C. to
The 52 °C. The
temperature temperature
was kept was(b)
at kept at
temperature was Figure°C. 16. Under s, the temperature
temperature step steps
change: from (a) 47 to 52 °C. The
Experimental temperature
results of MPPTwas kept at
control; using
◦ C at 1052 ◦ ◦
52
52 °C s, °C
thenatat
1015.0
at
VUFLC, s, (c)
then
15.0 ititat
s,s,FLC 15.0(d)
stepped
stepped
and s, from
itINC.
stepped
from 52°CCfrom
52 toto606052
°C.C.°C to 6016b–d
Figure
Figure °C. Figure
16b–d show
show 16b–d
thethe showwith
results
results the
withresults
VUFLC,
VUFLC, with VUFLC,
FLC,
FLC, and FLC, respectively.
and INC,
INC, and INC, respectively.
respectively.
The As
Theshown
The experimental
experimental in Figure
experimental
results
results show
show 16a,
results
that
that the show
theirradiation
the that the
proposed
proposed intensity
proposed wasVUFLC-temperature
VUFLC-temperature
VUFLC-temperaturea constantmethod 1000 W/m
method had
had2. Before 5.0 s, the set
method
a afaster had a faster
faster
tracking temperature
tracking speed
speed andsmooth
tracking
and speedwas
smooth and 47 °C. At in
transition
smooth
transition 5.0 in s,the
the
transition
the temperature
temperature
temperaturein the step steps
step from
change
temperature
change 47
in to
instep 5216b.
Figure
Figurechange °C.16b.The
in temperature
However,
Figure
However, 16b. the
the MPP was keptthe
However, at
fluctuations 52
MPP fluctuations
MPP °C fluctuations
and atoscillations
10
ands, oscillations
then at and15.0 s, it at
occurredstepped
oscillations
occurred temperature from 52transition
atoccurred
temperature at°Ctemperature
totransition
60 °C. Figure
points points 16b–d
transition
with show
withFLC
the the
the and
points FLCwithresults
andthe
INC INCwithand
FLC
where VUFLC,INC
wherewere
there there
FLC,were
where
both and both
INC,
there
power power
were
losses inlosses
respectively.
both power
Figure in16c,d.
Figure
losses 16c,d.
in Figure
Furthermore, Furthermore,
16c,d.
the VUFLC thehadVUFLC
Furthermore, better had
the betterstability
VUFLC
tracking tracking
had betteras tracking
stability
well as well
The
stability
as a more asas
robust a and
more
experimental
well asrobust
lower a more anderror
results
static lower
show
robust static
and
than that error
the
lower
the others. than
proposed
static the
Theerror others.
maximumthan the The
VUFLC-temperature maximum
others.
power method
The
magnitudes power
maximum
andhad apower
faster
magnitudes
MPPT and MPPT
tracking
magnitudes
effectivenesses speed effectivenesses
and
with andMPPT smooth
different with different
transition
effectivenesses
control in
with
algorithms control
the different
are algorithms
temperature
listed control step
in Table are listed in
change
algorithms
5. Table
inare
Figure 5. 16b.
listed However,
in Table 5. the
MPP fluctuations and oscillations occurred at temperature transition points with the FLC and INC
where there were Table
Table both5. Experimental
Experimental
Tablelosses
power comparison
comparison
5. Experimental of
of MPPT
in Figurecomparison
16c,d. methods.
MPPTFurthermore,
methods.
of MPPT methods. the VUFLC had better tracking
stability as well as a more robust and lowerThe static error
Experimental than the others. The maximum power
Results
Condition (1000 W/m 2) 2 The The Experimental
Experimental Results Results
magnitudes Condition
and MPPT
Condition (1000 W/m
W/m ) 2
(1000effectivenesses ) INC with different control FLCalgorithms are listedVUFLC
Proposed in Table 5.
INC INC FLC Proposed VUFLC Proposed VUFLC
FLC
47 °C 47 °CMaximumMaximum power
Maximum
Table 5.
power
power 1481.2
Experimental comparison 1485.3
of MPPT methods. 1488.9
47 ◦ C (W) 1481.2 1481.21485.3 1485.3
1488.9 1488.9
(W)
(W)
Tracking Tracking The Experimental Results
Condition Tracking
(1000 W/m2efficiency
) 98.23 98.92 99.93
efficiency (%) 98.2398.23
INC 98.92 98.92
99.93
FLC
99.93
Proposed VUFLC
efficiency(%) (%)
52 °C 47 MaximumMaximum power power
52 ◦°C
52 °C
C Maximum power 1452.1 1481.2
1452.11455.8 1455.8 1485.3
1455.8 1460.0 1488.9
1460.0
(W) (W) 1452.1 1460.0
(W)
Tracking Tracking
Tracking efficiency 98.01 98.74 99.90
efficiency (%) 98.0198.23
98.01 98.74 98.92
98.74
99.90 99.93
99.90
efficiency(%) (%)
Maximum power
52 °C Maximum
Maximum power power 1403.9 1407.7 1411.9
(W) 1403.9 1452.1
1403.91407.7 1455.8
1407.7
1411.9 1460.0
1411.9
60 °C ◦ (W)
(W)
60
60 °CC Tracking
Tracking
Tracking efficiency 97.87 97.8798.01 98.21 98.21 99.87
efficiency (%) (%) 97.87 98.74
98.21
99.87 99.90
99.87
efficiency (%)
Maximum power
It can be seen that when the temperature changed, the1403.9 proposed VUFLC-temperature 1407.7 controller1411.9
It can be seen that when(W) the temperature changed, the proposed VUFLC-temperature controller
had a better dynamic 60 °Cperformance than the conventional FLC and INC control algorithms, and it was
had a better dynamic performance Tracking than the conventional FLC and INC control algorithms, and it was
more effective at tracking and reducing the MPP oscillation. 97.87 98.21 99.87
more effective at tracking efficiency
and reducing (%) the MPP oscillation.
6. Conclusions
6. Conclusions
It can be seen that when the temperature changed, the proposed VUFLC-temperature controller
In thishad
paper, an advanced
a better dynamic MPPT VUFLC-temperature
performance method FLC
than the conventional was proposed for a photovoltaic
and INC control algorithms, and it was
In this paper, an advanced MPPT VUFLC-temperature method was proposed for a photovoltaic
system, which could dynamically adjust the universe of the fuzzy controller and consider the effects
Electronics 2018, 7, 355 16 of 18

It can be seen that when the temperature changed, the proposed VUFLC-temperature controller
had a better dynamic performance than the conventional FLC and INC control algorithms, and it was
more effective at tracking and reducing the MPP oscillation.

6. Conclusions
In this paper, an advanced MPPT VUFLC-temperature method was proposed for a photovoltaic
system, which could dynamically adjust the universe of the fuzzy controller and consider the effects
of temperature changes. The output characteristics of PV cells were discussed, and according to the
effects of temperature, the universe control factors were proposed and designed. Compared to the fixed
universe of conventional fuzzy control, the new VUFLC-temperature MPPT method had a dynamically
adjusted control factor according to the temperature change value, which could improve the MPPT
tracking speed and accuracy. Different experiments were carried out. The simulation and experimental
results verified the effectiveness and advantages of the proposed VUFLC-temperature MPPT method.
Compared to the traditional methods, the proposed controller had a better tracking control performance
under environmental changes in photovoltaic power generation systems, especially with temperature
variations. The experimental results of the control system are basically consistent with the theoretical
calculations when the temperature condition changes. As shown in Figure 14, the theoretical calculation
and actual error is only 0.08 W. There is almost no power loss and control overshoot in Figure 16. It has
the largest power generation when the temperature changes in Table 5, which is about 4 W higher
than other control methods. The proposed control method not only improves the MPP tracking speed,
it has the fastest tracking speed in all comparison control algorithms under the same simulation and
experimental conditions, but also has higher tracking efficiency, which can improve tracking efficiency
by approximately 1%.

Author Contributions: Y.W. proposed the main idea, designed the control system, performed the experiments
and wrote the paper. Y.Y., G.F and B.Z. contributed to the discussion of this research. H.W., H.T., L.F. and X.C.
double-checked and revised the whole manuscript.
Funding: This work was supported in part by the project of the Jiangsu Overseas Research and Training
Program for the University, Science and Technology Project of Jiangsu Province Construction System, Science and
Technology Planning Project of Suzhou City, and the Qinlan Project of Suzhou Vocational University.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Metry, M.; Shadmand, M.B.; Balog, R.S.; Abu-Rub, H. MPPT of Photovoltaic Systems Using Sensorless
Current-Based Model Predictive Control. IEEE Trans. Ind. 2017, 53, 1157–1167. [CrossRef]
2. Kreeumporn, W.; Ngamroo, I. Optimal Superconducting Coil Integrated Into PV Generators for Smoothing
Power and Regulating Voltage in Distribution System With PHEVs. IEEE Trans. Appl. Supercond. 2016, 26,
5402805. [CrossRef]
3. Reisi, A.R.; Moradi, M.H.; Jamasb, S. Classification and comparison of maximum power point tracking
techniques for photovoltaic system: A review. Renew. Sustain. Energy Rev. 2013, 19, 433–443. [CrossRef]
4. Ram, J.P.; Babu, T.S.; Rajasekar, N. A comprehensive review on solar PV maximum power point tracking
techniques. Renew. Sustain. Energy Rev. 2017, 67, 826–847. [CrossRef]
5. Li, X.; Li, Y.; Seem, J.E.; Lei, P. Detection of Internal Resistance Change for Photovoltaic Arrays Using
Extremum-Seeking Control MPPT Signals. IEEE Trans. Control Syst. Technol. 2016, 24, 325–333. [CrossRef]
6. Robles Algarín, C.; Sevilla Hernández, D.; Restrepo Leal, D. A Low-Cost Maximum Power Point Tracking
System Based on Neural Network Inverse Model Controller. Electronics 2018, 7, 4. [CrossRef]
7. Shebani, M.M.; Iqbal, T.; Quaicoe, J.E. Comparing bisection numerical algorithm with fractional short circuit
current and open circuit voltage methods for MPPT photovoltaic systems. In Proceedings of the 2016 IEEE
Electrical Power and Energy Conference (EPEC), Ottawa, ON, Canada, 12–14 October 2016; pp. 1–5.
Electronics 2018, 7, 355 17 of 18

8. Masoum, M.A.S.; Dehbonei, H.; Fuchs, E.F. Theoretical and experimental analyses of photovoltaic systems
with voltageand current-based maximum power-point tracking. IEEE Trans. Energy Convers. 2002, 7, 514–522.
[CrossRef]
9. Deopare, H.; Deshpande, A. Modeling and simulation of Incremental conductance Maximum Power Point
tracking. In Proceedings of the 2015 International Conference on Energy Systems and Applications, Pune,
India, 30 October–1 November 2015; pp. 501–505.
10. Amri, B.; Ashari, M. The comparative study of Buck-boost, Cuk, Sepic and Zeta converters for maximum
power point tracking photovoltaic using P&O method. In Proceedings of the 2015 2nd International
Conference on Information Technology, Computer, and Electrical Engineering (ICITACEE), Semarang,
Indonesia, 16–18 October 2015; pp. 327–332.
11. Lin, W.-M.; Hong, C.-M.; Chen, C.-H. Neural-Network-Based MPPT Control of a Stand-Alone Hybrid Power
Generation System. IEEE Trans. Power Electron. 2011, 26, 3571–3581. [CrossRef]
12. Karatepe, E.; Hiyama, T. Artificial neural network-polar coordinated fuzzy controller based maximum power
point tracking control under partially shaded conditions. IET Renew. Power Gener. 2009, 3, 239–253.
13. Modani, D.; Shrivastava, V. Improvement in efficiency of PV module using soft computing based MPPT.
In Proceedings of the 2016 3rd International Conference on Computing for Sustainable Global Development
(INDIACom), New Delhi, India, 16–18 March 2016; pp. 2273–2278.
14. Narendiran, S.; Sahoo, S.K.; Das, R.; Sahoo, A.K. Fuzzy logic controller based maximum power point tracking
for PV system. In Proceedings of the 2013 Fourth International Conference on Computing, Communications
and Networking Technologies (ICCCNT), Tiruchengode, India, 4–6 July 2016; pp. 29–34.
15. Da Rocha, N.M.M.; Martins, D.C.; Passos, J.C. MPPT method based on temperature control of the photovoltaic
cells. In Proceedings of the 2016 12th IEEE International Conference on Industry Applications (INDUSCON),
Curitiba, Brazil, 20–23 November 2016; pp. 1–8.
16. Francisco, R.; Cruz, D. An Optimized Maximum Power Point Tracking Method Based on PV Surface
Temperature Measurement. In Sustainable Energy—Recent Studies; Gebremedhin, A., Ed.; InTech: Vienna,
Austria, 2012.
17. Coelho, R.F.; Concerand, F.M.; Martins, D.C. A MPPT approach based on temperature measurements applied
in PV systems. In Proceedings of the 2010 IEEE International Conference on Sustainable Energy Technologies
(ICSET), Kandy, Sri Lanka, 6–9 December 2010; pp. 1–6.
18. Park, M.; Yu, I.A. Study on the optimal voltage for MPPT obtained by surface temperature of solar cell.
In Proceedings of the 30th Annual Conference of IEEE Industrial Electronics Society, Busan, South Korea,
2–6 Novemver 2004; Volume 30, pp. 2040–2045.
19. Ali, A.M.; Mustafa, S.S.; Mutlag, A.H. Design optimization of solar power system with respect to temperature
and sun tracking. In Proceedings of the 2016 Al-Sadeq International Conference on Multidisciplinary in IT
and Communication Science and Applications (AIC-MITCSA), Baghdad, Iraq, 9–10 May 2016; pp. 1–5.
20. Mellit, A.; Kalogirou, S.A. MPPT-based artificial intelligence techniques for photovoltaic systems and its
implementation into field programmable gate array chips: Review of current status and future perspectives.
Energy 2014, 70, 1–21. [CrossRef]
21. Mellita, A.; Kalogirou, S.A. Artificial intelligence techniques for photovoltaic applications: A review.
Prog. Energy Combust. Sci. 2008, 34, 574–632. [CrossRef]
22. Subudhi, B.; Pradhan, R. A Comparative Study on Maximum Power Point Tracking Techniques for
Photovoltaic Power Systems. IEEE Trans. Sustain. Energy 2013, 4, 89–98. [CrossRef]
23. Algazar, M.M.; AL-monier, H.; EL-halim, H.A.; El Kotb Salem, M.E. Maximum power point tracking using
fuzzy logic control. Electr. Power Energy Syst. 2012, 39, 21–28. [CrossRef]
24. Chin, C.S.; Neelakantan, P.; Yoong, H.P.; Teo, K.T.K. Fuzzy Logic Based MPPT for Photovoltaic Modules
Influenced by Solar Irradiation and Cell Temperature. In Proceedings of the 2011 UkSim 13th International
Conference on Computer Modelling and Simulation, Cambridge, UK, 30 March–1 April 2011; pp. 376–381.
25. Rezk, H.; Eltamaly, A.M. A comprehensive comparison of different MPPT techniques for photovoltaic
systems. Sol. Energy 2015, 112, 1–11. [CrossRef]
26. Li, H.-X. Variable universe stable adaptive fuzzy control of a nonlinear system. Comput. Math. Appl. 2002, 44,
799–815. [CrossRef]
27. Qin, L.; Hu, J.; Li, H.; Chen, W. Fuzzy Logic Controllers for Specialty Vehicles Using a Combination of Phase
Plane Analysis and Variable Universe Approach. IEEE Access 2017, 5, 1579–1588. [CrossRef]
Electronics 2018, 7, 355 18 of 18

28. Shan, W.; Ma, Y.; Newcomb, R.W.; Jin, D. Analog Circuit Implementation of a Variable Universe Adaptive
Fuzzy Logic Controller. IEEE Trans. Circuits Syst. II Exp. Briefs 2008, 10, 976–980. [CrossRef]
29. Tang, S.; Sun, Y.; Chen, Y.; Zhao, Y.; Yang, Y.; Szeto, W. An Enhanced MPPT Method Combining
Fractional-Order and Fuzzy Logic Control. IEEE J. Photovolt. 2017, 7, 640–650. [CrossRef]
30. Rekioua, D.; Matagne, E. Optimization of Photovoltaic Power Systems; Springer: London, UK, 2012.
31. Datta, M.; Senjyu, T. Fuzzy Control of Distributed PV Inverters/Energy Storage Systems/Electric Vehicles
for Frequency Regulation in a Large Power System. IEEE Trans. Smart Grid 2013, 4, 479–488. [CrossRef]
32. Ding, K.; Bian, X.G.; Liu, H.H.; Peng, T. A MATLAB-Simulink-Based PV Module Model and Its Application
Under Conditions of Nonuniform radiation. IEEE Trans. Energy Convers. 2012, 27, 864–872. [CrossRef]
33. Jiang, Y.; Qahouq, J.A.A.; Batarseh, I. Improved solar PV cell Matlab simulation model and comparison.
In Proceedings of the 2010 IEEE International Symposium on Circuits and Systems, Paris, France,
30 May–2 June 2010; pp. 2770–2773.
34. Villalva, M.G.; Gazol, J.R.; Filho, E.R. Modeling and circuit-based simulation of photovoltaic arrays.
In Proceedings of the 2009 Brazilian Power Electronics Conference, Bonito-Mato Grosso do Sul, Brazil,
27 September–1 October 2009; pp. 1244–1254.
35. Tan, Y.T.; Kirschen, D.S.; Jenkins, N. A Model of PV Generation Suitable for Stability Analysis. IEEE Trans.
Energy Convers. 2004, 19, 748–755. [CrossRef]
36. Ni, J.; Luo, J.; Chen, Z.; Chen, J. Variable Universe Adaptive Fuzzy Control for Liquid Lever. In Proceedings
of the 2008 International Symposium on Computational Intelligence and Design, Wuhan, China,
17–18 October 2008; Volume 1, pp. 58–161.
37. Cao, D.-Y.; Zeng, S.-P.; Li, J.-H. Variable universe fuzzy expert system for aluminum electrolysis.
Trans. Nonferrous Met. Soc. China 2011, 21, 429–436. [CrossRef]
38. Messai, A.; Mellit, A.; Guessoum, A.; Kalogirou, S.A. Maximum power point tracking using GA optimized
fuzzy logic controller and its FPGA implementation. Sol. Energy 2011, 85, 265–277. [CrossRef]
39. Tousi, S.M.R.; Moradi, M.H.; Basir, N.S.; Nemati, M. A functionbased maximum power point tracking
method for photovoltaic systems. IEEE Trans. Power Electron. 2016, 31, 2120–2128. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
International Journal of Research in Engineering and Science (IJRES)
ISSN (Online): 2320-9364, ISSN (Print): 2320-9356
www.ijres.org Volume 6 Issue 8 Ver. II ǁ 2018 ǁ PP. 89-94

Energy Comparison of MPPT techniques for PV Systems

Satyabrata Mohapatra, Satya Prakash Sahoo, Sibasweta Mishra,
Sidhant Pradhan, S B Pati
Department of Electrical and Electronics Engineering, , Gandhi Institute For Technology (GIFT), Bhubaneswar

Abstract: - Many maximum power point tracking techniques for photovoltaic systems have been developed to
maximize the produced energy and a lot of these are well established in the literature. These techniques vary in
many aspects as: simplicity, convergence speed, digital or analogical implementation, sensors required, cost,
range of effectiveness, and in other aspects. This paper presents a comparative study of ten widely-adopted
MPPT algorithms; their performance is evaluated on the energy point of view, by using the simulation tool
Simulink®, considering different solar irradiance variations.
Key-Words: - Maximum power point (MPP), maximum power point tracking (MPPT), photovoltaic (PV),
comparative study, PV Converter.

I. INTRODUCTION
Solar energy is one of the most important renewable energy sources. As opposed to conventional
unrenewable resources such as gasoline, coal, etc..., solar energy is clean, inexhaustible and free. The main
applications of photovoltaic (PV) systems are in either stand-alone (water pumping, domestic and street lighting,
electric vehicles, military and space applications) [1-2] or grid-connected configurations (hybrid systems, power
plants) [3].
Unfortunately, PV generation systems have two major problems: the conversion efficiency of electric
power generation is very low (9÷17%), especially under low irradiation conditions, and the amount of electric
power generated by solar arrays changes continuously with weather conditions.
Moreover, the solar cell V-I characteristic is nonlinear and varies with irradiation and temperature. In
general, there is a unique point on the V-I or V-P curve, called the Maximum Power Point (MPP), at which the
entire PV system (array, converter, etc…) operates with maximum efficiency and produces its maximum output
power. The location of the MPP is not known, but can be located, either through calculation models or by search
algorithms. Therefore Maximum Power Point Tracking (MPPT) techniques are needed to maintain the PV
array’s operating point at its MPP.
Many MPPT techniques have been proposed in the literature; examples are the Perturb and Observe (P&O)
methods [4-7], the Incremental Conductance (IC) methods [4-8], the Artificial Neural Network method [9], the
Fuzzy Logic method [10], etc...
These techniques vary between them in many aspects, including simplicity, convergence speed, hardware
implementation, sensors required, cost, range of effectiveness and need for parameterization.
The P&O and IC techniques, as well as variants thereof, are the most widely used.
In this paper, ten MPPT algorithms are compared under the energy production point of view: P&O, modified
P&O, Three Point Weight Comparison [12], Constant Voltage (CV) [13], IC, IC and CV combined [13], Short
Current Pulse [14], Open Circuit Voltage [15], the Temperature Method [16] and methods derived from it [16].
These techniques are easily implemented and have been widely adopted for low-cost applications. Algorithms
such as Fuzzy Logic, Sliding Mode [11], etc…, are beyond the scope of this paper, because they are more
complex and less often used.
The MPPT techniques will be compared, by using Matlab tool Simulink®, created by MathWorks,
considering different types of insulation and solar irradiance variations. The partially shaded condition will not
be considered: the irradiation is assumed to be uniformly spread over the PV array.
The PV system implementation takes into account the mathematical model of each component, as well as actual
component specifications. In particular, without lack of generality, we will focus our attention on a stand-alone
photovoltaic system constructed by connecting the dc/dc Single Ended Primary Inductor Converter (SEPIC) [17-
18] between the solar panel and the dc load as reported in Fig.1.

www.ijres.org 89 | Page
 Energy Comparison of MPPT techniques for PV Systems

II. PV ARRAY
A mathematical model is developed in order to simulate the PV array. Fig. 2 gives the equivalent influenced
circuit of a single solar cell, where IPV and VPV are the PV array’s current and voltage, respectively, Iph is the
35
cell’s photocurrent, Rj represents the nonlinear resistance of the p-n junction, and Rsh and Rs are the intrinsic
shunt and series resistances of the cell.

Since Rsh is very large and Rs is very small, these terms can be neglected in order to simplify the electrical
model. The following equation then describes the PV panel [8]:

where ns and np are the number of cells connected in series and the in parallel, q=1.602·10-19 C is the electron
charge, k=1.3806·10-23 J·K-1 is Boltzman’s constant, A=2 is the p-n junction’s ideality factor, T is the cell’s
temperature (K), Iph is the cell’s photocurrent (it depends on the solar irradiation and temperature), and Irs is the
cell’s reverse saturation current (it depends on temperature).
The PV panel here considered is a typical 50W PV module composed by ns=36 series-connected polycrystalline
cells (np=1). Its main specifications are shown in Table 1 while Fig. 2 and Fig. 3 show the power output
characteristics of the PV panel as functions of irradiance and temperature, respectively. These curves are
nonlinear and are crucially influenced by solar radiation and temperature.
The PV array is composed of three strings in parallel, each string consisting of 31 PV panels in series.
The total power is 4650W.

Table 1. Electrical characteristics of PV panel with an irradiance level of 1000 W/m2

Symbol Quantity Value
PMPP Maximum Power 50 W
VMPP Voltage at PMPP 17.3 V
IMPP Voltage at IMPP 2.89 A
ISC Short-Circuit Current 3.17 A
VOV Open-Circuit Voltage 21.8 V
TSC Temperature coefficient (0.065±0.015)%/°C
of ISC
TOC Temperature -(80±10) mV/°C
coefficient of
VOC
www.ijres.org 90 | Page
 Energy Comparison of MPPT techniques for PV Systems

Fig. 3. V-P panel characteristics for three different irradiance levels. Each point represents the MPP of related
curve.

III. MPPT CONTROL ALGORITHM

As known the output power characteristics of the PV system as functions of irradiance and temperature
curves are nonlinear and are crucially influenced by solar irradiation and temperature. Furthermore, the daily
solar irradiation diagram has abrupt variations during the day, as shown in Fig. 5. Under these conditions, the
MPP of the PV array changes continuously; consequently the PV system’s operating point must change to
maximize the energy produced. An MPPT technique is therefore used to maintain the PV array’s operating point
at its MPP. There are many MPPT methods available in the literature; the most widely-used techniques are
described in the following sections, starting with the simplest method.

www.ijres.org 91 | Page
 Energy Comparison of MPPT techniques for PV Systems

3.1 Constant Voltage Method

The Constant Voltage (CV) algorithm is the simplest MPPT control method. The operating point of the PV
array is kept near the MPP by regulating the array voltage and matching it to a fixed reference voltage Vref. The
Vref value is set equal to the VMPP of the characteristic PV module (see Table 1) or to another calculated best
fixed voltage. This method assumes that individual insulation and temperature variations on the array are
insignificant, and that the constant reference voltage is an adequate approximation of the true MPP. Operation is
therefore never exactly at the MPP and different data has to be collected for different geographical regions.
The CV method does not require any input. However, measurement of the voltage VPV is necessary in
order to set up the duty-cycle of the dc/dc SEPIC by PI regulator, as shown in the block diagram of Fig. 6.
It is important to observe that when the PV panel is in low insulation conditions, the CV technique is
more effective than either the P&O method or the IC method (analyzed below) [13]. Thanks to this
characteristic, CV is sometime combined together with other MPPT techniques.

3.2 Short-Current Pulse Method

The Short-Current Pulse (SC) method achieves the MPP by giving the operating current Iop to a current-
controlled power converter. In fact, the optimum operating current Iop for maximum output power is proportional
to the short-circuit current ISC under various conditions of irradiance level S as follows:
Iop S   k  ISC S  (2)
where k is a proportional constant. Eq. (2) shows that Iop can be determined instantaneously by detecting ISC. The
relationship between Iop and ISC isstill proportional, even though the temperature varies from 0°C to 60°C. The
proportional parameter is estimated to be approximately 92% [14].
Therefore, this control algorithm requires measurements of the current ISC. To obtain this measurement,
it is necessary to introduce a static switch in parallel with the PV array, in order to create the short-circuit

www.ijres.org 92 | Page
 Energy Comparison of MPPT techniques for PV Systems

condition. It is important to note that during the short-circuit VPV=0 consequently no power is supplied by the PV
system and no energy is generated. As in the previous technique, measurement of the PV array voltage VPV is
required for the PI regulator (see Fig. 7) in order to obtain the Vref value able to generate the current Iop.

3.3 Open Voltage Method

The Open Voltage (OV) method is based on the observation that the voltage of the maximum power point is
always close to a fixed percentage of the open-circuit voltage. Temperature and solar insulation levels change
the position of the maximum power point within a 2% tolerance band.
In general, the OV technique uses 76% of the open-circuit voltage VOV as the optimum operating voltage Vop (at
which the maximum output power can be obtained).
This control algorithm requires measurements of the voltage VOV (see Fig. 8). Here again it is necessary to
introduce a static switch into the PV array; for the OV method, the switch must be connected in series to open
the circuit. When IPV=0 no power is supplied by the PV system and consequently the total energy generated by
the PV system is reduced. Also in this method measurement of the voltage VPV is required for the PI regulator.

3.3 Perturb and Observe Methods

The P&O algorithms operate by periodically perturbing (i.e. incrementing or decrementing) the array
terminal voltage or current and comparing the PV output power with that of the previous perturbation cycle. If
the PV array operating voltage changes and power increases (dP/dVPV>0), the control system moves the PV array
operating point in that direction; otherwise the operating point is moved in the opposite direction. In the next
perturbation cycle the algorithm continues in the same way.
A common problem in P&O algorithms is that the array terminal voltage is perturbed every MPPT
cycle; therefore when the MPP is reached, the output power oscillates around the maximum, resulting in power
loss in the PV system. This is especially true in constant or slowly-varying atmospheric conditions.
Furthermore, P&O methods can fail under rapidly changing atmospheric conditions (see Fig. 9).
Starting from an operating point A, if atmospheric conditions stay approximately constant, a perturbation V the
voltage V will bring the operating point to B and the perturbation will be reversed due to a decrease in power.
However, if the irradiance increases and shifts the power curve from P1 to P2 within one sampling period, the
operating point will move from A to C. This represents an increase in power and the perturbation is kept the
same. Consequently, the operating point diverges from the MPP and will keep diverging if the irradiance
steadily increases.

There are many different P&O methods available in the literature. In this paper we consider the classic, the
optimized and the three-points weight comparison algorithms.
In the classic P&O technique (P&Oa), the perturbations of the PV operating point have a fixed magnitude. In our
analysis, the magnitude of perturbation is 0.37% of the PV array VOV (around 2V)

www.ijres.org 93 | Page
 Energy Comparison of MPPT techniques for PV Systems

In the optimized P&O technique (P&Ob), an average of several samples of the array power is used to
dynamically adjust the perturbation magnitude of the PV operating point.
In the three-point weight comparison method (P&Oc), the perturbation direction is decided by comparing the PV
output power on three points of the P-V curve. These three points are the current operation point (A), a point B
perturbed from point A, and a point C doubly perturbed in the opposite direction from point B.
All three algorithms require two measurements: a measurement of the voltage VPV and a measurement of the
current IPV (see Fig. 10).

IV. COSTS COMPARISON

To complete our analysis a simple discussion about the cost of the MPPT technique is presented [20]. A
satisfactory MPPT costs comparison can be carried out by knowing the technique (analogical or digital) adopted
in the control device, the number of sensors, and the use of additional power component, considering the other
costs (power components, electronic components, boards, etc…) equal for all the devices.
The MPPT implementation typology greatly depends on the end-users’ knowledge, with analogical
circuit, SC, OV, or CV are good options, otherwise with digital circuit that require the use of microcontroller,
P&O, IC, and temperature methods are enough easily to implement. Moreover it is important to underline that
analogical implementations are generally cheaper than digital (the microcontroller and relative program are
expensive). To make all the cost comparable between them, the computation cost comparison is formulated
taking into account the present spread of MPPT methods.
The number of sensors required to implement the MPPT technique also affects the final costs. Most of
the time, it is easier and more reliable to measure voltage than current and the current sensors are usually more
expensive and bulky. The irradiance or temperature sensors are very expensive and uncommon.

V. CONCLUSION
This paper has presented a comparison among ten different Maximum Power Point Tracking techniques
in relation to their performance and implementation costs. In particular, fourteen different types of solar
insulation are considered, and the energy supplied by a complete PV array is calculated; furthermore, regarding
the MPPT implementation costs, a cost comparison is proposed taking into consideration the costs of sensors,
microcontroller and additional power components.
A ranking of the ten methods has been proposed. Taking into account the analysis results along with
hardware and computational costs, the P&Ob and ICa methods receive the best rankings.

REFERENCES:
[1]. S. Leva, D. Zaninelli, Technical and Financial Analysis for Hybrid Photovoltaic Power Generation Systems, WSEAS Transactions
on Power Systems, vol.5, no.1, May 2006, pp.831- 838
[2]. S. Leva, D. Zaninelli, R. Contino, Integrated renewable sources for supplying remote power systems, WSEAS Transactions on Power
Systems, vol.2, no.2, February 2007, pp.41-48
[3]. J.Schaefer, Review of Photovoltaic Power Plant Performance and Economics, IEEE Trans. Energy Convers., vol. EC-5,pp. 232-
238, June, 1990.
[4]. N.Femia, D.Granozio, G.Petrone, G.Spaguuolo, M.Vitelli, Optimized One-Cycle Control in Photovoltaic Grid Connected
Applications, IEEE Trans. Aerosp. Electron. Syst., vol. 2, no 3, July 2006.
[5]. W. Wu, N. Pongratananukul, W. Qiu, K. Rustom, T. Kasparis and I. Batarseh, DSP- based Multiple Peack Power Tracking for
Expandable Power System, Proc. APEC, 2003, pp. 525-530.
[6]. C. Hua and C. Shen, Comparative Study of Peak Power Tracking Techniques for Solar Storage System, Proc. APEC, 1998, pp. 679-
685.
[7]. D.P.Hohm and M.E.Ropp, Comparative Study of Maximum Power Point Tracking Algorithms Using an Experimental,
Programmable, Maximum Power Point Tracking Test Bed, Proc. Photovoltaic Specialist Conference, 2000, pp. 1699-1702.
[8]. K.H.Hussein, I.Muta, T.Hoshino and M.osakada Maximum Power Point Tracking: an Algorithm for Rapidly Chancing Atmospheric
Conditions, IEE Proc.-Gener. Transm. Distrib., vol. 142, no.1, pp. 59-64, January, 1995.
[9]. X.Sun, W.Wu, Xin Li and Q.Zhao, A Research on Photovoltaic Energy Controlling System with Maximum Power Point Tracking,
Power Conversion Conference, 2002, pp. 822-826.
[10]. T.L. Kottas, Y.S.Boutalis and A. D. Karlis, New Maximum Power Point Tracker for PV Arrays Using Fuzzy Controller in Close
Cooperation with Fuzzy Cognitive Network, IEEE Trans. Energy Conv., vol.21, no.3, 2006.
[11]. A. El Jouni, R. El-Bachtiri and J. Boumhidi, Sliding Mode Controller for the Maximum Power Point Tracking of a Photovoltaic
Pumping System, WSEAS Transactions on Power Systems, vol.1, no.10, pp. 1675-1680, 2006.
[12]. Y.T.Hsiao and C.H.Chen, Maximum Power Tracking for Photovoltaic Power System, Proc. Industry Application Conference, 2002,
pp. 1035-1040.

www.ijres.org 94 | Page
This article was downloaded by: [University of Prince Edward Island]
On: 21 November 2014, At: 01:22
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of International Council on Electrical

Engineering
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tjee20

Simulation and Analysis of Existing MPPT Control

Methods in a PV Generation System
a a b b
Seok-II Go , Seon-Ju Ahn , Joon-Ho Choi, Won-Wook Jung , Sang-Yun Yun & II-Keun
b
Song
a
Dept. of Electrical Engineering, Chonnam National University, Korea (, )
b
Transmission & Distribution Lab., KEPCO Research Institute, Korea (, , )
Published online: 10 Sep 2014.

To cite this article: Seok-II Go, Seon-Ju Ahn, Joon-Ho Choi, Won-Wook Jung, Sang-Yun Yun & II-Keun Song (2011)
Simulation and Analysis of Existing MPPT Control Methods in a PV Generation System, Journal of International Council on
Electrical Engineering, 1:4, 446-451

To link to this article: http://dx.doi.org/10.5370/JICEE.2011.1.4.446

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. Taylor & Francis, our agents, and our licensors make no representations
or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content.
Versions of published Taylor & Francis and Routledge Open articles and Taylor & Francis and Routledge
Open Select articles posted to institutional or subject repositories or any other third-party website are
without warranty from Taylor & Francis of any kind, either expressed or implied, including, but not limited to,
warranties of merchantability, fitness for a particular purpose, or non-infringement. Any opinions and views
expressed in this article are the opinions and views of the authors, and are not the views of or endorsed by
Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified
with primary sources of information. Taylor & Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused
arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-conditions

It is essential that you check the license status of any given Open and Open Select article to
confirm conditions of access and use.
 446 Journal of International Council on Electrical Engineering Vol. 1, No. 4, pp. 446~451, 2011

Simulation and Analysis of Existing MPPT Control Methods

in a PV Generation System

Seok-Il Go*, Seon-Ju Ahn*, Joon-Ho Choi†, Won-Wook Jung**, Sang-Yun Yun**
and Il-Keun Song**

Abstract – Maximum Power Point Tracking (MPPPT) techniques are employed in PV systems to make
full utilization of PV array output power which depends on solar irradiation and temperature. Recently,
many MPPT algorithms and control schemes of PV system have been proposed, but different control
methods and different PID controller parameters lead to different output efficiency and dynamic response.
Therefore, in this paper, a PSCAD/EMTDC user defined module of the solar array is established to easily
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

simulate its inherent characteristics with the basic specification data. Moreover, we also compare the
advantages and weakness of the existing MPPT control methods by simulations to give a reference to the
Grid-Connected Photovoltaic System (GCPS) designers.

Keywords: Photovoltaic (PV) System, Maximum Power Point Tracking (MPPT) techniques,
PSCAD/EMTDC, Grid-Connected Photovoltaic System (GCPS)

1. Introduction established to easily simulate its inherent characteristics

with the basic specification data. We also compare and
As the conventional energy sources are rapidly depleted, analyze the advantages and weaknesses of the existing
the importance of solar Photovoltaic (PV) energy has been MPPT control methods through time-domain simulation
emerging as an alternative energy resource. Since it is clean, studies in the PSCAD/EMTDC software environment.
pollution-free, and inexhaustible, researches on the PV
power generation system have received much attention,
particularly, on many terrestrial applications. [1]. Maximum 2. PV Array Characteristics and Modeling
Power Point Tracking (MPPT) is one of the most important
and well-known problem for all PV systems. An efficient The basic working principle of the PV array is the
MPPT algorithm is very important to increase the efficiency photovoltaic effect of semiconductor PN junction. The
of PV system. complex physics of a PV module can be represented by the
Recently, many MPPT algorithms and control schemes of equivalent electrical circuit shown in Fig.1.
PV generation system have been proposed [2-8]. As for the
non-Artificial Intelligence method of the MPPT algorithms,
Constant Voltage Tracking (CVT) method, Fraction Open-
Circuit Voltage (OCV) method, Hill climbing and Perturb
and Observe (P&O) method, Incremental Conductance
(INC-CON) method, Variable Step Size (VSS) method
based on INC-CON method, and hybrid method which
combined above algorithms are introduced.
In this paper, an Electro-Magnetic Transient Program
(EMTP) user defined module of the solar array is

† Corresponding Author: Dept. of Elec. Eng., Chonnam National Fig. 1. PV module equivalent electrical circuit.
University, Korea ([email protected])
* Dept. of Electrical Engineering, Chonnam National University,
In Fig.1, ISC is the short circuit current of a moduleunder
Korea ([email protected], [email protected])
a given solar irradiance and temperature. ID is the diode
** Transmission & Distribution Lab., KEPCO Research Institute,
Korea ({wwjung, drk9034, songilk}@kepco.co.kr) current, which can be given by the classical diode current
Received: August 9, 2011; Accepted: September 28, 2011 expression. The series resistance RS represents the intrinsic
 Seok-Il Go, Seon-Ju Ahn, Joon-Ho Choi, Won-Wook Jung, Sang-Yun Yun and Il-Keun Song 447

resistance to the current flow, and RSH represents the Table 1. Parameters for PSCAD/EMTDC PV Module
equivalent parallel resistance. Since the value of RSH is Parameter Label Value
usually very large, it can be ignored in the simulation P-N junction ideal constant n 1.5
analysis. Constant Coulomb constant q 1.6e-19C
Boltzmann constant k 1.38e-23J/K
The load current is therefore given by the expression:
Reference temperature Tref 25 oC
Open-circuit voltage VOC 21.7V
I  I SC  I D  I SH PV Short-circuit current ISC 3.35A
(1) module Maximum power Pmax 53W
  V  IRS   V  IRS
= I SC  I o exp    1  Maximum power voltage Vmax 17.4V
  nVT   RSH Maximum power current Imax 3.05A
Number of cell in a module m 36
PV array Number of modules in parallel NP 20
where I is the output terminal current, Io is the diode Number of modules in series NS 50
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

saturation current [A], and V is the terminal voltage of a

module [V]. n is the ideal constant of diode, VT is the
thermal potential of a module, and its value is m×(kT/q) of VMPP/VOC depends on the solar cell parameters, but a
where, k is the Boltzmann constant (1.38e-23[J/K]), T is the commonly used value is 76%, so the CVT method is that
cell temperature [K], q is the magnitude of the electrical the reference MPP voltage is set to 0.76 Voc constantly for
charge on the electron (1.6e-19[C]), and m is the number of tracking the maximum power. The advantages of the CVT
cells in series in a module. method are in its simplicity and good starting
The characteristics of a PV module are nonlinear and characteristics. However, the applicability of the system is
each curve has only one Maximum Power Point (MPP). poor, when the external environment has been changing.
Moreover, the output current of a PV module is mainly Due to these characteristics, the CVT method is often
affected by irradiation variation, whereas the output voltage combined with other MPPT techniques.
of PV module is mainly affected by temperature variation.
Since the range of output power variation according to the 3.2 P&O Method
temperature is very small, the irradiation is the main factor
of output power variation of a PV array [9]. The Perturb and Observe (P&O) method is one of the
PSCAD/EMTDC (Power System Computer Aided most commonly used methods in practice. The P&O
Design /Electro Magnetic Transients including DC System) algorithms operate by periodically perturbing, i.e.
is an industry standard simulation tool for studying the incrementing or decrementing, the array terminal voltage
transient behavior of electrical apparatus and networks. Its and comparing the PV output power with that of the
graphic-based user interface enables users to graphically previous perturbation cycle. If the PV array operating
assemble the circuit, run the simulation, analyze the results, voltage changes and power increases, the control system
and manage the data in a completely integrated graphical moves the PV array operating point in that direction.
environment [10]. Otherwise the operating point is moved in the opposite
The PV array model is described and implemented by direction.
using Fortran Language in a User Defined Module (UDM) The logic of this algorithm and the flowchart are
with the reference to PSCAD. The main required explained in Fig. 2. The operating voltage of the PV system
parameters and used data of the PV module and PV array is perturbed by a small increment of V, and this resulting
are listed in Table 1. change in P. If P is positive, the perturbation of the
operating voltage needs to be in the same direction of the
increment. On the contrary, if P is negative, the obtained
3. MPPT Techniques system operating point moves away from the MPPT and the
operating voltage needs to move in the opposite direction of
3.1 Constant Voltage Tracking Method the increment.
A common shortcoming of this method is that the array
The constant voltage tracking (CVT) method makes use terminal voltage is perturbed every MPPT cycle. Therefore,
of the fact that when the temperature is constant, the MPP is when the MPP is reached, the output power oscillates
mostly distributed at a fixed voltage value by analyzing the around the maximum, resulting in a power loss in the PV
P-V curve of the PV array at different irradiations. The ratio system. Furthermore, it sometimes fails to find the MPP
 448 Simulation and Analysis of Existing MPPT Control Methods in a PV Generation System

under the continuously increasing or decreasing irradiation

conditions. START

Sample V(n), I(n)

I  I (n)  I ( n  1)
V  V ( n )  V ( n  1)

V  0
YES
NO

I V   I V I  0
YES YES
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

I V   I V I  0
YES NO NO YES

Vref=Vref+dV Vref=Vref-dV Vref=Vref-dV Vref=Vref+dV

Return

Fig. 2. P&O method flowchart. Fig. 3. INC-CON method flowchart.

conductance ( Δ I/ Δ V). Vref is the reference voltage at
3.3 INC-CON Method
which the PV array is forced to operate. Once the MPP is
The Incremental Conductance (INC-CON) algorithm is reached, the operation of the PV array is maintained at this
based on the observation that the following equations (2) point unless a change in ΔI is noted, indicating a change
and (3) hold for the MPP. The principle of this method is to in atmospheric conditions and the MPP. The flowchart of
judge whether the system work at MPP or work at the left this method is as shown in Fig. 3.
or the right. The main advantages of the INC-CON algorithm are the
fine control the system obtains, and its high stability under
dP d ( IV ) dI I rapidly changing atmospheric conditions. However, its
  I V  I V 0 (2) deficiency is that the tracking steps of the system are
dV dV dV V
constant, so that the tracking speed and the accuracy
I I
  (3) contradict with each other. It also requires a complex
V V control circuit.

Therefore, by analyzing the derivative, one can test 3.4 Variable Step Size (VSS) Method
whether the PV generator is operating at its MPP or far
from it using equations (4) - (6). All the above conventional tracking methods use fixed,
small iteration steps, which is determined by the accuracy
dP and tracking speed requirements. If the step size is
 0 for V  VMPP (4)
increased to speed up the tracking, the accuracy of tracking
dV
dP suffers and vice versa. To overcome the above limitation, a
 0 for V  VMPP (5) variable step size method has been proposed in [11]. It
dV
offers the fast response time and good steady-state
dP
 0 for V  VMPP (6) performance.
dV In this method, the large step is adopted to move the PV
system to the MPP quickly when the irradiation changes
The MPP can thus be tracked by comparing the abruptly. The small one is adopted to decrease the power
instantaneous conductance (I/V) to the incremental oscillation when the PV module works around the MPP.
 Seok-Il Go, Seon-Ju Ahn, Joon-Ho Choi, Won-Wook Jung, Sang-Yun Yun and Il-Keun Song 449

Table 2. Simulation case descriptions

START Temperature Irradiation Time
Case Statement
(oC) (W/m2) (s)
1 MPPT control start 25 1000 5
2 External condition change 20 800 8
Detect Voc 3 External condition change 10 600 11
Sample V(n)

Vset=0.78Voc
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

NO Variable step size

V(n)>Vset INC-CON/P&O
method

YES

Decrease output
Voltage

Fig. 4. CVT+INC-CON (P&O) + VSS method flowchart. (a)

The step size can be calculated by using equations (7) and (8).

Vstep  e  V  e  dV (7)
P ( n)  P (n  1) P dP
e   (8)
V ( n)  V (n  1) V dV

3.5 CVT + INC-CON (P&O) + VSS Method

The INC-CON method with variable step-size has good

performance in tracking but makes the starting process (b)
more complex. The CVT method shows a better
performance in the starting process. The control algorithm
is simple, as it only needs to judge whether the output
voltage of the PV array is bigger than the voltage
instruction. The voltage is changed in only one direction,
leading to an increasing power in one direction without
oscillation. The control flowchart is as shown in Fig. 4.

4. Simulation Results

The MPPT control methods introduced in section 3 has

been tested by the PSCAD/EMTDC simulation. The
simulation scenario is as summarized in Table 2, and the (c)
simulation results are shown in Fig. 5.
From the simulation results, we can observe that the Fig. 5. Simulation results: (a) case 1, (b) case 2, (c) case3.
 450 Simulation and Analysis of Existing MPPT Control Methods in a PV Generation System

P&O and INC-CON algorithms have very similar 1991.

performances and dynamic response times, and they are [5] M. A. S. Masoum, H. Dehbonei, and E. F. Fuchs,
superior to the CVT methods. However, both methods have “Theoretical and experimental analyses of photovol-
shortcoming of steady-state oscillation due to perturbations. taic systems with voltage and current-based maximum
The VSS method shows improved dynamic response time power-point tracking”, IEEE Trans. Energy Convers.,
and increased total power output compared to the P&O and vol. 17, no. 4, pp. 514–522, Dec. 2002.
INC-CON methods. However, it still takes relatively large [6] T. Esram and P. L. Chapman, “Comparison of
amount of time to get to the stable state when the MPPT photovoltaic array maximum power point tracking
first starts at 5 sec. The modified VSS method, which uses techniques”, IEEE Trans. Energy Convers., vol. 22, no.
the CVT method during the start process, shows the best 2, pp. 439–449, Jun. 2007.
performance in both the response time and the total power [7] K. H. Hussein, I. Muta, T. Hoshino, and M. Osakada,
output. “Maximum photovoltaic power tracking: an algorithm
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

for rapidly changing atmospheric conditions,” IEE

Proc. Generation, Transmission and Distribution, vol.
5. Conclusion 142, pp. 59-64, 1995.
[8] Desai and Patel, H.P., and Patel, H.K. “Maximum
In this paper, an electro-magnetic transient model of two power point algorithm in PV generation an overview”,
stage grid connected PV systems was presented by using in Proceedings of power electronics and drive system,
the UDM models of the PV array and the MPPT controller 2007.
using PSCAD. Various MPPT techniques have been [9] C. Zhang, D. Zhao, J. Wang, and G. Chen, “A modifi-
reviewed and simulated with different temperatures and ed MPPT method with variable perturbation step for
irradiation environments. From the simulation results, the photovoltaic system”, in Proceedings of Power
validity of the developed modeling has been confirmed. Electronics and Motion Control Conference 2009, pp.
Moreover, the advantages and weaknesses of the existing 2096-2099, 2009.
MPPT control methods were also analyzed by the [10] PSCAD/EMTDC User’s Guide, Manitoba HVDC Re-
simulation results. search Center, Jan. 2003.
The summary of MPPT techniques can be beneficial in [11] F. Liu, S. Duan, B. Liu and Y. Kang, “A variable step
understanding the existing MPPT control algorithms for size INC MPPT method for PV systems”, IEEE Trans.
engineers of PV industry. The discussion and performance Industrial Electronics, vol. 55, no.7, pp. 2622–2628,
summary of different MPPT methods would serve as a July 2008.
useful guide in choosing the right MPPT method for
specific PV systems in the practical applications.
Seok-Il Go received B.S. degree in
Electrical Engineering from Chonnam
References National University, Gwangju, Korea in
2010. His research interests are dis-
[1] G.J. Yu, Y.S. Jung, J.Y. Choi, I. Choy, J.H. Song and tributed generations, distribution net-
G.S. Kim, “A Novel Two-Mode MPPT Control works, Micro-Grid, and Smart-Grid.
Algorithm based on Comparative Study of Existing
Algorithms”, in Proceedings of Photovoltaic Specia-
lists Conference, 2002, pp. 1531-1534.
[2] R. Faranda, S. Leva, V. Maugeri, “MPPT techniques for Seon-Ju Ahn received the B.S., M.S.
PV systems: energetic and cost comparison”, in and Ph.D. degrees in Electrical
Proceedings of IEEE PES GM2008, July 2008. Engineering from Seoul National
[3] J.Schaefer, “Review of Photovoltaic Power Plant University, Seoul, Korea in 2002, 2004
Performance and Economics”, IEEE Trans. Energy and 2009. Currently, He is a Full-time
Convers, vol. EC-5, pp. 232-238, June, 1990. lecturer at Chonnam National Univer-
[4] Z. Salameh, F. Dagher, W. Lynch, “Step-down sity, Korea. His research interests are
maximum power point tracker for photovoltaic power quality, distributed energy resources, Micro-Grid,
systems”, Solar Energy, vol. 46, no 5, pp. 279–282, Smart-Grid, and real-time simulation.
 Seok-Il Go, Seon-Ju Ahn, Joon-Ho Choi, Won-Wook Jung, Sang-Yun Yun and Il-Keun Song 451

Joon-Ho Choi received the B.S., M.S.

and Ph.D. degrees in Electrical Eng-
ineering from Soongsil University,
Seoul, Korea in 1996, 1998 and 2002.
Currently, He is an Associate Professor
at Chonnam National University, Korea.
His research interests are distributed
generations, distribution networks, stability, and Smart-Grid.

Won-Wook Jung received his B.S.

degrees in Electrical Engineering from
Downloaded by [University of Prince Edward Island] at 01:22 21 November 2014

Chungnam National University, Daejeon,

Korea, in 2003, and the M.S. degree in
Electrical engineering from Seoul
National University, Seoul, Korea, in
2005. His research interests are dis-
tributed generation, distribution networks, and Smart-Grid.

Sang-Yun Yun received the B.S., M.

S. and Ph.D. degrees in Electrical E
ngineering from Soongsil University,
Seoul, Korea in 1996, 1998 and 20
02. His research interests are smart
distribution design and network secu
rity analysis.

Il-Keun Song received the B.S., M.S.

and Ph.D. degrees in Electrical Eng-
ineering from Soongsil University,
Seoul, Korea in 1984, 1986 and 1997.
His research interests are smart dis-
tribution management system and life
time estimation of power equipments.
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012 3627

An Improved Particle Swarm Optimization

(PSO)–Based MPPT for PV With Reduced
Steady-State Oscillation
Kashif Ishaque, Zainal Salam, Member, IEEE, Muhammad Amjad, and Saad Mekhilef, Member, IEEE

Abstract—This paper proposes an improved maximum power financial schemes, for example, the feed-in tariff [7] and sub-
point tracking (MPPT) method for the photovoltaic (PV) system sidized policies [8], have been introduced by various countries,
using a modified particle swarm optimization (PSO) algorithm. resulting in rapid growth of the industry. To optimize the uti-
The main advantage of the method is the reduction of the steady-
state oscillation (to practically zero) once the maximum power lization of large arrays of PV modules, maximum power point
point (MPP) is located. Furthermore, the proposed method has the tracker (MPPT) is normally employed in conjunction with the
ability to track the MPP for the extreme environmental condition, power converter (dc–dc converter and/or inverter). The objec-
e.g., large fluctuations of insolation and partial shading condition. tive of MPPT is to ensure that the system can always harvest the
The algorithm is simple and can be computed very rapidly; thus, maximum power generated by the PV arrays. However, due to
its implementation using a low-cost microcontroller is possible. To
evaluate the effectiveness of the proposed method, MATLAB sim- the varying environmental condition, namely temperature and
ulations are carried out under very challenging conditions, namely solar insolation, the P –V characteristic curve exhibits a max-
step changes in irradiance, step changes in load, and partial shad- imum power point (MPP) that varies nonlinearly with these
ing of the PV array. Its performance is compared with the con- conditions—thus posing a challenge for the tracking algorithm.
ventional Hill Climbing (HC) method. Finally, an experimental rig To date, various MPP tracking methods have been proposed
that comprises of a buck–boost converter fed by a custom-designed
solar array simulator is set up to emulate the simulation. The soft- [9]. These techniques vary in complexity, accuracy, and speed.
ware development is carried out in the Dspace 1104 environment Each method can be categorized based on the type of the control
using a TMS320F240 digital signal processor. The superiority of variable it uses: 1) voltage, 2) current, or 3) duty cycle. For the
the proposed method over the HC in terms of tracking speed and voltage- and current-based techniques, two approaches are used.
steady-state oscillations is highlighted by simulation and experi- The first one is the observation of MPP voltage VM P or current
mental results.
IM P with respect to the open circuit voltage VOC [10] and short
Index Terms—Buck–boost converter, Hill Climbing (HC), max- circuit current ISC , respectively [11]. Since this method ap-
imum power point tracking (MPPT), partial shading, particle proximates a constant ratio, its accuracy cannot be guaranteed.
swarm optimization (PSO), photovoltaic (PV) system.
Consequently, the tracked power would most likely be below the
real MPP, resulting in significant power loss [12]. The second
I. INTRODUCTION approach is to obtain the information on the actual operating
point of the PV array (i.e., voltage and current) and these points
OLAR photovoltaic (PV) is envisaged to be a popular
S source of renewable energy due to several advantages, no-
tably low operational cost, almost maintenance free and envi-
are updated according to the variation in environmental condi-
tions. The most popular technique is the perturb and observe
(P&O) method. It is based on the perturbation of voltage (or
ronmentally friendly. Despite the high cost of solar modules, current) using the present P and previous Pold operating power,
PV power generation systems, in particular the grid-connected respectively. If P is improved, the direction of perturbation is
type, have been commercialized in many countries because of retained; otherwise, the direction is reversed accordingly.
its potential long-term benefits [1]–[6]. Furthermore, generous Despite the simplicity of the algorithm, the performance of
P&O method is heavily dependent on the tradeoff between
the tracking speed and the oscillations that occurs around the
MPP [13]. A small perturbation reduces the oscillations but at
Manuscript received September 14, 2011; revised September 21, 2011;
accepted January 8, 2012. Date of current version April 20, 2012. Recom- the expense of tracking speed, or vice versa. Another major
mended for publication by Associate Editor M. Liserre. drawback of P&O is that during rapid fluctuations of insolation,
K. Ishaque was with the Universiti Teknologi Malaysia, Johor Bahru the algorithm is very likely to lose its direction while tracking
81310, Malaysia. He is now with the Department of Electronics Engineer-
ing, Karachi Institute of Economics and Technology, Karachi 75190, Pakistan the true MPP. Several improvements are proposed to address this
(e-mail: [email protected]). issue—mainly by considering adaptive perturbation. However,
Z. Salam (corresponding author) and M. Amjad are with the Uni- these techniques are not fully adaptive and hence are not very
versiti Teknologi Malaysia, Johor Bahru 81310, Malaysia (e-mail:
[email protected]; [email protected]). effective [14]. Moreover, under special condition such as partial
S. Mekhilef is with the University of Malaya, Kuala Lumpur 50603, Malaysia shading and modules irregularities, these methods often fail to
(e-mail: [email protected]). track the true MPP because the PV curves are characterized by
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. multiple peaks (several local and one global). Since the P&O
Digital Object Identifier 10.1109/TPEL.2012.2185713 algorithm could not distinguish the correct peak, its usefulness

0885-8993/$31.00 © 2012 IEEE

3628 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

under such conditions diminishes rapidly [12]. Another method system. Since the method specifically focused on the grid side,
is the incremental conductance (IC) which is based on incre- it is of little interest because the concern of the majority of the
mentally comparing the ratio of derivative of conductance with MPPT work is on the input side.
the instantaneous conductance [15]. Although IC does not suffer Interestingly, one important feature of the PSO which is ig-
the lost of tracking direction, it inherits the same problems as nored by researchers [23]–[26] is the fact that once the particles
P&O, namely the inevitable tradeoff between the MPPT speed find the MPP, their associated velocity becomes very low or
and oscillation. Several improved IC techniques are proposed; practically null. This advantage is overlooked because the pre-
for instance, the authors in [4] managed to improve the oscil- vious work has not attempted to use PSO in conjunction with the
lations at MPP but during rapid fluctuation of environmental direct duty cycle control. If the two are combined, at MPP, the
conditions, the tracking speed reduces significantly. duty cycle can be maintained at constant value and this in turn
The third category of MPPT techniques is characterized by diminishes the steady-state oscillations that typically exist in
the duty cycle control. In the literature, it is known as Hill conventional MPPT techniques. Consequently, power loss due
Climbing (HC) or direct control method. As the latter name to the oscillation is eliminated and system efficiency increases.
implies, HC operates by directly updating the converter’s duty To capitalize that particular advantage, this paper proposes an
cycle [16]. The approach greatly simplifies the control structure improved MPPT method based on a modified PSO algorithm.
as it eliminates the need for the proportional integral (PI) or hys- It will be shown that the proposed MPPT has a similar structure
teresis controller. In principle, it works on the same concept as to the conventional HC; therefore, a direct duty cycle control
P&O, but instead of perturbing the voltage or current, it updates can be utilized. The main feature of the proposed method is the
the operating point of the PV array by perturbing the duty cy- absence of steady-state oscillation at MPP. It also has the ability
cle. With the absence of the PI loop, the implementation of HC to track the MPP for the extreme environmental condition, e.g.,
is greatly simplified. Consequently, this method is extensively large fluctuations of insolation and partial shading condition.
used in PV systems [17]. However, it also suffers with same Compared to other conventional MMPT techniques, it has a
drawbacks inherited by P&O. faster tracking speed. Furthermore, the algorithm is simple and
In an effort to overcome aforementioned disadvantages, sev- can be computed very rapidly; thus, its implementation using a
eral pieces of research have used artificial intelligence approach low-cost controller is possible.
such as fuzzy logic controller (FLC) [18] and neural network The remainder of this paper is organized as follows.
(NN) [19]. Although these methods are effective in dealing with Section II discusses the modeling of the PV module and array
the nonlinear characteristics of the I–V curves, they require ex- based on the work published in [27]. This would be the basis
tensive computation. For example, FLC has to deal with fuzzi- for the simulation work that ensues. In Section III, the conven-
fication, rule base storage, inference mechanism, and defuzzi- tional HC method is briefly introduced. Section IV describes
fication operations. For NN, the large amount of data required the overview of the PSO and how it is applied to track the MPP.
for training are a major source of constraint. Furthermore, as In addition, the modified PSO algorithm used to improve the
the operating conditions of the PV system vary continuously, tracking performance is also outlined. Section V describes the
MPPT has to respond to changes in insolation and temperature overview of the overall simulation and experimental set-up. For
variations in real time. Clearly, a low-cost processor cannot be simulation, a comprehensive PV simulator developed in [28] is
employed in such a system. utilized. To validate the idea, the algorithm is implemented us-
An alternative approach is to employ evolutionary algorithm ing a prototype buck–boost converter fed by a custom-designed
(EA) techniques. Due to its ability to handle nonlinear objective solar array simulator. The software development is carried out
functions [20], [21], EA is envisaged to be very effective to deal in the DSpace 1104 environment with a TMS320F240 digital
with the MPPT problem. Among the EA techniques, particle signal processor. Section VI compares the results obtained using
swarm optimization (PSO) is highly potential due to its simple proposed PSO and HC methods. Finally, the conclusion is made
structure, easy implementation, and fast computation capabil- in the last section.
ity [22]. Since PSO is based on search optimization, in princi-
ple, it should be able to locate the MPP for any type of P−V II. MODELING OF THE PV MODULE AND ARRAY
curve regardless of environmental variations. Realizing these
advantages, several researchers have employed this technique A. Modeling PV Module
to improve the MPP tracking [23]–[26]. In [23] and [24], the Among various modeling methods of the PV module, the
authors have added various extra coefficients in the conventional two-diode model, as depicted in Fig. 1(a), is known to be
PSO searching scheme equations, thus increasing the computa- the more accurate one. The output current of the module can
tional burden of the algorithm. The authors in [25] formulated be described as
an analytical expression of the objective functions based on PV  
current, solar insolation, and temperature; then, PSO is utilized V + IRs
I = IPV − Id1 − Id2 − (1)
to track the MPP. However, it is not shown that the expression Rp
is equivalent to the real operating power; hence, tracking of the
where
correct MPP is not certain—raising the questions about the ef-    
fectiveness of this method. In [26], the authors employed PSO V + IRs
Id1 = Io1 exp −1 (2)
to optimize the parameters for the LC filter in the microgrid a1 VT 1
ISHAQUE et al.: IMPROVED PARTICLE SWARM OPTIMIZATION (PSO)−BASED MPPT FOR PV WITH REDUCED STEADY-STATE OSCILLATION 3629

Fig. 2. P−V curves for MSX-60 in 4 × 1 configuration for different irradiation

levels.

Fig. 1. (a) Two-diode model of the PV cell. (b) Series parallel combination of
the PV array. TABLE I
PARAMETERS OF THE MSX-60 PV MODULE AT STC: TEMPERATURE = 25 ◦ C,
AIR MASS = 1.5, AND INSOLATION = 1000 W/m2

and
   
V + IRs
Id2 = Io2 exp −1 (3)
a2 VT 2
where IPV is the current generated by the incidence of light;
and Io 1 and Io 2 are the reverse saturation currents of diode 1
and diode 2, respectively. The Io 2 term is introduced to com-
pensate for the recombination loss in the depletion region as
described in [29]. Other variables are defined as follows: VT 1
and VT 2 (both equal to Ns kT/q) are the thermal voltages of the
PV module having Ns cells connected in series, q is the elec- B. Modeling of the PV Array
tron charge (1.60217646 × 10−19 C), k is the Boltzmann con- In a typical installation of a large PV power generation sys-
stant (1.3806503 × 10−23 J/K), and T is the temperature of the tem, the modules are configured in a series–parallel structure
p-n junction in Kelvin. Variables a1 and a2 represent the diode (i.e., Nss × Npp modules), as depicted in Fig. 1(b). To handle
ideality constants, respectively. Although greater accuracy can such cases, the output current equation in (4) has to be modified
be achieved using this model than the single-diode model, it as follows:
requires the computation of seven parameters, namely IPV , Io 1 ,  
Io 2 , Rp , Rs , a1 , and a2 . V + IRs Γ
I = Npp {IPV − Io (Ip + 2)} − (7)
Recently, a fast and simple two-diode model has been pro- Rp Γ
posed in [27]. In this model, the simplified current equation is
developed and is given as where
     
V + IRs V + IRs Γ V + IRs Γ
I = IPV − Io (Ip + 2) − (4) Ip = exp + exp (8)
Rp VT Nss (p − 1)VT Nss

where and
   
V + IRS V + IRs Nss
Ip = exp + exp (5) Γ= (9)
VT (p − 1)VT Npp
and
where IPV , I0 , Rp , Rs , p are the parameters of the individual
p = 1 + a2 . (6) module. Fig. 2 shows the P−V curves for a commercial PV
module (MSX-60) configured in a 4 × 1 PV array. The param-
The model only requires five parameters to be computed with eters of this particular module under the standard test condition
no loss of accuracy. (STC) are shown in Table I.
3630 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

Fig. 4. Movement of particles in the optimization process.

using
Fig. 3. Flowchart of the conventional HC method.
xki +1 = xki + Φki +1 (11)

III. CONVENTIONAL HC METHOD where the velocity component Φi represents the step size. The
velocity is calculated by
To obtain the maximum power from the PV modules, MPPT
   
is normally employed. Over the years, various MPPT methods Φki +1 = wΦki + c1 r1 Pb esti − xki + c2 r2 Gb est − xki
are proposed; for example, P&O, IC, HC, NN, and FLC [4], [9], (12)
[12], [15]–[19], [22]. In particular, the conventional HC method where w is the inertia weight, c1 and c2 are the acceleration
is interesting as the duty cycle of the power converter can be coefficients, r1 , r2 ∈ U (0, 1), Pb esti is the personal best position
varied directly [16]. This can be explained with the help of a of particle i, and Gb est is the best position of the particles in
flowchart as shown in Fig. 3. The algorithm periodically updates the entire population. Fig. 4 shows the typical movement of
the duty cycle d(k) by a fixed step size Φ with the direction of in- particles in the optimization process.
creasing power. The perturbation direction is reversed if P (k) < If position is defined as the actual duty cycle while velocity
P(k − 1), an indication that the tracking is not moving toward shows the perturbation in the present duty cycle, then (11) can
the MPP. This can be described by the following equation: be rewritten as

dold + Φ if P > Pold dki +1 = dki + Φki +1 . (13)
dnew = (10)
dold − Φ if P < Pold . From (10) and (13), it can be seen that both HC and PSO
algorithms have an equivalent structure. However, for the case
A clear advantage of this algorithm is that the MPPT algorithm
of PSO, resulting perturbation in the present duty cycle depends
does not require proportional (P) or proportional integral (PI)
on Pb esti and Gb est . If the present duty cycle is far from these
action, which is normally employed to control the duty cycle
two duty cycles, the resulting change in the duty cycle will
with reference to voltage or current. In this case, the duty cycle
also be large, and vice versa. Therefore, PSO can be thought
directly feeds the power converter.
of as an adaptive form of HC. In the latter, the perturbation in
the duty cycle is always fixed but in PSO it varies according
IV. PSO-BASED MPPT to the position of the particles. With proper choice of control
parameters, a suitable MPPT controller using PSO can be easily
A. General Overview of PSO
designed.
PSO is a stochastic, population-based EA search method,
modeled after the behavior of bird flocks [30]. The PSO algo- B. Application of PSO for MPPT
rithm maintains a swarm of individuals (called particles), where
each particle represents a candidate solution. Particles follow To illustrate the application of the PSO algorithm in tracking
a simple behavior: emulate the success of neighboring parti- the MPP using the direct control technique, first a solution vector
cles and its own achieved successes. The position of a particle of duty cycles with Np particles is determined, i.e.
is, therefore, influenced by the best particle in a neighborhood
xki = dg = [d1 , d2 , d3 , . . . , dj ]
Pb est as well as the best solution found by all the particles in
the entire population Gb est . The particle position xi is adjusted j = 1, 2, 3, . . . , N p. (14)
ISHAQUE et al.: IMPROVED PARTICLE SWARM OPTIMIZATION (PSO)−BASED MPPT FOR PV WITH REDUCED STEADY-STATE OSCILLATION 3631

Fig. 5. PSO particle movements in searching for the MPP.

The objective function is defined as

P (dki ) > P (dki −1 ). (15)
To start the optimization process, the algorithm transmits three
duty cycles di (i = 1, 2, 3) to the power converter. In Fig. 5,
duty cycles d1 , d2 , and d3 are marked with triangular, circular,
and square points, respectively. These duty cycles served as
the Pb esti in the first iteration. Among these, d2 is the Gb est
that gives the best fitness value (which is the array power),
as illustrated by Fig. 5(a). In the second iteration, the resulting
velocity is only due to the Gb est term. The (Pb esti − d (i)) factor
in (12) is zero. Furthermore, the velocity of Gb est particle (d2 ) is
zero due to the (Gb est − d (2)) factor in (12) is zero. This results
in a zero velocity and accordingly the duty cycle is unchanged.
As a result, this particle will not contribute in the exploration
process. To avoid such situation, a small perturbation in duty
cycle is allowed, as shown in Fig. 5(b), to ensure the change in Fig. 6. MPPT tracking by PSO during partial shading.
fitness value. Fig. 5(c) shows the particles movement in the third
iteration. Due to the fact that all the duty cycles in the previous
same signs. Since all conventional MPPT methods are based
iteration attain a better fitness value, the velocity direction of
on the slope and sign value of dP/dV, the algorithm could not
these particles remains unchanged and subsequently they move
distinguish the local (P 1, P 2, and P 4) and global peaks (P 3)
toward Gb est along the same direction. In the third iteration, all
correctly. It is very likely that the MPPT is being forced to trap
duty cycles (di , i = 1, 2, 3) arrive at MPP with a low value of
into the local peak, resulting in reduced output power and thus
velocity. In the subsequent iteration, due to very low velocity,
greatly deteriorates the efficiency of the PV system.
the value of the duty cycle is approaching a constant. Therefore,
On the other hand, since the PSO method works on the basis
the operating point will be maintained and the oscillation around
of search technique, the global peak can be tracked without any
the MPP diminishes.
difficulty. Fig. 6 depicts the tracking capability of PSO during
partial shading. Similar to the previous P−V curve (see Fig. 5),
C. Tracking During Partial Shading
the proposed method transmits three duty cycles, which serve
When the PV array is operating in a uniform solar insolation, as Pb est particles. It can be seen that the voltage and current
the resulting P−V characteristic curve of the array exhibits a contributing to these initial duty cycles (Pb esti ) are away from
single MPP. However, under partial shading, the P−V curves the global peak (P 3). But in the later phases of iterations, it
are characterized by multiple peaks, i.e., with several local and successfully finds the global peak, P 3.
one global peak as depicted in Fig. 6. In this example, the
I−V curve is characterized by four stairs, while the P−V curve
is characterized by four peaks. The latter are labeled as P 1, D. Modified PSO Structure
P 2, P3, and P 4. It can be observed that the time derivative In the case of slow variation in the solar insolation, a proper
of power dP/dV is zero for the global as well as all the local initialization of duty cycles in PSO is very critical. In this case,
peaks. Furthermore, the slope at its right and left sides has the a change in the duty cycle from the previous one should be
3632 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

can be obtained using

1
dnew = dold − (Pold,M PP − PM PP ) (16)
K1
where dold is the previous Gb est duty cycle and K1 =
ΔPM PP /Δd is the slope of the linear segment.
An attractive advantage of (15) is that with the change in
the environmental condition, the new duty cycles will be varied
according to the change in operating power but its value will
always be very close to the new best duty cycle. Hence, the
reinitialization of duty cycles will not result in unnecessary
exploring of the P−V curve. This allows for the new MPP to be
tracked very rapidly.
It should be noted that the aforementioned analyses are carried
out with decreasing insolation, i.e., from λ = 1 to λ = 0.1. The
reducing insolation will always result in the load line being
Fig. 7. Relationship between the G b e st duty cycle and P M M P . on the left of the maximum PV array voltage VM PP of the I−V
curve. Consequently, a larger change in the operating power will
occur. However, this is not true for the case when the insolation
small to track the MPP. Thus, due to the initialization, when increases from a low to high value, for example, from λ = 0.4
the change in the duty cycle is large, the particles will have to to λ = 1. This is due to the fact that for an increasing insolation,
search a large area of the P−V curve. However, MPP will still the load line will always be at the right side of VM PP . Since the
be tracked at the expense of large fluctuations in the operating difference between VM PP and VOC is not large, a low change
point. Consequently, certain amount of energy will be wasted in operating power is observed. Therefore, if the same value of
during the exploration process. Another serious problem needs K1 is used for this case, dold will not be properly scaled. As a
to be considered is when the change in the insolation is small result, PSO will utilize more iteration to reach at MPP. To avoid
(as discussed earlier) but occurs very rapidly. In this scenario, such problem, different pairs of K1 values are used for each of
the tracking needs to be fast enough to follow the rapid change. the two cases. The value of K1 is selected accordingly using the
However, the large change in the duty cycles does not allow for following relationship:
the duty cycles to follow the new MPP very accurately [22]. On ⎧
⎨ K1 if ΔP > 0
the other hand, a large change in the operating point can also K1 = K 1 (17)
occur due to a large change in insolation, for example, during ⎩ if ΔP < 0
the partial shading condition [31]. In this case, if the change 2
in the duty cycle is small, the convergence toward the MPP where
could be slow. This could be more critical for the case of partial ΔP = (P − Pold ) . (18)
shading. As duty cycles are not allowed to explore a larger area
of the P−V curve, the final MPP could settle at a local instead Note that ΔP > 0 and ΔP < 0 indicate decreasing and increas-
of global peak. ing insolation, respectively.
To address these two issues, the conventional PSO need to be To perturb the new duty cycle which is obtained through
modified. The duty cycles are initialized in two phases. First, (16), d1 and d3 are equally displaced in positive and negative
the previous duty cycles are decreased or increased linearly directions, respectively by a factor of K2 i.e.,
(according to the change in array power) by a factor K1 . It can
dki,new = [d1 − K2 , d2 , d3 + K2 ] for K2 ≥ 0.05. (19)
be noted that once the PSO reaches MPP, all the three duty
cycles are at almost the same value due to the zero velocity. In The value of 0.05 in (19) is selected so that the fluctuations in the
order to search the P−V curve for the new MPP, the second step operating power of the PV array will not be too large. However,
involves the perturbation of two extreme duty cycles (d1 and d3 ) in some special cases, such as partial shading in which there
in positive and negative directions with a constant value of K2 . exist multiple peaks (several local with one global), this number
Fig. 7 proposes a systematic method to estimate value of can be increased based on the operating voltage (in the range of
K1 . In this figure, a relationship between the array maximum 30−85% of VOC voltage of the PV array [12]). This is to allow
power PM PP and the corresponding duty cycle Gb est is plot- the PSO algorithm to explore a wider range of the I−V curve so
ted. Gb est is the corresponding duty cycle which is responsible that global peak could be tracked.
for the dc–dc converter to operate at PM PP . The response is It should also be noted that even if the duty cycle computed
obtained by reducing the solar insolation from λ = 1 to λ = through (16) is nowhere near to the final best duty cycle, but due
0.1with a step size of 0.1. It can also be seen that there exists to the perturbation factor K2 , at least one of di (i = 1, 2, 3) will
a quadratic relationship between PM PP and db est . More impor- still be close to the best duty cycle. Hence, (16) and (19) always
tantly, an approximate linear correlation, shown by the dotted ensure a fast tracking. Fig. 8 shows the complete flowchart of
line in Fig. 7, between the change in array power and duty cycle the proposed method.
ISHAQUE et al.: IMPROVED PARTICLE SWARM OPTIMIZATION (PSO)−BASED MPPT FOR PV WITH REDUCED STEADY-STATE OSCILLATION 3633

Fig. 8. Complete flowchart of the proposed method.

E. Advantages of the Proposed PSO Method V. SIMULATION RESULTS

The proposed method offers several advantages compared to Fig. 9 shows the MATLAB–Simulink simulation model of
other MPPT techniques. the PV system used in this study. The buck–boost dc/dc con-
1) Equation (10) shows that the perturbation in duty cycle is verter is utilized due to several reasons [32], namely 1) it ex-
computed by two difference terms: the difference between hibits superior characteristics with respect to the performance
the previous duty cycle di (k) and the local best particles of PV array’s MPP; and 2) it follows the MPP at all times,
Pb esti , and the difference between the previous duty cycle regardless of the solar insolation, the array temperature, and
di (k) and the global best particle Gb est . Thus, the power the connected load. The converter is designed for continuous
converter tracks the two best Pb esti and Gb est at the same inductor current mode with the following specifications: C1 =
time. As a result, the tracking spaces are searched to obtain 470 μF, C2 = 220 μF, L = 1 mH, and 50-kHz switching fre-
an optimal solution with a faster speed. quency. The PV modules utilized are the MSX60 configured in
2) Once the particle reaches MPP, the velocity of particles is a 4 × 1 array. The key specifications of the module are shown in
practically zero. Hence, at steady state no oscillations will Table I.
be seen. These steady-state oscillation (normally present To implement the PSO algorithm, the following control pa-
in HC) are very critical because it is one of the major rameters are used in both simulation and experiment: C1 = 1.2,
reasons for the reduced MPPT efficiency [17]. C2 = 1.6, w = 0.4, and Pthr = 1.5%. These parameters are ob-
3) In the case of rapid fluctuations in the environmental con- tained in [22]. To avoid large fluctuations in the operating point
ditions, the HC method can lose the direction of new MPP of the PV array, the value of K2 is chosen to be 0.05 as sug-
and tracking could be driven into a wrong direction. This gested in (19). The array is configured as 4 × 1 (series–parallel),
is one of the major problems of the HC method [18]. with each module rated at 60 W. The total power of the array is
However, the proposed method works on three duty cy- 240 W at STC. Fig. 7 is used to estimate the value of K1 . Using
cles. Since the operating power information is obtained (16), the approximated values of K1 are computed as 675 and
from all three duty cycles, it never loses the direction of 1350 for increasing and decreasing insolation, respectively.
MPP—in rapid fluctuations. Simulations are carried out using the comprehensive PV sys-
4) In the condition of partial shading, the P−V characteristic tem simulator developed in [28]. The sampling time for the
curve is characterized by multiple peaks. As a result, the MPPT controller is very crucial, i.e., the PV system must reach
HC method is most likely to trap at local maxima. On the the steady state (MPP) before the next sampling begins. In ac-
other hand, the PSO method works based on a searching cordance to the work published in [32], the sampling interval is
scheme. Hence, it can still track the global peak correctly. chosen to be 0.1 s.
3634 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

Fig. 9. MATLAB–Simulink simulation model for the PV system with the buck–boost converter and MPPT controller.

Fig. 10. I−V and P−V curves that are used in the simulation.

To evaluate the performance of the proposed method, com-

parison is made with the conventional HC. Three challenging
scenarios are imposed to the system: 1) large step change in
(uniform) solar insolation; 2) step change in load; and 3) partial
shading conditions. These are discussed in subsequent sections.

A. Large Step Change in (Uniform) Solar Insolation

For the case of uniform change in insolation, the insolation is
stepped from low to high, and then to low again. The initial level
is set at λ = 0.4 kW/m2 . At t = 2 s, the insolation is suddenly
stepped up to λ = 1.0 kW/m2 . Finally at t = 6 s, it is stepped
down to λ = 0.4 kW/m2 . The temperature is kept constant at
25 ◦ C. The corresponding I–V and P−V curves for these step
changes are depicted in Fig. 10. The initial operating points for Fig. 11. Tracking voltage, current, duty cycle, and power. (a) HC method.
(b) Proposed method.
the voltage and current are labeled as point A. When a step-up
of insolation occurs, the operating point shifts from point A to
B. As a negative step change occurs, the point shifts from B 0.1 s. This choice of perturbation value is based on the tradeoff
to C. between the speed of MPP tracking and the oscillation that re-
Fig. 11(a) shows the simulation results for the voltage, cur- sults from the HC algorithm. As can be observed, for both step
rent, duty cycle, and power, respectively obtained using the changes in insolation, the tracking is very slow. The case is more
HC method. A fixed perturbation of 0.015 is imposed at every crucial when the insolation changes from λ = 1.0 to λ = 0.4
ISHAQUE et al.: IMPROVED PARTICLE SWARM OPTIMIZATION (PSO)−BASED MPPT FOR PV WITH REDUCED STEADY-STATE OSCILLATION 3635

around the MPP diminishes permanently until another variation

in insolation takes place. The absence of oscillation is the main
feature of the proposed method.

B. Load Variation and Partial Shading

Fig. 13(a) and (b) shows the output power tracked by HC
and the proposed methods, respectively, with respect to load
variation and partial shading. The PV is initially operated at peak
power (240 W). At t = 2 s, a 50% step change in load is imposed,
causing a sudden drop in the operating power. Thereafter, the
HC is forced to track the MPP for the new load condition; it
requires ten MPPT cycles to reach the MPP. In addition, the
oscillations around the MPP are very significant.
At t = 4 s, partial shading occurs. Consequently, the P−V
Fig. 12. Duty cycle variations during the exploring phase. curves are characterized by four peaks labeled by P 1−P4.
Among these, P 1−P 3 are the local while P 4 (170 W) are the
global peaks. For HC, when the operating point changes due
because as the operating point moves from point B to C, the PV
to partial shading, it enters the vicinity of P 3 (144 W). It will
array PV decreases significantly. Due to the fixed perturbation
eventually climb to this local peak and recognize it as the MPP.
in HC, more MPPT cycles are required to reach the MPP. This
The difference between the true global P 4 and the perceived
is the reason for the sluggish response seen in Fig. 11(a).
peak P 3 is 26 W or 11% of the output power. This is a very
More importantly, it can be clearly seen that the oscillations
significant loss for a PV system.
(in the form of ripples) around the MPP are very significant.
In contrast, both the sudden load variation and partial shading
When the insolation is maintained at λ = 1, the operating power
conditions are elegantly handled by the proposed method as
oscillates between 240 and 231 W—a variation of 9 W. Although
depicted in Fig. 13(b). When sudden load change occurs, the
it is known that the oscillation can be reduced by employing a
correct duty cycles are recalculated within five MPPT cycles.
smaller duty step for the perturbation, such action would slow
Furthermore, when the partial shading is imposed on the PV
down the MPPT speed— particularly when the environmental
array, the algorithm explores the P−V curve and successfully
fluctuation is quite large. Furthermore, if the insolation changes
tracks the true global peak (P 4) within seven MPPT cycles.
very rapidly, the HC method may not be able to track the changes
Table II qualitatively summarizes the performance of both
adequately.
methods. It can be seen that in general the proposed method
Fig. 11(b) shows the simulation results for the tracking volt-
is superior to HC with regard to dynamic performance (due to
age, current, duty cycle and power using the proposed PSO
load change) and tracking speed. Moreover, it is very effective
method. Additionally, a separate figure of the duty cycle is
in handling the partial shading conditions. The most crucial
shown in Fig. 12 to provide a clearer insight on the workability
advantage of the proposed method is its ability to eliminate
of the proposed algorithm. When a step change of insolation
the steady-state oscillations to practically zero value. However,
occurs at t = 2 s, it results in a negative change in the operat-
execution wise, it takes slightly longer time to update the new
ing voltage, i.e., P −Pold . Accordingly, new particles, i.e., the
duty cycle. This is to be expected because the PSO algorithm is
three duty cycles, are computed using (16) and (17). It can be
relatively more complex than HC. Nevertheless, both methods
observed that one of the duty cycle d3 is very near to the final
can still be easily implemented in low-cost microcontrollers.
best duty cycle, as shown in Fig. 12. Using these duty cycles,
searches are initiated to locate the new MPP. The exploration
continuously changes the duty cycles, causing fluctuations on VI. EXPERIMENTAL RESULTS
the voltage and current, as can be clearly observed in Fig. 11(b). Hardware wise, the algorithms for the MPPT controller (both
At the eighth sampling interval, the algorithm correctly tracks the proposed and HC methods) are implemented by the Texas
the GP as follows: Vm p = 68.4 V and Im p = 3.5 A, respec- Instrument’s TMS320F240 DSP. The platform for the software
tively. These accurately correspond to the MPP (240 W) shown development is the DS1104 DSspace. The overall experimental
in Fig. 10. set-up is based around the custom-designed PV array simulator,
When a negative step insolation is imposed at t = 6 s, it re- the PVAS2 [33]. The power stage of the PVAS2 is constructed
sults in a large change in operating power. In a similar way (to using 20 linear MOSFETs connected in series to emulate the
the positive step change described earlier), new duty cycles are ripple-free PV voltage. Due to the linear power supply design of
computed using (16) and (17). Once again the algorithm demon- PVAS2, switching frequency ripple does not exist, hence guar-
strates its effectiveness by locating d2 to be very near to the final anteeing noninterference between the PVAS2 with the power
best duty cycle. Accordingly, the exploration begins until it finds converter switching circuit. The I−V curve is generated based
the correct MPP (67 V, 1.35 A, in Fig. 10) at the ninth sampling on a voltage-controlled current source concept. It consists of an
interval. Furthermore, it can be seen that once the algorithm has integrated RAM, which holds the I–V curves and a micropro-
successfully locates the MPP and has stabilized, the oscillation cessor unit that provides a complete communication, monitoring
3636 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

Fig. 13. Tracking performance under load variation and partial shading conditions. (a) HC method. (b) Proposed method.

TABLE II
COMPARISON OF PROPOSED PSO AND HC METHODS

Fig. 14. (a) PVAS unit. (b) Experimental set-up of the MPPT system.

and measurement capabilities. It provides a real-time simulation

capability of any time series of insolation, temperature, fill fac-
tor, or any arbitrary I–V curves measured under real testing
conditions. Unlike the conventional PV array simulators with
diode strings, the usage of digital control and precisely defined
characteristics allows the exact knowledge of MPP for every
time step. Hence, a high-precision dynamic MPPT measure- Fig. 15. Experimental waveforms: tracking voltage, current, duty cycle, and
ment can be easily achieved. The photographs of the PVAS2 power. (a) HC method. (b) Proposed method.
and the overall experimental set-up are shown in Fig. 14(a) and
(b), respectively.
ISHAQUE et al.: IMPROVED PARTICLE SWARM OPTIMIZATION (PSO)−BASED MPPT FOR PV WITH REDUCED STEADY-STATE OSCILLATION 3637

Fig. 15(a) and (b) shows the experimental results for the [10] M. A. Masoum, H. Dehbonei, and E. F. Fuchs, “Theoretical and experi-
tracking voltage, current, duty cycle, and power for HC and mental analyses of photovoltaic systems with voltage and current-based
maximum power point tracking,” IEEE Power Eng. Rev., vol. 22, no. 8,
proposed methods, respectively. The experiments are conducted pp. 62–62, Oct. 2002.
under the same conditions as described in the simulation, i.e., [11] T. Noguchi, S. Togashi, and R. Nakamoto, “Short-current pulse-based
Section V-A. It can be seen that the experimental results match maximum-power-point tracking method for multiple photovoltaic-and-
converter module system,” IEEE Trans. Ind. Electron., vol. 49, no. 1,
very closely to the simulation. For each operating condition, the pp. 217–223, Feb. 2002.
MPP is attained in a relatively short time and exhibits almost [12] H. Patel and V. Agarwal, “Maximum power point tracking scheme for PV
zero oscillation in steady state. Hence, the correctness of the systems operating under partially shaded conditions,” IEEE Trans. Ind.
Electron., vol. 55, no. 4, pp. 1689–1698, Apr. 2008.
proposed algorithm is validated. [13] S. L. Brunton, C. W. Rowley, S. R. Kulkarni, and C. Clarkson, “Maximum
power point tracking for photovoltaic optimization using ripple-based
extremum seeking control,” IEEE Trans. Power Electron., vol. 25, no. 10,
VII. CONCLUSION pp. 2531–2540, Oct. 2010.
[14] S. Jain and V. Agarwal, “A single-stage grid connected inverter topology
In this paper, a PSO with the capability of direct duty cycle for solar PV systems with maximum power point tracking,” IEEE Trans.
is used to track the MPP of a PV system. It is shown that the Power Electron., vol. 22, no. 5, pp. 1928–1940, Sep. 2007.
proposed MPPT controller exhibits an adaptive form of the HC [15] K. H. Hussein, I. Muta, T. Hoshino, and M. Osakada, “Maximum pho-
tovoltaic power tracking: an algorithm for rapidly changing atmospheric
method. To improve the tracking speed, a simple and efficient conditions,” IEE Proc.-Gen., Transmiss., Dist., vol. 142, no. 1, pp. 59–64,
method is proposed to reinitialize the particles to search for the Jan. 1995.
new MPP, resulting in superior dynamic response. The results [16] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “Optimization of
perturb and observe maximum power point tracking method,” IEEE Trans.
indicate that the proposed controller outperforms the HC and Power Electron., vol. 20, no. 4, pp. 963–973, Jul. 2005.
gives a number of advantages: 1) it has a faster tracking speed; [17] A. Safari and S. Mekhilef, “Simulation and hardware implementation of
2) it exhibits zero oscillations at the MPP; 3) it could locate the incremental conductance MPPT with direct control method using cuk
converter,” IEEE Trans. Ind. Electron., vol. 58, no. 4, pp. 1154–1161,
MPP for any environmental variations including partial shading Apr. 2011.
condition and large fluctuations of insolation; and 4) the algo- [18] B. N. Alajmi, K. H. Ahmed, S. J. Finney, and B. W. Williams, “Fuzzy-
rithm can be easily developed using a low-cost microcontrollers. logic-control approach of a modified hill-climbing method for maximum
power point in microgrid standalone photovoltaic system,” IEEE Trans.
Power Electron., vol. 26, no. 4, pp. 1022–1030, Apr. 2011.
ACKNOWLEDGMENT [19] A. K. Rai, N. D. Kaushika, B. Singh, and N. Agarwal, “Simulation
model of ANN based maximum power point tracking controller for so-
The authors would like to thank the Universiti Teknologi lar PV system,” Solar Energy Mater. Solar Cells, vol. 95, pp. 773–778,
2011.
Malaysia for providing the facilities to carry out this research. [20] K. Ishaque and Z. Salam, “An improved modeling method to determine
The vote number for the research fund is 68704. the model parameters of photovoltaic (PV) modules using differential
evolution (DE),” Solar Energy, vol. 85, pp. 2349–2359, 2011.
[21] K. Ishaque, Z. Salam, H. Taheri, and A. Shamsudin, “A critical evaluation
REFERENCES of EA computational methods for Photovoltaic cell parameter extraction
based on two diode model,” Solar Energy, vol. 85, pp. 1768–1779, 2011.
[1] L. Bangyin, D. Shanxu, and C. Tao, “Photovoltaic DC-building-module- [22] M. Miyatake, F. Toriumi, T. Endo, and N. Fujii, “A Novel maximum
based BIPV system-concept and design considerations,” IEEE Trans. power point tracker controlling several converters connected to photo-
Power Electron., vol. 26, no. 5, pp. 1418–1429, May 2011. voltaic arrays with particle swarm optimization technique,” in Proc. Eur.
[2] Z. Li, S. Kai, X. Yan, F. Lanlan, and G. Hongjuan, “A modular grid- Conf. Power Electron. Appl., 2007, pp. 1–10.
connected photovoltaic generation system based on DC bus,” IEEE Trans. [23] V. Phimmasone, Y. Kondo, T. Kamejima, and M. Miyatake, “Evaluation of
Power Electron., vol. 26, no. 2, pp. 523–531, Feb. 2011. extracted energy from PV with PSO-based MPPT against various types of
[3] J. L. Agorreta, M. Borrega, Lo, x, J. pez, and L. Marroyo, solar irradiation changes,” presented at the Int. Conf. Electrical Machines
“Modeling and control of N-paralleled grid-connected inverters and Systems, Incheon, Korea, 2010.
with LCL filter coupled due to grid impedance in PV plants,” [24] V. Phimmasone, T. Endo, Y. Kondo, and M. Miyatake, “Improvement of
IEEE Trans. Power Electron., vol. 26, no. 3, pp. 770–785, Mar. the maximum power point tracker for photovoltaic generators with particle
2011. swarm optimization technique by adding repulsive force among agents,”
[4] J. Young-Hyok, J. Doo-Yong, K. Jun-Gu, K. Jae-Hyung, L. Tae-Won, and in Proc. Int. Conf. Electr. Mach. Syst., 2009, pp. 1–6.
W. Chung-Yuen, “A real maximum power point tracking method for mis- [25] Y. Liu, D. Xia, and Z. He, “MPPT of a PV system based on the particle
matching compensation in PV array under partially shaded conditions,” swarm optimization,” in Proc. 4th Int. Conf. Electric Utility Deregulation
IEEE Trans. Power Electron., vol. 26, no. 4, pp. 1001–1009, Apr. 2011. and Restruct. Power Technol., 2011, pp. 1094–1096.
[5] Y. Bo, L. Wuhua, Z. Yi, and H. Xiangning, “Design and analysis of a grid- [26] M. A. Hassan and M. A. Abido, “Optimal design of microgrids in au-
connected photovoltaic power system,” IEEE Trans. Power Electron., tonomous and grid-connected modes using particle swarm optimization,”
vol. 25, no. 4, pp. 992–1000, Apr. 2010. IEEE Trans. Power Electron., vol. 26, no. 3, pp. 755–769, Mar. 2011.
[6] E. Serban and H. Serban, “A control strategy for a distributed power gen- [27] K. Ishaque, Z. Salam, and H. Taheri, “Simple, fast and accurate two-
eration microgrid application with voltage- and current-controlled source diode model for photovoltaic modules,” Solar Energy Mater. Solar Cells,
converter,” IEEE Trans. Power Electron., vol. 25, no. 12, pp. 2981–2992, vol. 95, pp. 586–594, 2011.
Dec. 2010. [28] K. Ishaque, Z. Salam, and H. Taheri, “Accurate MATLAB simulink PV
[7] R. Ian H, “Envisaging feed-in tariffs for solar photovoltaic electricity: system simulator based on a two-diode model,” J. Power Electron., vol. 11,
European lessons for Canada,” Renew. Sustainable Energy Rev., vol. 9, pp. 179–187, 2011.
pp. 51–68, 2005. [29] S. Chih-Tang, R. N. Noyce, and W. Shockley, “Carrier generation and
[8] L. Y. Seng, G. Lalchand, and G. M. Sow Lin, “Economical, environmen- recombination in P-N junctions and P-N junction characteristics,” Proc.
tal and technical analysis of building integrated photovoltaic systems in IRE, vol. 45, no. 9, pp. 1228–1243, Sep. 1957.
Malaysia,” Energy Policy, vol. 36, pp. 2130–2142, 2008. [30] R. Eberhart and J. Kennedy, “A new optimizer using particle swarm
[9] A. K. Abdelsalam, A. M. Massoud, S. Ahmed, and P. N. Enjeti, theory,” in Proc. 6th Int. Symp. Micro Mach. Human Sci., 1995, pp. 39–
“High-performance adaptive perturb and observe MPPT technique for 43.
photovoltaic-based microgrids,” IEEE Trans. Power Electron., vol. 26, [31] K. Ishaque, Z. Salam, H. Taheri, and Syafaruddin, “Modeling and sim-
no. 4, pp. 1010–1021, Apr. 2011. ulation of photovoltaic (PV) system during partial shading based on a
3638 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 8, AUGUST 2012

two-diode model,” Simul. Modelling Pract. Theory, vol. 19, pp. 1613– Muhammad Amjad received the B.Sc. and M.Sc.
1626, 2011. degrees in Electrical Engineering from the Univer-
[32] D. Sera, R. Teodorescu, J. Hantschel, and M. Knoll, “Optimized maximum sity of Engineering and Technology, Lahore, Pak-
power point tracker for fast-changing environmental conditions,” IEEE istan, in 1998 and 2006, respectively. He is currently
Trans. Ind. Electron., vol. 55, no. 7, pp. 2629–2637, Jul. 2008. working toward the Ph.D. degree in Electrical Engi-
[33] User’s Manual, “Programmable Photovoltaic Array Simulator PVAS1,” neering from Universiti Teknologi Malaysia, Johor
Arsenal Research, AIT Austrian Institute of Technology, Australia, 2007. Bahru, Malaysia.
He was a Lecturer at the University College of
Engineering and Technology, TheIslamia University
of Bahawalpur, Punjab, Pakistan, for 10 years. His re-
search interests include modeling of dielectric barrier
discharge (DBD) chamber and power electronic converter for DBD applications.

Kashif Ishaque received the B.E. degree in Indus-

trial Electronics Engineering from the Institute of In-
dustrial Electronics Engineering, NEDUET, Karachi,
Pakistan, in 2007, and the Master of Engineering Sci-
ence and Ph.D. degrees from Universiti Teknologi
Malaysia Johor Bahru, Malaysia, in 2009 and 2012
respectively.
He is currently an Assistant Professor at the De-
partment of Electronics Engineering, PAF Karachi
Institute of Economics and Technology (PAF-KIET), Saad Mekhilef (M’01) received the B.Eng. degree in
Karachi. He is the author or coauthor of more than 30 Electrical Engineering from the University of Setif,
publications in international journals and proceedings. His research interests in- Setif, Algeria, in 1995, and the Master of Engineer-
clude photovoltaic modeling and control, intelligent control, nonlinear systems ing Science and Ph.D. degrees from the University of
control, and optimization techniques such as genetic algorithm, particle swarm Malaya, Kuala Lumpur, Malaysia, in 1998 and 2003,
optimization, and differential evolution. respectively.
He is currently a Professor at the Department of
Electrical Engineering, University of Malaya, Kuala
Lumpur. He is the author or coauthor of more than
150 publications in international journals and pro-
ceedings. He is actively involved in industrial con-
sultancy, for major corporations in the power electronics projects. His research
interests include power conversion techniques, control of power converters, re-
Zainal Salam (M’11) received the B.S., M.E., and newable energy, and energy efficiency.
Ph.D. degrees from the University of California, Oak-
land; Universiti Teknologi Malaysia (UTM), Johor
Bahru, Malaysia; and the University of Birmingham,
U.K., in 1985, 1989, and 1997, respectively.
He was a Lecturer at UTM for 24 years and is
now a Professor of Power Electronics in the School
of Electrical Engineering. He has been involved on
several research and consulting projects in the area of
battery-powered converters. He is currently the Di-
rector of the Inverter Quality Control Center, UTM,
where he is responsible for testing PV inverters that are to be connected to the
local utility grid. His research interests include all areas of power electronics,
renewable energy, power electronics, and machine control.
 Renewable and Sustainable Energy Reviews 19 (2013) 475–488

Contents lists available at SciVerse ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A review of maximum power point tracking techniques of PV system for

uniform insolation and partial shading condition
Kashif Ishaque a,b, Zainal Salam b,n
a
Karachi Institute of Economics and Technology, Karachi 75190, Pakistan
b
Faculty of Electrical Engineering, Universiti Teknologi Malaysia, UTM 81310, Skudai, Johor Bahru, Malaysia

a r t i c l e i n f o abstract

Article history: This paper presents a review on the state-of-the-art maximum power point tracking (MPPT) techniques
Received 7 June 2012 for PV power system applications. The main techniques that will be deliberated are the Perturb and
Accepted 9 November 2012 Observe, Incremental Conductance and Hill Climbing. The coverage will also encompass their variations
Available online 14 December 2012
and adaptive forms. In addition, the more recent MPPT approaches using soft computing methods such
Keywords: as Fuzzy Logic Control, Artificial Neural Network and Evolutionary Algorithms are included. Whilst the
Global peak (GP) paper provides as thorough treatment of MPPT at normal (uniform) insolation, its focus will be on the
Maximum power point tracking (MPPT) applications of the abovementioned techniques during partial shading conditions. It is envisaged that
Partial shading this review work will be a source of valuable information for PV professionals to keep abreast with the
Photovoltaic (PV) system
latest progress in this area, as well as for new researchers to get started on MPPT.
Uniform insolation
& 2012 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
2. MPPT control structure with power converters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
3. Partial shading condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
4. Categorization of MPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
5. Conventional MPPT under uniform insolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.1. Perturb and observe (P & O) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.1.1. Principle of operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.1.2. Previous important work on P&O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.2. Incremental conductance (IC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.2.1. Principle of operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
5.2.2. Previous work on IC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
5.3. Hill climbing (HC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
5.3.1. Operation of HC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
5.3.2. Previous work on HC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
5.4. IC with direct duty cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
6. Soft computing MPPT under uniform insolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
6.1. Fuzzy logic control (FLC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
6.1.1. Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
6.1.2. Previous work using FLC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.2. Artificial neural network (ANN). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.2.1. ANN operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.2.2. Previous work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.3. Evolutionary algorithm (EA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.3.1. PSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.3.2. Other EA techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482

n
Corresponding author. Tel.: þ6075536187; fax: þ 6075566272.
E-mail addresses: [email protected] (K. Ishaque), [email protected] (Z. Salam).

1364-0321/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.rser.2012.11.032
476 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

7. MPPT during partial shading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482

7.1. Modified P&O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
7.2. Modified IC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
7.3. Modified HC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
7.4. ANN and FLC in partial shading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
7.5. PSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
7.6. Other EA methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
8. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
Acknowledgment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485

1. Introduction shading effect by improvising the conventional MPPT controllers.

However, with recent availability of vast and low cost computing
Despite the relatively high cost of solar modules, PV power power, new methods based on soft computing techniques have
systems have been commercialized in many countries for its long been attracting considerable interests. This approach, which
term economic prospects and more crucially, the concerns over includes the Fuzzy Logic Controller, Artificial Neural Network and
the environment [1]. These systems, which could be grid- Evolutional Algorithm, offers numerous opportunities for more
connected or stand-alone, are being installed in wide-ranging robust and flexible MPPT schemes, in a way that is not possible
power capacities and by using various silicon technologies. using conventional techniques. More importantly, these methods
Regardless of the type and size, one critical component of any are well suited to cater for the partial shading conditions.
PV is the effectiveness of its maximum power point tracker In a short span of time, the body of knowledge related to this
(MPPT). This area has been and is still attracting immense interest subject has grown tremendously. Clearly, it is difficult to keep
from PV research communities as well as industrial players track with the literatures unless a single reference with concise
because it is the most economical way to improve the overall and accurate summaries of the various MPPT schemes is made
PV system efficiency. available. The most recent reviews [4,5] are carried out over five
Conceptually, MPPT is a simple problem—it is basically an years ago, and obviously progress since then has been consider-
operating point matching between the PV array and power able. Therefore, it seems appropriate at this juncture to provide a
converter. However, because of the non-linear I–V characteristics comprehensive and up to date review of the subject. In particular,
of the PV curve and the consequence of the varying environ- the MPPT for partial shading conditions is not fully documented;
mental conditions (particularly insolation and temperature), justifiably, there is a critical need to carry out this exercise
tracking the correct maximum power point (MPP) can sometimes even more.
be a challenging task. The tracking becomes more complicated Although this paper will mainly deal with the latest work on
when the entire PV array does not receive uniform insolation—a MPPT, it is envisaged that some background knowledge would be
condition known as partial shading [2]. This issue has received helpful for certain group of readers to follow through this paper
much attention lately, partly due to the proliferations of building- effectively. Accordingly, information on the principles of opera-
integrated (BIPV) in urban areas. The BIPV, which is commonly tion, limitations and advantages of the conventional MPPT meth-
installed on rooftop, often lends itself to shading because of ods (and their variations) are provided. However, since the
unfavourable physical constraints such as space limitations and number of papers related to one particular method is quite large,
unforeseen structural development. Notwithstanding, partial it is imperative that only works with significant contributions are
shading is also being experienced in large PVPS installations, cited. Papers that refer to previous works with minor modifica-
i.e., solar farms. Typically, it is caused by the clouds that strike on tions or improvements may not be included in the reference list.
certain spots of the solar array, while other parts are left In that regard, apologies are offered to the respective authors.
uniformly irradiated. Such meteorological condition is quite
common in tropical and near-equatorial regions. Another source
of partial shading-like characteristics is exhibited by module 2. MPPT control structure with power converters
irregularities; a common example would be the presence of
cracks on one or more modules of the PV array. The aim of employing MPPT is to ensure that at any environ-
When a PV system is subjected to partial shading, its P–V mental condition (particularly solar insolation and temperature),
curves exhibit multiple peaks with several local peaks and a maximum power is extracted from the PV modules. This is
unique global peak (GP). The conventional MPPT algorithm, such achieved by matching the PV’s MPP with the corresponding
as Perturb and Observe, Hill Climbing and Incremental Conduc- power converter’s operating voltage and current. Fig. 1 shows
tance (to name a few) is likely to be trapped at one of the local the general block diagram of MPPT in conjunction with a dc–dc
peak, simply because its algorithm could not differentiate the converter. Although a stand-alone dc–dc system is depicted here,
local with the GP. Consequently, it oscillates around the vicinity of the application can be extended to a grid connected PV system by
the local peak and remains there indefinitely—a significant adding other power electronic devices such as inverter and grid
reduction in the PV power yield will be experienced [3]. components. Basically, the MPPT works by sensing the current
Over the years, numerous MPPT techniques have been pro- and voltage of the PV array; using this information the array
posed; this is evident from the growth in number of related papers power is calculated and compared with the present value of MPP.
published in various journals and proceedings [4,5]. Despite the Accordingly, the duty cycle of the converter is adjusted using a PI
fact that these methods are designed for the same objectives, they or hysteresis controller to match the MPP—which in turn, forces
differ markedly in terms of complexity, convergence speed, steady the converter to extract the maximum power from the array.
state oscillations, cost, effectiveness and flexibility. Furthermore, An alternative control structure is characterized by directly
numerous researches have been carried out to address the partial updating the duty cycle of the power converter; this is known in
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 477

literature as the direct duty cycle MPPT control. In this scheme, In normal condition, i.e., when the solar insolation on the
the PI block in Fig. 1 is eliminated and duty cycle is computed entire PV array is uniform, as shown in Fig. 2(a), the P–V curve
directly in the MPPT algorithm. This scheme offers number of exhibits the typical unique MPP (curve 1 of Fig. 2(c)). During
advantages: (1) it simplifies the tracking structure, (2) it reduces partial shading, as the third PV module being less illuminated
the computation time and (3) no tuning effort is needed for the PI (shown by shaded block in Fig. 2(b)), the difference in insolation
gains. In short, it replaces the sophisticated MPPT control with a between two modules activates the bypass diode of module 3.
more simplified structure while maintaining similar optimal As a result, two stairs current waveform is created on the I–V
results. curve. Consequently, the corresponding P–V curve is character-
ized by several local peaks and one global peak (GP), as depicted
by curve 2. Furthermore, if the bypass diode is removed, the PV
array exhibits only a single peak (curve 3); but this is achieved at
3. Partial shading condition the expense of a significant reduction in power. Therefore, in
general, the bypass diode is always installed to improve the
Fig. 2 shows a PV array in a typical series-parallel configura- power throughput of the PV array, despite the complication that
tion. In this example, the modules are connected in strings, with arises during partial shading [6].
three modules per string. When one of the modules in the string
experiences less illumination due to shading, its voltage drops;
thus it behaves as a load instead of a generator. A hot spot ensued 4. Categorization of MPPT techniques
and typically a bypass diode is connected in parallel with each PV
module to protect the shaded module from being damaged. In this paper, the selected MPPT are systematically categorized
Additionally, a blocking diode is connected at the end of each into two main groups. First, which is by far the most popular, is
string (combination of series modules in one current path) to referred to as the conventional MPPT. Three main methods
provide the protection against reverse current caused by the (together with their important variations), namely the Perturb
voltage mismatch between the parallel-connected strings. and Observe (P&O), Incremental Conductance (IC) and the Hill
Climbing (HC), will be discussed in detail. The other group is
based on soft computing techniques. They are becoming more
important lately due to the availability of vast and cheap
computing power. The techniques that fall into this category
include the Fuzzy Logic Controller (FLC), Artificial Neural Network
(ANN) and Evolutionary Algorithms (EA). The last, in particular, is
of interests due to its natural suitability to adapt for multi-modal
problem imposed by the partial shading condition. They shall be
discussed comprehensively in this paper.
Besides these, there are other MPPT described in literature:
Fractional Short Circuit Current (FSC) [7–10], Fractional Open
Circuit Voltage (FOV) [11–14] and Ripple Correlation Control
(RCC) [15–19]. These MPPTs have limited accuracy but they do
have their own distinct advantages, namely fewer sensors and
simpler algorithm; thus they offer a reliable and lower cost
solution for certain applications. Other techniques include current
sweep method [20], DC-link capacitor drop control [21,22], load
Fig. 1. A typical voltage or current based MPPT system. current and load voltage minimization [23–27], dP/dV or dP/dI

Blocking Curve 1
1.2
Diode Shaded I-V curve
with bypass diode
0.9 Curve 2
Ipv (A)

3 Shaded 3
PV Module PV
Module 3 Module 3 0.6 Shaded I-V curve
without bypass diode
0.3 Curve 3
Bypass Vpv (V)
0
2 Diode 2 0 15 30 45 60
PV PV
Module 2 Module 2 60

40
Ppv (A)

1 1
PV PV 20
Module 1 Module 1

Vpv (V)
0
0 15 30 45 60

Fig. 2. Operation of PV array (a) under uniform insolation (b) under partial shading (c) the resulting I–V and P–V curve for (a) and (b).
478 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

feedback control [28–32], linear current control [33], state based using a simple auxiliary resonant circuit, an MPPT for soft-
MPPT [34], best fixed voltage algorithm [35], linear reoriented switching boost converter is proposed. To track the MPP under
coordinate method [36] and slide control method [37]. They are rapid changes in insolation, authors in [42] propose a three-point
not as popular and hence have been omitted for brevity. weight comparison P&O method. The slope of the F is decided
based on the comparison of the actual operating power to two
preceding ones. Other efforts to solve this issue are carried out by
5. Conventional MPPT under uniform insolation exploiting the sampling rate of MPPT. For instance, the sampling
rate is optimized in [43], while authors in [44], simply use a high
5.1. Perturb and observe (P & O) sampling rate.
Work in [45] suggests an adaptive P&O scheme for a grid-
5.1.1. Principle of operation connected three-phase inverter. In the beginning, the F is set to
The working principle of P&O is depicted by the flowchart in be 10% of the VOC. Each successive perturbation is set to 50% of the
Fig. 3. The algorithm introduces a perturbation (F) in the operat- preceding one until the value of F is 0.5% of the VOC. Although the
ing voltage and current of PV array and subsequently the change method exhibits somewhat better performance, it is still not fully
in the operating power is observed. The increment in operating adaptive due to the predetermined perturbation steps. Patel and
power implies that the converter is getting closer to the MPP. Agarwal [46] proposed a threshold-based MPPT scheme for
Accordingly, in the next sampling cycle, the direction (slope) of single-stage grid-connected PV system. The system operates in
perturbation is retained and the reference voltage/current is the continuous conduction mode and employs only voltage
further increased by an amount F. Note that once the vicinity of sensor. The array current is estimated using the inductor current.
MPP is reached, with each new perturbation (which is changed in
sign alternately), the algorithm will go back and forth around the
MPP. Consequently, it will never stick exactly at MPP; rather, it
5.2. Incremental conductance (IC)
oscillates around that point indefinitely.
5.2.1. Principle of operation
5.1.2. Previous important work on P&O The idea behind IC is to incrementally compare the ratio of
Since F can be constant or variable, two types of P&O are derivative of conductance with the instantaneous conductance
reported in the literature, i.e., the fixed step and adaptive, [47]. It is derived from the fact that at MPP, the derivative of
respectively. For the fixed step P&O, authors in [38] proposed a power with respect to voltage (dP/dV) is in fact zero, i.e.,
two-stage grid connected inverter; the first stage is comprised of
a two-switch buck-boost converter that performs the MPPT and dP dðVIÞ dI
¼ ¼ IþV ¼0 ð1Þ
sinusoidal waveform generation. Hashimoto et al. [39] proposed a dV dV dV
novel MPPT control algorithm for a half bridge inverter with a Eq. (1) can be rearranged in the following form:
control unit consisting of a multistage buck-boost converter. The
proposed MPPT enables each of the PV modules to generate its I dI DI
 ¼ ffi ð2Þ
maximum power by simply detecting only the total output power V dV DV
of the PV system. Authors in [40] exploit the capability of multi-
objective optimization technique to design the one-cycle con- where DI and DV are the increments of PV current and voltage,
troller for single-stage inverter. The optimization technique is the respectively. The basic rules for IC can thus be derived from the
basis to design the P&O based method. In another work [41], P–V characteristics shown in Fig. 4 and can be written as:
8 dI
>
> ¼  VI , At MPP
< dV
Initialization dI I
dV 4  V , Left of MPP ð3Þ
>
>
: dI o  I , Right of MPP
dV V

Sense VPV(k) and IPV(k)

Using the rules in (3), the basic flow chart for IC is depicted in
Fig. 5. It can be noticed that the reference signal is based on
Calculate Power voltage, V*. Since the rules in (3) are derived using P–V curve, the
P(k) = VPV(k)×IPV(k) current cannot be used as the final output. Instead, the P–I
characteristics curve is utilised. To use IC as a current based

No
P(k) > P(k-1)

yes Complement
slope sign
* =V(k-1)
V(k) * + Φ×slope
OR
* =I(k-1)
I(k) * + Φ×slope

P(k-1) = P(k)

Fig. 3. Flow chart of conventional Perturb and Observe (P&O) method. Fig. 4. P–V characteristics for the basis of IC method.
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 479

marginal error e in (3), i.e., (dI/dVþI/V)o e. Furthermore, in [53],

a totally different approach is proposed; IC is implemented using
the mixed-signal circuit. A major portion of the MPPT algorithm is
realized using analog circuits which exploit the linear region of a
field-effect transistor.
For adaptive IC, numerous works have been reported [54–57].
For instance, Mei et al. [54] suggest an improved IR method,
where two modes of perturbation step are used, i.e., fixed and
variable. In another work related to adaptive IC, Li and Wang [55],
vary the value of F in original IC method with respect to dP/dV
using the following equality
 
 PðkÞPðk1Þ 
F ¼ N  ð7Þ
VðkÞV ðk1Þ
In (7), N is the scaling factor, which is tuned using a trial and
error method. Authors in [56] propose a three-phase grid-con-
nected PV system with modified IC method and reactive power
compensation. Similar to [55], a threshold value, based on the
change in solar insolation, is used to adjust F. In the case of
increasing insolation i.e., DP is positive, F is selected to be smaller
and consequently a smooth tracking towards MPP is achieved. By
investigating the characteristics of PV array in the voltage and
current source region, authors in [57] propose another type of
adaptive IC. If r ¼V(k)/I(k) and rs ¼ dV(k)/dI(k), then (8) outlines the
basic idea of their method, i.e.,
8
< 1 o r=r s o 1, Voltagesourceregion
>
Fig. 5. The basic flow chart of the voltage based IC method. r=r s ¼ 1, At MPP ð8Þ
>
: 1 o r=r o 0,
s Currentsourceregion

MPPT, Eq. (1) can be modified as: Using (8), the reference value of the MPPT algorithm is expres-
dP dðVIÞ dV sed by:
¼ ¼ V þI ¼0 ð4Þ
dI dI dI r ref ðkþ 1Þ ¼ r ref ðkÞ þ Mur ð9Þ
Accordingly, the rules in (3) will be transformed to
M ¼ 1 þ r=r s ð10Þ
8 dV
>
> ¼  VI , At MPP where rref is a reference value of a constant resistance and ur is a
< dI
dV
dI
4  VI , Left of MPP ð5Þ unity conversion factor for the resistance domain.
>
>
: dV o  V , Right of MPP
dI I
5.3. Hill climbing (HC)
Using (5), the flow chart in Fig. 5 can be transformed into
current based IC by interchanging V with I, DV with DI and V(k)* 5.3.1. Operation of HC
with I(k)*. Subsequently, the method will be characterized by In principle, the HC operation is very similar to P&O; the only
incremental resistance (IR), not IC. difference is, instead of perturbing the voltage or current, it
updates the operating point of the PV array by perturbing the
duty cycle, d(k) by a fixed step-size (F) in the direction of
5.2.2. Previous work on IC
increasing power. The perturbation direction is reversed if
Considering its superior dynamic performance, many authors
P(k)oP(k 1); an indication that the tracking is not moving
[48–53] employ IC in its original form. For instance, in [48] an
towards the MPP. This can be described by the following equa-
interface circuit using IC for solar cells without battery storage is
tion:
proposed. In two separate works [49,50], the IC method is applied (
to track the MPP in a single-stage grid inverter system using a dðkÞ þ F, if PðkÞ 4P ðk1Þ
dðkÞ ¼ ð11Þ
two-mode scheme. If the insolation is sufficient, it operates in dðkÞF, if PðkÞ o Pðk1Þ
solar generation mode, otherwise in active power line condition-
ing mode. In [51], a hybrid approach is proposed to track the MPP Due to the fact that the duty cycle is changed directly without
of the module-integrated converter. The authors utilize a linear a PI controller, the HC method is sometimes known as the direct
function to segmentize the I–V curves into MPPs and non-MPPs duty cycle technique. As with the case of P&O, this method can be
region. characterized by fixed or adaptive perturbation step.
Another effective way to utilize the IC is to generate an error,
e, using the instantaneous conductance and the incremental 5.3.2. Previous work on HC
conductance [52]. Mathematically, it can be written as: Considering the fact that the parallel capacitance connected at
the output affects the characteristics of the PV array, Kasa et al.
dI I
þ ¼e ð6Þ [58] estimate F by the following equation:
dV V
1
From (6), it can be seen that the value of e is zero at MPP. F ¼ K M  C PV  ð12Þ
^
C PV ðkÞ
The value of e is usually selected based on the trade-off between
the accuracy of (3) and the allowable oscillations around MPP. where C^ PV(k) is the estimated value of capacitance CPV. It is
Considering the advantages of (6), authors in [52] introduce reported that the power from the PV array can be transferred to
480 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

the utility grid with the nearly unity power factor at the Although fast and accurate, the method is not generic because for
maximum power point. Koutroulis et al. [59] use HC for buck different PV system, the perturbation step F and marginal error e
converter battery charging PV system, while in [60], the authors need to be recomputed based on trial and error approach.
simplify it for a fly back inverter using only voltage sensor. Jain Various adaptive IC are proposed to obtain the best trade-off
and Agarwal [61] apply the conventional HC scheme on a high between MPP oscillations and dynamic performance. In [70],
gain inverter, operating in discontinuous current mode. This based on a threshold value of operating power e, a two step
allows for the low PV array voltages (typically 50–100 V range) based direct duty cycle IC method is proposed. Small value of F is
to be boosted up to the level of grid voltage (115 V or 230 V ac used in the vicinity of the MPP while big step is used in other
systems). In another work [62], HC is used in a parallel connected regions. Using the same idea of [67], Liu et al. [71] propose
MPPT system for stand-alone PV power generation. The proposed another auto tuning scheme for the scaling factor M in (17).
structure reduces the negative influence of dc–dc converter The value of M is constrained to the following inequality:
losses, hence increasing the efficiency of power generation. Kwon  
 dP 
et al. [63] propose a simple three-level boosting MPPT control for M o DDMAX =  ð17Þ
dV Fixed Step ¼ DDMAX
a three phase grid connected PV system. The proposed MPPT
converter, which utilizes the HC algorithm, reduces the reverse If (17) is not satisfied, the algorithm works with a fixed step
recovery losses of the diodes and thereby increases the overall size. Additionally, for smooth start, a simple start program is
efficiency. introduced to the MPPT method. Another work carried out by [72]
Chiang et al. [64] propose a modification to (12) to formulate presents an alternative form of adaptive IC; F is calculated using
an adaptive HC algorithm. The perturbation step F is computed  
 DI IðkÞ 
using the following relationship: F ¼ 7 M þ ¼ 7 Cf ðkÞ ð18Þ
DV VðkÞ
DP=Dd As duty cycle always fall in a particular range, M is allowed to
F¼ ð13Þ
P=d be selected from the following inequalities:
In a separate work, Xiao and Dunford [65] exploit the P–D DDMIN DDMAX
curve for the PV array. Using the value of (dP/dD), they tuned the oMo ð19Þ
f MAX ðkÞ f MAX ðkÞ
perturbation value F using
where
DP  
FðkÞ ¼ M ð14Þ  DI Ið1Þ 
Fðk1Þ f MAX ¼  þ ð20Þ
DV Vð1Þ
If the change in array power (DP) is large due to rapid
environmental variations, F(k) is forced likewise to ensure
improved dynamic performance. When DP is small, the system 6. Soft computing MPPT under uniform insolation
is near to MPP and accordingly, the value of F(k) will also be
small. Wolfs and Tang [66] propose another modification to (14) 6.1. Fuzzy logic control (FLC)
using the relation of dP/dV as follows:
DP 6.1.1. Operation
F¼M ð15Þ
In FLC, a mathematical model of the system is not required.
DV
This in itself is a significant advantage because the uncertainties
However, since the value of M in (15) primarily decides the
such as un-modelled physical quantities, non-linearity and unpre-
performance of MPPT, manual tuning of this scaling factor can be
dictable changes in operating point can be excellently dealt with
extremely sensitive to initial operating conditions. To overcome
[73,74]. However, it does require the designer to have some prior
this issues, Pandey et al. [67] propose a fully adaptive HC method
knowledge of how the output responds qualitatively to the
to automatically tune M under all operating conditions. This is
inputs. The typical process structure of an FLC is shown in
calculated using
  Fig. 6. The controller has three functional blocks namely fuzzifica-
DV MAX DDMAX tion, rules inferences and defuzzification. In addition, it has a rule
M¼   ð16Þ
DP MAX  table in which the designed rules are stored. The process in which
the FLC performs the calculation is called rule inference.
where, DVMAX and DPMAX are the maximum change in array
The inputs to a FLC-based MPPT are usually an error E and a
voltage and power, respectively, with respect to maximum
change in error DE. Since dP/dV vanishes at the MPP, both inputs
change in duty cycle DDMAX.
can be calculated as follows:
As the HC method works on the same principle like P&O, its
dynamic performance suffers similar problem to the latter, i.e., PðkÞP ðk1Þ PðkÞPðk1Þ
EðkÞ ¼ or EðkÞ ¼ ð21Þ
the divergence of MPP. The three point sampling technique [42] VðkÞV ðk1Þ IðkÞIðk1Þ
can also be applied in HC to overcome this problem.
DE ¼ EðkÞEðk1Þ ð22Þ

5.4. IC with direct duty cycle Once E and DE are computed and converted to the linguistic
variables, the FLC output can also be defined linguistically in
Another class of direct duty cycle control is inspired by the IC terms of voltage (DV), current (DI) or duty cycle (DD). The
method. The duty cycle of the converter is varied according to the
rules given in (3). Like other techniques, this method can also be
employed in its original form or adaptive. Xuesong et al. [68]
utilize the IC algorithm depicted by the flowchart in Fig. 5.
Despite the absence of PI control loop, it is shown through
simulations that this method can satisfactorily track the MPP.
Using the same concept of marginal error e as reported in [52],
Safari and Saad [69] use a Cuk converter as the MPPT converter. Fig. 6. Fuzzy logic control (FLC) process structure.
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 481

linguistic variables for the output signal are usually assigned Layer h Layer j Layer k
based on innate knowledge of the particular PV system being
1
employed.

6.1.2. Previous work using FLC

In [75], an FLC with seven linguistic variables is used, while in
1 1
[76] and [77], the membership functions are made less symmetric Solar IMP
2
to give more significance to the specific linguistic variables. In an Insolation (W/m2)
effort to overcome tracking deterioration with respect to changes
in solar insolation experienced by [78], authors in [79] introduce
the array power variation and duty cycle as the inputs for the FLC. 2 2
Temperature VMP
Although, it performs satisfactorily in varying environmental (°C)
conditions, it could not reduce the steady state error in the array j
power. Extending the idea of [78] and [79], authors in [80] j=20
propose a 3-input FLC based MPPT. The information derived from
array current, power and duty cycle of converter are selected as Fig. 7. The ANN structure for [90].
the inputs to the FLC. Some improvements are seen in [81], where
the authors employ the Fuzzy Cognitive Networks (FCN) to
enhance the performance of [78]. The voltage, current, tempera- ANN technique, as shown in Fig. 9, to further improve the P&O
ture, solar insolation and control variable (current) are used as the performance. It can be seen that the insolation and temperature
nodes of the FCN. The weights of the interconnection nodes are are used as the input for the network together with 20 neurons in
determined using a wide range of different climatic conditions. the hidden layer. In another work, [91], the authors improve the
Once the FCN is trained, it can be mounted on any PV system. conventional IC method by combining it with a trained ANN.
The proposed method results in better tracking speed, but at the Most ANN methods [88–91] employ back-propagation (BP)
expense of additional switch and a sensor. training algorithm. Larger number of hidden nodes yield more
In [82], the authors exploit the advantages of the FLC to improve accurate results, but at the expense of longer computational time.
the performance of the conventional HC method. The FLC is devel- Hence, in rapid fluctuations of environment, the ANN MPPT may not
oped by translating the HC algorithm into 16 fuzzy rules. In addition, be able to respond quickly enough to cope with the fast changes.
several authors utilise the idea of adaptive FLC. For instance, in [83], Moreover, the BP algorithm remembers the new value but forgets
an adaptive FLC is proposed to facilitate the constant tuning of the the old, leading to the poor memory retainement. Therefore, in
membership functions and the rule based table in order to achieve general, ANN is unsuitable for low cost microprocessors Fig. 7.
optimum performance. In another work, [84], the scaling factor of
both fuzzy inputs and output are automatically tuned to achieve the 6.3. Evolutionary algorithm (EA)
better dynamic performance of MPPT.
Authors in [85] propose a single input fuzzy logic controller Another soft computing approach that is gaining attention
(SI-FLC) by applying the ‘‘signed distance method’’ [86]. The input recently is the evolutionary algorithm (EA). It is a stochastic
to SI-FLC consists of only one variable known as ‘‘distance’’. This is method that appears to be very efficient in optimizing real-valued
in contrast to the conventional FLC, which requires an error and non-linear and multi-modal objective functions [92,93]. Since the
the derivative (change) of the error as its inputs. The reduction in technique is based on search optimization, in principle, it should
the number of inputs simplifies the rule table to one-dimensional, be able to locate the MPP regardless of environmental variations.
allowing it to be treated as a single-input single-output controller. Various EA methods for MPPT are found in the literature; the
most popular ones are Particle Swarm Optimization (PSO) [94],
6.2. Artificial neural network (ANN) Genetic Algorithm (GA) and Differential Evolution (DE)

6.2.1. ANN operation 6.3.1. PSO

The ability of ANN to recognize and estimate unknown para- PSO is a population-based search method, modelled after the
meters inspired its application for MPP tracking [87]. The input behaviour of bird flocks [95]. The algorithm maintains a swarm of
variables to the ANN can be PV array parameters such as VOC and individuals (called particles), where each particle represents a
ISC, insolation and temperature, or any combination of these. candidate solution. Particles follow a simple behaviour: emulate
The output is usually a reference signal, which can be either the the success of neighbouring particles, and achieve their own
voltage, current or duty cycle. To accurately identify the MPP, the successes. The position of a particle is therefore influenced by
weights associated to the neurons have to be carefully computed the best particle in a neighbourhood, as well as the best solution
through a comprehensive training process. Once trained, the ANN found by the particle. Particles positions, xi, are adjusted using:
can be used as the MPP estimator which will give the reference
value (VMP or IMP) to the MPPT controller. xi k þ 1 ¼ xi k þvi k þ 1 ð23Þ
Note that in (23) where the velocity component, vi, represents
6.2.2. Previous work the step size. The velocity is calculated by:
Typically, the ANN is used to estimate the MPP with respect to
vi k þ 1 ¼ wvi k þ c1 r 1 fP best i xki g þ c2 r 2 fGbest xki g ð24Þ
environmental variations. For example, in [88], the ANN is
utilized to identify the MPP using a gradient descent algorithm where w is the inertia weight, c1 and c2 are the acceleration
training. Other authors used the capability of ANN to improve the coefficients, r1, r2AU(0, 1), Pbesti is the personal best position of
P&O or IC method; Alabedin et al. [89] develop an ANN based P&O particle i, and Gbest is the neighbourhood best position of particle i.
controller. Compared to the FLC based MPPT, the proposed The inertia weight w plays an important role in balancing the global
method results in improved performance in dealing with the search and local search. A large w facilitates a global search while a
fluctuations in the array power. Authors in [90] propose a novel small inertia weight improves a local search. It can be a positive
482 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

7. MPPT during partial shading

Define the Generation Size Gmax, Population Size
NP, Learning Factors C1 ,C2 , Inertia Weight w In this section, the performance conventional MPPT (P&O, IC,
and HC) and the soft computing techniques (FLC, ANN and EA)
under partial shading are reviewed.
Initialize Population

7.1. Modified P&O

For each Iteration G
To aid the discussion, Fig. 9 is used. In this figure, the I–V and
P–V curves are shown: (1) for uniform insolation (solid trace), and
Evaluate Fitness function (2) for partially shaded condition (dotted trace). Under uniform
insolation only one peak exists, i.e., GP1; the conventional P&O
almost always successfully track this point. For partial shading,
Calculate Velocity of five peaks labelled by A–E emerge; clearly, peak E is the true MPP
Particles
(GP) while the others are local. When shading occurs, the array
operating point shifts from 1 to 2; the perturbation sign for the
Calculate Position of P&O reverses and control is increased. Eventually it will recognize
Particles point D as the GP, which is obviously wrong.
The above discussion highlights the limitation of the conven-
tional P&O during partial shading and hence the need for remedial
effort. In [100] a two-mode modified P&O, outlined by the flowchart
G=G+1 G>Gmax
No in Fig. 10 is proposed. The complete algorithm is divided into two
parts: (1) the main program and (2) the GP tracking routine.
Fig. 8. The basic flow chart of PSO method. To avoid scanning the whole P–V curve, two boundary conditions
(Vmin and Vmax) are introduced; these values are based on the
voltage in the P–V curve of the deployed PV array. Despite this
simplification, for certain shading conditions, the algorithm still
needs to scan almost 80% of the P–V curve to ensure none of the
potential peak GP is missed. As a result, the tracking speed is greatly
compromised. The scanning of P–V curve is terminated when the
highest local peak, i.e., the true GP, is found.
In another work [101], an alternative P&O using the compar-
ison of two instantaneous power values is proposed. Mathemati-
cally, the comparison can be written as
Pm ðtÞP ref ðtÞ
oe ð25Þ
P m ðt1Þ
In (25), [Pm (t)] and [Pref (t)] are the instantaneous measured
power and instantaneous maximum power reference, respec-
tively. Despite the good GP tracking capability, it introduces
various new coefficients (which are not shown here for simpli-
city) that complicate the overall MPPT process. Moreover, the
method is tested only on one shading condition; thus, its effec-
Fig. 9. The I–V and P–V curves used for evaluating the performance of various tiveness is not fully validated. Authors in [102] propose a voltage
MPPT methods. sweep method to periodically change the voltage of PV array from
its maximum value (near to VOC) to a minimum value (near to ISC).
At each sampling cycle, the operating voltage and current are
constant or a positive decreasing linear function of iteration index j. measured and stored in the microcontroller memory. After
Fig. 8 depicts the basic flow chart of the PSO method. identifying the region of global maxima, the P&O method is
Authors in [96] employ the PSO algorithm to predict the MPP executed to maintain the operation at the GP.
of PV system. At the beginning of the evolution, the conventional
PSO is applied for the global search. Once it arrives at the vicinity 7.2. Modified IC
of the MPP, a special algorithm is triggered to quickly obtain the
local optimal point. Such scheme effectively improves the speed Since the global and local peaks exhibit the same derivative
of local search. In general, PSO is more suitably used for partial characteristics, i.e., (dP/dV ¼0) or (dP/dI ¼0), the original IC algo-
shading conditions, as shall be described in the following sections. rithm is unable to recognize the true MPP. Accordingly, several
authors propose modified IC schemes during partial shading. For
example, Kobayashi et al. [103] implement a two-stage IC
method. In the first stage, the PV array is forced to operate into
6.3.2. Other EA techniques the neighbourhood of the GP using the values of the maximum
In [97], a hybrid GA-ANN MPPT is proposed. The GA is used to voltage and current, i.e.,
obtain the optimized values of array power and voltage for
V MP
different insolation and temperatures conditions. The results are RMP ¼ k ð26Þ
IMP
used for the offline training of the ANN. In a similar manner,
authors in [98] propose another GA-ANN based MPPT, while in where k is the correction factor, while VMP and IMP are approxi-
[99], GA is used to optimize the FLC based MPPT. mated to be 80% of VOC and 90% of ISC, respectively. Subsequently,
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 483

Fig. 10. Flow chart to track the GP under partial shading condition using modified P&O [100].

in the second stage, adjustment is made to move the operating 7.3. Modified HC
point towards the GP based on the IC method. In another attempt,
authors in [104] propose a linear function to track the GP under Since the HC method is based on the P&O principle, the
partial shading. Mathematically, it can be written as: algorithm also traps at the local minima. In [105], Lei et al.
proposed a duty cycle sweep method based on modified HC to
V grid cater for partial shading condition. The initial value of the duty
Vn ¼  IðkÞ ð27Þ
Iout cycle is computed using the following equation:
sffiffiffiffiffiffiffiffiffiffiffi
where Vgrid and Iout is the output voltage and current of the grid, RMP
respectively. The linear function is activated whenever shading D ¼ 1 ð30Þ
RLoad
condition occurs. It is detected using the following relationships:
where RLoad is estimated using the rating of PV array. The sweep
VðkÞV ðk1Þ o V thr ð28Þ
duty cycle is carried out for the range of 0–90%. Later, the
conventional HC is improved by dividing the duty cycle perturba-
IðkÞIðk1Þ tion (F) by two whenever the direction of the F is increased.
o Ithr ð29Þ
Iðk1Þ Similar to [100], this method also needs to scan over 80% of the PV
curve. In another work, authors in [106] implemented a multiple-
Once the PV array is subjected to partial shading, (27) shifts
input boost converter for micro-inverters based on modified HC.
the operating point to the reference value (V*). Then, the IC
method is employed to track the GP. Although quite effective,
the method can only be applied to a grid-connected system. 7.4. ANN and FLC in partial shading
Interestingly, it is observed in [104] that under rapid fluctuation
of insolation, the performance of the modified IC deteriorates By carefully analyzing (26), it can be deduced that the operat-
more rapidly compared to the conventional method. ing principle of FLC is identical to the IC; they both work on the
484 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

basis of dP/dV information. A zero value of E(k) implies that the

algorithm is at the MPP, i.e., dP/dV¼0. However, in partial S
shading, all peaks (local and GP) exhibit the same dP/dV char-
acteristic. Hence, FLC too cannot guarantee the true GP.
Accordingly, a modified FLC based on three stages operation is
proposed in [107]. The key steps involved: (1) the scanning of P–V
curve (in uniform or partial shading), (2) storing of all possible
local peaks and (3) performing the perturbation and observation
of the array power. Three variables, given by the following A1 A10 A11 A20
equations, are fed to the FLC inputs:
DP ¼ PðkÞP ðk1Þ ð31Þ Left observation Right observation
nodes nodes
DI ¼ IðkÞIðk1Þ ð32Þ
Fig. 11. Bayesian network for MPPT information fusion.
DP M ¼ PM ðkÞP ðk1Þ ð33Þ
where DP and DI are the changes in array power and current in
two subsequent samples, respectively. DPM is the difference
between the stored GP (PM) and the currently measured power. a Bayesian fusion technique to combine PSO with IC methods, as
Using these inputs, thirty-four FLC rules for the duty cycle signal shown in Fig. 11. The network has N¼ 20 nodes represented by
(DD) are formulated based on the HC algorithm. [A1, A2,..., AN]. They are divided into two sectors, left and right. Ten
In addition, efforts have been made to combine the conven- nodes [A1, A2,..., A10] correspond to 10 points of the I–V curve,
tional MPPT with ANN. Syafarudin et al. [108] propose a three estimated by the PSO method. The remaining 10 nodes refer to
layer feed-forward ANN, combined with an FLC. The method is the 10 points, estimated by the IC method. Finally, the symbol S
tested on different solar cell technologies, namely mono- denotes the estimated location after fusion process. Despite its
crystalline silicon, thin-film and amorphous silicon. In another effectiveness, the complexity of the method is the major con-
work by the same authors [109], a Fuzzy wavelet network to straint for implementing it using a low-cost microprocessor.
identify the GP of non-crystalline Si PV modules is proposed. A Ishaque et al. [118] adopt the PSO algorithm in the direct
data set which consists of 200 input–output pairs is used to train control structure, where the duty cycles are used as the positions
and test the proposed ANN network. However, for both works, it of the PSO particles. The main idea is shown by flow chart in
is clearly recognized that obtaining the reliable training data set Fig. 12. If a drastic change in operating power is observed (which
for partial shading is the most challenging tasks. Although implies the occurrence of partial shading), the algorithm begins
shading due to towers, trees and roofs can be modelled easily, the search by sending duty cycles to the power converter. The
the conditions due to the uncertainty of passing clouds are information (voltage and current) obtained from these duty cycles
extremely difficult to estimate. Since the training is carried out are used to compute the Pbest and Gbest values. Using the velocity
using predefined shading scenarios, the effectiveness of the ANN and position equations, the new duty cycles are computed based
algorithm could not be guaranteed. using the following equations:

7.5. PSO Fi k þ 1 ¼ wFi k þc1 r1 fP best i xi k g þ c2 r 2 fGbest xi k g ð36Þ

Since partially shaded I–V curve exhibits multimodal charac- kþ1 k

di ¼ di þ F i k þ 1 ð37Þ
teristics, PSO methods are envisaged very effective to track the GP
under this condition. Miyatake et al. [110,111] utilize the con-
Compared with conventional HC method, this method is able
ventional PSO to track the GP in a constant bus voltage applica-
to handle severe partial shading conditions with excellent
tion. Authors in [112] formulate an analytical expression of the
dynamic tracking speed. Moreover, the steady state oscillations
objective functions based on PV current, insolation and tempera-
are found to be exceptionally low.
ture; then the conventional PSO is utilized to track the MPP. In
More recently, an improved version of [118] (proposed by the
another work, current based PSO method [113] is proposed. The
same authors) is published in [119]. The main idea is to remove
current through the series inductor in boost converter is used as
the random number in the accelerations factor of the conven-
the reference signal to generate the PWM signals for the switch-
tional PSO velocity equation. Hence it is appropriately named as
ing converter. Phimmasone et al. [114,115] introduce a repulsive
deterministic the PSO (DPSO). The modification is described as
term in the velocity equation of the PSO as follows:
follows:
vi k þ 1 ¼ wvi k þ c1 r 1 fP best i xi k g þc2 r 2 fGbest xi k g
3 vi k þ 1 ¼ wvi k þ fPbest i xi k g þ fGbest xi k g
c3 r 3 fcent k xi k g=f9cent k xi k 9 þ dg ð34Þ
+ f or 0 ov oV max ð38Þ
where c3 is the coefficient of repulsive particle. The function vi k þ 1 ¼ wvi k þ fGbest þ Pbest i 2xi k g
‘‘cent’’ is the centre of all particles, described by
X
N
xi k The maximum change in velocity (Vmax) is restricted to a
cent k ¼ ð35Þ particular value, which is determined based on the critical study
i¼1
N
of P–V characteristics during partial shading. The proposed
Another work by [116] propose an Adaptive Perceptive PSO method is shown to have following advantages: (1) consistent
(APPSO); however, due to an additional dimensional search space, solution is achieved even for a small number of particles (2) only
the number of particles is increased significantly compared to its the inertia weight need to be tuned and (3) simpler MPPT
original counterpart [111]. This increases the computation bur- structure compared to the conventional PSO. Despite using only
den, thus making real-time implementation more difficult. three particles, it is experimentally shown that the method
Further improvement is seen in [117], where the authors utilize performs very satisfactorily under very severe shading conditions.
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 485

Fig. 12. Basis idea of PSO based direct control MPPT method [118].

7.6. Other EA methods particularly the selection of appropriate evolutionary control

parameters such as population size, mutation factor etc. This
Authors in [120] propose another EA method, namely differ- may be resolved with the assistance of other optimization
ential evolution (DE). The method is tested on three P–V curves, technique such as ANN. There is also a need to consider the
each exhibits three peaks. However, for each curve presented, the viability of EA methods when it comes to real time implementa-
GP is on the right side, which eventually turns out to be only one tion. The way forward is to find an effective method such that the
shading condition. Hence, these trivial tests cannot be considered search space can be narrowed, so that the computational time is
sufficient to prove the effectiveness of the method if other shading reduced. Furthermore, there are several more EA techniques,
conditions occur. which are yet to be exploited fully. One example is the differential
Using the similar searching methodology of EA methods, evolution (DE). Although it has been successfully applied in the
another method that exploits the capability of dividing rectangles area of PV cell modelling, it has not been used for MPPT-except
(DIRECT) algorithm is developed in [121]. It is shown that the P–V for one very basic work done by [120].
characteristics curve can be represented by the Lipschitz func-
tion.* The algorithm operates either in the global and local mode.
Once the occurrence of partial shading is confirmed, the global Acknowledgment
mode is activated and DIRECT algorithm is applied to locate the
vicinity of the GP. Thereafter, the algorithm switches to the local
This work was supported and financed by the Ministry of
mode where a conventional P&O with a relatively small perturba-
Higher Education, Malaysia, under the Research University Grant
tion is employed to maintain the operating point at the GP. Once
Programme. The project was managed by the Research Manage-
again, despite the improved tracking capability, the complexity of
ment Centre, Universiti Teknologi Malaysia under the vote num-
the method is the major issue.
ber Q.J130000.2423.00G40.

References
8. Summary

[1] Cacciato M, Consoli A, Attanasio R, Gennaro F. Soft-switching converter with

Judging by the amount of recent work, it can be concluded that HF transformer for grid-connected photovoltaic systems, industrial electro-
the MPPT is still a very active research area. There are rooms for nics. IEEE Transactions on 2010;57:1678–86.
improvement, particularly when dealing with partial shading [2] Ishaque K, Salam Z, Taheri H, Syafaruddin. Modeling and simulation of
photovoltaic (PV) system during partial shading based on a two-diode
conditions. For uniform insolation, there is an increasing trend
model. Simulation Modelling Practice and Theory 2011;19:1613–26.
in combining the conventional methods with soft computing [3] R Bruendlinger, B Bletterie, M Milde, H Oldenkamp,. Maximum power point
techniques. Examples of successful schemes have been demon- tracking performance under partially shaded PV array conditions, In: Proc.
21st EUPVSEC, Germany, 2006, p. 2157–2160.
strated by various works described in this paper. Such ‘‘hybrid
[4] Esram T, Kimball JW, Krein PT, Chapman PL, Midya P. Dynamic maximum
MPPT’’ is quite attractive due to the availability of the vast power point tracking of photovoltaic arrays using ripple correlation control,
computational power using relatively low cost microprocessors. power electronics. IEEE Transactions on 2006;21:1282–91.
As for the partial shading, it appears that the EA methods are [5] Salas V, Olı́as E, Barrado A, Lázaro A. Review of the maximum power point
tracking algorithms for stand-alone photovoltaic systems. Solar Energy
the most promising. The combination of direct duty cycle method Materials and Solar Cells 2006;90:1555–78.
and EA can be an interesting option to explore. The PSO, which [6] Ishaque K, Salam Z, Syafaruddin. A comprehensive MATLAB Simulink PV
has a very similar structure to HC has proven to be a viable system simulator with partial shading capability based on two-diode
model. Solar Energy 2011;85:2217–27.
option. However, there are still potential areas of concerns for EA, [7] Hart GW, Branz HM, Cox Iii CH. Experimental tests of open-loop maximum-
power-point tracking techniques for photovoltaic arrays. Solar Cells
    1984;13:185–95.
*
A Lipschitz function f is such that f ðxÞ ¼ f ðyÞ r C x ¼ y for all x and y, where [8] S Yuvarajan, X Shanguang, Photo-voltaic power converter with a simple
C is a constant independent of x and y. maximum-power-point-tracker, In: circuits and systems, 2003. ISCAS ‘03.
486 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

proceedings of the 2003 international symposium on, 2003, p. III-399– [33] P. Ching-Tsai, C Jeng-Yue, C Chin-Peng, H Yi-Shuo, A fast maximum power
III-402 vol. 393. point tracker for photovoltaic power systems, In: industrial electronics
[9] N Mutoh, T Matuo, K Okada, M Sakai, Prediction-data-based maximum- society, 1999. IECON ‘99 Proceedings. The 25th annual conference of the
power-point-tracking method for photovoltaic power generation systems, IEEE, 1999, p. 390–393 vol. 391.
In: power electronics specialists conference, 2002. pesc 02. 2002 IEEE 33rd [34] Solodovnik EV, Shengyi L, Dougal RA. Power controller design for maximum
Annual, 2002, p. 1489–1494 vol. 1483. power tracking in solar installations, power electronics. IEEE Transactions
[10] T Noguchi, S Togashi, R Nakamoto, Short-current pulse based adaptive on 2004;19:1295–304.
maximum-power-point tracking for photovoltaic power generation system, [35] PCM de Carvalho, RST Pontes, DS Oliveira, Jr, DB Riffel, RGV de Oliveira, SB
In: industrial electronics, 2000. ISIE 2000. Proceedings of the 2000 IEEE Mesquita, Control method of a photovoltaic powered reverse osmosis plant
international symposium on, 2000, p. 157–162 vol. 151. without batteries based on maximum power point tracking, In: transmis-
[11] K Kobayashi, H Matsuo, Y Sekine, A novel optimum operating point tracker sion and distribution conference and exposition: Latin America, 2004 IEEE/
of the solar cell power supply system, In: power electronics specialists PES, 2004, p. 137–142.
conference, 2004. PESC 04. 2004 IEEE 35th annual, 2004, p. 2147–2151 [36] E.I. rtiz-Rivera, P Fang, A novel method to estimate the maximum power for
Vol. 2143. a photovoltaic inverter system, In: power electronics specialists conference,
[12] DJ Patterson, Electrical system design for a solar powered vehicle, In: power 2004. PESC 04. 2004 IEEE 35th annual, 2004, p. 2065–2069 Vol. 2063.
electronics specialists conference, 1990. PESC ‘90 record., 21st annual IEEE, [37] Z Miao, W Jie, Z Hui, The application of slide technology in PV maximum
1990, p. 618–622. power point tracking system, In: intelligent control and automation,
[13] N Hyeong-Ju, L Dong-Yun, H Dong-Seok, An improved MPPT converter with 2004. WCICA 2004. Fifth world congress on, 2004, p. 5591–5594 Vol. 5596.
current compensation method for small scaled PV-applications, In: IECON [38] K Chomsuwan, P Prisuwanna, V Monyakul, Photovoltaic grid-connected
02 [Industrial electronics society, IEEE 2002 28th annual conference of the], inverter using two-switch buck-boost converter, In: photovoltaic specialists
2002, p. 1113–1118 vol. 1112. conference, 2002. Conference record of the twenty-ninth IEEE, 2002,
[14] JHR Enslin, DB Snyman, Simplified feed-forward control of the maximum p. 1527–1530.
power point in PV installations, In: industrial electronics, control, instru- [39] O Hashimoto, T Shimizu, G Kimura, A novel high performance utility
mentation, and automation, 1992. Power electronics and motion control., interactive photovoltaic inverter system, In: industry applications confer-
Proceedings of the 1992 international conference on, 1992, ence, 2000. Conference record of the 2000 IEEE, 2000, p. 2255–2260
p. 548–553 vol. 541. vol. 2254.
[15] P Midya, PT Krein, RJ Turnbull, R Reppa, J Kimball, Dynamic maximum [40] Fortunato M, Giustiniani A, Petrone G, Spagnuolo G, Vitelli M. Maximum
power point tracker for photovoltaic applications, In: power electronics power point tracking in a one-cycle-controlled single-stage photovoltaic
specialists conference, 1996. PESC ‘96 record., 27th annual IEEE, 1996, inverter, industrial electronics. IEEE Transactions on 2008;55:2684–93.
p. 1710–1716 vol. 1712. [41] Sang-Hoon P, Gil-Ro C, Yong-Chae J, Chung-Yuen W. Design and application
[16] Yan Hong L, Hamill DC. Simple maximum power point tracker for photo- for PV generation system using a soft-switching boost converter with SARC,
voltaic arrays. Electronics Letters 2000;36:997–9. industrial electronics. IEEE Transactions on 2010;57:515–22.
[17] L Yan Hong, DC Hamill, Synthesis, simulation and experimental verification [42] H Ying-Tung, C China-Hong, Maximum power tracking for photovoltaic
of a maximum power point tracker from nonlinear dynamics, In: power
power system, In: industry applications conference, 2002. 37th IAS annual
electronics specialists conference, 2001. PESC. 2001 IEEE 32nd annual, 2001,
meeting. Conference record of the, 2002, p. 1035–1040 vol. 1032.
p. 199–204 vol. 191.
[43] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimization of perturb and
[18] L Stamenic, M Greig, E Smiley, R Stojanovic, Maximum power point tracking
observe maximum power point tracking method, power electronics. IEEE
for building integrated photovoltaic ventilation systems, In: photovoltaic
Transactions on 2005;20:963–73.
specialists conference, 2000. Conference record of the twenty-eighth IEEE,
[44] NS D’Souza, LAC Lopes, L Xuejun, An intelligent maximum power point
2000, p. 1517–1520.
tracker using peak current control, In: power electronics specialists con-
[19] Kimball JW, Krein PT. Discrete-time ripple correlation control for maximum
ference, 2005. PESC ‘05. IEEE 36th, 2005, p. 172.
power point tracking, power electronics. IEEE Transactions on 2008;23:
[45] A Al-Amoudi, L Zhang, Optimal control of a grid-connected PV system for
2353–62.
maximum power point tracking and unity power factor, In: power electro-
[20] M Bodur, M Ermis, Maximum power point tracking for low power photo-
nics and variable speed drives, 1998. Seventh international conference on
voltaic solar panels, In: electrotechnical conference, 1994. Proceedings.,
(Conf. Publ. No. 456), 1998, p. 80–85.
7th Mediterranean, 1994, p. 758–761 vol. 752.
[46] Patel H, Agarwal V. MPPT scheme for a PV-fed single-phase single-stage
[21] M Matsui, T Kitano, X De-hong, Y Zhong-qing, A new maximum photo-
grid-connected inverter operating in CCM with only one current sensor,
voltaic power tracking control scheme based on power equilibrium at DC
energy conversion. IEEE Transactions on 2009;24:256–63.
link, In: industry applications conference, 1999. Thirty-fourth IAS annual
[47] Wasynezuk O. Dynamic behavior of a class of photovoltaic power systems,
meeting. Conference record of the 1999 IEEE, 1999, p. 804–809 vol. 802.
power apparatus and systems. IEEE Transactions on, PAS-102 1983:3031–7.
[22] T Kitano, M Matsui, X De-hong, Power sensor-less MPPT control scheme
[48] Harada K, Zhao G. Controlled power interface between solar cells and AC
utilizing power balance at DC link-system design to ensure stability and
response, In: industrial electronics society, 2001. IECON ‘01. The 27th source, power electronics. IEEE Transactions on 1993;8:654–62.
annual conference of the IEEE, 2001, p. 1309–1314 vol. 1302. [49] Yeong-Chau K, Tsorng-Juu L, Jiann-Fuh C. Novel maximum-power-point-
[23] HJ Beukes, JHR Enslin, Analysis of a new compound converter as MPPT, tracking controller for photovoltaic energy conversion system, Industrial
battery regulator and bus regulator for satellite power systems, In: power Electronics. IEEE Transactions on 2001;48:594–601.
electronics specialists conference, 1993. PESC ‘93 record., 24th annual IEEE, [50] GJ Yu, YS Jung, JY Choi, I Choy, JH Song, GS Kim, A novel two-mode MPPT
1993, p. 846–852. control algorithm based on comparative study of existing algorithms, In:
[24] CR Sullivan, MJ Powers, A high-efficiency maximum power point tracker for photovoltaic specialists conference, 2002. Conference record of the twenty-
photovoltaic arrays in a solar-powered race vehicle, In: power electronics ninth IEEE, 2002, p. 1531–1534.
specialists conference, 1993. PESC ‘93 record., 24th annual IEEE, 1993, [51] H Koizumi, K Kurokawa, A Novel Maximum power point tracking method
p. 574–580. for PV module integrated converter, In: power electronics specialists
[25] AS Kislovski, R Redl, Maximum-power-tracking using positive feedback, In: conference, 2005. PESC ‘05. IEEE 36th, 2005, p. 2081–2086.
power electronics specialists conference, PESC ‘94 record., 25th annual IEEE, [52] W Wenkai, N Pongratananukul, Q Weihong, K Rustom, T Kasparis,
1994, p. 1065–1068 vol. 1062. I Batarseh, DSP-based multiple peak power tracking for expandable power
[26] J Arias, FF Linera, J Martin-Ramos, AM Pernia, J Cambronero, A modular PV system, In: applied power electronics conference and exposition, 2003.
regulator based on microcontroller with maximum power point tracking, APEC ‘03. Eighteenth annual IEEE, 2003, p. 525–530 vol. 521.
In: industry applications conference, 2004. 39th IAS annual meeting. [53] P Mattavelli, S Saggini, E Orietti, G Spiazzi, A simple mixed-signal MPPT
Conference record of the 2004 IEEE, 2004, p. 1178–1184 vol. 1172. circuit for photovoltaic applications, In: applied power electronics
[27] Shmilovitz D. On the control of photovoltaic maximum power point tracker conference and exposition (APEC), 2010 twenty-fifth annual IEEE, 2010,
via output parameters, electric power applications. IEE Proceedings p. 953–960.
2005;152:239–48. [54] Qiang M, Mingwei S, Liying L, Guerrero JM. A. Novel improved variable step-
[28] R Bhide, SR Bhat, Modular power conditioning unit for photovoltaic size incremental-resistance MPPT method for PV systems, industrial elec-
applications, In: power electronics specialists conference, 1992. PESC ‘92 tronics. IEEE Transactions on 2011;58:2427–34.
record., 23rd annual IEEE, 1992, p. 708–713 vol. 701. [55] L Jiyong, W Honghua, A novel stand-alone PV generation system based on
[29] H Sugimoto, H Dong, A new scheme for maximum photovoltaic power variable step size INC MPPT and SVPWM control, In: power electronics and
tracking control, In: power conversion conference—Nagaoka 1997. motion control conference, 2009. IPEMC ‘09. IEEE 6th international, 2009,
Proceedings of the, 1997, p. 691–696 vol. 692. p. 2155–2160.
[30] Chiang SJ, Chang KT, Yen CY. Residential photovoltaic energy storage [56] Wu L, Zhao Z, Liu J. A. single-stage three-phase grid-connected photovoltaic
system, industrial electronics. IEEE Transactions on 1998;45:385–94. system with modified MPPT method and reactive power compensation,
[31] Bleijs JAM, Gow A. Fast maximum power point control of current-fed DC-DC energy conversion. IEEE Transactions on 2007;22:881–6.
converter for photovoltaic arrays. Electronics Letters 2001;37:5–6. [57] L Jae , B HyunSu, C Bo Hyung, Advanced incremental conductance MPPT
[32] H Cong-Ling, W Jie, Z Miao, Y Jin-Ming, L Jin-Peng, Application of adaptive algorithm with a variable step size, In: power electronics and motion
algorithm of solar cell battery charger, In: electric utility deregulation, control conference, 2006EPE-PEMC 2006. 12th international, 2006, p. 603–607.
restructuring and power technologies, 2004. (DRPT 2004). Proceedings of [58] T Noguchi, H Matsumoto, Maximum-power-point tracking method of
the 2004 IEEE international conference on, 2004, p. 810–813 Vol. 812. photovoltaic power system using single transducer, In: industrial
 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488 487

electronics society, 2003. IECON ‘03. The 29th annual conference of the [84] TF Wu, CH Yang, YK Chen, ZR Liu, Photovoltaic inverter systems with self-
IEEE, 2003, p. 2350–2355 Vol. 2353. tuning fuzzy control based on an experimental planning method, In:
[59] Koutroulis E, Kalaitzakis K, Voulgaris NC. Development of a microcontroller- industry applications conference, 1999. Thirty-fourth IAS annual meeting.
based, photovoltaic maximum power point tracking control system, power Conference record of the 1999 IEEE, 1999, p. 1887–1894 vol. 1883.
electronics. IEEE Transactions on 2001;16:46–54. [85] I Purnama, L Yu-Kang, C Huang-Jen, A fuzzy control maximum power point
[60] Kasa N, Iida T, Chen L. Flyback inverter controlled by sensorless current tracking photovoltaic system, In: fuzzy systems (FUZZ), 2011 IEEE interna-
MPPT for photovoltaic power system, industrial electronics. IEEE Transac- tional conference on, 2011, p. 2432–2439.
tions on 2005;52:1145–52. [86] Byung-Jae C, Seong-Woo K, Byung Kook K. Design and stability analysis of
[61] Jain S, Agarwal V, Single-Stage Grid A. Connected inverter topology for solar single-input fuzzy logic controller, systems, man, and cybernetics, Part B:
PV systems with maximum power point tracking, power electronics. IEEE Cybernetics. IEEE Transactions on 2000;30:303–9.
Transactions on 2007;22:1928–40. [87] Syafaruddin E, Karatepe T. Hiyama, Polar coordinated fuzzy controller based
[62] Gules, De Pellegrin Pacheco J, Hey HL, Imhoff J. A maximum power point real-time maximumpower point control of photovoltaic system. Renewable
tracking system with parallel connection for PV stand-alone applications, Energy 2009;34:10.
industrial electronics. IEEE Transactions on 2008;55:2674–83. [88] Veerachary M, Yadaiah N. ANN based peak power tracking for PV supplied
[63] Jung-Min K, Bong-Hwan K, Kwang-Hee N. Three-phase photovoltaic system DC motors. Solar Energy 2000;69:343–50.
with three-level boosting MPPT control, power electronics. IEEE Transac- [89] AMZ Alabedin, EF El-Saadany, MMA Salama, Maximum power point tracking
tions on 2008;23:2319–27. for Photovoltaic systems using fuzzy logic and artificial neural networks, In:
[64] C Mao-Lin, H Chih-Chiang, L Jong-Rong, Direct power control for distributed power and energy society general meeting, 2011 IEEE, 2011, p. 1–9.
PV power system, In: power conversion conference, 2002. PCC Osaka 2002. [90] X Jinbang, S Anwen, Y Cheng, R Wenpei, Y Xuan, ANN based on IncCond
Proceedings of the, 2002, p. 311–315 vol. 311. algorithm for MPP tracker, In: bio-inspired computing: theories and
[65] X Weidong, WG Dunford, A modified adaptive hill climbing MPPT method applications (BIC-TA), 2011 sixth international conference on, 2011,
for photovoltaic power systems, In: power electronics specialists confer- p. 129–134.
ence, 2004. PESC 04. 2004 IEEE 35th annual, 2004, p. 1957–1963 Vol. 1953. [91] MA Islam, MA Kabir, Neural network based maximum power point tracking
[66] PJ Wolfs, L Tang, A single cell maximum power point tracking converter of photovoltaic arrays, In: TENCON 2011–2011 IEEE region 10 conference,
without a current sensor for high performance cehicle solar arrays, In: 2011, p. 79–82.
power electronics specialists conference, 2005. PESC ‘05. IEEE 36th, 2005, [92] Ishaque K, Salam Z. An improved modeling method to determine the model
p. 165–171. parameters of photovoltaic (PV) modules using differential evolution (DE).
[67] Pandey A, Dasgupta N, Mukerjee AK. High-performance algorithms for drift Solar Energy 2011;85:2349–59.
avoidance and fast tracking in solar mppt system, energy conversion. IEEE [93] Ishaque K, Salam Z, Taheri H, Shamsudin A. A critical evaluation of EA
Transactions on 2008;23:681–9. computational methods for Photovoltaic cell parameter extraction based on
[68] X Zhou, D Song, Y Ma, D Cheng, The simulation and design for MPPT of PV two diode model. Solar Energy 2011;85:1768–79.
system based on incremental conductance method, In: information engi- [94] K Ishaque, Z Salam, A Shamsudin, Application of particle swarm optimiza-
neering (ICIE), 2010 WASE international conference on, 2010, p. 314–317. tion for maximum power point tracking of PV system with direct control
[69] Safari A, Mekhilef S. Simulation and hardware implementation of incre- method, In: 37th annual conference on IEEE industrial electronics society,
mental conductance MPPT with direct control method using Cuk converter, 2011, p. 1214–1219.
industrial electronics. IEEE Transactions on 2011;58:1154–61. [95] R Eberhart, J Kennedy, A new optimizer using particle swarm theory, In:
[70] P Wang, H Ding, C Diao, S Qi, An improved MPPT algorithm based on micro machine and human science, 1995. MHS ‘95., Proceedings of the sixth
traditional incremental conductance method, In: power electronics systems international symposium on, 1995, p. 39–43.
and applications (PESA), 2011 4th international conference on, 2011, p. 1–4. [96] F Qiang, T Nan, A. Complex-method-based PSO algorithm for the maximum
[71] Fangrui L, Shanxu D, Fei L, Bangyin L, Yong K. A variable step size INC MPPT power point tracking in photovoltaic system, In: information technology
method for PV systems, industrial electronics. IEEE Transactions on and computer science (ITCS), 2010 second international conference on,
2008;55:2622–8. 2010, p. 134–137.
[72] D Menniti, A Burgio, N Sorrentino, A Pinnarelli, G Brusco, An incremental [97] R Ramaprabha, V Gothandaraman, K Kanimozhi, R Divya, BL Mathur,
conductance method with variable step size for MPPT: design and imple- Maximum power point tracking using GA-optimized artificial neural net-
mentation, In: electrical power quality and utilisation, 2009. EPQU 2009. work for Solar PV system, In: electrical energy systems (ICEES), 2011 1st
10th international conference on, 2009, p. 1–5. international conference on, 2011, p. 264–268.
[73] Ishaque K, Abdullah S, Ayob S, Salam Z. Single input fuzzy logic controller [98] J Long, Z Chen, Research on the MPPT algorithms of photovoltaic system
for unmanned underwater vehicle. Journal of Intelligent & Robotic Systems based on PV neural network, In: control and decision conference (CCDC),
2010;59:87–100. 2011 Chinese, 2011, p. 1851–1854.
[74] Ishaque K, Abdullah SS, Ayob SM, Salam Z. A simplified approach to design [99] Messai A, Mellit A, Guessoum A, Kalogirou SA. Maximum power point
fuzzy logic controller for an underwater vehicle. Ocean Engineering tracking using a GA optimized fuzzy logic controller and its FPGA imple-
2011;38:271–84. mentation. Solar Energy 2011;85:265–77.
[75] AMA Mahmoud, HM Mashaly, SA Kandil, H El Khashab, MNF Nashed, Fuzzy [100] Patel H, Agarwal V. Maximum power point tracking scheme for PV systems
logic implementation for photovoltaic maximum power tracking, In: operating under partially shaded conditions, industrial electronics. IEEE
industrial electronics society, 2000. IECON 2000. 26th annual confjerence Transactions on 2008;55:1689–98.
of the IEEE, 2000, p. 735–740 vol. 731. [101] Carannante G, Fraddanno C, Pagano M, Piegari L. Experimental performance
[76] Veerachary M, Senjyu T, Uezato K. Neural-network-based maximum- of MPPT algorithm for photovoltaic sources subject to inhomogeneous
power-point tracking of coupled-inductor interleaved-boost-converter- insolation. IEEE Transaction on Industrial Electronics 2009;56:7.
supplied PV system using fuzzy controller, industrial electronics. IEEE [102] E Koutroulis, F Blaabjerg, A New Technique for tracking the global max-
Transactions on 2003;50:749–58. imum power point of PV arrays operating under partial-shading conditions,
[77] N Khaehintung, K Pramotung, B Tuvirat, P Sirisuk, RISC-microcontroller IEEE Journal of Photovoltaics, in press (2012).
built-in fuzzy logic controller of maximum power point tracking for solar- [103] Kobayashi K, Takano I, Sawada Y. A study of a two stage maximum power
powered light-flasher applications, in: Industrial Electronics Society, 2004. point tracking control of a photovoltaic system under partially shaded
IECON 2004. 30th Annual Conference of IEEE, 2004, p. 2673–2678 Vol. 2673. insolation conditions. Solar Energy Materials and Solar Cells 2006;90:2975–88.
[78] W Chung-Yuen, K Duk-Heon, K Sei-Chan, K Won-Sam, K Hack-Sung, A new [104] Young-Hyok J, Doo-Yong J, Jun-Gu K, Jae-Hyung K, Tae-Won L, Chung-Yuen
maximum power point tracker of photovoltaic arrays using fuzzy control- W. A. real maximum power point tracking method for mismatching
ler, In: power electronics specialists conference, PESC ‘94 record., compensation in PV array under partially shaded conditions, power elec-
25th annual IEEE, 1994, pp. 396–403 vol. 391. tronics. IEEE Transactions on 2011;26:1001–9.
[79] MG Simoes, NN Franceschetti, M Friedhofer, A fuzzy logic based photo- [105] M Lie, S Yaojie, L Yandan, B Zhifeng, T Liqin, S Jieqiong, A. high performance
voltaic peak power tracking control, In: industrial electronics, 1998. MPPT control method, In: materials for renewable energy & environment
Proceedings. ISIE ‘98. IEEE international symposium on, 1998, p. 300–305 (ICMREE), 2011, p. 195–199.
vol. 301. [106] SV Dhople, JL Ehlmann, A Davoudi, PL Chapman, Multiple-input boost
[80] Masoum MAS, Sarvi M. Design, simulation and implementation of a fuzzy- converter to minimize power losses due to partial shading in photovoltaic
based maximum power point tracker under variable insolation and tem- modules, In: energy conversion congress and exposition (ECCE), 2010,
perature conditions. Iranian Journal of Science and Technology 2005;29:6. p. 2633–2636.
[81] Kottas TL, Boutalis YS, Karlis AD. New maximum power point tracker for PV [107] BN Alajmi, KH Ahmed, SJ Finney, BW Williams, A maximum power point
arrays using fuzzy controller in close cooperation with fuzzy cognitive tracking technique for partially shaded photovoltaic systems in microgrids,
networks, Energy Conversion. IEEE Transactions on 2006;21:793–803. industrial electronics, IEEE transactions on, in press (2011).
[82] Alajmi BN, Ahmed KH, Finney SJ, Williams BW. Fuzzy-logic-control [108] Syafaruddin E, Karatepe T. Hiyama, artificial neural network-polar coordi-
approach of a modified hill-climbing method for maximum power point nated fuzzy controller based maximum power point tracking control under
in microgrid standalone photovoltaic system, power electronics. IEEE partially shaded conditions. IET Renewable Power Generation 2008;3:15.
Transactions on 2011;26:1022–30. [109] Syafaruddin E, Karatepe T. Hiyama, Fuzzy wavelet network identification of
[83] N Patcharaprakiti, S Premrudeepreechacharn, Maximum power point track- optimum operating point of non-crystalline silicon solar cells, Computers
ing using adaptive fuzzy logic control for grid-connected photovoltaic &amp. Mathematics with Applications 2012;63:68–82.
system, In: power engineering society winter meeting, 2002. IEEE, 2002, [110] M Miyatake, T Inada, I Hiratsuka, Z Hongyan, H Otsuka, M Nakano, Control
p. 372–377 vol. 371. characteristics of a fibonacci-search-based maximum power point tracker
488 K. Ishaque, Z. Salam / Renewable and Sustainable Energy Reviews 19 (2013) 475–488

when a photovoltaic array is partially shaded, In: power electronics and irradiation changes, In: electrical machines and systems (ICEMS), 2010,
motion control conference, 2004. IPEMC 2004. The 4th international, 2004, p. 487–492.
p. 816–821 Vol. 812. [116] Roy Chowdhury S, Saha H. Maximum power point tracking of partially
[111] M Miyatake, F Toriumi, T Endo, N Fujii, A Novel maximum power point shaded solar photovoltaic arrays. Solar Energy Materials and Solar Cells
tracker controlling several converters connected to photovoltaic arrays with 2010;94:1441–7.
particle swarm optimization technique, In: power electronics and applica- [117] F Keyrouz, S Georges, Efficient multidimensional maximum power point
tracking using bayesian fusion, In: electric power and energy conversion
tions, 2007 European conference on, 2007, p. 1–10.
systems (EPECS), 2011, p. 1–5.
[112] Y Liu, D Xia, Z He, MPPT of a PV system based on the particle swarm
[118] K Ishaque, Z Salam, M Amjad, S Mekhilef,. An improved particle swarm
optimization In: electric utility deregulation and restructuring and power
optimization (PSO)-based MPPT for PV with reduced steady state oscilla-
technologies (DRPT), 2011, p. 1094–1096.
tion, IEEE transactions on power electronics, 27 (2012) 3627–3638.
[113] K Ishaque, Z Salam, H Taheri, A Shamsudin, Maximum power point track-
[119] K Ishaque, Z Salam, A deterministic particle swarm optimization maximum
ing for PV system under partial shading condition via particle swarm power point tracker for photovoltaic system under partial shading condi-
optimization, In: applied power electronics colloquium (IAPEC), 2011 IEEE, tion, IEEE Transaction on Industrial Electronics, In press (2012).
2011, p. 5–9. [120] H Taheri, K Ishaque, Z Salam, A. Novel maximum power point tracking
[114] V Phimmasone, T Endo, Y Kondo, M Miyatake, Improvement of the control of photovoltaic system under partial and rapidly fluctuating shadow
maximum power point tracker for photovoltaic generators with particle conditions using differential evolution, In: IEEE symposium on industrial
swarm optimization technique by adding repulsive force among agents, In: electronics & applications (ISIEA), 2010, p. 82–87.
electrical machines and systems, 2009, p. 1–6. [121] Tat Luat N, Kay-Soon L. A global maximum power point tracking scheme
[115] V Phimmasone, Y Kondo, T Kamejima, M Miyatake, Evaluation of extracted employing DIRECT search algorithm for photovoltaic systems, industrial
energy from PV with PSO-based MPPT against various types of solar electronics. IEEE Transactions on 2010;57:3456–67.
 Computers & Operations Research ] (]]]]) ]]]–]]]

Contents lists available at SciVerse ScienceDirect

Computers & Operations Research

journal homepage: www.elsevier.com/locate/caor

Multiobjective cuckoo search for design optimization

Xin-She Yang a,n, Suash Deb b
a
Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK
b
Department of Computer Science & Engineering, C.V. Raman College of Engineering, Bidyanagar, Mahura, Janla, Bhubaneswar 752054, India

a r t i c l e i n f o abstract

Many design problems in engineering are typically multiobjective, under complex nonlinear con-
Keywords: straints. The algorithms needed to solve multiobjective problems can be significantly different from the
Cuckoo search methods for single objective optimization. Computing effort and the number of function evaluations
Metaheuristic may often increase significantly for multiobjective problems. Metaheuristic algorithms start to show
Multiobjective their advantages in dealing with multiobjective optimization. In this paper, we formulate a new cuckoo
Optimization search for multiobjective optimization. We validate it against a set of multiobjective test functions, and
then apply it to solve structural design problems such as beam design and disc brake design. In
addition, we also analyze the main characteristics of the algorithm and their implications.
Crown Copyright & 2011 Published by Elsevier Ltd. All rights reserved.

1. Introduction form the so-called Pareto front. In order to get the sense of the
unknown Pareto front, we have to generate many different
Engineering design often concerns multiple design objectives solution points, and therefore computational effort will increase
under complex, highly nonlinear constraints. In reality, different depending on the number of approximate points, complexity of
objectives often conflict each other, and sometimes, truly optimal the problem and the way of handling solution diversity. Ideally,
solutions may not exist at all, and some compromise and the solutions obtained on the Pareto front should distribute
approximations are often needed [8,11,23]. In addition to these relatively uniformly and un-biased. However, there is no techni-
challenges and complexity, a design problem is subjected to que to ensure that this can be achieved in practice.
various design constraints, limited by design codes or standards, From the implementation point of view, algorithms work well
material properties and choice of available resources and costs for single objective optimization usually cannot directly work for
[11,15]. Even for global optimization problems with a single multiobjective problems, unless under special circumstances such
objective, if the design functions are highly nonlinear, global as combining multiobjectives into a single objective using some
optimality is not easy to reach. Furthermore, many real-world weighted sum method. Substantial modifications are needed to
problems are often NP-hard, which means there is no known make an algorithm work. In addition to these difficulties, a further
efficient algorithm which can be used for a given problem. challenge is how to generate solutions with enough diversity so
Therefore, for a given problem, heuristic choices of algorithms that new solutions can sample the search space efficiently.
and techniques are usually used in practice, in combination with Furthermore, real-world optimization problems always involve
the problem-specific knowledge as some guidance. some degree of uncertainty or noise. For example, materials
On the other hand, metaheuristic algorithms are very powerful properties for a design product may vary significantly, an optimal
in dealing with this kind of optimization, and there are many design should be robust enough to allow such inhomogeneity and
review articles and excellent textbooks [6,9,11,36,38,39,42]. In also provides good choice for decision-makers or designers.
contrast with single objective optimization, multiobjective pro- Despite these challenges, multiobjective optimization has many
blems are typically much difficult and complex [9,17,18]. In a powerful algorithms with many successful applications [1,4,10,
single objective, we have to find the optimal solution which is 15,22,39,31,40,26]. In addition, metaheuristic algorithms start to
often a single point in the solution space, except the case where emerge as a major player for multiobjective global optimization,
there are multiple, equally optimal points. For a multiobjective they often mimic the successful characteristics in nature, especially
optimization problem, there are multiple optimal solutions which biological systems [2,42]. Many new algorithms are emerging with
many important applications [1,20,27,33,39,49].
Recently, a new metaheuristic search algorithm, called Cuckoo
n
Corresponding author. Present address: Mathematics and Scientific Computing,
Search (CS), has been developed by Yang and Deb (2009) [44,45].
National Physical Laboratory, Teddington TW11 0LW, UK. Preliminary studies show that it is very promising and could
E-mail addresses: [email protected], [email protected] (X.-S. Yang). outperform existing algorithms such as PSO. In this paper, we will

0305-0548/$ - see front matter Crown Copyright & 2011 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.cor.2011.09.026

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
2 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]]

extend CS to solve multiobjective problems and formulate a Here m 40 is a minimum step and g is a scale parameter. Clearly,
multiobjective cuckoo search (MOCS) algorithm. We will first as s-1, we have
validate it against a subset of multiobjective test functions. Then, rffiffiffiffiffiffi
g 1
we will apply it to solve design optimization problems in Lðs, g, mÞ  : ð2Þ
2p s3=2
engineering, including bi-objective beam design and a design of
a disc brake. Meanwhile, we will also discuss the unique features This is a special case of the generalized Lévy distribution.
of the proposed algorithm as well as topics for further studies. In general, Lévy distribution should be defined in terms of
Fourier transform
b
FðkÞ ¼ exp½a9k9 , 0 o b r 2, ð3Þ
2. Multiobjective cuckoo search
where a is a scale parameter. The inverse of this integral is not
In order to extend the Cuckoo Search for single optimization to easy, as it does not have analytical form, except for a few
solve multiobjective problems, let us briefly review the interest- special cases.
ing breed behavior of certain cuckoo species. Then, we will For the case of b ¼ 2, we have
outline the basic ideas and steps of the proposed algorithm. 2
FðkÞ ¼ exp½ak , ð4Þ
whose inverse Fourier transform corresponds to a Gaussian
2.1. Cuckoo behavior distribution [43]. Another special case is b ¼ 1, and we have
FðkÞ ¼ exp½a9k9, ð5Þ
Cuckoo are fascinating birds, not only because of the beautiful
sounds they can make, but also because of their aggressive repro- which corresponds to a Cauchy distribution
duction strategy. Some species such as the ani and Guira cuckoos lay 1 g
their eggs in communal nests, though they may remove others’ eggs pðx, g, mÞ ¼ , ð6Þ
p g2 þ ðxmÞ2
to increase the hatching probability of their own eggs [28]. Quite a
number of species engage the obligate brood parasitism by laying where m is the location parameter, while g controls the scale of
their eggs in the nests of other host birds (often other species). There this distribution.
are three basic types of brood parasitism: intraspecific brood For the general case, the inverse integral
Z
parasitism, cooperative breeding, and nest takeover. Some host birds 1 1 b
can engage direct conflict with the intruding cuckoos. If a host bird LðsÞ ¼ cosðksÞexp½a9k9  dk, ð7Þ
p 0
discovers the eggs are not its owns, it will either throw these alien
can be estimated only when s is large. We have
eggs away or simply abandons its nest and builds a new nest
elsewhere. Some cuckoo species such as the New World brood- abGðbÞsinðpb=2Þ
LðsÞ- , s-1: ð8Þ
parasitic Tapera have evolved in such a way that female parasitic p9s91 þ b
cuckoos are often very specialized in the mimicry in color and
pattern of the eggs of a few chosen host species [28]. This reduces Here GðzÞ is the Gamma function
Z 1
the probability of their eggs being abandoned and thus increases
GðzÞ ¼ t z1 et dt: ð9Þ
their reproductivity. 0

In the case when z¼n is an integer, we have GðnÞ ¼ ðn1Þ!

2.2. Efficiency of Lévy flights Fig. 1 shows an example of drawing 100 steps which obey a
Lévy distribution, while Fig. 2 shows their path during a Lévy
In nature, animals search for food in a random or quasi- flight. It is worth pointing out that Lévy flights are more efficient
random manner. In general, the foraging path of an animal is than Brownian random walks in exploring unknown, large-scale
effectively a random walk because the next move is based on the search space. There are many reasons to explain this efficiency,
current location/state and the transition probability to the next and one of them is due to the fact that the variance of Lévy flights
location. Which direction it chooses depends implicitly on a s2 ðtÞ  t3b , 1r b r 2, ð10Þ
probability which can be modeled mathematically. For example,
2
various studies have shown that the flight behavior of many increases much faster than the linear relationship (i.e., s ðtÞ  t)
animals and insects has demonstrated the typical characteristics of Brownian random walks.
of Lévy flights [7,34,29].
A recent study by Reynolds and Frye shows that fruit flies or
Drosophila melanogaster, explore their landscape using a series of
straight flight paths punctuated by a sudden 90o turn, leading to a
Lévy-flight-style intermittent scale-free search pattern [34]. Even
light can be related to Lévy flights [5]. Subsequently, such
behavior has been applied to optimization and optimal search,
and preliminary results show its promising capability [29].
Broadly speaking, Lévy flights are a random walk whose step
length is drawn from the Lévy distribution, often in terms of a
1b
simple power-law formula LðsÞ  9s9 where 0 o b r 2 is
an index. Mathematically speaking, a simple version of Lévy
distribution can be defined as [46]
8 rffiffiffiffiffiffi  
>
< g g 1
exp  if 0 o m o so 1,
Lðs, g, mÞ ¼ 2p 2ðsmÞ ðsmÞ3=2 ð1Þ
>
:
0 if s r 0: Fig. 1. Distribution of 100 consecutive steps.

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]] 3

similarity of solutions to the other solution. This implies that the

mutation is a vectorized operator via the combination of Lévy
flights and differential quality of the solutions. These unique
features work in a combination can ensure the efficiency of the
proposed algorithm.
Based on these three rules, the basic steps of the Multi-
objective Cuckoo Search (MOCS) can be summarized as the
pseudo code shown in Fig. 3.
When generating new solutions xðt þ 1Þ for, say cuckoo i, a Lévy
flight is performed

xðti þ 1Þ ¼ xðtÞ
i
þ a  Levyð
 bÞ, ð11Þ

where a 4 0 is the step size which should be related to the scales

of the problem of interest. In most cases, we can use a ¼ Oð1Þ. In
order to accommodate the difference between solution quality,
we can also use

a ¼ a0 ðxðtÞ
j
xðtÞ
i
Þ, ð12Þ
Fig. 2. 2D Lévy flights in 100 steps.
where a0 is a constant, while the term in the bracket corresponds
to the difference of two randomly solutions. This mimics that fact
2.3. Multiobjective cuckoo search algorithm that similar eggs are less likely to be discovered and thus new
solutions are generated by the proportionality of their difference.
In the original Cuckoo Search for single objective optimization The product  means entry-wise multiplications. Lévy flights
by Yang and Deb [44], the following three idealized rules are essentially provide a random walk while their random steps are
used: drawn from a Lévy distribution for large steps

 Each cuckoo lays one egg at a time, and dumps it in a randomly   u ¼ t 1b ,
Levy ð0 o b r 2Þ, ð13Þ
chosen nest.
which has an infinite variance with an infinite mean. Here the
 The best nests with high quality of eggs (solutions) will carry
consecutive jumps/steps of a cuckoo essentially form a random
over to the next generations.
walk process which obeys a power-law step-length distribution
 The number of available host nests is fixed, and a host can
with a heavy tail.
discover an alien egg with a probability pa A ½0; 1. In this case,
In addition, a fraction pa of the worst nests can be abandoned
the host bird can either throw the egg away or abandon the
so that new nests can be built at new locations by random walks
nest so as to build a completely new nest in a new location.
and mixing. The mixing of the eggs/solutions can be performed by
random permutation according to the similarity/difference to the
For multiobjective optimization problems with K different
host eggs.
objectives, then we modify the first and last rules to incorporate
Obviously, the generation of step size s samples is not trivial
multiobjective needs:
using Lévy flights. A simple scheme discussed in detail by Yang
[46,45] can be summarized as
 Each cuckoo lays K eggs at a time, and dumps them in a
randomly chosen nest. Egg k corresponds to the solution to the u
s ¼ a0 ðxðtÞ ðtÞ

j xi Þ  LevyðbÞ  0:01 ðxðtÞ ðtÞ
j xi Þ, ð14Þ
kth objective. 9v9
1=b

 Each nest will be abandoned with a probability pa and a new

nest with K eggs will be built, according to the similarities/ where u and v are drawn from normal distributions. That is
differences of the eggs. Some random mixing can be used to
generate diversity. u  Nð0, s2u Þ v  Nð0, s2v Þ, ð15Þ

For simplicity, this last assumption can be approximated by a

fraction pa of the n nests being replaced by new nests (with new
random solutions at new locations). For the maximization of
objectives, the quality or fitness of a solution can simply be
proportional to each objective function, and a non-dominated
solution should be sought.
Mathematically speaking, the first rule can be converted into a
randomization process, so that a new solution can be randomly
generated either by a random walk or by Lévy flight. At the same
time, a localized random permutation is carried out over solu-
tions, which can be considered as a form of crossover. For each
nest, there can be K solutions which are generated in the same
way as (11). In essence, the second rule corresponds to elitism so
that the best solutions are passed onto the next generation, and
such selection of the best helps to ensure the algorithm converge
properly. In addition, the third rule can be considered as the
mutation so that the worst solutions are discarded with a
probability and new solutions are generated, according to the Fig. 3. Multiobjective cuckoo search (MOCS).

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
4 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]]

with The stopping criterion can be defined in many ways. We can

( )1=b either use a given tolerance or a fixed number of iterations. From
Gð1 þ bÞsinðpb=2Þ the implementation point of view, a fixed number of iterations is
su ¼ , sv ¼ 1: ð16Þ
G½ð1 þ bÞ=2b2ðb1Þ=2 not only easy to implement, but also suitable to compare the
closeness of Pareto front of different functions. So we have set the
Here G is the standard Gamma function.
fixed number iterations as 5000, which is sufficient for most
It is worth pointing out that the unique features of the MOCS
problems. If necessary, we can also increase it to a larger number.
algorithm are: exploration by Lévy flight, mutation by a combina-
In order to generate more optimal points on the Pareto front,
tion of Lévy flights and vectorized solution difference, crossover
we can do it in two ways: increase the population size n or run
by selective random permutation, and elitism. Cuckoo search
the program a few more times. Through simulations, we found
essentially uses a good combination of all these basic compo-
that to increase of n typically leads to a longer computing time
nents, and thus it is potentially more powerful than algorithms
than to re-run the program a few times. This may be due to the
using one or some of these components. For example, particle
fact that manipulations of large matrices or longer vectors usually
swarm optimization use an update of velocity vector which
take longer. So to generate 200 points using a population size
consists a term of eðxi g n Þ, which is the difference of the current
n¼50 requires to run the program four times, which is easily
solution xi and the current global best g n . This is the main way for
done within a few minutes. Therefore, in all our simulations, we
randomization, which limits the steps that are proportional to the
will use the fixed parameters: n¼50, b ¼ 1:5 and pa ¼0.5.
solution difference. In cuckoo search, however, the randomization
can be more efficient as the steps obey a Lévy distribution which
can be approximated by a power law; therefore, the steps consist
of many small steps and occasionally large-step, long-distance 3.2. Multiobjective test functions
jumps. Comparing with PSO, this long jumps may increase the
search efficiency of cuckoo search significantly in some cases, There are many different test functions for multiobjective
especial for multimodal, nonlinear problems. optimization [13,48,50,51], but a subset of a few widely used
functions provides a wide range of diverse properties in terms
2.4. Pareto front Pareto front and Pareto optimal set. To validate the proposed
MOCS, we have selected a subset of these functions with convex,
A solution vector, u ¼ ðu1 ,: :,un ÞT A F , is said to dominate another non-convex and discontinuous Pareto fronts. We also include
vector v ¼ ðv1 , . . . ,vn ÞT if and only if ui r vi for 8iA f1, . . . ,ng and functions with more complex Pareto sets. To be more specific in
(i A f1, . . . ,ng : ui o vi : In other words, no component of u is larger this paper, we have tested the following five functions:
than the corresponding component of v, and at least one component
is smaller. Similarly, we can define another dominance relationship  Schaffer’s Min–Min (SCH) test function with convex Pareto
$ by front [37,47]
u$v () u!v3u ¼ v: ð17Þ f 1 ðxÞ ¼ x2 , f 2 ðxÞ ¼ ðx2Þ2 , 103 rx r 103 : ð20Þ
It is worth pointing out that for maximization problems, the
dominance can be defined by replacing  with . Therefore,  ZDT1 function with a convex front [50,51]
a point xn A F is called a non-dominated solution if no solution qffiffiffiffiffiffiffiffiffiffi
can be found that dominates it [9]. f 1 ðxÞ ¼ x1 , f 2 ðxÞ ¼ gð1 f 1 =g Þ,
The Pareto front PF of a multiobjective can be defined as the
set of non-dominated solutions so that Pd
9 i¼2 xi
g ¼ 1þ , x1 A ½0; 1, i ¼ 1, . . . ,30, ð21Þ
PF ¼ fsA S9(s0 A S : s0 !sg, ð18Þ d1
where d is the number of dimensions. The Pareto-optimality is
or in term of the Pareto optimal set in the search space
reached when g ¼1.
PF n ¼ fx A F 9(x0 A F : f ðx0 Þ!f ðxÞg, ð19Þ  ZDT2 function with a non-convex front
T  2
where f ¼ ðf 1 , . . . ,f K Þ . f
To obtain a good approximation to Pareto front, a diverse f 1 ðxÞ ¼ x1 , f 2 ðxÞ ¼ g 1 1 ,
g
range of solutions should be generated using efficient techniques
[19,16,22]. In the present approach, Lévy flights ensure the good
diversity of the solutions, as we can see from later simulations.  ZDT3 function with a discontinuous front
" sffiffiffiffiffi #
f1 f1
f 1 ðxÞ ¼ x1 , f 2 ðxÞ ¼ g 1  sinð10pf 1 Þ ,
g g
3. Numerical results
where g in functions ZDT2 and ZDT3 is the same as in function
3.1. Parametric studies ZDT1. In the ZDT3 function, f1 varies from 0 to 0.852 and f2
from 0.773 to 1.
The proposed multiobjective cuckoo search is implemented in  LZ function [24,49]
Matlab, and computing time is within a few seconds to less than a   
minute, depending on the problem of interest. We have tested it 2 X jp 2
f 1 ¼ x1 þ xj sin 6px1 þ ,
using a different range of parameters such as population size (n), 9J1 9 j A J d
1
Lévy exponent b, and discovery probability pa. By varying
  
n ¼ 5; 10,15; 20, to 50; 100,150; 200,250; 300,400; 500, b ¼ 0:5,1, pffiffiffiffiffi 2 X jp 2
1:5,2 and pa ¼ 0:1,0:2, . . . ,0:5, . . . ,0:9, we found that the best para- f 2 ¼ 1 x1 þ þ xj sin 6px1 þ , ð22Þ
9J 2 9 j A J d
2
meters for most applications are: n¼25 to 50, pa ¼0.25 to 0.5 and
b ¼ 1 or 1.5. In addition, the parameter a0 can be in the range of where J1 ¼ fj9j is odd and 2 r j rdg and J 2 ¼ fj9j is
peven
ffiffiffiffiffi and
0.01–0.5, and the value a0 ¼ 0:1 works well for most applications. 2 r j r dg. This function has a Pareto front f 2 ¼ 1 f 1 with a

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]] 5

Pareto set In order to compare the performance of the proposed MOCS

  with other established multiobjective algorithms, we have care-
jp
xj ¼ sin 6px1 þ , j ¼ 2; 3, . . . ,d, x1 A ½0; 1: ð23Þ fully selected a few algorithms with available results from the
d
literature. In case of the results are not available, we have tried to
implement the algorithms using well-documented studies and
then generated new results using these algorithms. In particular,
After generating 200 Pareto points by MOCS, the Pareto front pffiffiffiffiffi we have used other methods for comparison, including vector
generated by MOCS is compared with the true front f 2 ¼ 1 f 1 evaluated genetic algorithm (VEGA) [37], NSGA-II [12], multi-
of ZDT1 (see Fig. 4). objective differential evolution (MODE) [41,3], differential evolu-
Let us define the distance or error between the estimate Pareto tion for multiobjective optimization (DEMO) [35], multiobjective
front PFe to its correspond true front PFt as bees algorithms (Bees) [30], and strength Pareto evolutionary
algorithm (SPEA) [12,25]. The performance measures in terms of
X
N
generalized distance Dg are summarized in Table 2 for all the
Ef ¼ JPF e PF t J2 ¼ ðPF ej PF t Þ2 , ð24Þ
j¼1
above major methods.

where N is the number of points. The convergence property can

be viewed by following the iterations. Fig. 5 shows the exponen-
4. Design optimization
tial-like decrease of Ef as the iterations proceed (left) and the
logarithmic scale (right). We can see clearly that our MOCS
Design optimization, especially design of structures, has many
algorithm indeed converges almost exponentially. The results
applications in engineering and industry. As a result, there are
for all the functions are summarized in Table 1, and the estimated
many different benchmarks with detailed studies in the literature
Pareto fronts and true fronts of other functions are shown in
[21,30–32]. Among the well-known benchmarks are the welded
Figs. 6 and 7.
beam design, and disc brake design. In the rest of this paper, we
will solve these two design benchmarks using MOCS.
1
True Pareto front
MOCS 4.1. Design of a welded beam
0.8
Multiobjective design of a welded beam is a classical bench-
mark which has been solved by many researchers [10,18,32]. The
0.6 problem has four design variables: the width w and length L of
the welded area, the depth d and thickness h of the main beam.
2
f

The objective is minimize both the overall fabrication cost and the
0.4 end deflection d.

Table 1
0.2 Summary of results.

Functions Errors (1000 iterations) Errors (5000 iterations)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 SCH 4.9E  09 2.1E  40
f1 ZDT1 1.2E  6 3.3E  32
ZDT2 7.3E  6 4.5E  25
ZDT3 2.2E  5 1.7E  19
Fig. 4. Pareto front of ZDT1: a comparison of the front found by MOCS and the LZ 9.1E  7 2.3E  21
true front.

4
x 10
6 0
10

4 −10
10

−20
2 10

1
−30
10
0
0 200 400 600 800 1000 0 1000 2000 3000 4000 5000

Fig. 5. Convergence of the proposed MOCS. The least-square distance from the estimated front to the true front of ZDT1 for the first 1000 iterations (left) and the
logarithmic scale for 5000 iterations (right).

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
6 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]]

4 1
True Pareto front True Pareto front
3.5 MOCS MOCS
0.8
3

2.5 0.6
f2

f2
2

1.5 0.4

1
0.2
0.5

0 0
0 1 2 3 4 0 0.2 0.4 0.6 0.8 1
f f
1 1

Fig. 6. Pareto fronts of test functions SCH and ZDT2.

1 1
True Pareto front True Pareto front
MOCS MOCS
0.8
0.5

0.6
2

0
f

0.4

−0.5
0.2

−1 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
f f
1 1

Fig. 7. Pareto fronts of test functions ZDT2 and LZ.

Table 2 x 10−3
Comparison of Dg for n¼ 50 and t¼ 500 iterations.

Methods ZDT1 ZDT2 ZDT3 SCH LZ 8

VEGA 3.79E  02 2.37E  03 3.29E  01 6.98E  02 1.47E  03

NSGA-II 3.33E  02 7.24E  02 1.14E  01 5.73E  03 2.77E  02
MODE 5.80E 03 5.50E 03 2.15E  02 9.32E  04 3.19E  03 6
DEMO 1.08E 03 7.55E  04 1.18E  03 1.79E  04 1.40E 03
Bees 2.40E 02 1.69E  02 1.91E  01 1.25E  02 1.88E  02
2

SPEA 1.78E  03 1.34E  03 4.75E  02 5.17E  03 1.92E  03

f

MOCS 1.17E  07 2.23E  05 2.88E  05 1.25E  07 4.19E  05 4

The problem can be written as

2
minimise f 1 ðxÞ ¼ 1:10471w2 L þ0:04811dhð14:0 þ LÞ, minimize f 2 ¼ d,
ð25Þ
subject to 0
0 10 20 30 40
g 1 ðxÞ ¼ wh r 0, f1

g 2 ðxÞ ¼ dðxÞ0:25 r0, Fig. 8. Pareto front for the bi-objective beam design.

g 3 ðxÞ ¼ tðxÞ13; 600 r0,

g 7 ðxÞ ¼ 6000PðxÞ r 0, ð26Þ
g 4 ðxÞ ¼ sðxÞ30; 000 r0,
where
g 5 ðxÞ ¼ 0:10471w2 þ 0:04811hdð14 þ LÞ5:0 r0,  
504; 000 L
sðxÞ ¼ , Q ¼ 6000 14 þ ,
g 6 ðxÞ ¼ 0:125w r0, hd
2 2

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]] 7

2 F
10 g 3 ðxÞ ¼ 0:4 r0,
MOCS 3:14ðR2 r 2 Þ
Bees
0 DEMO 2:22 103 FðR3 r 3 Þ
10 MODE g 4 ðxÞ ¼ 1 r 0,
ðR2 r 2 Þ2
NSGA−II
SPEA
−2
10 VEGA 0:0266FsðR3 r 3 Þ
g 5 ðxÞ ¼ 900 r0: ð29Þ
Dg

ðR2 r 2 Þ

10
−4 The simple limits are
55 rr r 80, 75 r R r110, 1000 rF r 3000, 2 rs r 20: ð30Þ

−6 The Pareto front of 50 solution points after 1000 iterations

10
obtained by MOCS is shown in Fig. 10, where we can see that the
results are smooth and are the same or better than the results
0 200 400 600 800 1000 obtained in [32].
iterations The comparison of the convergence rates is plotted in the
logarithmic scales in Fig. 11. We can see from these two figures
Fig. 9. Convergence comparison for the beam design.
that the convergence of MOCS are of the highest and exponential
" # in both cases. This again suggests that MOCS provides better
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
1 L2 ðw þdÞ2 solutions in a more efficient way.
D¼ L2 þ ðwþ dÞ2 , J ¼ 2wL þ ,
2 6 2
30
65; 856 QD
d¼ 3
, b¼ ,
30; 000hd J
25
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
6000 abL 2
a ¼ pffiffiffi , tðxÞ ¼ a2 þ þb ,
2wL D 20
3
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
dh d 30=48
P ¼ 0:61423 106 1 : ð27Þ
f2

6 28 15

The simple limits or bounds are 0:1 rL,d r 10 and 0:125 r

w,hr 2:0. 10
By using the MOCS, we have solved this design problem. The
approximate Pareto front generated by the 50 non-dominated
solutions after 1000 iterations is shown in Fig. 8. This is consistent 5
with the results obtained by others [32,30]. In addition, the
results are more smooth with fewer iterations.
0
In order to see how the proposed MOCS performance for the 0 0.5 1 1.5 2 2.5 3
real-world design problems, we also solved the same problems f1
using other available multiobjective algorithms. The comparison
of the convergence rates is plotted in the logarithmic scales in Fig. 10. Pareto front for the disc brake design.
Fig. 9. By comparison, the convergence rates of MOCS are of the
highest in an exponentially decreasing way.
2
4.2. Design of a disc brake 10
MOCS
Bees
Design of a multiple disc brake is another benchmark for DEMO
0
multiobjective optimization [18,27,32]. The objectives are to 10
MODE
minimize the overall mass and the braking time by choosing NSGA−II
optimal design variables: the inner radius r, outer radius R of the SPEA
−2
discs, the engaging force F and the number of the friction surface S. 10 VEGA
Dg

This is under the design constraints such as the torque, pressure,

temperature, and length of the brake. This bi-objective design
problem can be written as −4
10
Minimize f 1 ðxÞ ¼ 4:9 105 ðR2 r 2 Þðs1Þ,
9:82 106 ðR2 r 2 Þ
f 2 ðxÞ ¼ , ð28Þ −6
FsðR3 r 3 Þ 10

subject to
0 200 400 600 800 1000
g 1 ðxÞ ¼ 20ðRrÞ r0, iterations

g 2 ðxÞ ¼ 2:5ðs þ 1Þ30 r0, Fig. 11. Convergence comparison for the disc brake design.

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
8 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]]

The simulations for these benchmarks and test functions [11] Deb K. Multi-objective optimization using evolutionary algorithms. New
suggest that MOCS is a very efficient algorithm for multiobjective York: John Wiley & Sons; 2001.
[12] Deb K, Pratap A, Agarwal S, Mayarivan T. A fast and elitist multiobjective
optimization. It can deal with highly nonlinear problems with algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation 2002;6:
complex constraints and diverse Pareto optimal sets. 182–97.
[13] Deb EZK, Thiele L. Comparison of multi objective evolutionary algorithms:
empirical results, Technical Report 70, Computer Engineering and Networks
Laboratory (TIK), Swiss Federal Institute of Technology (ETH) Zurich, Glor-
5. Conclusions iastrasse 35, CH-8092 Zurich, Switzerland, February 1999.
[14] Erfani T, Utyuzhnikov S. Directed search domain: a method for even
generation of Pareto frontier in multiobjective optimization. Engineering
Multiobjective optimization problems are typically very diffi- Optimization 2011;43(5):467–84.
cult to solve. In this paper, we have successfully formulated a new [15] Farina M, Deb K, Amota P. Dynamic multiobjective optimization problems:
algorithm for multiobjective optimization, namely, multiobjective test cases, approximations, and applications. IEEE Transactions on Evolu-
tionary Computation 2004;8:425–42.
cuckoo search, based on the recently developed cuckoo search [16] Figueira JR, Tavares G, Wiecek MM. Labeling algorithms for multiple
algorithm. The proposed MOCS has been tested against a subset of objective integer knapsack problems. Computers & OR 2010;37(4):700–11.
well-chosen test functions, and then been applied to solve design [17] Floudas CA, Pardalos PM, Adjiman CS, Esposito WR, Gumus ZH, Harding ST,
Klepeis JL, Meyer CA, Scheiger CA. Handbook of test problems in local and
optimization benchmarks in structural engineering. Results sug- global optimization. Springer; 1999.
gest that MOCS is an efficient multiobjective optimizer. [18] Gong WY, Cai ZH, Zhu L. An effective multiobjective differential evolution
In comparison with other algorithms, cuckoo search performs algorithm for engineering design. Structural and Multidisciplinary Optimiza-
tion 2009;38:137–57.
well for almost all these test problems. This superiority can be [19] Gujarathi AM, Babu BV. Improved strategies of multi-objective differential
attributed to the fact that cuckoo search uses a combination of evolution (MODE) for multi-objective optimization. In: Proceedings of fourth
vectorized mutation, crossover by permutation and Lévy flights Indian international conference on artificial intelligence (IICAI-09), December
16–18, 2009.
and selective elitism among the best solutions. In addition, the
[20] Kennedy J, Eberhart RC. Particle swarm optimization. In: Proceedings of
not-so-good solutions can be replaced systematically by new IEEE international conference on neural networks, Piscataway, NJ, 1995,
solutions, and new solutions are often generated by preferring pp. 1942–8.
quality solutions in the solution sets. Thus, the mechanism of the [21] Kim JT, Oh JW, Lee IW. Multiobjective optimization of steel box girder bridge.
In: Proceedings of seventh KAIST-NTU-KU trilateral seminar/workshop on
overall search moves is more subtle and balanced, compared with civil engineering, Kyoto, December 1997.
the simple mechanism used in particle swarm optimization. [22] Konak A, Coit DW, Smith AE. Multiobjective optimization using genetic
Obviously, a more detailed validation study over a wide range algorithms: a tutorial. Reliability Engineering and System Safety 2006;91:
992–1007.
of test functions will be carried out in the follow-up work in the [23] Leifsson L, Koziel S. Multi-fidelity design optimization of transonic airfoils
near future. using physics-based surrogate modeling and shape-preserving response
Additional test and comparison of the proposed are highly prediction. Journal of Computer Science 2010;1:98–106.
[24] Li H, Zhang QF. Multiobjective optimization problems with complicated
needed. In the future work, we will focus on the parametric Pareto sets, MOEA/D and NSGA-II. IEEE Transactions on Evolutionary Com-
studies for a wider range of test problems, including discrete and putation 2009;13:284–302.
mixed type of optimization problems. We will try to test the [25] Madavan NK. Multiobjective optimization using a Pareto differential evolu-
tion approach. In: Congress on evolutionary computation (CEC’2002), vol. 2.
diversity of the Pareto front it can generate so as to identify the
New Jersey: IEEE Service Center; 2002. p. 1145–50.
way to improve this algorithm to suit for a diverse range of [26] Marler RT, Arora JS. Survey of multi-objective optimization methods for
problems. There are a few efficient techniques to generate diverse engineering. Structural and Multidisciplinary Optimization 2004;26:
Pareto fronts [14,16,19], and some combination with these 369–95.
[27] Osyczka A, Kundu S. A genetic algorithm-based multicriteria optimization
techniques may improve MOCS even further. method. In: Proceedings of first world congress on structural and multi-
Further research can also emphasize the performance compar- disciplinary optimization. Elsevier Science; 1995. p. 909–14.
ison of this algorithm with other popular methods for multi- [28] Payne RB, Sorenson MD, Klitz K. The cuckoos. Oxford University Press; 2005.
[29] Pavlyukevich I. Lévy flights, non-local search and simulated annealing.
objective optimization [9,10,38]. In addition, hybridization with Journal of Computational Physics 2007;226:1830–44.
other algorithms may also prove to be fruitful. [30] Pham DT, Ghanbarzadeh A. multi-objective optimisation using the bees
algorithm. In: Third international virtual conference on intelligent produc-
tion machines and systems (IPROMS 2007): Whittles, Dunbeath, Scotland,
2007.
References
[31] Rangaiah G. Multi-objective optimization: techniques and applications in
chemical engineering. World Scientific Publishing; 2008.
[1] Abbass HA, Sarker R. The Pareto differential evolution algorithm. Interna- [32] Ray L, Liew KM. A swarm metaphor for multiobjective design optimization.
tional Journal on Artificial Intelligence Tools 2002;11(4):531–52. Engineering Optimization 2002;34(2):141–53.
[2] Adamatzky A, Bull L, De Lacy Costello B, Stepney S, Teuscher C. Unconven- [33] Reyes-Sierra M, Coello CAC. Multi-objective particle swarm optimizers: a
tional computing 2007. UK: Luniver Press; 2007. survey of the state-of-the-art. International Journal of Computational Intelli-
[3] Babu BV, Gujarathi AM. Multi-objective differential evolution (MODE) for gence Research 2006;2(3):287–308.
optimization of supply chain planning and management. In: IEEE Congress on [34] Reynolds AM, Frye MA. Free-flight odor tracking in Drosophila is consistent
Evolutionary Computation (CEC 2007), pp. 2732–9. with an optimal intermittent scale-free search. PLoS One 2007;2:e354.
[4] Banks A, Vincent J, Anyakoha C. A review of particle swarm optimization. Part [35] Robič T, Filipič B. DEMO: differential evolution for multiobjective optimiza-
II: hydridisation combinatorial, multicriteria and constrained optimization, tion. In: Coello CA, editor. EMO 2005, Lecture Notes in Computer Sciences,
and indicative applications. Natural Computing 2008;20:109–24. vol. 3410; 2005. p. 520–33.
[5] Barthelemy P, Bertolotti J, Wiersma DS. A Lévy flight for light. Nature 2008; [36] Sayadi MK, Ramezanian R, Ghaffari-Nasab N. A discrete firefly meta-heuristic
453:495–8. with local search for makespan minimization in permutation flow shop
[6] Blum C, Roli A. Metaheuristics in combinatorial optimization: overview and scheduling problems. International Journal of Industrial Engineering Compu-
conceptural comparison. ACM Computing Surveys 2003;35:268–308. tations 2010;1:1–10.
[7] Brown C, Liebovitch LS, Glendon R. Lévy flights in Dobe Ju/’hoansi foraging [37] Schaffer JD. Multiple objective optimization with vector evaluated genetic
patterns. Human Ecology 2007;35:129–38. algorithms. In: Proceedings of the first international conference on genetic
[8] Cagnina LC, Esquivel SC, Coello CA. Solving engineering optimization pro- algorithms; 1985. p. 93–100.
blems with the simple constrained particle swarm optimizer. Informatica [38] Talbi E-G. Parallel combinatorial optimization. John Wiley and Sons; 2006.
2008;32:319–26. p. 330.
[9] Coello CAC. An updated survey of evolutionary multiobjective optimization [39] Talbi E-G. Metaheuristics: from design to implementation. John Wiley and
techniques: state of the art and future trends. In: Proceedings of 1999 Sons; 2009. p. 624.
congress on evolutionary computation, CEC99, 1999. doi: 10.1109/CEC. [40] Tantar E, Dhaenens C, Figueira JR, Talbi E-G. A priori landscape analysis in
1999.781901. guiding interactive multi-objective metaheuristics. In: IEEE congress on
[10] Deb K. Evolutionary algorithms for multi-criterion optimization in engineer- evolutionary computation 2008; 2008. p. 4104–11.
ing design. In: Evolutionary algorithms in engineering and computer science. [41] Xue F. Multi-objective differential evolution: theory and applications. PhD
Wiley; 1999. p. 135–61. thesis. Rensselaer Polytechnic Institute; 2004.

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
 X.-S. Yang, S. Deb / Computers & Operations Research ] (]]]]) ]]]–]]] 9

[42] Yang XS. Nature-Inspired Metaheuristic Algorithms. Luniver Press; 2008 of the 2003 congress on evolutionary computation CEC2003, vol. 4. Canberra
(first edition, 2008; second edition, 2010). Australia: IEEE Press; December 2003. p. 2400–5.
[43] Yang XS. Introduction to computational mathematics. World Scientific [48] Zhang QF, Zhou AM, Zhao SZ, Suganthan PN, Liu W, Tiwari S. Multiobjective
Publishing; 2008. optimization test instances for the CEC 2009 special session and competition.
[44] Yang XS, Deb S. Cuckoo search via Lévy flights. In: Proceedings of world Technical Report CES-487, University of Essex, Nanyang Technological
congress on nature & biologically inspired computing (NaBIC 2009 India). University, and Clemson University; April 2009.
USA: IEEE Publications; 2009. p. 210–4. [49] Zhang QF, Li H. MOEA/D: a multiobjective evolutionary algorithm based on
[45] Yang XS, Deb S. Engineering optimization by cuckoo search. Int J Math decomposition. IEEE Transactions on Evolutionary Computation 2007;11:712–31.
Modelling Num Opt 2010;1(4):330–43. [50] Zitzler E, Thiele L. Multiobjective evolutionary algorithms: a comparative
[46] Yang XS. Engineering optimisation: an introduction with metaheuristic case study and the strength Pareto approach. IEEE Transactions on Evolu-
applications. John Wiley and Sons; 2010. tionary Computation 1999;3:257–71.
[47] Zhang LB, Zhou CG, Liu XH, Ma ZQ, Liang YC. Solving multi objective [51] Zitzler E, Deb K, Thiele L. Comparison of multiobjective evolutionary algo-
optimization problems using particle swarm optimization. In: Proceedings rithms: empirical results. Evolution Computing 2000;8:173–95.

Please cite this article as: Yang X-S, Deb S. Multiobjective cuckoo search for design optimization. Computers and Operations Research
(2011), doi:10.1016/j.cor.2011.09.026
CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019 173

A Novel MPPT Method Based on Cuckoo Search

Algorithm and Golden Section Search Algorithm
for Partially Shaded PV System
Une nouvelle méthode PPPM basée sur un
algorithme de recherche Coucou et un algorithme
de recherche de la section d’or pour un système PV
partiellement ombragé
Dimas Aji Nugraha, K. L. Lian , Senior Member, IEEE, and Suwarno, Senior Member, IEEE
Abstract— Partial shading is a common and difficult problem to be solved in a photovoltaic (PV) system.
Numerous efforts have been introduced to mitigate this problem. Some commonly used approaches are deploying
some metaheuristic (MH) algorithm to track the multiple-peak P–V curve of partially shaded PV system.
Cuckoo search (CS) is a new optimization algorithm based on the MH approach. It has been used to solve an
optimization problem in many applications, including the maximum power point tracking (MPPT) problem.
The CS algorithm performs well in tracking the global maximum power point (GMPP). However, just like any
other MH algorithm, there is still a dilemmatic trading between their accuracy and the tracking time needed to
find GMPP. This paper proposes a new MPPT algorithm by combining the CS algorithm with golden section
search (GSS) to take beneficial features from both the algorithms. To validate the proposed algorithm, it is
evaluated with various cases of partial shading. The simulation and experimental results show a noticeable
performance improvement compared with the original CS algorithm and other MH algorithms.
Résumé— L’ombrage partiel est un problème commun et difficile à résoudre dans un système photovoltaïque
(PV). De nombreux efforts ont été déployés pour atténuer ce problème. Certaines approches couramment
utilisées utilisent un algorithme métaheuristique (MH) pour suivre la courbe P - V à plusieurs pics d’un système
photovoltaïque partiellement ombré. La recherche Coucou (RC) est un nouvel algorithme d’optimisation basé
sur l’approche MH. Il a été utilisé pour résoudre un probléme d’optimisation dans de nombreuses applications,
notamment le problème de la poursuite du point de puissance maximale (PPPM). L’algorithme RS exécute bien
la poursuite du point de puissance maximale globale (PPMG). Cependant, comme pour tout autre algorithme
MH, il existe toujours un échange dilemmatique entre leur précision et le temps de poursuite nécessaire pour
trouver le GMPP. Cet article propose un nouvel algorithme PPPM en combinant l’algorithme RC avec la
recherche de la section d’or (GSS) pour tirer parti des caractéristiques bénéfiques des deux algorithmes. Pour
valider l’algorithme proposé, il est évalué avec différents cas d’ombrage partiel. La simulation et les résultats
expérimentaux montrent une amélioration notable des performances par rapport à l’algorithme RS original et
à d’autres algorithmes MH.
Index Terms— Cuckoo search (CS), golden section search (GSS), maximum power point tracking (MPPT),
optimization, partial shading, photovoltaic (PV) system.

Manuscript received October 30, 2018; accepted April 29, 2019. Date of I. I NTRODUCTION
current version July 23, 2019. This work was supported in part by MOST
under Grant NSC 106-2221-E-011-100 and in part by the Taiwan Building
Technology Center from the Featured Areas Research Center Program within
the framework of the Higher Education Sprout Project by the Ministry
I N earlier years, the utilization of solar energy was inhibited
by the high investment cost required on photovoltaic (PV)
panels. However, significant reduction on the production and
of Education, Taiwan. This paper was presented in part at the 18th IEEE installation cost of PV panels in recent years has initiated not
Canada Electrical Power and Energy Conference (EPEC 2018), Toronto, ON,
Canada, 2018. (Corresponding author: K. L. Lian.)
only solar energy farming through rooftop PV panel systems
D. A. Nugraha is with the Department of Electrical Engineering, National but also a trend of large-scale PV system installation in many
Taiwan University of Science and Technology, Taipei 10607, Taiwan, also countries all over the world [1], [2].
with the School of Electrical Engineering and Informatics, Institut Teknologi
Bandung, Bandung 40116, Indonesia, and also with PT PLN (Persero),
A common yet difficult problem in a PV system imple-
Jakarta 12160, Indonesia (e-mail: [email protected]). mentation was partial shading. Especially for a large PV
K. L. Lian is with the Department of Electrical Engineering, National system, a partial shading problem is an inevitable issue.
Taiwan University of Science and Technology, Taipei 10607, Taiwan (e-mail:
[email protected]).
Uniform solar irradiance is hardly maintained by the PV arrays
Suwarno is with the School of Electrical Engineering and Infor- because of overshadowing caused by clouds or nearby objects.
matics, Institut Teknologi Bandung, Bandung 40116, Indonesia (e-mail: The occurrence of a shaded PV cell will limit the current
[email protected]).
Associate Editor managing this paper’s review: Alireza Safaee.
and significantly decrease the power generated by the whole
Digital Object Identifier 10.1109/CJECE.2019.2914723 PV panel. This condition causes a hot spot problem in the
0840-8688 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
174 CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019

shaded area and could accelerate the materials’ degradation

in the affected area. Therefore, a bypass diode is usually
installed to minimize the effect of the shaded cells to the
whole systems’ performance. However, it will change the
current–voltage (I –V ) characteristic of the system so that its
power–voltage (P–V ) curve will exhibit a multiple-peak form
with several local maximum points (LMPs).
The existence of LMPs in the P–V curve will make
the global maximum power point (GMPP) become harder
to attain. A PV system aided with a conventional max- Fig. 1. ODM of PV panel.
imum power point tracking (MPPT) algorithm, such as
Perturb & Observe (P&O) [3], incremental conductance
(INC) [4], Newton quadratic interpolation (NQI) [5], rithm, it still suffers from dilemmatic trading between their
and golden section search (GSS) [6], will be easily trapped accuracy and the tracking time needed to find the GMPP.
at the LMP. As discussed in [7] and [8], conventional MPPT Considering the potential performance of the CS algorithm,
algorithms usually perform a deterministic approach to find a new hybrid MPPT algorithm is proposed in this paper by
the maximum point. Since the partial shading pattern itself combining the CS algorithm with a deterministic approach
cannot be well predicted, those deterministic approaches will called golden section search (GSS). Essentially, the proposed
tend to fail the GMPP tracking. method consists of two parts. In the first part, the CS algo-
Efforts to mitigate partial shading have been classified rithm is deployed just to find the correct tracking area that
into two categories. The first category includes some efforts contains GMPP. Then, by using a properly designed transition
through a hardware fixture approach. This category is focused mechanism, the tracking will be switched to the second part
on hardware system development. Its objective is to modify where the GSS algorithm is employed to find the exact GMPP.
the system’s architecture to adapt the partial shading con- The GSS is a fast and reliable optimization algorithm. This
dition. In this category, some approaches, such as adaptive algorithm allows fast tracking and provides accurate results
reconfiguration [9], distributed MPPT configuration [10], [11], by bracketing the maximum point inside a certain tracking
multilevel dc-link inverter system [12], and power electronics area. The tracking area will be shrunk iteratively by using
equalizer [13], have been introduced and successfully imple- a special ratio called golden ratio until GMPP is located.
mented to minimize a partial shading effect on the PV system. By combining CS and GSS, the proposed method can attain
However, these approaches are not popular because they are the GMPP quickly without compromising the tracking speed.
costly and difficult to be implemented on the existing systems. The remainder of this paper is organized as follows.
In the second category, engineers mitigate the partial Section II will discuss PV modeling and the occurrence of
shading problem by using advanced control strategies. The the multiple-peak problem. In Section III, the conventional
approaches focus on tracking ability development. The objec- CS and GSS algorithm are reviewed. Section IV describes
tive of these approaches is to develop the tracking methods the proposed method, which is a hybrid algorithm of CS and
such that the GMPP can be found even if the P–V curve GSS. Section V will provide the tracking performance for both
possesses a multiple-peak problem. Soft computing techniques simulation and experimental results of the proposed method.
are often involved in this category. Many efforts have been Finally, a conclusion is given in Section VI.
introduced by involving artificial intelligence (AI) principle,
such as artificial neural network (ANN) and fuzzy logic control
II. PV M ODEL AND PARTIAL S HADING P ROBLEM
(FLC), as discussed in [14]–[16]. These approaches can yield a
good tracking performance. However, GMPP is found through The one-diode model (ODM) is the most commonly used
complicated calculation and procedure. Moreover, the accu- model for a PV cell due to its simplicity [22]. The ODM
racy of the results is highly dependent on the training data. represents a PV cell by using a simple circuit, which consists
Alternatively, engineers have employed metaheuristic (MH) of one current source, one diode, and two resistors that are
MPPT algorithms to attain the GMPP in the multiple-peak connected in series (Rs ) and parallel (R p ), as shown in Fig. 1.
P–V curve. MH algorithms can avoid being trapped at LMPs To explain the system behavior under partial shading con-
by utilizing some randomness. Particle swarm optimization ditions, a PV system composed of two identical PV cells,
(PSO) [17], ant colony optimization (ACO) [18], and simu- as shown in Fig. 2, is considered. Cell 1 is assumed to receive
lated annealing (SA) [19], [20] are some of the examples that solar irradiance at the level of 1000 W/m2 (unshaded cell),
have been implemented for MPPT. while cell 2 is assumed to receive solar irradiance only at the
Lately, a new MH algorithm called cuckoo search (CS) was level of 200 W/m2 (shaded cell). The individual V –I and P–V
introduced as an alternative solution to solve the multiple-peak characteristics of the cells are shown in Fig. 3. Prior to bypass
problem in a PV system [21]. This algorithm was first intro- diodes’ installation, cell 2 will be forced to operate in its
duced to solve the multiobjective optimization problem by reverse bias region and cell 1 will be forced to operate at some
mimicking parasitic behavior of cuckoo bird species. This working point such that I1 = I2 = Iout and Vout = V1 + V2 ,
algorithm shows a promising result compared with other as shown in Fig. 4 [23]. In this condition, the current of the
MH-based algorithms. However, just like any other MH algo- unshaded cell will be limited by the shaded cell since they
NUGRAHA et al.: NOVEL MPPT METHOD BASED ON CS ALGORITHM AND GSS ALGORITHM 175

Fig. 4. Partially shaded PV system without bypass diodes.

Fig. 2. Partially shaded PV system.

Fig. 5. Characteristic curve of partially shaded PV system with bypass diode.

(a) V –I characteristic curve. (b) P–V characteristic curve.

Fig. 3. Characteristic curve of PV panel. (a) V –I characteristic curve.

(b) P–V characteristic curve.
III. R EVIEW OF MPPT A LGORITHM
A. Cuckoo Search Algorithm
are in series condition. Consequently, the possible total power The CS algorithm was first proposed by Yang and Deb [24].
output will be significantly decreased. The CS algorithm is an MH method inspired by the behavior of
When bypass diodes are installed, cells 1 and 2 can have a cuckoo bird species in their reproduction process. Cuckoos are
different current level. The parameter values will apply as fol- one of the parasitic organisms which lay their eggs inside the
lows, I1 + I D1 = I2 + I D2 and Vout = V1 +V2 . In the mentioned other birds’ nests instead of building their own nests. Cuckoo
condition, bypass diode 2 will be ON and accommodate the birds will fly randomly from one nest to another nest in order
excess current when the unshaded cell produces current higher to search for the host nest candidate. Then, it will choose the
than the value possessed by the shaded cell. When bypass best nest such that their eggs have the best chance to hatch
diode 2 is ON, V2 will be equal to the forward bias voltage of and generate a new generation of cuckoo. In some conditions,
the diode. When I D2 = 0, the bypass diode 2 will be turned off cuckoo bird will make some efforts to boost the hatching
and I1 = I2 . Due to the occurrence of this switching activity, chance by strategically laying their eggs in a good position
the PV characteristic curve will have two LMPs, as shown and sometimes dropping the host bird eggs outside the nest.
in Fig. 5. The installation of bypass diodes enables the system Some species of cuckoo even evolves to be able to produces
to produce higher power output but makes the maximum similar eggs to other specific bird species. Nevertheless, there
point tracking become harder. Hence, a new MPPT method is still a possibility that the host bird finds the alien eggs and
is proposed in this paper to mitigate the problem. abandons their nest. In this case, the cuckoo’s eggs will fail
176 CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019

to hatch. Based on this natural behavior, the CS optimization

algorithm is developed. For a single objective problem such
as the MPPT problem, there are three idealized characteristic
rules for implementing this method [25]. Those rules are listed
in the following.
1) Each cuckoo bird will only lay one egg in the randomly
chosen host nest.
2) Only the best nest that contains high-quality organism
will carry over the next generation of cuckoo birds.
3) There are a fixed number of host nests in the ecosystem
and the probability of alien eggs discovered by the host
bird has a Pa  [0,1] probability. In case that the host
bird finds the alien eggs, the nest will be destroyed or
abandoned. The host birds then build the new nest in
another place.
From the optimization perspective, cuckoo birds represent
the particles assigned to find the solution, while cuckoo bird
eggs represent the solution for the current iteration process car-
ried out from the optimization problem. In the CS algorithm,
the random step in the tracking process is characterized using
a certain distribution. Many animals, including cuckoo birds,
perform a quasi-random flight or movement in nature. This
behavior shows a typical characteristic of Lévy flight. Lévy
flight means that the step length of the movement or flight
will follow the Lévy distribution, allowing the CS algorithm
to do such a “long jump” among their small flights. This “long
jump” is an important feature, which allows the CS algorithm
to avoid LMP and, in the same time, shortens the tracking
time required to reach GMPP.
The CS algorithm has been implemented for the MPPT
problem in [21]. The flowchart of the CS algorithm for MPPT Fig. 6. Flowchart of the CS algorithm for MPPT.
can be seen in Fig. 6.
In general, Lévy flight is characterized by using the An arbitrarily small number is used as a convergence criterion
following: to ensure the optimum solution.
xji+1 = xji + α ⊕ Lévy (1)
B. Golden Section Search Algorithm
and the operator ⊕ represents the entrywise multiplication
for the multidimensional problem. For MPPT, it can be The GSS algorithm is a method designed for solving an
simplified to optimization problem from a unimodal function. In [6], GSS is
employed to solve a single-peak MPPT problem. GSS locates
V ji+1 = V ji + α · Lévy = V ji + s (2) the maximum point by bracketing a certain tracking area. This
tracking area then iteratively shrinks along the time until the
where
  tracking area is converged and the maximum point is found.
u The tracking process can be explained in Fig. 7. Consider a PV
s≈K· (Vbest − V j ) (3)
|v| β
1
curve represented as an unknown unimodal function of P(V ).
Initially, let the initial tracking area be in the range between
V ji represents the voltage of the j th particle at the i th iteration Vg1 and Vg4 . Maximum point of P(V ) can be located by
cycle, β is the power law index that sets to be 1.5 in this paper, evaluating the fitness values of two points inside the tracking
and K is the step coefficient. u and v follow the distribution: area, Vg2 and Vg3 . If P(Vg3 ) > P(Vg2 ), then the maximum
    point is located somewhere between Vg2 and Vg4 [P(V ), solid
u ≈ N 0, σu2 ; v ≈ N 0, σv2 (4)
 1 line], and the range between Vg2 and Vg4 then will be the
(1 + β) × sin(π + β/2) β new tracking area. Otherwise, the maximum point is located
σu =    β−1  ; σv = 1 (5) somewhere between Vg1 and Vg3 [P(V ), dashed line], and the
 (1+β)
2 ×β× 2 2 range between Vg1 and Vg4 then will be the new tracking area.
and  is the integral gamma function. Next iteration is done by inserting a new point inside the new
Lévy flights are deployed by all particles in each iteration tracking area and then evaluating its fitness. The tracking area
cycle until they find the GMPP. The tracking process will can be shrunk by using the mechanism described previously
be stopped if all the particles converge to one solution. until the tracking area becomes sufficiently narrow such that
NUGRAHA et al.: NOVEL MPPT METHOD BASED ON CS ALGORITHM AND GSS ALGORITHM 177

GSS with probability acceptance level as formulated in the

following:
   
P2 Pave − Pmax
Acc2 = exp = exp (9)
T T
where P2 is defined as the difference between the average
result of all particles deployed and the best result. In the
original SA, P1 represents the difference between the cur-
rent state and the current best result. Since the CS deploys
several particles to search for the GMPP, it is better to
consider every particle in representing the current state of the
system.
Besides defining the transition method, ensuring that the
Fig. 7. Tracking scheme of the GSS algorithm. GSS is able to track the GMPP is another important issue.
In the initial state of the GSS, the system needs to define
the maximum point can be located accurately. To optimize the the tracking area. The GSS can only find the GMPP if it is
tracking process, the inserted point has to follow the following: inside the tracking area. This condition then will be ensured
by evaluating the configuration of the best solution from CS
a+b = b+c (6)
and its adjacent particles in the P–V curve. Assuming that all
a c b+c
= = = ϕ. (7) particles resulted from the CS are close enough to the GMPP,
b b a there are three possible configurations, as shown in Fig. 8,
The optimum ratio ϕ derived by solving (6) and (7) yields to be considered.
to ϕ = 1.61803398. This number is called golden ratio The first configuration is that the best solution is the
and is used to determine the new point inside the tracking rightmost particle, as shown in Fig. 8(a). In this configura-
area. tion, the best solution will only have left neighbor particles.
The second configuration is that the best solution is the
leftmost particle, as shown in Fig. 8(b). In this configuration,
IV. P ROPOSED M ETHOD
the best solution will only have the right neighbor particles.
The main idea of the proposed method is to combine the These first two configurations cannot be used to locate the
CS and GSS algorithms in order to gain the benefits of GMPP by using the GSS algorithm because the GMPP is not
each algorithm. At the starting point, the CS algorithm with located inside the tracking area.
five particles is deployed to find the correct region where The last configuration is that the best solution is located
the GMPP is located. After the correct region is identified, between other particles. The best solution has the left and right
then the GSS algorithm is activated to search for the exact neighbor particles, as shown in Fig. 8(c). In this configuration,
GMPP. The idea is to accelerate the tracking process of the the GMPP must be located inside the area between the left
CS algorithm by switching to the GSS algorithm, which is a and right neighbors, enabling the system to take the range
faster algorithm, after the correct region has been identified. between those two points as the initial tracking area for GSS.
A proper transition mechanism is required to firmly determine The particles from the CS algorithm should be sorted by their
when the system needs to switch from CS to GSS. Therefore, voltage such that
the proposed algorithm adopts the acceptance mechanism from
V1 < V2 < V3 < V4 < V5 .
the SA algorithm [20] to decide whether the system is ready to
switch into the GSS. In the SA algorithm, a directed random Therefore, the best solution cannot be V1 or V5 because
walk is performed in the tracking process. In every step, the left or right neighbor point does not exist. This then
the system always accepts the result if it is better. Nevertheless, becomes an additional requirement between the CS and the
if the result is worse, then the system will only accept it with GSS.
probability equal to the acceptance level as formulated in the Finally, the three transition requirements are summarized as
following: follows.
    1) CS part cannot find any better result [Pmax (i ) 
P1 P(n) − P(n − 1)
Acc1 = exp = exp (8) Pmax (i − 1), and i is the iteration cycle of the CS
T T algorithm].
where P represents the difference between current power 2) Transition acceptance level or (9) is satisfied.
[P(n)] and the previous result [P(n − 1)] and T represents 3) The best solution from CS part has both left and right
the temperature. A similar rule is adopted for designing the neighbors and the best solution is neither V1 nor V5 .
transition mechanism between the CS and GSS, but some As long as the system cannot satisfy all the three require-
modifications on the acceptance level formula are required. ments, it will continue the tracking process by using the CS
In the proposed method, the system needs to maintain the algorithm. In case that the CS converges before all of those
CS algorithm as long as the CS yields a better result; requirements are satisfied, V5 and V1 will be used as a right
otherwise, the system will have a chance to switch into and left limit of the tracking area. This step is necessary to
178 CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019

Fig. 8. Configurations of the particles. (a) Best solution is the rightmost particle. (b) Best solution is the leftmost particle. (c) Best solution is lied between
other particles.

Fig. 9. Flowchart of the proposed method.

ensure the tracking accuracy by using the GSS. Thus, T and V. E VALUATION R ESULT
its decaying constant (K temp ) are set such that the transition
A. Simulation Result
acceptance can be satisfied in sufficient time before the system
converged by the CS part. The flowchart of the proposed The proposed method is first evaluated through MATLAB
method is given in Fig. 9. simulations. Six PV curves with different GMPPs have been
Fig. 10 shows a tracking process example from the proposed generated to validate the proposed method. The validation
method that shows the transition between the CS and the curves consist of one single-peak PV curve and five various
GSS. Fig. 10(a) shows how all the particles approach the multiple-peak curves with varying GMPP locations. The gen-
GMPP. In the initial state, all particles are spread away eral parameters of the validation curves are listed in Table I.
from each other with initial value V j0 . Using the Lévy flight, Fig. 11 shows all the PV curves under study.
the particles start to approach the GMPP. In the end of For comparison, the PSO algorithm and the original CS
the second iteration, some particles are close enough to GMPP method are also used to track the GMPP for all of these curves.
and all the switching requirements are fulfilled. The system All algorithms then will be compared in terms of tracking time
then is switched to GSS with initial tracking area marked and tracking efficiency. To have a fair comparison, all initial
as gray region. This tracking area will be shrunk iteratively values are set at [0.90, 0.75, 0.60, 0.45, 0.30] of the open
until the GMPP is found. The transition between CS and circuit of voltage 220 V in the earlier state of the iteration.
GSS along the tracking process can be seen in Fig. 10(b). These initial values are positioned with equal space across the
By using the transition strategy introduced in this paper, P–V curves.
the MH and deterministic MPPT algorithms can be combined Next step is to select the parameters utilized in the validation
satisfactorily. process. There are three parameters that need to be set in
NUGRAHA et al.: NOVEL MPPT METHOD BASED ON CS ALGORITHM AND GSS ALGORITHM 179

TABLE II
CS T RACKING R ESULT W ITH VARIOUS K VALUES

TABLE III
P ROPOSED M ETHOD R ESULT W ITH VARIOUS Tini VALUES

TABLE IV
Fig. 10. Examples of the proposed method. (a) Tracking process. (b) Tracking PARAMETER U TILIZED
scheme.

TABLE I
VALIDATION C URVES ’ PARAMETER

higher value of K temp is still preferable in the proposed method

to prevent the condition when the CS already converged before
the transition process. Hence, K temp = 0.9 is chosen. The
last parameter that needs to be set is Tini . Similar to K , Tini
is also selected through tentative simulation of the proposed
method involving several values of Tini. Based on the result
that is available in Table III, Tini = 20 is selected for the best
the proposed method, and they are step coefficient of the CS
performance.
(K ), initial temperature (Tini ), and temperature decay constant
The utilized parameter for all the algorithms used in this
(K temp ). Through extensive simulations, the parameters’ values
experiment is summarized in Table IV. Each algorithm has
are chosen as follows: K = 0.85, Tini = 20, and K temp = 0.9.
been run at least 100 times so as to carry out meaningful
Since K is the same step coefficient parameter from the
statistical comparison. Average values and their respective
original CS, it is selected by running the original CS algorithm
standard deviation are used to measure the algorithms’ per-
in simulation to track PV curves. Several values of K from
formance. Tracking results of all the algorithms for each
0.6 to 1 with an interval step of 0.05 have been evaluated.
of the PV curves are summarized in Table V. The over-
The results from various K values can be seen in Table II.
all result shows that all the algorithms perform well and
By assuming that the average changing rate of the shading
reach a very high average accuracy above 99.80% with a
pattern is longer than the average tracking speed, the tracking
relatively small value of standard deviations. Nevertheless,
accuracy of the controller is prioritized in this step over the
the tracking time needed by the proposed method is signif-
tracking time. K = 0.85 is considered as the best option and
icantly lower compared with other MH-based algorithm. This
will be used in both the original CS and the proposed method.
in fact can be seen in every single curve. The statistical
In the proposed method, the transition acceptance criterion
result implies that the proposed method exhibits the best
Acc2 has an increasing trend along the time because of
performance.
decreasing value of P. The CS makes that all particles move
to approach the best solution during the tracking process,
and the difference between the average result and the best B. Experimental Result
result becomes smaller. K temp acts as a compensator to this A similar experimental setup as used in [26] is utilized to
increasing trend and prevents premature transition. However, evaluate the performance of all three algorithms. In this setup,
180 CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019

TABLE V
S IMULATION R ESULT

Fig. 11. Validation curves. (a) Curve 1. (b) Curve 2. (c) Curve 3. (d) Curve 4. (e) Curve 5. (f) Curve 6.
TABLE VI
E XPERIMENTAL R ESULT

PV panels are realized by using a programmable PV emulator

from AMETEK (ETS600X8C-PVF). This emulator then is set
to perform partial shading patterns that exhibit similar V –I
and P–V characteristics utilized in the MATLAB simulation.
The algorithms are executed by a 32-bit digital signal
processor (DSP-TMS320F28035) controller that provides gat-
ing signals to the dc–dc converter. The output terminals of
the dc–dc converter are connected to a resistive load that
is regulated by a programmable dc voltage source (Chroma
620120P-600-8). The voltage level in the load terminal
is kept at 450 V. The schematic of the setup is shown
Fig. 12. Schematic of the experimental setup.
in Fig. 12.
All the algorithms have been run multiples times for each
of the PV curves listed in Fig. 11. The tracking result can other algorithms. It reaches average accuracy at 99.77% with
be seen in Table VI. Similar to the MATLAB simulation 0.28 standard deviation, while the PSO and original CS reach
result, all the algorithms are able to perform good accuracy. average accuracy at 99.76% and at 99.71% with 0.27 and 0.29
The proposed method yields a better result compared with standard deviation.
NUGRAHA et al.: NOVEL MPPT METHOD BASED ON CS ALGORITHM AND GSS ALGORITHM 181

Fig. 13. Experimental result. (a) Curve 1. (b) Curve 2. (c) Curve 3. (d) Curve 4. (e) Curve 5. (f) Curve 6.

A significant tracking time gap between the proposed Fig. 11, the proposed method yields the best performance in
method and the other MH algorithms demonstrates the drastic terms of tracking accuracy and tracking time.
improvement gained by the proposed method. The proposed
method yields average tracking time at 2.95 s with 0.44 VI. C ONCLUSION
standard deviation, while the PSO and original CS at 4.33 In this paper, a new MPPT algorithm is proposed. The
and 3.97 s with 0.38 and 0.24 standard deviation. In the MH proposed method is arranged by combining two different
perspective, usually, there is a tradeoff between accuracy and MPPT algorithms, the CS and GSS. The CS is employed to
tracking time. Fast tracking time tends to lead the solution with avoid being trapped in a local maximum. Then, the tracking
less accuracy and vice versa. It is because the system needs process is switched to the GSS algorithm to find the exact
to experience as many as potential solutions before deciding GMPP. The proposed method can maintain an accurate result
the best solution in an MH approach. while significantly reducing the tracking time. The validation
In the proposed algorithm, the system starts the tracking process shows that the proposed method can cut more than
process by utilizing the MH approach (CS part) to avoid local 25% of the tracking time compared with other methods.
minima but then switches to the deterministic approach (GSS
part), which eliminates random search of MH and is able R EFERENCES
to find the GMPP very quickly. This is the reason why the [1] N. Kannan and D. Vakeesan, “Solar energy for future world: A review,”
proposed result can keep the high accuracy and, in the same Renew. Sustain. Energy Rev., vol. 62, pp. 1092–1105, Sep. 2016.
time, reduce the tracking time required to reach solution. [2] T. Covert, M. Greenstone, and C. R. Knittel, “Will we ever stop
using fossil fuels?” J. Econ. Perspect., vol. 30, no. 1, pp. 38–117,
Fig. 13 shows the best tracking process of various algorithms Feb. 2016. [Online]. Available: http://www.aeaweb.org/articles?id=10.
for the PV curves listed in Fig. 11. As can be seen from 1257/jep.30.1.117
182 CANADIAN JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING, VOL. 42, NO. 3, SUMMER 2019

[3] R. Jiang, Y. Han, and S. Zhang, “Wide-range, high-precision and low- [22] H. A. B. Siddique, P. Xu, and R. W. De Doncker, “Parameter extrac-
complexity MPPT circuit based on perturb and observe algorithm,” tion algorithm for one-diode model of PV panels based on datasheet
Electron. Lett., vol. 53, no. 16, pp. 1141–1142, Aug. 2017. values,” in Proc. Int. Conf. Clean Elect. Power (ICCEP), Alghero, Italy,
[4] G. C. Hsieh, H. I. Hsieh, C. Y. Tsai, and C. H. Wang, “Pho- Jun. 2013, pp. 7–13.
tovoltaic power-increment-aided incremental-conductance MPPT with [23] M. Seyedmahmoudian et al., “State of the art artificial intelligence-based
two-phased tracking,” IEEE Trans. Power Electron., vol. 28, no. 6, MPPT techniques for mitigating partial shading effects on PV systems—
pp. 2895–2911, Jun. 2013. A review,” Renew. Sustain. Energy Rev., vol. 64, pp. 435–455, Oct. 2016.
[5] Y. Liu, L. Chen, L. Chen, H. Xin, and D. Gan, “A Newton quadratic [24] X. Yang and S. Deb, “Cuckoo Search via Lévy flights,” in Proc.
interpolation based control strategy for photovoltaic system,” in Proc. World Congress Nature Biol. Inspired Comput. (NaBIC), Dec. 2009,
Int. Conf. Sustain. Power Gener. Supply, Sep. 2012, pp. 1–6. pp. 210–214.
[6] R. Shao, R. Wei, and L. Chang, “A multi-stage MPPT algorithm for PV [25] X.-S. Yang and S. Deb, “Multiobjective cuckoo search for design
systems based on golden section search method,” in Proc. IEEE Appl. optimization,” Comput. Oper. Res., vol. 40, no. 6, pp. 1616–1624, 2013.
Power Electron. Conf. Expo. (APEC), Mar. 2014, pp. 676–683. [26] D. F. Teshome, C. H. Lee, Y. W. Lin, and K. L. Lian, “A modified firefly
[7] M. S. Ngan and C. W. Tan, “A study of maximum power point tracking algorithm for photovoltaic maximum power point tracking control under
algorithms for stand-alone photovoltaic systems,” in Proc. IEEE Appl. partial shading,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 5,
Power Electron. Colloq. (IAPEC), Apr. 2011, pp. 22–27. no. 2, pp. 661–671, Jun. 2017.
[8] M. A. Danandeh and S. M. G. Mousavi, “Comparative and comprehen-
sive review of maximum power point tracking methods for PV cells,”
Renew. Sustain. Energy Rev., vol. 82, pp. 2743–2767, Feb. 2018.
[9] Y. Liu, Z. Pang, and Z. Cheng, “Research on an adaptive solar
photovoltaic array using shading degree model-based reconfigura- Dimas Aji Nugraha received the B.Sc. degree in
tion algorithm,” in Proc. Chin. Control Decis. Conf., May 2010, electrical engineering from the Institut Teknologi
pp. 2356–2360. Bandung, Bandung, Indonesia, in 2011. He is cur-
[10] N. Femia, G. Lisi, G. Petrone, G. Spagnuolo, and M. Vitelli, “Distributed rently pursuing the M.Sc. degree in electrical engi-
maximum power point tracking of photovoltaic arrays: Novel approach neering with the Institut Teknologi Bandung and
and system analysis,” IEEE Trans. Ind. Electron., vol. 55, no. 7, with the National Taiwan University of Science and
pp. 2610–2621, Jul. 2008. Technology, Taipei, Taiwan.
[11] J. Jiahui, Q. Yanhui, and C. Daolian, “A distributed maximum power He joined PT PLN (Persero), Jakarta, Indonesia,
point tracking flyback type PV grid-connected inverter,” in Proc. 43rd in 2012.
Annu. Conf. IEEE Ind. Electron. Soc., Oct./Nov. 2017, pp. 7713–7717.
[12] I. Abdalla, J. Corda, and L. Zhang, “Multilevel DC-link inverter and
control algorithm to overcome the PV partial shading,” IEEE Trans.
Power Electron., vol. 28, no. 1, pp. 14–18, Jan. 2013.
[13] L. F. L. Villa, T.-P. Ho, J.-C. Crebier, and B. Raison, “A power elec-
tronics equalizer application for partially shaded photovoltaic modules,”
IEEE Trans. Ind. Electron., vol. 60, no. 3, pp. 1179–1190, Mar. 2013. K. L. Lian (SM’14) received the B.A.Sc. (Hons.),
[14] M. M. Algazar, H. Al-Monier, H. A. El-Halim, and M. E. El K. Salem, M.A.Sc., and Ph.D. degrees in electrical engineering
“Maximum power point tracking using fuzzy logic control,” Int. J. Elect. from the University of Toronto, Toronto, ON,
Power Energy Syst., vol. 39, no. 1, pp. 21–28, 2012. Canada, in 2001, 2003, and 2007, respectively.
[15] N. Khaehintung, K. Pramotung, B. Tuvirat, and P. Sirisuk, “RISC- He was a Visiting Research Scientist with the
microcontroller built-in fuzzy logic controller of maximum power Central Research Institute of the Electric Power
point tracking for solar-powered light-flasher applications,” in Proc. Industry, Kanagawa, Japan, from 2007 to 2009.
30th Annu. Conf. IEEE Ind. Electron. Soc., vol. 3, Nov. 2004, He is currently an Associate Professor with
pp. 2673–2678. the National Taiwan University of Science and
[16] K. Huang, W. Li, and X. Huang, “MPPT of solar energy generating Technology, Taipei, Taiwan.
system with fuzzy control and artificial neural network,” in Proc. Int.
Conf. Inf. Technol. Comput. Eng. Manage. Sci., vol. 1, Sep. 2011,
pp. 230–233.
[17] Y.-H. Liu, S.-C. Huang, J.-W. Huang, and W.-C. Liang, “A particle
swarm optimization-based maximum power point tracking algorithm for
PV systems operating under partially shaded conditions,” IEEE Trans.
Energy Convers., vol. 27, no. 4, pp. 1027–1035, Dec. 2012. Suwarno (SM’16) received the B.Sc. and M.Sc.
[18] S. K. Sahoo, M. Balamurugan, S. Anurag, R. Kumar, and V. Priya, degrees in electrical engineering from the
“Maximum power point tracking for PV panels using ant colony Institut Teknologi Bandung, Bandung, Indonesia,
optimization,” in Proc. Innov. Power Adv. Comput. Technol., Apr. 2017, in 1988 and 1991, respectively, and the Ph.D.
pp. 1–4. degree from Nagoya University, Nagoya, Japan,
[19] S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by in 1996.
simulated annealing,” Science, vol. 220, no. 4598, pp. 671–680, 1983. From 2011 to 2015, he was the Dean of the
[20] M. A. Azam, S. Abdullah-Al-Nahid, M. A. Kabir, and School of Electrical Engineering and Informatics,
S. M. H. Chowdhury, “Microcontroller based maximum power tracking Institut Teknologi Bandung, where he is currently
of PV using stimulated annealing algorithm,” in Proc. Int. Conf. Inform. a Professor. He is also the Head of the Power
Electron. Vis. (ICIEV), May 2012, pp. 298–303. Engineering Research Group.
[21] J. Ahmed and Z. Salam, “A maximum power point tracking (MPPT) for Dr. Suwarno is the Editor-in-Chief of the International Journal on
PV system using cuckoo search with partial shading capability,” Appl. Electrical Engineering and Informatics (www.ijeei.org) and an editor for
Energy, vol. 119, pp. 118–130, Apr. 2014. several international journals.
 Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A review on MPPT techniques of PV system under partial shading condition MARK

a,⁎ a a b
Alivarani Mohapatra , Byamakesh Nayak , Priti Das , Kanungo Barada Mohanty
a
School of Electrical Engineering, KIIT University, Bhubaneswar 751024, India
b
Department of Electrical Engineering, National Institute of Technology, Rourkela 769008, India

A R T I C L E I N F O A BS T RAC T

Keywords: This paper presents a concise and an organized review of various maximum power point tracking (MPPT)
Maximum power point tracker (MPPT) algorithms implemented in the photovoltaic (PV) generation system useable under partial shading condition.
Partial shading condition Various algorithms, PV modeling techniques, PV array configurations and controller topologies have been
MPPT optimization technique widely explored till date. But, every technique always has its advantages as well as disadvantages simulta-
neously; as a result, a proper literature review is essential while designing a PV generation system (PGS) under
partial shading condition. In this paper, the detailed review of MPPT algorithms has been done. The review on
MPPT techniques has been classified into mainly four essential groups. The first among them includes all the
new MPPT optimization algorithms, the second group includes the hybrid MPPT algorithms, the third category
includes new modeling approach, and the fourth category includes the various converter topologies. This paper
provides an accessible reference to undertake mass research works in PV systems in the near future under
partial shading condition.

1. Introduction reappraises the various techniques to extract the maximum amount of

power from the shaded PV arrays and is destined to foster more
The economic development of a country depends largely on its researches in global PV-based power systems. In the former review
efficacious electricity supply. In India, the rural electrification program papers in [7] and [8] the authors primarily had put stress on the
started in the year 1950s with the aim of promoting economic conventional MPPT algorithms under both uniform and non-uniform
development and improving the quality of life in rural areas. irradiance. In [9], the author mainly accounted the various converter
Installations of photovoltaic generation systems (PGS) for clean topologies, but a complete realization of all the recent advancements in
electricity generation directly from sunlight can help to upgrade the MPPT for PGS is not being done till date. The conventional MPPT
domestic, healthcare, agriculture, education and the enterprise sectors. techniques and circuit topologies are not included in this paper. The
In this modern era, photovoltaic (PV) technology has the capability to organization of the paper is as follows: Section 2 provides a brief
establish a strong expanse in electrifying every corner of the world. introduction to the PV characteristics under partial shading condition.
In PGS, it is quite essential to extract the maximum amount of Section 3 focuses on new MPPT optimization algorithms under partial
available power from PV panels without getting affected due to the change shading condition [10–33]. Section 4 discusses the hybridization of earlier
in irradiance during all day long. But due to partial shading conditions, MPPT algorithms as discussed in [34–39] and the effectiveness of these
the power output of PV array reduces drastically and thus the efficiency hybrid techniques in tracking global maximum power point (GMPP)
decreases, configuration complexity increases, and cost increases. In the under partial shading condition. Section 5 reviews on the new modeling
case of uniform irradiance, the PV array characteristics curve exhibits only approach of PGS as discussed in [40,41]. Section 6 provides the knowl-
one maximum power point which is being tracked using anyone of the edge on various modern PV circuit topologies as discussed in [42–46] that
long-familiar maximum power point tracking (MPPT) techniques [1–5]. are quite successful in eliminating the problems of partial shading of PV
But PV arrays do not get uniform solar radiation throughout the day. arrays and Section 7 concludes this paper.
During partial shading conditions, the conventional MPPT techniques
predominantly fail as multiple maxima points occur due the presence of 2. PV characteristics in both uniform and non-uniform
the bypass diodes, used to forefend hot spots formation in the PV strings. irradiance condition
To handle the multiple maxima during partial shading conditions, many
modern optimized MPPT techniques are proposed [6]. This paper The output of a PGS is directly affected by the change in solar

⁎
Corresponding author.
E-mail address: [email protected] (A. Mohapatra).

http://dx.doi.org/10.1016/j.rser.2017.05.083
Received 4 December 2016; Received in revised form 21 February 2017; Accepted 17 May 2017
Available online 02 June 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Operation of Unshaded cell

3.1. Grey wolf optimization
shaded cell in

Current I
reverse bias region Operation of unshaded
cell The grey wolf optimization (GWO) is a meta-heuristic approach
strongly inspired by optimizing the attacking technique used by the
String current
grey wolves while hunting. This technique is quite capable of imitating
the leadership hierarchical order and also the hunting proficiency of
grey wolves. There are mainly four types of grey wolves-alpha (α), beta
(β), delta (δ) and omega (ω) which are being employed in order to
Shaded cell simulate the leadership hierarchy properly. In the mathematical model
of this bio-inspired technique, the fittest solution is assumed to be α.
Then, β and δ are considered to be the second and third best solution,
Breakdown Bias voltage
Voltage V ω is denoted as the rest of the candidate solutions. There are mainly
voltage
three steps for GWO, such as hunting, chasing and tracking of the prey
Fig. 1. Current -Voltage characteristic of PV cell during reverse biased region. by forming a group, then encircling the prey and then finally attacking
the prey. This overall hunting mechanism is implemented while
irradiance and also the change in temperature. When the PV strings designing the GWO for executing optimization problems in MPPT for
receive uniform insolation from the sun, the power-voltage (P-V) curve PV modules. The hunting technique of the grey wolves is guided by α
shows a unique peak. But when PV array is exhibited to partial shading, clans which are termed as leaders and are followed by the β clans. The
it shows multiple peaks on the P-V curve among them there is one main duty of the δ and ω is to take care of all the wolves that are
global maximum power point (GMPP) and others are called as local wounded in the entire pack. The GWO in [10] is being combined with
maxima power points (LMPPs). the direct duty cycle control (DCC) to keep duty cycle constant at MPP
It becomes quite difficult to choose the GMPP from the LMPPs. to reduce steady state oscillations. The flowchart of GWO algorithm is
This is because a snippet of the PV array is able to receive uniform shown in Fig. 4. The total hunting mechanism of grey wolves described
irradiance and operates at the optimum efficiency at the partial here is modeled by the help of the following equations:
shading. The shaded cells (which receive less irradiance or no ⎯→
⎯ ⎯⎯⎯→ ⎯⎯⎯→ ⎯⎯⎯→
irradiance) operate along with a reverse biased voltage in order to give E = C. XP (t ) − XP (t ) (2)
the same current as given by the unshaded cells because a constant ⎯→
⎯ ⎯⎯⎯→ ⎯→
⎯ ⎯→
⎯
amount of current should flow in every module in the series config- X (t + 1) = XP (t ) − F . E (3)
uration of PV modules. The insolation level is proportional to short Here t denotes the current iteration; E, F and C represent the
circuit current of PV cell. Fig. 1 shows how the shaded cells are coefficient vectors. Xp represent the position vector of the hunting prey
operating in reverse biased voltage region for providing the same and X denotes the position vector for the Grey wolf. The vector F and C
current as provided by the unshaded cells. Fig. 2 shows the conduction are computed as follows:
of bypass diode takes place under shading condition when Eq. (1) is ⎯→
⎯
satisfied. F = 2→
a. →
r1 − →
a (4)
n →
V2 − ∑ Vi ≥ VDO, i≠2 C = 2.→
r2 (5)
i =1 (1) Where a decreases linearly from 2 to 0 and → r1 and →
r2 vector values in
Where VDO is the forward voltage drop of the diode [6]. [0,1]. At MPP the duty cycle is normalized at a fixed value to reduce the
The bypass diodes as shown in Fig. 3(b) are used to provide an steady-state oscillation and power loss that exists in conventional
alternate path to the current flow if the partial shading condition occurs MPPT optimization algorithms. In order to implement the GWO
in the PV array. The P-V curve shown in Fig. 3(c) depicts the multiple MPPT, the duty cycle d is considered as the grey wolf. Therefore Eq.
maxima during partial shading condition. As the conventional MPPT (3) modified as Eq. (6) and the GWO fitness function is calculated as
optimization algorithms fail to differentiate between the GMPP and the Eq. (7).
LMPPs, so many new modern MPPT optimization algorithms are di (k + 1) = di (k ) − F . E (6)
developed using the evolutionary algorithms, differential algorithms,
artificial neural networks, artificial intelligence techniques, new topol- P (dik ) > P (dik−1) (7)
ogy of converters, new reconfiguration of PV modules, and new PV Here P denotes power, d as duty cycle, i is the number of the current
modeling techniques. individual grey wolves, k represents the iteration count. The major
advantages of GWO technique are higher tracking efficiency with
3. Maximum power point tracking optimization algorithms elimination of transient and steady state oscillations.
under partial shading condition
3.2. Firefly algorithm with an updated β coefficient
Various MPPT optimization techniques are being addressed in a The author in [12] has proposed a simplified firefly algorithm (SFA)
stochastic order in this section. with an updated β coefficient which is mostly used in order to track
MPPT of a PV system under the condition of partial shading. The SFA
algorithm is a modification of the firefly algorithm (FA) [11]. In FA, the
first position of the firefly is randomized with the use of variables γ (the
light absorption coefficient) and α (random coefficient) while in the
SFA the initial position of the firefly is selected in between 0 and 1 and
the above two variables are hence not required. The optimization
equation of SFA is represented as:
Xit +1 = Xit + β (Xj − Xi ) (8)
where Xi and Xj signifies the position of i (less brighter firefly) and j
Fig. 2. Bypass Diode conduction when one cell is shaded. (more brighter firefly) and β is the firefly attractiveness factor. In

855
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 3. A single PV string showing (a) uniform irradiance (b) non-uniform irradiance patterns (c) the P-V peaks consisting of GMPP and LMPPs.

MPPT, the objective function of SFA is to get the desired photovoltaic

output power and the firefly position represents the duty cycle D. In the
paper [12], the value of β coefficient is updated in every iteration for
faster convergence and accurate result.

3.3. Ant colony optimization

The ant colony optimization (ACO) is a probabilistic algorithm as

discussed in [13,14] which helps to find out the optimized output based
on the food searching behavior of the ants. This method is a modified
form of PSO method. After sensing the power from the PV module, it
provides the optimized duty cycle which is further being applied to the
converter. ACO is able to reduce the number of LMPPs of the I-V curve
and is used in both centralized type as well as distributed type MPPT
controllers.

3.4. Artificial bee colony based algorithm

Artificial bee colony (ABC) algorithms is mainly a bio-inspired

method discussed in [15,16] which is simple, uses very few controlled
parameters and the algorithm convergence criteria are not dependent
on initial conditions of the system. It is a swarm based meta-heuristic
algorithm capable of solving multidimensional as well as multimodal
optimization problems very easily. The artificial bees, classified mainly
into three important groups- they are the employed bees, the onlooker
bees and last the scouts. The bee which currently searches the food or
exploits the food production source is called as an employed bee, a bee
that waits in the hive to make decisions to choose a food source is called
as an onlooker and the scout bee is used to carry the random search for
a new food source. All three groups work together by communication
and coordination to get the optimal solution in lesser time. Here duty
cycle is the food position and maximum power as the food source of
ABC algorithm. The ABC optimization algorithm as shown in Fig. 5 is
divided into four phases as discussed in [16].
For implementing ABC in MPPT for PV system, the duty cycle for
the DC to DC converter is calculated as follows
de = d min + rand [0, 1](d max − d min ) (9)

newde = de + ϕe (de − dk ) (10)

where de (current duty cycle), d min (minimum value of duty cycle), d max
(maximum value of duty cycle), ϕe (is a constant) between [−1,1] and dk
(previous duty cycle). The ABC tracks MPP with good accuracy and
efficiency under partial shading conditions.

3.5. Deterministic particle swarm optimization

The deterministic particle swarm optimization (DPSO) is a mod-

ified PSO algorithm which has better tracking capability than conven-
tional PSO. In conventional PSO [47–49], if a change in the duty cycle
Fig. 4. Flowchart of Grey wolf optimization technique. for two successive iteration count is low, then the subsequent change in

856
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 6. Extremum seeking controller.

modulated load resistor (DMLR) along with a hysteretic modified self-

control circuit. The boost converter is used to push the panel operating
point to a load operating point. The improved curve tracer as proposed
in [19] has mainly two operating modes, the first one is the fixed
resistor (FR) and the second is the modulated resistor (MR) mode. In
the FR mode, the curve tracer acts like a conventional boost V-I curve
tracer whereas in the MR modes, DMLR modulates equivalent
resistance of the load switch. The improved curve tracer has a simple
structure; has low implementation cost and it has no tracing limitation.

3.7. Improved extremum-seeking based MPPT

The extremum-seeking control (ESC) as discussed in [20–23]

mainly consists of a control strategy applied to the nonlinear system
that has local minimum and maximum points. The algorithm as shown
in Fig. 6 mainly employs the injection of a small perturbing signal
called as the dither signal (α sin(ωt )). The proposed ESC-based algo-
Fig. 5. Flowchart of artificial bee colony optimization.
rithm operates in a cyclic process to iterate the array voltage according
to the count of the maximum possible power peaks. The main theory
duty cycle results in more iteration number to reach the final solution. behind the proposed method is to find the GMPP by comparing the
Also, in PSO, if a particle is away from the best position then a large powers as well as the power gradients in every segment after the end of
change in velocity is required which may make the particle to escape each of the cycle. As a result, this method is used to identify a non-
from the GMPP. The DPSO proposed in [17,18] makes the optimiza- global MPP segment; the process continues till the GMPP is calculated.
tion simpler by removing the number of acceleration factor in velocity
equation and limiting the velocity factor according to the distance 3.8. Simulated annealing method
between two peaks. The algorithm operates usually in two modes, the
local and the global mode. The global mode gets activated when the The simulated annealing (SA) optimization method follows the
partial shading occurs. At global mode, the algorithm shifts to DPSO process of metal annealing. For finding the main GMPP of a given PV
subroutine and during local mode variable step size perturbation is system, SA discussed in [24] uses parameters such as an initial
used. For global mode, the range of duty cycle is calculated as follows: temperature, the final temperature and the nominal cooling rate. For
each temperature, the algorithm does several perturbations at the
ηbbRL min operating voltage and measure the corresponding energy. The energy
d min =
RPV max + ηbbRL min (11) that is measured is compared with current reference energy. If the new
operating point has greater energy, then it will be accepted as the new
ηbbRL max operating point. If the new operating point has less energy than the
d max =
RPV min + ηbbRL max (12) reference operating point, then it may still be accepted depending upon
the acceptance probability Pr as follows:
where d min represents the minimum duty cycle, d max represents the
maximum duty cycle, ηbb represents the converter's efficiency, RL min as ⎡ P − Pi ⎤
Pr = exp ⎢ k ⎥
well as RL max are the minimum and maximum values for the load ⎣ Tk ⎦ (13)
connected at output, RPV min as well as RPV max represents the reflective
where, Pk denotes the power at the recent voltage, Pi denotes the power
impedances of the PV array. The DPSO has a greater advantage than
at the earlier operating point and Tk denotes the current system
normal PSO as the number of iteration is reduced, tuning effort is
temperature. The SA algorithm comprises of a cooling schedule of
reduced and the problem of random search is also minimized.
either static or adaptive type. The geometric cooling schedule is given
as
3.6. Improved curve tracer
Tk = αTk−1 (14)

Curve Tracer is used mainly in the MPPT techniques which are where Tk denotes the temperature at kth step, Tk−1 denotes the
software based as discussed in [19]. It consists of two stages; one is temperature at (k − 1)th step and α (cooling rate) denotes a constant
mainly the boost converter's stage, and the other stage contains a duty- (α < 1). The flow chart of SA method is shown in Fig. 7.

857
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 7. Flowchart of the simulated annealing (SA) method.

3.9. Variable step Newton-Raphson (VSNR) method through model 3.11. Optimal P & O control technique for MPPT based on least square
predictive control support vector machines algorithm

The variable step newton-raphson (N-R) method uses the model This control strategy as discussed in [29] combines the perturb and
predictive control (MPC) which is mainly an alternative to the classical observe (P & O) method with the least square support vector machines
control methods. It has many advantages like fast dynamic response, algorithm (LS-SVM). The LS-SVM method is employed for perturba-
simple concept, easy implementation, have high speed, good reliability tion of voltage when the irradiation is changed. The major merit of LS-
and also avoids unacceptable oscillations. In the proposed scheme in SVM identification technique in the PV application is that nonlinea-
[25], MPC is applied using a boost converter. The main purpose of the rities are not considered as shown in Fig. 8. As the MPP changes due to
MPC is to estimate the future behavior of the controlled variables to the change in illumination, the output voltage starts increasing slowly
minimize the cost function. The cost function is being evaluated twice by a proper voltage step size δV at every period of P & O control. When
for each switching states. It is being proved that despite the variations δV is gradually small due to the minimized oscillations, the conver-
in output voltage, the proposed model predictive control circuit with gence rate is reduced to get the optimal result.
boost converter is successful in regulating the load current.

3.12. Dynamic population size differential evolution (DynNP-DE)

algorithm

3.10. Variable step size perturb and observe method In this technique as discussed in [30] at the beginning higher size of
population is being needed to make thorough evaluation of the
It is a modified form of the conventional perturb and observe (P & function's landscape. The entire population of a community is con-
O) method as discussed in [26]. At the start of the algorithm the sidered at first and later on by dividing the population into multiple
reference voltage (Vref ) is set at about 0.80*Voc , where Voc is the open groups the best individuals among all is being sorted. The fittest
circuit voltage of the array. It has two mode of operation i.e voltage candidate among all iterations (current and previous) will survive for
search mode and MPP search mode. Voltage search mode bring the next iteration. To make the optimization simpler, the population size
operating point near the reference voltage and MPP search mode do for next generation is mainly reduced to half of the initial size. In
the fine tuning near MPP. The peak is tracked first by applying MPPT, the control parameter (voltage or the duty cycle) is being
modified P & O algorithm. If there is an indication of partial shading randomly generated in between specified upper and lower limit. The
GMPP tracking is performed. In this method, unnecessary tracking of size of the population is then reduced for doing the mutation, crossover
GMPP is avoided. The proposed algorithm takes shorter time to track and selection process. The DynNP- DE has good accuracy in searching
the MPP as compared to [3,46,50]. global MPP with higher convergence speed and gives the optimal
result.

858
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 8. Flowchart of optimal perturb and observe method.

3.13. Chaotic search

The chaos word can be defined easily in a simple term as

randomness that occurs due to the sensible nature of system for a
small change in initial conditions. Hence, this predefined feature in
[31] is used to search for the optimization of MPPT in PGS. In PGS,
voltage is the optimization variable and power is the fitness function.
There are two methods for optimization, single carrier and the multiple
carriers. Single carrier search is always not feasible as it takes a much
longer time to converge to the MPP while the multiple carriers (dual-
carrier) is able to improve the precision, efficiency and also the system
robustness. By consecutive iterations, the searching zone decreases and
stops when it reaches the threshold value and provide a faster
convergence.

3.14. Cuckoo search

The cuckoo search (CS) method in [32,33] is a bio-inspired parasitic

reproduction scheme of the cuckoo birds. In CS, searching steps for a nest
is characterized by levy flight mechanism. A levy flight is mainly random
walk from which the step sizes are being extracted for levy distributions.
Due to levy's flight mechanism, step sizes for CS are relatively larger than
the normal PSO. It accelerates faster convergence. As the particles move
nearer to the MPP, the step size consecutively gets smaller and finally
reduced to zero. The flow chart for CS is shown in Fig. 9.

4. Hybridization of conventional MPPT algorithms

Fig. 9. Flowchart of cuckoo search method. This section provides the importance of hybrid MPPT algorithms in
comparison to conventional MPPT algorithms for MPP tracking under
partial shading condition.

859
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 10. Flowchart of hybrid GWO and P & O algorithm.

4.1. Hybrid GWO and P & O MPPT algorithm 4.2. Hybrid PSO-PI based MPPT algorithm using adaptive sampling
time strategy
The hybrid MPPT as discussed in [34] combines GWO with P & O to
extract maximum power from the PV array efficiently under partial The hybrid PSO-PI MPPT controller as discussed in [35] uses
shading condition. The GWO technique [10] has been hybridized with particle swarm optimization(PSO) approach to locate the global peak
the conventional P & O method in which GWO works in the initial (GP). Once, the GP is traced, the algorithm is switched to the
phase of MPPT and later on P & O operates for faster convergence to proportional-integral(PI) mode to track the slow change in the GP
GMPP. By doing this the search space of GWO has been reduced and location. An adaptive sampling time strategy (ASTS) is applied to
computational burden decreases. In this method, the position of wolves accelerate the convergence towards the GP. The hybrid PSO-PI
denotes the duty cycle of the converter used. The use of PI controller in improves the performance of the standalone PSO controller. For the
the MPPT implementation is completely eliminated. The major ad- PSO controller, the objective function f is chosen to be output power of
vantage of this method when compared to the normal GWO and P & O PV array, pibest is the best power, the equation is given by-
method is that it has higher tracking capability, faster convergence
speed and higher efficiency. The flowchart of hybrid GWO and P & O f (xik ) > f ( pibest ) (15)
algorithm is shown in Fig. 10.
An initial vector x1 of four agents [V1, V2, V3, V4] is defined-

860
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

x1 = [V1, V2, V3, V4] (16) Ppv, new − Ppv, last

> threshold
Ppv, last (20)
These agents are applied successively as a reference voltage to the
converter. At GP, the derivative of P (t ) with respect to V (t ) becomes where Ppv, new is the new generated power of PV module and Ppv, last is the
zero. Hence, a control variable e (t ) is defined as: last generated power of PV module. The threshold is a small value
dP (t ) nearly 2%. The optimization technique takes a constant threshold value
e (t ) = irrespective of any change in irradiation.
dV (t ) (17)

Since, e (t ) disappears at the MPP, the objective of the PI controller 4.4. P & O combined with PSO
is to nullify the slope e (t )
The hybridization of P & O along with PSO is proposed in [37,38].
Vmp (t ) = Kp. e (t ) + KI . ∫ e (t ) dt (18) At the beginning of the algorithm PSO is used for global search and
then P & O is used at the final stage. PSO method as discussed in [51–
The inequality (19) must be satisfied to re-initialize the search
54] is employed for searching the GMPP. The hybrid method finds the
process.
GMPP in a shorter time than normal PSO method. In [37] different
⎧
⎪
Pi +1 − Pi
≥ 10% shading cases are being tested to verify the efficacy of the proposed
⎨ Pi method. The boost converter with interleaved topology is used to
⎪
⎩ Vi +1 − Vi ≤ 0.5V (19) reduce ripple current, improve reliability and increase the efficiency.
Where i+1 is the actual value and i is the previous value. The proposed method tracks the GMP easily and has faster conver-
In the hybrid PSO-PI method, the tracking speed is increased and gence time and also better dynamic response than normal PSO method.
the tracking error is greatly reduced.
4.5. Hybrid DEPSO method
4.3. Hybridized simulated annealing and perturb and observe
The DEPSO is a hybridization of differential evolutionary (DE)
method (P & O)
algorithm and PSO. The DEPSO algorithm as discussed in [39] is able
to avoid local optima by combining DE operator with PSO. As the
The hybridized simulated annealing and perturb and observe
insolation changes, there is sharp fluctuation in power with respect to
method takes the advantages of both SA and P & O as proposed in
the location of qth particle. The condition given in Eq. (21) must be
[36]. Since, the conventional P & O method is incapable to locate the
satisfied for the reinitialization of the algorithm. The condition shows
global maxima and SA method is unsuccessful to perform the
the minimum variation in output power to run the algorithm to find
continuous search, so both of them are combined in the hybrid
new MPP.
MPPT method. SA is used for global search and P & O is used for local
search. As a result, the performance of tracking is easily improved as J (Xq +1) − J (Xq )
compared with the independent working of each algorithm as shown in > ΔP
J (X q ) (21)
Fig. 11. When the P & O method is tracking around MPP, only a small
change in power should occur for each small step. The equation is given Where J (Xq ) is the output power of PV panel. The proposed DEPSO is
by- successful in distinguishing the GMPP from LMPPs during mismatch-

Fig. 11. Flowchart of hybrid SA and P & O algorithm.

861
 A. Mohapatra et al.

Table 1
Comparison of various optimization techniques applied for MPPT.

Name of technique and reference Year of publication Control strategy Input parameters Output parameters Cost Applications (Stand-alone/ Converter used (DC/DC or
number Grid) (DC/AC)

Grey wolf Optimization Technique [10] 2016 Bio-Inspired Vpv, Ipv Duty Cycle High Stand-alone DC-DC
Evolutionary Algorithm
β -Firefly Algorithm [11] 2014 Bio-Inspired Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
Evolutionary Algorithm
Ant Colony Optimization [14] 2014 Probabilistic Algorithm Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
Artificial Bee Colony Optimization [16] 2015 Bio-Inspired Evolutionary Algorithm Vpv Duty Cycle High Stand-alone DC-DC
Deterministic Particle Swarm Optimization 2014 Modified PSO Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
[18]
Improved Curve Tracer [19] 2014 Software Based Vpv, Ipv Duty Cycle High Stand-alone DC-DC
Improved Extremum Seeking Control [22] 2014 Cyclic Algorithm Vpv, Ppv Vout High Both Both
Simulated Annealing Algorithm [24] 2016 Metal Annealing Technique Tk Pmax High Stand-alone DC-DC

862
Variable-Step Newton-Raphson Method 2013 Modified P & O Vpv, Ipv Pmax Low Stand-alone DC-DC
[25]
Variable Step Perturb and observe method 2016 Modified P & O Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
[26]
Optimal Perturb and Observe method [29] 2015 Modified P & O Vpv Vout Low Stand-alone DC-DC
Dynamic Population size Differential 2014 Differential Algorithm Vpv Pmax Low Stand-alone DC-DC
Evolution [30]
Chaotic Search [31] 2011 Bio-Inspired Evolutionary Algorithm Vpv, Ipv Duty Cycle High Stand-alone DC-DC
Cuckoo Search [32,33] 2013 Bio-Inspired Evolutionary Algorithm Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
Hybrid GWO and P & O [34] 2016 Bio-Inspired Computational Vpv, Ipv Duty Cycle High Stand-alone DC-DC
Algorithm
Hybrid PSO and PI method [35] 2015 Adaptive Sampling Time Strategy Vpv, Ipv Duty Cycle High Stand-alone DC-DC
Hybrid SA and P & O method [36] 2015 Metal Annealing Technique Vpv, Tk Pmax High Stand-alone DC-DC
Hybrid PSO and P & O [37,38] 2015 Evolutionary Algorithm Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
Hybrid DEPSO method [39] 2015 Hybrid Evolutionary Algorithm Vpv, Ipv Duty Cycle Low Stand-alone DC-DC
Renewable and Sustainable Energy Reviews 80 (2017) 854–867
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Table 2
Summary of merits and demerits of various MPPT methods.

Name of technique Merits Demerits

Grey wolf Optimization Higher tracking efficiency, no transient and steady state oscillations, robust, Computational complexity, Large search space, high cost
Technique fewer parameters needed for adjustment
β -Firefly Algorithm Faster convergence, accurate results, high tracking efficiency, never fall on Beta coefficient is updated in every iterations which is difficult
LMPP to do, give poor result than other swam based algorithm
Ant Colony Optimization Convergence does not depend upon the initial position of the sample, Simple Four evolutionary parameters required to be optimized at once,
control, low cost, robust to various shading pattern which is difficult to do simultaneously, complex calculation
Artificial Bee Colony Simple, uses fewer control parameters, convergence is independent of initial Slow tracking, complex, May fall on LMPP because of fewer
Optimization conditions control parameters
Deterministic Particle Swarm Convergence speed and accuracy is improved, Excessive amount of calculation,computationally complex,
Optimization Removes the random number in the accelerations factor of the dependent on initial conditions
conventional PSO velocity equation. MPPT structure is simple compared to
conventional PSO. Efficiency 99.5%
Improved Curve Tracer Simple structure, implementation cost is low, no tracing limitation near Voc May fall at LMPP
Improved Extremum Seeking It does not require a system model, higher overall efficiency, Robust, Need large number of samples before finding all the peaks. Thus
Control implementation is inexpensive tracking speed is low with steady state oscillations.
Simulated Annealing Converge accurately to GMPP, fewer parameters required Computational complexity, oscillation at MPP, reinitialization
Algorithm required with change in weather condition
Variable-Step Newton- Good dynamic response, flexible, less fluctuation around MPP Complex calculation
Raphson Method
Variable Step Perturb and Quickly tracks global peak among local peaks Two stage tracking, oscillation around MPP
observe method
Optimal Perturb and Observe Reduced oscillation, faster convergence, LS-SVM constructed offline with Complex control structure, May fall at LMPP
method reduced training data
Dynamic Population size Small population size, Faster convergence Complex calculation, May fall on LMPP because of reduction of
Differential Evolution population size dynamically
Chaotic search Search efficiency increases for multiple MPP Track maximum power quickly Complex algorithm. Tracking speed depend upon chosen step
and accurately under sudden change in insolation. Robust and reliable size.
Cuckoo Search Efficient randomization, convergence speed is very high, robust and fewer Complex calculation, tracking time depend upon levy flight,
tuning parameter required. High efficiency deterioration of convergence speed and accuracy.
Hybrid GWO and P & O Superior tracking performance, reduced oscillation, reduction of search Difficult control structure, costly, difficult to implement
space, computational overhead decreases, tracking efficiency nearly 99.9%
Hybrid PSO and PI method Tracking speed increase, tracking error reduced Oscillations around MPP, difficult control structure, costly,
proper tuning of KP and KI required
Hybrid SA and P & O method SA is used for global search and P & O used for local search, tracking Reinitialization required with change in weather condition,
performance improved, fast tracking difficult to optimize the parameters of SA, suitable threshold
detection is difficult
Hybrid PSO and P & O Search space is reduced, faster convergence to GMPP, reduced oscillations in Complex control structure, hardware implementation is costly,
output power, voltage and current Convergence cannot be guaranteed if GMPP located outside the
search zone
Hybrid DEPSO method Reliable, system independent, accurate tracking under partial shading Complex calculation, large number of parameters need to be
condition with high tracking speed selected for optimization

ing conditions. The DEPSO is system independent MPPT technique V = (Iph + Is ) Rsh − (Rs + Rsh ) I − a × Lambert (W )
and presence of random numbers helps this algorithm to keep its meta- ⎛ Rsh .(Iph + Is − I ) ⎞
Rsh *Is ⎜⎝ a ⎟
heuristic approach and also find the GMPP in any partially shaded W= e ⎠
a (22)
condition.
As discussed above, comparison of various optimization techniques
where Iph , Is , Rs , Rsh and Ns are photovoltaic current, saturation current,
applied for MPPT of PV system under partial shading condition is given
series resistance, shunt resistance and the number of series PV cells in
in Table 1. Merits and demerits of different optimization MPPT
PV module. The method is successful in determining the power peaks
techniques are presented in Table 2.
without doing simulation of entire power curve and is able to save
computational time.

5. New PV modeling approach under partial shading

condition
5.2. Sub-module integrated converter based PV system

5.1. Fast power peaks estimator during partially shaded PV systems

The sub-module integrated converter based PV system is proposed
in [41] to reduce power loss by reducing the effect of I-V mismatch
The author in [40] proposed a new modeling method for PV array
among the modules comprising the PV array. PV modules as proposed
under a partially shaded condition to track the MPP. Due to the
in [62–66] is based on ideal single diode model. In the proposed
complexity and more time consuming problems, no model based MPPT
technique in [41] a comprehensive control strategy is developed to
was developed earlier for partially shaded PV system. The homoge-
coordinate the control of distributed MPPT (DMPPT), PV sub-module
neous irradiance PV systems discussed in [55–61] represent the
voltage regulation (PSVR) and dc-link voltage regulation (DCLVR)
modeling of PV system. The proposed modeling approach as discussed
under partial shading conditions. Taking into account several case
in [40] rely on three governing rules for the identification of power
studies, the effectiveness of the proposed model is verified and
peaks in partially shaded PV system. The Lambert model as discussed
simulated. The author has taken 42 sub-modules to verify the efficiency
in [40] is based on the following equation:
of system under partial shade and various PV mismatch conditions.

863
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

Fig. 13. Topology of multilevel PV inverter.

compromising the output AC voltage distortion. The topology as

discussed in [74–77] i.e. three-level neutral point clamped (NPC)
inverter comprising of two PV groups has been used in [44]. The
topology of multilevel PV inverter system is shown in Fig. 13. For
controlling each dc unit separately, a dc unit voltage feedback control
strategy which is based on the zero-sequence injection is implemented.
For high non-uniform irradiance condition, system stability and power
generation get affected. Improvisations of algorithms as well as new
circuit topologies are required for extending the utilization of MLI in
PGS.
Fig. 12. Topology of CHB photovoltaic system.
6.4. Shunt-series compensation
6. Modern PV circuit topologies
The use of dc to dc converter discussed in [78–82] is called as
6.1. Cascaded H-Bridge (CHB) photovoltaic system DMPPT is the most efficient MPPT extraction method. In the shunt-
series compensation technique as discussed in [45], a current-compen-
The modularity of CHB converter [67–73] has been used to sated converter is connected in shunt with each PV string. Also, a
enhance the MPP tracking performance of PV system. The author in voltage-compensating converter as discussed in [71] is connected in
[42] has used a 7-level CHB converter for a 3.3 kW peak solar PV series along with each PV string. This makes each PV module to operate
system for MPP tracking. The effectiveness of the proposed MPPT at exact MPP and deliver maximum power. A combined shunt series
scheme is compared under different scenarios. The advantages of this compensating technique is proposed in [45], for operating each module
method are that tracking speed is faster with good steady state at MPP. The compensation of the voltage difference between the strings
performance. As no extra sensors are required, hence model complexity is called as "compensating power dedicated converter". For the unified
is low. The major advantage is that during partial shading or with shunt and series compensating nature of the converter, in the proposed
change in temperature, the control decision is only made by present scheme [45], the compensated voltage is produced by compensating
quantities and do not depend on previous quantities as a result the converter. When the shunt-series compensation are unified together, it
misled of MPPT does not occur. The topology of CHB photovoltaic makes each PV module to be able to operate at exact MPP and delivers
system is shown in Fig. 12. maximum power to the given load but fails during unlikely shade
condition.
6.2. DC-DC converter topology with a direct control method
6.5. Variable interleaving cascaded DC-DC converter
The DC-DC push-pull converter with direct control MPP algorithm
is proposed in [43]. It can extract power directly from the MPP In the variable interleaving algorithm, as proposed in [46], the DC
algorithm easily and is able to track the MPPs very accurately with link voltage is used. Under partial shading condition, the first harmonic
change in irradiance. For maintaining MPP, the converter control frequency component remains in the DC link while applying the fixed
requires two loops. The push-pull converter proposed in [43] requires interleaving technique. But in the variable interleaving technique, the
less components and uses the direct control scheme. From the first harmonic component in the DC link voltage can be completely
experimental results, it is being concluded that the proposed control eliminated. Due to the presence of interleaving technique, it is able to
system is capable of tracking available PV panel output power for every reduce the components of the cascaded DC-DC converter. The effi-
time, hence reduces the power loss and system cost. ciency of this converter is near about 99% which is much more than
discussed in [44,46,83–88] and due to small size of the converter the
6.3. Multi-level PV inverter with PV groups for independent MPPT system is light and low cost. The topology of variable interleaving
control cascaded DC-DC converter is shown in Fig. 14. Different converter
topologies with their efficiency are given in Table 3.
As discussed in [44], series-connected PV modules are divided into
multiple segments. As a result, each segment became the input dc 7. Challenges and further studies
source for each unit of the multi-level inverter (MLI). The voltage of
every DC unit in the MLI is easily adjustable independently without The primary challenge in tracking maximum power from partially

864
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

invasive weed optimization (IWO), gravitational search algorithm

(GSA), biogeography-based optimization (BBO), grenade explosion
method (GEM), seeker optimization algorithm (SOA), krill herd
optimization (KHO), water cycle algorithm (WCA), harmony search
algorithm (HS), and evolution strategy may be used for better tracking
and recommended for future research. Under partial shading condition
evolutionary algorithm (EA) plays a promising role. However, there is
scope for improving EA, by selecting appropriate control parameters
and narrowing down the search space, so that computational time and
complexity can be reduced. Control structure should be simple and
inexpensive so that it can be designed with ease. A number of hybrid
techniques should come up, which can take the advantages of both
conventional and soft computing MPPT techniques. Different meta-
heuristic optimization techniques are available for MPP tracking for
different PV array rating, size and different shading conditions. So it is
very difficult to categorize them by common benchmarked indices.
However, it is still possible to classify them according to their tracking
speed, algorithm complexity, cost, hardware implementation, steady
state oscillation and their applications. To reduce the cost, complexity
in hardware implementation and measurement error, use of current
sensorless MPPT is encouraged. Finding simpler, faster and cheaper
Fig. 14. Topology of variable interleaving cascaded DC-DC converter. MPP tracker with better efficiency and reliability is recommended.

Table 3 8. Conclusion
Comparison of various converter topologies.

Converter topology with reference number Efficiency This review article provides brief descriptions of all the modern
MPPT algorithms those are being used in software and hardware
Buck Converter [83] 92% platform. It deals with the MPPT optimization techniques those are
Boost Converter [84] 97–98.2% mainly focused on partial shading conditions of PGS. This review has
Buck-Boost Converter [85] 99.5%
Cuk Converter [86] > 96%
included many recent hybrid techniques apart from the new MPPT
SEPIC Converter [87] > 90% algorithms. PV modeling approach under partial shading condition is
Flyback Converter [88] > 92% discussed for better tracking with easy identification of power peaks.
Variable Interleaving Cascaded DC-DC Converter [46] 99% Many modern PV circuit topologies considered to enhance the MPP
Multilevel PV Converter [44] > 90%
tracking performance of the PV system. Merits and demerits of
different optimization techniques are discussed to choose a suitable
shaded PV array is that it's accurate mathematical model under MPPT under partial shading condition. From the various methods
shading condition is not available. It is tough to predict the global discussed in this paper it is very difficult to conclude which method is
peak beforehand under partial shading condition. PV module para- the better one. The choice of MPPT depends upon the application,
meters changes when insolation and temperature vary, which is not hardware availability, cost, convergence time, accuracy and reliability
considered while modeling it to avoid the complexity of the model. of the system. Considering the importance of MPPT under partial
Under partial shading condition, I-V mismatch among the modules shading condition, it can be concluded that there is lots of research
forming the array causes significant energy loss of the array. To avoid scope to find a suitable MPPT which can improve the output efficiency
hotspot formation and reverse current flow, bypass and blocking diodes of PGS. This review is expected to provide a very beneficial tool to all
are used which causes multiple maxima point in the P-V characteristic. the researchers working on PV system and also to all the industries
To mitigate partial shading effect on PV array characteristics, many excelled in generating an efficient, clean and sustainable energy to the
array reconfiguration techniques such as series-parallel (SP), total mankind.
cross tied (TCT), bridge-linked (BL), honey-comb (HC), hybrid array
reconfiguration are available in the literature [89,90]. Unfortunately Acknowledgments
none of the discussed methods have considered the array reconfigura-
tion in their proposed MPPT. Many of the swarm based techniques The authors are thankful for kind supports from the Institution of
need large number of iterations to locate the MPP which increases the Engineers (PG2016014), (India) for providing the fund under research
computational complexity. The major challenges in all swarm based and development grant in aid scheme to carry out the research work.
techniques are selection of search space, population size, initial
condition and control parameters. Selection of control parameters References
varies from problem to problem and it affects the computational
[1] Esram T, Chapman PL. Comparison of photovoltaic array maximum power point
behavior of the optimization algorithm. Hybrid techniques give better
tracking techniques. IEEE Trans Energy Convers 2007;22:439–49. http://
result than individual MPPT techniques but it has complicated control dx.doi.org/10.1109/TEC.2006.874230.
structure. [2] Subudhi B, Pradhan R. A comparative study on maximum power point tracking
After reviewing the MPPT techniques under partial shading condi- techniques for photovoltaic power systems. IEEE Trans Sustain Energy
2013;4:89–98.
tion, the following points are proposed for further studies. [3] Lyden S, Haque ME, Gargoom A, Negnevitsky M. Review of maximum power point
Although many metaheuristic optimization techniques have already tracking approaches suitable for PV systems under partial shading conditions. In:
been explored and used for MPPT of PV system under PSC, still many Proceedings of the Australas univ power eng conf; 2013. p. 1–6. http://dx.doi.org/
10.1109/AUPEC.2013.6725344.
techniques are there to explore such as teaching learning based [4] El-Khozondar HJ, El-Khozondar RJ, Matter K, Suntio T. A review study of
optimization (TLBO), JAYA optimization, bat swarm optimization photovoltaic array maximum power tracking algorithms. Renew Wind Water Sol
(BSO), shuffled frog leaping (SFL), evolutionary programming (EP), 2016;3:3. http://dx.doi.org/10.1186/s40807-016-0022-8.
[5] Engineering IP, Conference O, Jewel T. The application of soft computing

865
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

techniques to improve the performance of maximum power point tracker for PV [30] Tajuddin MFN, Ayob SM, Salam Z. Global maximum power point tracking of PV
system during partial shading. In: Proceedings of the 2014 8th int power eng optim system using dynamic population size differential evolution (DynNP-DE) algo-
conf (PEOCO 2014); 2014. p. 237–42. http://dx.doi.org/10.1109/PEOCO.2014. rithm. In: Proceedings of the 2014 IEEE conf energy conversion (CENCON 2014);
6814432. 2014. p. 254–9. http://dx.doi.org/10.1109/CENCON.2014.6967511.
[6] Bidram A, Davoudi A, Balog RS. Control and circuit techniques to mitigate partial [31] Zhou L, Chen Y, Guo K, Jia F. New approach for MPPT control of photovoltaic
shading effects in photovoltaic arrays. IEEE J Photovolt 2012;2:532–46. http:// system with mutative-scale dual-carrier chaotic search. IEEE Trans Power Electron
dx.doi.org/10.1109/JPHOTOV.2012.2202879. 2011;26:1038–48. http://dx.doi.org/10.1109/TPEL.2010.2078519.
[7] Ishaque K, Salam Z. A review of maximum power point tracking techniques of PV [32] Ahmed J, Salam Z. A soft computing MPPT for PV system based on Cuckoo Search
system for uniform insolation and partial shading condition. Renew Sustain Energy algorithm. In: Proceedings of the int conf power eng energy electr drives; 2013. p.
Rev 2013;19:475–88. http://dx.doi.org/10.1016/j.rser.2012.11.032. 558–62. http://dx.doi.org/10.1109/PowerEng.2013.6635669.
[8] Liu L, Meng X, Liu C. A review of maximum power point tracking methods of PV [33] Ahmed J, Salam Z. A Maximum Power Point Tracking (MPPT) for PV system using
power system at uniform and partial shading. Renew Sustain Energy Rev Cuckoo Search with partial shading capability. Appl Energy 2014;119:118–30.
2016;53:1500–7. http://dx.doi.org/10.1016/j.rser.2015.09.065. http://dx.doi.org/10.1016/j.apenergy.2013.12.062.
[9] Rajesh R, Carolin Mabel M. A comprehensive review of photovoltaic systems. [34] Mohanty S, Subudhi B, Ray PK. A grey wolf assisted perturb & observe MPPT
Renew Sustain Energy Rev 2015;51:231–48. http://dx.doi.org/10.1016/ algorithm for a photovoltaic power system. IEEE Trans Energy Convers 2016;8969.
j.rser.2015.06.006. http://dx.doi.org/10.1109/TEC.2016.2633722.
[10] Mohanty S, Subudhi B, Ray PK. A new MPPT design using grey Wolf optimization [35] Kermadi, Mostefa, Berkouk EM. A hybrid PSO-PI based maximum power point
technique for photovoltaic system under partial shading conditions. IEEE Trans tracking algorithm using adaptive sampling time strategy. In: Proceedings of the
Sustain Energy 2016;7:181–8. http://dx.doi.org/10.1109/TSTE.2015.2482120. 4th int conf electr eng (ICEE, 2015); 2015. p. 1–6.
[11] Sundareswaran K, Peddapati S, Palani S. MPPT of PV systems under partial shaded [36] Lyden S, Haque ME. A hybrid simulated annealing and perturb and observe
conditions through a colony of flashing fireflies. IEEE Trans Energy Convers method for maximum power point tracking in PV systems under partial shading
2014;29:463–72. http://dx.doi.org/10.1109/TEC.2014.2298237. conditions. In: Proceedings of the 2015 Australas univ power eng conf challenges
[12] Safarudin YM, Priyadi A, Purnomo MH, Pujiantara M. Maximum power point futur grids (AUPEC 2015); 2015. http://dx.doi.org/10.1109/AUPEC.2015.
tracking algorithm for photovoltaic system under partial shaded condition by 7324803.
means updating β firefly technique. In: Proceedings of the 6th int conf inf technol [37] Lian KL, Jhang JH, Tian IS. A maximum power point tracking method based on
electr eng (ICITEE 2014); 2014. p. 1–4. http://dx.doi.org/10.1109/ICITEED.2014. perturb-and-observe combined with particle swarm optimization. IEEE J Photovolt
7007942. 2014;4:626–33. http://dx.doi.org/10.1109/JPHOTOV.2013.2297513.
[13] Jiang LL, Maskell DL, Patra JC. A novel ant colony optimization-based maximum [38] Sundareswaran K, Vignesh kumar V, Palani S. Application of a combined particle
power point tracking for photovoltaic systems under partially shaded conditions. swarm optimization and perturb and observe method for MPPT in PV systems
Energy Build 2013;58:227–36. http://dx.doi.org/10.1016/j.enbuild.2012.12.001. under partial shading conditions. Renew Energy 2015;75:308–17. http://
[14] Jiang LL, Maskell DL. A uniform implementation scheme for evolutionary dx.doi.org/10.1016/j.renene.2014.09.044.
optimization algorithms and the experimental implementation of an ACO based [39] Seyedmahmoudian M, Rahmani R, Mekhilef S, Maung Than Oo A, Stojcevski A,
MPPT for PV systems under partial shading. In: Proceedings of the IEEE symp Soon TK, et al. Simulation and hardware implementation of new maximum power
comput intell appl smart grid (CIASG 2014); 2014. p. 1–8. http://dx.doi.org/10. point tracking technique for partially shaded PV system using hybrid DEPSO
1109/CIASG.2014.7011560. method. IEEE Trans Sustain Energy 2015;6:850–62. http://dx.doi.org/10.1109/
[15] Sundareswaran K, Sankar P, Nayak PSR, Simon SP, Palani S. Enhanced energy TSTE.2015.2413359.
output from a PV system under partial shaded conditions through artificial bee [40] Mahmoud Y, El-Saadany EF. Fast power-peaks estimator for partially shaded PV
colony. IEEE Trans Sustain Energy 2015;6:198–209. http://dx.doi.org/10.1109/ systems. IEEE Trans Energy Convers 2016;31:206–17. http://dx.doi.org/10.1109/
TSTE.2014.2363521. TEC.2015.2464334.
[16] soufyane Benyoucef A, Chouder A, Kara K, Silvestre S, sahed OA. Artificial bee [41] Khan O, Xiao W. An efficient modeling technique to simulate and control
colony based algorithm for maximum power point tracking (MPPT) for PV systems submodule-integrated PV system for single-phase grid connection. IEEE Trans
operating under partial shaded conditions. Appl Soft Comput 2015;32:38–48. Sustain Energy 2016;7:96–107. http://dx.doi.org/10.1109/TSTE.2015.2476822.
http://dx.doi.org/10.1016/j.asoc.2015.03.047. [42] Farivar G, Agelidis VG, Member S, Hredzak B. A simple perturb and observe MPPT
[17] Ishaque ZS K. A deterministic particle swarm optimization maximum power point scheme for cascaded H-bridge based photovoltaic system; 2013.
tracker for photovoltaic system under partial shading condition. IEEE Trans Ind [43] Gaikwad DD, Chavan MS, Gaikwad MS Hardware implementation of DC-DC
Electron 2013;60:3195–206. converter for MPPT in PV applications. In: Proceedings of the 2014 IEEE glob conf
[18] Divya K, Sugumaran G. Intelligent control technique for MPPT of PV system. In: wirel comput networking (GCWCN 2014); 2015. p. 16–20. http://dx.doi.org/10.
Proceedings of the IEEE int caracas conf devices, circuits syst (ICCDCS 2014); 1109/GCWCN.2014.7030839.
2014. p. 1–5. http://dx.doi.org/10.1109/ICDCSyst.2014.6926164. [44] Xie L, Qi J, Weng G, Zhang Y Multi-level PV inverter with photovoltaic groups
[19] Ramaprabha R, Jubair SH, Suhas K, Lokesh A. Development of improved curve independent MPPT control. In: Proceedings of the 2014 17th int conf electr mach
tracer for photovoltaic system under partial shaded conditions. In: Proceedings of syst (ICEMS 2014); 2015. p. 829–34. http://dx.doi.org/10.1109/ICEMS.2014.
the 2014 int conf circuits, power comput technol (ICCPCT 2014); 2014. p. 585–9. 7013599.
http://dx.doi.org/10.1109/ICCPCT.2014.7054857. [45] Sharma P, Agarwal V. Maximum power extraction from a partially shaded PV array
[20] Lei P, Li Y, Seem J. Sequential ESC based global MPPT control for photovoltaic using shunt-series compensation. IEEE J Photovolt 2014;4:1128–37. http://
array with variable shading. IEEE Trans Sustain Energy 2011;2:348–58. dx.doi.org/10.1109/JPHOTOV.2014.2323698.
[21] Heydari-doostabad H, Keypour R, Khalghani MR, Khooban MH. A new approach in [46] Kashani MG, Mobarrez M, Bhattacharya S. Variable interleaving technique for
MPPT for photovoltaic array based on extremum seeking control under uniform photovoltaic cascaded DC-DC converters. In: Proceedings of the 40th annual
and non-uniform irradiances. Sol Energy 2013;94:28–36. http://dx.doi.org/ Industrial Electron Conf; 2014. p. 5612–7. http://dx.doi.org/10.1109/IECON.
10.1016/j.solener.2013.04.025. 2014.7049359.
[22] Elnosh A, Khadkikar V, Xiao W, Kirtely JL. An improved Extremum-Seeking based [47] Ishaque K, Salam Z, Taheri H, Shamsudin A. Maximum power point tracking for
MPPT for grid-connected PV systems with partial shading. IEEE Int Symp Ind PV system under partial shading condition via particle swarm optimization. In:
Electron 2014:2548–53. http://dx.doi.org/10.1109/ISIE.2014.6865021. Proceedings of the 2011 IEEE appl power electron colloquium (IAPEC 2011). 2;
[23] Leyva R, Olalla C, Zazo H, Cabal C, Cid-Pastor A, Queinnec I, et al. MPPT based on 2011. p. 5–9. http://dx.doi.org/10.1109/IAPEC.2011.5779866.
sinusoidal extremum-seeking control in PV generation. Int J Photoenergy [48] Liu YH, Huang SC, Huang JW, Liang WC. A particle swarm optimization-based
2012;2012. http://dx.doi.org/10.1155/2012/672765. maximum power point tracking algorithm for PV systems operating under partially
[24] Lyden S, Haque ME. A simulated annealing global maximum power point tracking shaded conditions. IEEE Trans Energy Convers 2012;27:1027–35. http://
approach for PV modules under partial shading conditions. IEEE Trans Power dx.doi.org/10.1109/TEC.2012.2219533.
Electron 2016;31:4171–81. http://dx.doi.org/10.1109/TPEL.2015.2468592. [49] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved particle swarm
[25] Hosseini SH, Farakhor A, Khadem Haghighian S Novel algorithm of MPPT for PV optimization (PSO)-based MPPT for PV with reduced steady-state oscillation. IEEE
array based on variable step Newton-Raphson method through model predictive Trans Power Electron 2012;27:3627–38. http://dx.doi.org/10.1109/
control. In: Proceedings of the 2013 13th int conf control autom syst (ICCAS 2013); TPEL.2012.2185713.
2013. p. 1577–82. http://dx.doi.org/10.1109/ICCAS.2013.6704181. [50] Liu CX, Liu LQ An improved perturbation and observation MPPT method of
[26] Ahmad J, Spertino F, Leo P Di, Ciocia A. A variable step size perturb and observe photovoltaic generate system. In: Proceedings of the 2009 4th IEEE conf ind
method based MPPT for partially shaded photovoltaic arrays 2. Characteristics of electron appl (ICIEA 2009); 2009. p. 2966–70. http://dx.doi.org/10.1109/ICIEA.
photovoltaic arrays under partial shading conditions; 2016. 10–2. 2009.5138752.
[27] Mohapatra A, Nayak B, Mohanty KB. Current based novel adaptive P & O MPPT [51] Jumpasri N, Pinsuntia K, Woranetsuttikul K, Nilsakorn T, Khan-ngern W.
algorithm for photovoltaic system considering sudden change in the irradiance. In: Improved particle swarm optimization algorithm using average model on MPPT for
Proceedings of the 2014 IEEE int. conf. power electron. drives energy syst. (PEDES, partial shading in PV array. In: Proceedings of the 2014 int electr eng congr; 2014.
2014); 2014. http://dx.doi.org/10.1109/PEDES.2014.7042032. p. 1–4. http://dx.doi.org/10.1109/iEECON.2014.6925856.
[28] Khaehintung N, Wiangtong T, Sirisuk P. FPGA implementation of MPPT using [52] Chen K, Tian S, Cheng Y, Bai L. An improved MPPT controller for photovoltaic
variable step-size P & O algorithm for PV applications. In: Proceedings of the 2006 system under partial shading condition. IEEE Trans Sustain Energy
int symp commun inf technol; 2006. p. 212–15. http://dx.doi.org/10.1109/ISCIT. 2014;5:978–85. http://dx.doi.org/10.1109/TSTE.2014.2315653.
2006.340033. [53] Badis A, Mansouri MN, Sakly A. PSO and GA-based maximum power point
[29] Dahhani O, Jouni A El, Sefriti B, Boumhidi I. Optimal perturb and observe control tracking for partially shaded photovoltaic systems. In: Proceedings of the 7th int
for MPPT based on least square support vector machines algorithm. Intell Syst renew energy congr (IREC 2016); 2016. http://dx.doi.org/10.1109/IREC.2016.
Comput Vis 2015:1–7. 7478923.

866
A. Mohapatra et al. Renewable and Sustainable Energy Reviews 80 (2017) 854–867

[54] Dhar S, Sridhar R, Mathew G. Implementation of PV cell based standalone solar [73] Mishra S, Sekhar PC. Ts fuzzy based adaptive perturb algorithm for MPPT of a grid
power system employing incremental conductance MPPT algorithm. In: connected single stage three phase VSC interfaced PV generating system. In:
Proceedings of the IEEE int conf circuit, power comput technol (ICCPCT 2013); Proceedings of the IEEE power energy soc gen meet; 2012. p. 1–7. http://dx.doi.
http://dx.doi.org/10.1109/ICCPCT.2013.6528926. org/10.1109/PESGM.2012.6345397.
[55] Boussaibo A, Kamta M, Kayem J, Toader D, Haragus S, Maghet A. Characterization [74] Sun D, Ge B, Liang W, Abu-Rub H, Peng FZ. An energy stored quasi-Z-source
of photovoltaic pumping system model without battery storage by MATLAB/ cascade multilevel inverter-based photovoltaic power generation system. IEEE
Simulink. In: Proceedings of the 2015 9th int symp adv top electr eng (ATEE 2015); Trans Ind Electron 2015;62:5458–67. http://dx.doi.org/10.1109/
2015. p. 774–80. http://dx.doi.org/10.1109/ATEE.2015.7133907. TIE.2015.2407853.
[56] Al-Masri H, Alhuwaishel F, Alismail F, Sabeeh S, Kanakri H. Investigation of MPPT [75] Liang W, Abu-Rub H, Liu Y, Ge B. State-of-charge balancing control for battery
for PV applications by mathematical model. In: Proceedings of the 15th 2015 IEEE energy stored quasi-Z source cascaded multilevel inverter based photovoltaic power
int conf environ electr eng (EEEIC 2015) – Conf Proc; 2015. p. 1800–5. http://dx. system. In: Proceedings of the 2015 IEEE energy convers congr expo (ECCE 2015);
doi.org/10.1109/EEEIC.2015.7165446. 2015. p. 8–13. http://dx.doi.org/10.1109/ECCE.2015.7309662.
[57] Lentine AL, Nielson GN, Okandan M, Sweatt WC, Cruz-campa JL, Gupta V. [76] Xiao B, Hang L, Mei J, Riley C, Tolbert LM, Ozpineci B. Modular cascaded H-bridge
Optimal cell connections for improved shading , reliability , and spectral perfor- multilevel PV inverter with distributed MPPT for grid-connected applications. IEEE
mance of microsystem enabled photovoltaic (Mepv) modules; 2010. p. 3048–54. Trans Ind Appl 2015;51:1722–31. http://dx.doi.org/10.1109/TIA.2014.2354396.
[58] Macabebe EQB, Sheppard CJ, van Dyk EE. Parameter extraction from I-V [77] Zhu H, Zhang D, Athab HS, Wu B, Gu YPV. isolated three-port converter and
characteristics of PV devices. Sol Energy 2011;85:12–8. http://dx.doi.org/ energy-balancing control method for PV-battery power supply applications. IEEE
10.1016/j.solener.2010.11.005. Trans Ind Electron 2015;62:3595–606. http://dx.doi.org/10.1109/
[59] Villalva M, Gazoli J, Filho E. Comprehensive approach to modeling and simulation TIE.2014.2378752.
of photovoltaic arrays. IEEE Trans Power Electron 2009;24:1198–208. http:// [78] Benavides ND, Esram T, Chapman PL Ripple correlation control of a multiple-
dx.doi.org/10.1109/tpel.2009.2013862. input dc-dc converter. In: Proceedings of the IEEE annu power electron spec Conf
[60] Villalva MG, Gazoli JR, Ruppert EF. Modeling and circuit-based simulation of (PESC Rec 2005); 2005. p. 160–4. http://dx.doi.org/10.1109/PESC.2005.
photovoltaic arrays. Braz J Power Electron 2009;14:1244–54. 1581618.
[61] Nayak BK, Mohapatra A, Mohanty KB. Parameters estimation of photovoltaic [79] Benavides ND, Chapman PL. Power budgeting of a multiple-input buck-boost
module using nonlinear least square algorithm: a comparative study. In: converter. IEEE Trans Power Electron 2005;20:1303–9. http://dx.doi.org/
Proceedings of the 2013 IEEE annu. India conf. (INDICON 2013); 2013. http://dx. 10.1109/TPEL.2005.857531.
doi.org/10.1109/INDCON.2013.6726120. [80] Debnath D, Chatterjee K. Two-stage solar photovoltaic-based stand-alone scheme
[62] Gao L, Dougal RA, Liu S, Iotova AP. Parallel-connected solar PV system to address having battery as energy storage element for rural deployment. IEEE Trans Ind
partial and rapidly fluctuating shadow conditions. IEEE Trans Ind Electron Electron 2015;62:4148–57. http://dx.doi.org/10.1109/TIE.2014.2379584.
2009;56:1548–56. http://dx.doi.org/10.1109/TIE.2008.2011296. [81] Lu DDC, Agelidis VG. Photovoltaic-battery-powered DC bus system for common
[63] El-dein MZS, Member S, Kazerani M, Member S, Salama MMA. An optimal total portable electronic devices. IEEE Trans Power Electron 2009;24:849–55. http://
cross tied interconnection for reducing mismatch losses in photovoltaic arrays; dx.doi.org/10.1109/TPEL.2008.2011131.
2013. 4. p. 99–107. [82] Uoya M, Koizumi H. A calculation method of photovoltaic array's operating point
[64] Zhang MZM, Wu JWJ, Zhao HZH. The application of slide technology in PV for MPPT evaluation based on one-dimensional Newton-Raphson method. IEEE
maximum power point tracking system. In: Proceedings of the fifth world congr Trans Ind Appl 2015;51:567–75. http://dx.doi.org/10.1109/TIA.2014.2326083.
intell control autom (IEEE Cat No04EX788); 2004. 6. p. 5591–4. http://dx.doi. [83] Peter PK, Agarwal V. On the input resistance of a reconfigurable switched capacitor
org/10.1109/WCICA.2004.1343805. DC-DC converter-based maximum power point tracker of a photovoltaic source.
[65] Rtiz-Rivera EI, Peng F. A novel method to estimate the maximum power for a IEEE Trans Power Electron 2012;27:4880–93. http://dx.doi.org/10.1109/
photovoltaic inverter system. In: Proceedings of the IEEE annu power electron spec TPEL.2012.2192452.
conf (PESC Rec); 2004. 3. p. 2065–9. http://dx.doi.org/10.1109/PESC.2004. [84] Jiang S, Cao D, Li Y, Peng FZ. Grid-connected boost-half-bridge photovoltaic
1355436. microinverter system using repetitive current control and maximum power point
[66] Hilloowala RM, Sharaf AM. A rule-based fuzzy logic controller for a PWM inverter tracking. IEEE Trans Power Electron 2012;27:4711–22. http://dx.doi.org/
in photo-voltaic energy conversion scheme. In: Proceedings of the IEEE ind appl 10.1109/TPEL.2012.2183389.
soc annu meet (Conf Rec 1992); 1992. p. 762–9. http://dx.doi.org/10.1109/IAS. [85] Wu TF, Kuo CL, Sun KH, Chen YK, Chang YR, Lee Y Der. Integration and operation
1992.244319. of a single-phase bidirectional inverter with two buck/boost MPPTs for DC-
[67] Hilloowala RM, Sharaf AM. A rule-based fuzzy logic controller for a PWM inverter distribution applications. IEEE Trans Power Electron 2013;28:5098–106. http://
in a stand alone wind energy conversion scheme. IEEE Trans Ind Appl dx.doi.org/10.1109/TPEL.2013.2245681.
1996;32:57–65. http://dx.doi.org/10.1109/28.485813. [86] Safari A, Mekhilef S. Simulation and hardware implementation of incremental
[68] Hilloowala RM, Sharaf AM. No Title. A rule-based fuzzy log. controll. A PWM conductance MPPT with direct control method using cuk converter. IEEE Trans
invert. A stand alone wind energy convers. scheme. In: Proceedings of the IEEE Ind Electron 2011;58:1154–61. http://dx.doi.org/10.1109/TIE.2010.2048834.
conference; 1993, p. 2066–73. http://dx.doi.org/10.1109/IAS.1993.299150. [87] Mamarelis E, Petrone G, Spagnuolo G. Design of a sliding-mode-controlled SEPIC
[69] Kim H, Kim S, Kwon C-K, Min Y-J, Kim C, Kim S-W. An energy-efficient fast for PV MPPT applications. IEEE Trans Ind Electron 2014;61:3387–98. http://
maximum power point tracking circuit in an 800-μW photovoltaic energy harvester. dx.doi.org/10.1109/TIE.2013.2279361.
IEEE Trans Power Electron 2013;28:2927–35. http://dx.doi.org/10.1109/ [88] Lee JH, Park JH, Jeon JH. Series-connected forward-flyback converter for high
TPEL.2012.2220983. step-up power conversion. IEEE Trans Power Electron 2011;26:3629–41. http://
[70] Uprety S, Member S, Lee H, Member S. A 0.4 W-to-21 W fast-transient global- dx.doi.org/10.1109/TPEL.2011.2162747.
search- algorithm based integrated photovoltaic energy harvester with 99% GMPPT [89] Malathy S, Ramaprabha R. Comprehensive analysis on the role of array size and
efficiency and 94% power efficiency. IEEE J Solid-State Circuits 2016;51:2153–67. configuration on energy yield of photovoltaic systems under shaded conditions.
[71] Zubair Mehmood, Yumna Bilal, Muniba Bashir AA. Performance analysis of MPPT Renew Sustain Energy Rev 2015;49:672–9. http://dx.doi.org/10.1016/
charge controller with single and series/parallel connected PV panels; 2016. p. j.rser.2015.04.165.
278–82. [90] ShamsEl-Dein MZ, Kazerani Mehrdad SMM. Optimal photovoltaic array reconfi-
[72] Pradhan R, Subudhi B, Ray PK. A real-time linearized maximum power point guration to reduce partial shading losses. IEEE Trans Sustain Energy
tracker for photovoltaic system. In: Proceedings of the int conf power electron drive 2013;4:145–53.
syst; 2013. p. 962–7. http://dx.doi.org/10.1109/PEDS.2013.6527157.

867
 Applied Energy 179 (2016) 523–530

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Photovoltaic maximum power point tracking under fast varying

of solar radiation
A. Belkaid a,b,⇑, I. Colak c, O. Isik c
a
Department of Electromechanics, University of Bordj BouArreridj, El-Anasser 34030, Bordj Bou Arreridj, Algeria
b
Automatic Laboratory of Setif (LAS), University of Setif 1, El Maabouda, Street of Bejaia, 19000 Setif, Algeria
c
Engineering and Architecture Faculty of Istanbul Gelisim University, Turkey

h i g h l i g h t s

An INC-MPPT technique is improved for PV plant.

Dynamic efficiency of the basic INC under a gradual irradiance profile is improved.

Energy conversion efficiency is increased by approximately 5%.

Stringent irradiance profile with different shapes is selected as in EN50530.

A robust tracker is proposed for variable irradiance, temperature and load.

a r t i c l e i n f o a b s t r a c t

Article history: Perturb and Observe (P&O) and Incremental Conductance (INC) are widely used as Maximum Power
Received 29 January 2016 Point Tracking (MPPT) techniques in Photovoltaic (PV) systems. But, they fail under rapidly varying of
Received in revised form 1 June 2016 sunlight levels. This paper proposes a new MPPT technique, which can make a distinction between per-
Accepted 10 July 2016
turbation in the reference voltage and sudden-changing of sunlight and thus optimize the PV system effi-
ciency. This method consists on a modified INC algorithm, which is used to fine-tune the duty cycle of the
DC/DC converter in order to avoid divergences of the maximum power point (MPP) when using basic INC
Keywords:
under fast varying of luminosity levels.
Photovoltaic system
Incremental Conductance
The proposed PV-MPPT system, which is composed by a step-up converter as the interface to feed the
Maximum power point tracker load, is tested by simulation within the Matlab/Simulink software by taking into account the luminosity,
Step-up converter the temperature and the load variation. The simulation results are satisfactory and demonstrate that the
Modeling improved INC technique can track the PV maximum power at diverse operating conditions with the most
Simulation excellent performance, the energy conversion efficiency is increased by approximately 5%.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction nected. In each operating condition, there is just one and unique
particular point that makes the PV array in functioning at its max-
The solar energy has many advantages compared with fossil-fuel imum power. In most PV applications (stand-alone system, hybrid
energy. It is inexhaustible, free of charge, naturalness, clean, no eco- system and public grid connected system), it is very important to
logical pollution, and with modular character, which allows con- exploit the peak power provided by the photovoltaic generator.
struction of the solar array at different power levels. On the other To achieve this goal, a power electronics device with an adequate
hand, it has an important disadvantage, which is the low efficiency MPPT technique is needed [1–5]. Many of these MPPT algorithms
of conversion of light to electrical energy [1]. Moreover, the power have been published in several references [6–9]. These methods
harvested by a PV generator depends on a number of factors, such vary in many aspects such as the necessary sensors, the complexity,
as the luminosity, the temperature, and the load in that it is con- the cost, the efficiency, the convergence rapidity, the accurate
tracking when temperature and/or luminosity vary, and equipment
required for the implementation or popularity [10]. A full analysis
⇑ Corresponding author at: Department of Electromechanics, University of Bordj of 30 different MPPT algorithms can be found in [6].
BouArreridj, El-Anasser 34030, Bordj Bou Arreridj, Algeria. Among these techniques, Perturb and Observe [11–17] and
E-mail addresses: [email protected] (A. Belkaid), [email protected]
Incremental Conductance [2,17–21] are algorithms most
(I. Colak), [email protected] (O. Isik).

http://dx.doi.org/10.1016/j.apenergy.2016.07.034
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
524 A. Belkaid et al. / Applied Energy 179 (2016) 523–530

commonly used in the literature due to their simplicity of imple- voltage and sudden-changing of sunlight and thus avoid diver-
mentation and independence to PV array parameters. Other tech- gence from the MPP. An additional contribution is the selecting
niques based on different principles are fractional short circuit of stringent profile for the irradiance variation as suggested by
current that estimates the optimal current by short circuit current EN 50530 [17,37]. This irradiance profile consists on different
[22], fractional open circuit voltage that estimates the optimal shapes such as the ramp up, the step down, the step up and the
voltage by the open circuit voltage [23], the neural network [24], ramp down. These diverse shapes are capable to confirm truthfully
the fuzzy logic [10,25] and the Sliding Mode Control-based MPPT the efficacy of the proposed tracker under static and dynamic
(SMC) [1,26–29]. Most of these algorithms are iterative, based on states. As a result, this work is done to improve the dynamic effi-
the disruption of the cyclic ratio of the converter by a small incre- ciency of PV conversion chain with a modified INC that can
ment. These techniques can track the MPP with great precision decrease the probability of diverging away from the peak power.
under stable conditions. But it still reveals compromise between To compare the proposed tracker with the basic one, the tracking
speed pursuit and the reliability of prosecution when the values efficiency parameter is taken into account.
of load or weather conditions change rapidly. A lot of researchers This paper is organized as follows. PV panel and boost chopper
made changes to these algorithms in the objective to improve models are given in Sections 2 and 3 respectively. Then Maximum
them. In Ref. [30], two different implementations for the INC algo- Power Point Tracking Algorithms are introduced in Section 4.
rithm were applied to PV pumping system; the first one is with ref- Subsequently, the basics of classical and modified INC techniques
erence voltage perturbation in combination with a PI regulator to are given. Thereafter, Section 5 depicts the results of simulation.
regulate the power converter duty cycle; the second one is without Finally, the conclusion section summarizes highlights the achieve-
PI controller in which the duty cycle is employed straight as the ments of this study.
control parameter. It was found that the direct duty cycle control
exhibits better energy efficiency than the first control by about 2. Equivalent circuit and characteristics of PV panel
2%. The INC direct duty cycle control was firstly implemented in
solar array using Cuk converter by Safari et al. [31]. This last The solar cell is the basic elementary component, which can
method is considered in the present paper as basic INC. It perturbs transform directly the sunlight into electricity. Assembly of cells
the duty ratio until the peak power is reached with a fixed step in series forms a module. The equivalent circuit of the PV module
size. However, it suffers from the low speed tracking and the losing is illustrated in Fig. 1 [20,21]. A series-parallel assemblage of mod-
tracking direction when rapid changes occurred in environmental ules forms an array. The PV panel can be described mathematically
conditions. For that reason, many researchers introduce the vari- by the following set of equations as in [38,39].
able step size in INC MPPT [32,33]. Basic and variable step size " ! #  
INC techniques are also not capable to respond precisely in fast- V pv þ ðNNps ÞRs Ipv V pv þ NNps Rs Ipv
Ipv ¼ N p Iph  Np Is exp 1    ð1Þ
changing of solar radiance [2]. That is why the authors of [2] have ns av t Ns
Rp
proposed a variant of variable step size INC method with a permit- Np




dI

ted error
Ipv þ V pv dVppvv
< 0:06 to mitigate inaccurate answers of where Ipv is the PV array output current (A), V pv is the PV array out-
the conventional INC (fixed step size) during sudden changes in put voltage (V), N s and N p are the number of PV modules connected
sunlight levels. But this last work should be checked with a profile in series and parallel, respectively, ns is the number of PV cells
of gradual change of irradiance. Therefore, this paper proposes a connected in series in one string, Rs and Rp are respectively the PV
new and simple INC to reduce the probability of losing tracking module series and parallel resistances (Ω), a is the p-n junction ide-
direction that may be occurred and thus to enhance the energy ality factor.
efficiency of the PV system. The photo-current Iph depends on solar irradiance G and tem-
The INC algorithm is among the Hill-Climbing techniques [34]. perature T as follows:
Through adjusting the duty ratio of the converter, it drives up G
the operating point on the power versus voltage curve to achieve Iph ¼ ½Isc þ ki ðT  T  Þ ð2Þ
G
the MPP which is at the top. This method offers advantages such
as simple implementation, fast convergence speed, high efficiency where ki is the short circuit current temperature coefficient, Isc is
and low cost [6,19,31], but also some disadvantages: The first one the short circuit current at standard test conditions (STC) that are:
is the incessant unwanted oscillating of the operating point around solar irradiance G ¼ 1000 W=m2 , cell temperature T  ¼ 298 K and
the MPP [35,36], once the tracking control reaches the vicinity of a spectral distribution AM 1.5.
the PV panel maximum power. As a result, this steady-state undu- The reverse saturation current Is varies with temperature
lation causes the energy losses. The second one is that the algo- according to the following expression:
rithm can lose its tracking direction, i.e. it can deviate away from Isc þ ki ðT  T  Þ
the MPP, when the change in the luminosity is rapid [19,35], espe- Is ¼   
 ð3Þ
cially if the irradiation varies according to a slope. For the reason exp V oc þknvs vðTT
t
Þ
1
that the algorithm is not capable to determine if the change in
the power is due to the perturbation of the voltage or due to the
change in the irradiance. This divergence also causes the power
loss and thus a decrease in the efficiency. Another problem to con-
sider is that many researchers test their algorithms with simple
irradiance profile (constant or step). Consequently, the improve-
ments in performance cannot be truthfully demonstrated. In order
to reduce the effect of these drawbacks, this paper proposes a new
MPPT control based on a modified incremental conductance. The
proposed tracker is similar to the old one, but it incorporates two
tests consisting in change in current and voltage with same signs.
The added part is for the fast-varying of the sunshine. Therefore,
the new INC can distinguish between perturbation in the reference Fig. 1. Solar module equivalent circuit.
 A. Belkaid et al. / Applied Energy 179 (2016) 523–530 525

Table 1 L iL
MSX 60 solar panel specifications [1].
I pv
Parameter Value D
STC power rating P max 60 W G
Open circuit voltage V oc 21.1 V u =1
G V pv
Short circuit current Isc 3.8 A
P C1 K C2 Vo R
T
Voltage at maximum power V opt 17.1 V V u=0
Current at maximum power Iopt 3.5 A
Temperature coefficient of V oc 0.08 V/°C
Temperature coefficient of Isc 0.003 A/°C

V pv
MPPT
+
PWM I pv

Fig. 4. Diagram schematic of the proposed photovoltaic system.

MPPT algorithm adjusts incessantly the DC–DC Boost converter

duty cycle. Comparing this duty ratio to a saw signal, a PWM signal
is generated. The control input u is the switch position; it takes 0
when the switch is open or 1 when the switch is closed. Based
on the Kirchhoff’s first and second rules, the dynamic model of
the converter can be written [1,5].
8
Fig. 2. PV module curves under diverse irradiance levels, temperature 25 °C: < didtL ¼ V pvLV o þ VLo  u
current-voltage (blue); power-voltage (red). (For interpretation of the references to   ð4Þ
colour in this figure legend, the reader is referred to the web version of this article.) : dV o ¼  V o þ iL  iL  u
dt RC 2 C2 C2

where V o and iL are the load voltage and the current across the
inductor, respectively.
The output voltage of the boost converter V o can be expressed
in function of the input voltage V pv and its duty cycle d:

Vo 1
¼ ð5Þ
V pv 1  d

Eq. (5) demonstrates that incrementing (decrementing) the voltage

reference V pv can be done by decrementing (incrementing) the con-
verter duty cycle d.

4. Maximum power point tracking algorithms

Fig. 3. PV module curves under diverse temperatures, irradiation 1000 W/m2:
current-voltage (blue); power-voltage (red). (For interpretation of the references to 4.1. Classical INC technique
colour in this figure legend, the reader is referred to the web version of this article.)

The INC algorithm tracks the MPP by comparing the sum of

incremental conductance plus instantaneous conductance of a PV
where V oc
is the open circuit voltage at STC conditions, kv is the panel with zero [41]. It is based on the fact that the slope of the
 dP 
open circuit voltage temperature coefficient, and v t ¼ kb T=q is the curve power vs. voltage dV ¼ I þ DDVI  V of the solar panel is zero
thermal voltage, kb ¼ 1:38065  1023 J=K is the Boltzmann’s con- at the MPP, positive on the left of it and negative on the right. If
stant and q ¼ 1:60218  1019 C is the charge of an electron. the slope is positive, the operating point must be moved to the
The model described by the equations above was implemented right by increasing the module voltage. If it is negative, the operat-
in Matlab software using the known reference specifications for a ing point must be moved to the left by decreasing the module volt-
60 W solar module (Table 1). The levels of solar irradiance and cell age. Finally, when the slope is null, the operating point is at the
temperature were varied and the corresponding characteristics MPP and thus the algorithm stop the voltage adjustment. A scheme
were plotted, as shown in Figs. 2 and 3. describing this algorithm is illustrated in Fig. 5 [17,42,43]. The
accuracy and rapidity in which the algorithm tracks the MPP
depends on the size of the increment of the reference voltage or
3. Modeling of the step-up chopper of the reference of the duty cycle.
Two main handicaps are reconciled to this method; the first one
The power converter selected in this study is a kind of step-up is the oscillating of the operating point around the MPP in steady
chopper, it is inserted in order to interface the PV output to the state, the second is that the algorithm can easily lose the pursuit
resistor as shown in Fig. 4. Its objective is to track the MPP of the of the MPP if the solar radiation varies quickly [14,44]. When the
PV array. A coil, a capacitor, a diode and an IGBT controlled with irradiation varies instantly as step, the MPP tracking sounds well.
a PWM signal, build this converter and it transforms input voltage But, if the irradiation changes according to a slope, the tracking will
to the output voltage as a step up transformer [40]. As shown in be bad. The algorithm is not able to determine if the change in the
Fig. 4, two sensors are required for measuring V pv and Ipv . The power is due to the perturbation of the voltage or due to the
526 A. Belkaid et al. / Applied Energy 179 (2016) 523–530

Start

Measure V ( k ), I ( k )

No Yes
ΔV = 0

Yes ΔI Yes
I+ V =0
ΔV ΔI = 0
No
No
ΔI
I+ V >0 ΔI > 0
ΔV
No No
Yes Yes

d = d + Δd d = d − Δd d = d + Δd d = d − Δd

Update
V (k ) = V (k − 1)
I (k ) = I (k − 1)

Return

Fig. 5. Flowchart of the basic INC algorithm.

change in solar radiation. This principle is shown in Fig. 6. That’s of the perturbation. One can say that the new algorithm must act
why we chose an irradiation profile with various shapes to check contrary to the old algorithm when the system functions under
the performance of the proposed method. fast-varying conditions. This principle is explained in Fig. 7, which
is given in the flowchart of the modified INC algorithm. The added
part with respect to the old algorithm is indicated in yellow color.
4.2. Modified INC technique
It consists of two tests consisting in change in current and voltage
with same signs. It indicates the fast-varying of sunlight.
In stable conditions of fixed irradiance level and temperature,
the current vs. voltage characteristic shows well that the change
in voltage leads a change in current with differing sign. If perturba- 5. Results of simulation and discussion
tion in voltage leads a perturbation with same sign in current, the
PV array is in sudden varying atmospheric conditions. Contrary to In this section, based on the modeling given above for different
the traditional method, the modified INC algorithm can distinguish components of the standalone PV system shown in Fig. 4, a
between these two operating conditions and thus avoid divergence Matlab/Simulink simulation is realized. The system is composed
in the case of the second circumstances by changing the direction of the boost converter that is located between an MSX 60 module

Fig. 6. Divergence of INC algorithm when the luminosity varies.

 A. Belkaid et al. / Applied Energy 179 (2016) 523–530 527

Start

Measure V ( k ), I ( k )

No Yes
ΔV = 0
Yes ΔI
I+ V =0
ΔV Yes
No ΔI = 0
Yes ΔI No
I+ V >0
ΔV > 0 ΔV
ΔI > 0
&ΔI > 0
No No
Yes No
ΔV < 0 Yes

&ΔI < 0 Yes

No
d = d + Δd 2 d = d − Δd d = d − Δd1 d = d + Δd d = d − Δd

Update
V (k ) = V (k − 1)
I (k ) = I (k − 1)

Return

Fig. 7. Flowchart of the modified INC algorithm.

as a power supply and a resistance as a load. The boost converter is Fig. 9 depicts the comparison of the peak power tracking
controlled by an IGBT gate signal with a switching frequency of between the modified and classical INC technique. It can be seen
f ¼ 10 kHz. The inductor and capacitors are used as filters: that the modified algorithm exhibits better performance than the
L ¼ 5 mH, C 1 ¼ 1000 lF, C 2 ¼ 470 lF. classical one, especially in dynamic response. The classical method
In fact, the article aims at exhibiting the performance and effi- presents a poorer performance than the new method under
ciency of the proposed new INC MPPT controller, and comparing decreasing solar radiation and an even lower performance when
it with the classical one. So, to reach this goal, a stringent profile solar radiation is increasing. It demonstrates that the basic INC pre-
was chosen to change the solar radiance, maintaining the temper- sents divergence from the MPP under slope changing irradiance
ature fixed at constant value of 25 °C and the resistive load at levels. The proposed MPPT can resolve this problem. One can say
R ¼ 30 X. The tests were done for duration of 3.2 s. The duty cycle that the new INC improve the performance of the conventional
increments used for the INC algorithm was about of Dd = 0.0002, one.
and for the modified method were Dd ¼ 0:0002, Dd1 ¼ 0:0005, Fig. 9(b) depicts the evolution of converter duty cycle under the
Dd2 ¼ 0:00009. irradiance change for these two algorithms. Part (c) of the same fig-
First of all, the traditional INC tracker was tested under the irra- ure illustrates the tracking efficiency comparison. The tracking effi-
diance profile change. Second, the proposed INC tracker was also ciency T eff is the average MPPT efficiency; it is evaluated as [34]:
tested under the same conditions. The irradiance profile and the Rt
matching simulation results with both algorithms are exposed in PMPP
T eff ¼ R0t  100 ½% ð6Þ
Fig. 8. The results shown in part (b) and (c) of Fig. 8, consist on P
0 pv
the waveforms of PV current Ipv ðAÞ, voltage V pv ðVÞ, power P pv ðWÞ
and output voltage V o ðVÞ. where P MPP is the power obtained by a given MPPT method and Ppv
From these results, it can be confirmed that the different gran- is the theoretical available power.
deurs have achieved similar values as given by the PV characteris- From Fig. 9(c), it was found that the average efficiency of the
tics (Fig. 2) for constant solar irradiance. In addition, the basic INC is about T eff ¼ 91:39% and for modified INC is about
waveforms of Ipv ðAÞ, Ppv ðWÞ are significantly affected by the T eff ¼ 96:40%. One notices that the tracking efficiency of the pro-
change of sunshine than V pv ðVÞ, which is only slightly affected. posed MPPT technique is increased by approximately five percent.
Also, by comparing V pv ðVÞ to V o ðVÞ, it can be confirmed that the According to the above-obtained results, it concludes that the
power converter used is a step-up voltage type. proposed method guarantees better dynamic response under fast
Under the standard tests conditions (G = 1000 W/m2, T = 25 °C), varying solar radiation when the change follows an upward or
the PV module generates an average of 60 W of power, 17.4 V of downward slope.
voltage and 3.5 A of current. These output values vary when the After, we test the proposed controller with two other parame-
irradiance level decreases or increases. It can be noticed that the ters influencing the PV system, which are the temperature and
PV voltage is just slightly affected by the change of the irradiance. the load.
528 A. Belkaid et al. / Applied Energy 179 (2016) 523–530

Fig. 9. Comparison of tracking performance between classical and modified INC

Fig. 8. Results of simulation: (a) irradiance profile, (b) with basic INC algorithm, and algorithm: (a) tracking of MPP, (b) duty cycle evolution, and (c) tracking efficiency.
(c) with modified INC algorithm.

For the temperature, the changing profile is as in Fig. 10(a). It

evolves as stair amount from 0 °C to 25 °C and then to 50 °C. The
irradiation is kept constant at 1000 W/m2 and the resistor is fixed
to 30 Ω. The obtained results are illustrated in Fig. 10(b). It can be
seen that the waveforms of V pv ðVÞ, P pv ðWÞ are reverse proportional
to the change of temperature and Ipv ðAÞ is only slightly affected.
The last test is for the load changing as shown in Fig. 11(a). The
matching results are presented in Fig. 11(b). They demonstrate that
the PV grandeurs the power, the voltage and the current are not
affected by the resistance modification.

6. Conclusion

In this paper, the modeling, simulation and control of a stand-

alone photovoltaic power electricity generation system has been
proposed by incorporating a power electronics converter to permit
the use of the PV array at its maximal power. The proposed PV-
MPPT system has been implemented under MATLAB-SIMULINK
environment.
A control strategy based on an improved INC technique was
developed in order to regulate the power output of a solar system,
which comprises photovoltaic generation, a boost DC-DC converter Fig. 10. (a) Temperature variation profile, and (b) results of simulation with
that is able to step-up the load voltage. proposed algorithm.
 A. Belkaid et al. / Applied Energy 179 (2016) 523–530 529

[12] Murtaza AF, Sher HA, Chiaberge M, Boero D, De Giuseppe M, Addoweesh KE. A
novel hybrid MPPT technique for solar PV applications using perturb & observe
and fractional open circuit voltage techniques. In: 15th intern. symposium on
Mechatronika, Prague, Czech, 5–7 December. p. 1–8.
[13] Bianconi E, Calvente J, Giral R, Mamarelis E, Petrone G, Andres Ramos-Paja C,
et al. Perturb and observe MPPT algorithm with a current controller based on
the sliding mode. Electr Power Energy Syst 2013;44:346–56.
[14] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimization of perturb and
observe maximum power point tracking method. IEEE Trans Power Electron
2005;20:963–73.
[15] Tafticht T, Agbossou K, Doumbia ML, Cheriti A. An improved maximum power
point tracking method for photovoltaic systems. Renewable Energy
2008;33:1508–16.
[16] Ahmed J, Salam Z. An improved perturb and observe (P&O) maximum power
point tracking (MPPT) algorithm for higher efficiency. Appl Energy
2015;150:97–108.
[17] Ishaque K, Salam Z, Lauss G. The performance of perturb and observe and
incremental conductance maximum power point tracking method under
dynamic weather conditions. Appl Energy 2014;119:228–36.
[18] Lin CH, Huang CH, Du YC, Chen JL. Maximum photovoltaic power tracking for
the PV array using the fractional-order incremental conductance method. Appl
Energy 2011;88:4840–7.
[19] Radjai T, Rahmani L, Mekhilef S, Gaubert JP. Implementation of a modified
incremental conductance MPPT algorithm with direct control based on a fuzzy
duty cycle change estimator using dSPACE. Sol Energy 2014;110:325–37.
[20] Abdulkadir M, Yatim AHM. Hybrid maximum power point tracking technique
based on PSO and incremental conductance. In: IEEE conference on energy
conversion, CENCON. p. 271–6.
[21] Morales-Acevedo A, Diaz-Bernabe JL, Garrido-Moctezuma R. Improved MPPT
adaptive incremental conductance algorithm. In: 40th annual conference of
the IEEE industrial electronics society, IECON. p. 5540–5.
Fig. 11. (a) Load variation profile, and (b) results of simulation with proposed [22] Kollimalla SK, Mishra MK. A new adaptive P&O MPPT algorithm based on FSCC
algorithm. method for photovoltaic system. In: International conference on circuits,
power and computing technologies, ICCPCT, 21st–23rd March 2013, Nagercoil,
India. p. 406–11.
Comparing with traditional INC technique used in the past, the [23] Lee Dong-Yun, Noh Hyeong-Ju, Hyun Dong-seok, Choy I. An improved MPPT
use of the proposed MPPT control improves the PV system perfor- converter using current compensation method for small scaled PV-
applications. 18th annual IEEE applied power electronics conference and
mance mainly in dynamic states, thus reducing the loss of power exposition, APEC ’03, vol. 1. p. 540–5.
and enhancing the tracking efficiency by five percent. The results [24] Rai AK, Kaushika ND, Singh B, Agarwal N. Simulation model of ANN based
of simulation are present. maximum power point tracking controller for solar PV system. Sol Energy
Mater Sol Cells 2011;95:773–8.
Drawbacks of this study are the slow time response at start-up [25] Alajmi BN, Ahmed KH, Finney SJ, Williams BW. Fuzzy logic-control approach of
in comparison to others techniques such as Sliding Mode Control a modified hill-climbing method for maximum power point in microgrid
and the lack of experimental work. standalone photovoltaic system. IEEE Trans Power Electron 2011;26
(4):1022–30.
Future work will continue by extending the principle of MPPT
[26] Belkaid A, Gaubert JP, Gherbi A. A fast and accurate maximum power point
technique proposed on other more efficient converter topologies tracking for photovoltaic power generation system. Optoelectron Adv Mater –
and taking into account the shading phenomenon in the test of Rapid Commun 2015;9(3–4):520–4. March – April.
developed MPPT system for an inhomogeneous illumination. [27] Farhat M, Barambones O, Sbita L. Real-time efficiency boosting for PV systems
using MPPT based on sliding mode. Energy Procedia 2015;75:361–6.
[28] Farhat M et al. A new maximum power point method based on a sliding mode
References approach for solar energy harvesting. Appl Energy 2016. http://dx.doi.org/
10.1016/j.apenergy.2016.03.055.
[29] Belkaid A, Gaubert JP, Gherbi A. An improved sliding mode control for
[1] Belkaid A, Gaubert JP, Gherbi A, Rahmani L. Maximum power point tracking for
maximum power point tracking in photovoltaic systems. J Contr Eng Appl
photovoltaic systems with boost converter sliding mode control. In: IEEE 23rd
Inform 2016;18(1):86–94. ISSN 1454-8658.
international symposium on industrial electronics (ISIE), Istanbul, Turkey, 1–4
[30] Elgendy MA, Zahawi B, Atkinson DJ. Assessment of the incremental
June. p. 556–61.
conductance maximum power point tracking algorithm. IEEE Trans Sustain
[2] Tey Kok Soon, Mekhilef Saad. Modified incremental conductance MPPT
Energy 2013;4(1):108–17.
algorithm to mitigate inaccurate responses under fast-changing solar
[31] Safari A, Mekhilef S. Implementation of incremental conductance method with
irradiation level. Sol Energy 2014;101:333–42.
direct control. In: Proceedings of the IEEE region 10th conference
[3] deCesare G, Caputo D, Nascetti A. Maximum power point tracker for portable
TENCON. Bali: IEEE Xplore Press; 2011. p. 944–8. http://dx.doi.org/
photovoltaic systems with resistive-like load. Sol Energy 2006;80:982–8.
10.1109/tencon.2011.6129249. Nov. 21–24.
[4] Chu CC, Chen CL. Robust maximum power point tracking method for
[32] Fangrui L, Shanxu D, Fei L, Bangyin L, Yong K. A variable step size INC MPPT
photovoltaic cells: a sliding mode control approach. Sol Energy
method for PV systems. IEEE Trans Industr Electron 2008;55(7):2622–8.
2009;83:1370–8.
[33] Her-Terng Y, Qin-Cheng L, Chin-Tsung H. Maximum power point tracking and
[5] Houssamo I, Locment F, Sechilariu M. Maximum power tracking for
optimal Li-ion battery charging control for photovoltaic charging system.
photovoltaic power system: development and experimental comparison of
Comput Math Appl 2012;64:822–32.
two algorithms. Renew Energy 2010;35:2381–7.
[34] Rizzo Santi Agatino, Scelba Giacomo. ANN based MPPT method for rapidly
[6] Bhatnagar AP, Nema BRK. Conventional and global maximum power point
variable shading conditions. Appl Energy 2015;145:124–32.
tracking techniques in photovoltaic applications: a review. J Renew Sustain
[35] Ishaque K, Salam Z, Amjad M, Mekhilef S. An improved particle swarm
Energy 2013;5:1–23.
optimization (PSO) based MPPT for PV with reduced steady state oscillation.
[7] Ishaque K, Salam Z. A review of maximum power point tracking techniques of
IEEE Trans Power Electron 2012;27(8):3627–38.
PV system for uniform insolation and partial shading condition. Renew Sustain
[36] Petreus D, Moga D, Rusu A, Patarau T, Daraban S. A Maximum Power Point
Energy Rev 2013;19:475–88.
tracker for photovoltaic system under changing luminosity conditions. In: IEEE
[8] Eltawil MA, Zhao Z. MPPT techniques for photovoltaic applications. Renew
international symposium industrial electronics. ISIE; 2010. p. 556–61.
Sustain Energy Rev 2013;25:793–813.
[37] Bizon Nicu. Global extremum seeking control of the power generated by a
[9] Subudhi B, Pradhan R. A comparative study on maximum power point tracking
photovoltaic array under partially shaded conditions. Energy Convers Manage
techniques for photovoltaic power systems. IEEE Trans Sustain Energy 2013;4
2016;109:71–85.
(1):89–98.
[38] Yu GJ, Jung YS, Choi JY, Kim GS. A novel two-mode MPPT control algorithm
[10] Salam Z, Ahmed J, Merugu BS. The application of soft computing methods for
based on comparative study of existing algorithms. Sol Energy
MPPT of PV system: a technological and status review. Appl Energy
2004;76:455–63.
2013;107:135–48.
[39] Colak I, Kabalci E, Bal G. Parallel DC-AC conversion system based on separate
[11] Kollimalla SK, Mishra MK. A new adaptive P&O MPPT algorithm based on FSCC
solar farms with MPPT control. In: 8th international conference on power
method for photovoltaic system. In: International conference on circuits,
electronics- ECCE Asia. Korea: The ShillaJeju; 2011.
power and computing technologies, Nagercoil, India, 21–23 mars. p. 406–11.
530 A. Belkaid et al. / Applied Energy 179 (2016) 523–530

[40] Colak I, Demirtas M, Kabalci E. Design, optimisation and application of a [43] Jain S, Agarwal V. Comparison of the performance of maximum power point
resonant dc link inverter for solar energy systems. Compel (Int J Comput Math tracking schemes applied to single-stage grid-connected photovoltaic systems.
Electr Electron Eng) 2014;33(5):1761–76. Electr Power Appl, IET 2007;1(5):753–62.
[41] Kim IS. Sliding mode controller for the single-phase grid-connected [44] Femia N, Petrone G, Spagnuolo G, Vitelli M. Optimizing sampling rate of P&O
photovoltaic system. Appl Energy 2006;83:1101–15. MPPT technique. In: Proc. IEEE PESC. p. 1945–9.
[42] Hussein KH, Muta I, Hoshino T, Osakada M. Maximum photovoltaic power
tracking: an algorithm for rapidly changing atmospheric conditions. IEE Proc
Gener, Trans Distrib 1995;142(1):59–64.