822 IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 9, NO.

3, MAY 2019

A Multi-Level Control and Optimization Scheme for

Islanded PV Based Microgrids: A Control
Frame Work
Philip Mathew , Sreedhar Madichetty , and Sukumar Mishra

Abstract—This paper proposes a multi-level control and opti- countries is the isolated geographical locations of rural areas
mization scheme, including grid control and node control, for an that make it extremely difficult to connect them to the main
islanded 48-V PV-based low-voltage dc (LVdc) microgrid that aims grid.
to overcome the drawbacks of centralized and decentralized con-
trol schemes. The analyzed microgrid includes a 20-kW rooftop
solar system as the main power source with distributed compen- A. Motivation
sation systems. The central supervisory controller is responsible
for updating grid characteristics and sending/receiving informa- Lack of access to electricity has a direct link to poor economic
tion to/from local node controllers, which are responsible for bus progress as these power-absent areas are unable to contribute to
voltage regulation and energy management. The control hierar- the country’s overall productivity [1]. Load shedding occurs dur-
chy features optimized and safe operation (charge and discharge) ing peak demand hours of the day when the amount of power
of storage devices in dc microgrids. The paper also demonstrates being generated is not sufficient to provide for the power be-
the application of battery-supercapacitor systems to absorb sys-
tem transients during load changes. The simulation showcases the ing drawn from the grid. In such scenarios, people staying in
continuous flow of information and decision processes via each these areas are left helpless during the blackout. With each MG
level of control, while simultaneously taking the constraints of each installed, the stress on the main grid significantly decreases re-
subsystem into consideration. The scheme has been simulated in sulting in less load shedding and better standards of living. Most
MATLAB/Simulink environment for various case studies to evalu- renewable sources produce dc, which, in the case of ac MGs,
ate system stability and robustness. Further, the proposed scheme
has been tested experimentally with its prototype and its results needs to be converted into ac, resulting in conversion losses [2].
are explored. In addition, loads in households and office spaces are mostly
dc driven. In this scenario, wherein both the energy generation
Index Terms—Battery supercapacitor management, control
frame work, hybrid energy storage (HES) systems, islanded low-
and consumption sides use dc, a dc MG is a much more feasible
voltage dc (LVdc) microgrid (MG), multi-level. option. Controlling the flow of power based on load demand,
left-over charge in the compensation system, and unit pricing in
I. INTRODUCTION the most optimized manner is an important part of MG control
PART from collectively moving the entire world toward a objectives. The concept of low-voltage ac/dc (LVdc/LVac) MGs
A greener tomorrow, microgrids (MGs) also play a vital role
in lighting up parts of the world that have never even seen a tube
has been proposed in recent years to enhance the reliability,
power quality, and efficiency of the distribution power systems
light work. At the country level, according to [1], India alone [3]. The installation of PV-based dc MGs is becoming more apt
has a little less than one-third of the global deficit, followed in areas where other sources of renewable energy (such as wind
by Nigeria and Ethiopia for electricity. The 20 highest access- and hydro) are not available.
deficit countries for electricity account for 80% of the global
deficit. An important reason for the lack of electricity in these B. Literature Review
An overview of the existing control strategies and architec-
Manuscript received December 1, 2018; revised January 14, 2019; accepted tures on dc MGs can be found in [4] and [5]. Lu and Guer-
January 30, 2019. Date of publication February 26, 2019; date of current ver-
sion April 19, 2019. This work was supported in part by the DST, Govt. rero proposed an alternative to decentralized high bandwidth
of India under its project “Identification and Demonstration of Cost Effec- commuication control schemes by implementing droop control
tive Technologies to Maximize Habitat Energy Self-Sufficiency” with file based on low bandwidth commuication between nodes to re-
no. TMD/CERI/BEE/2016/096(G) and NPDF scheme of SERB with file no.
SERB/NPDF/2017/000568. (Corresponding author: Sreedhar Madichetty.) duce data collision and line current sharing inaccuracy [6]. In
P. Mathew is with the Department of Electrical and Electronics Engineering, [7], Setiawan et al. proposed a multi-level control scheme for
National Institute of Technology, Karnataka 575025, India (e-mail:, joemat line-current sharing using bus-voltage regulation. However, both
[email protected]).
S. Madichetty and S. Mishra are with the Department of Electrical Engi- of these do not discuss the perfomance of the scheme when it
neering, Indian Institute of Technology, Delhi 110016, India (e-mail:,sreedhar. needs to handle optimization of compensating storage devices,
[email protected]; [email protected]). such as batteries and supercapacitors, as is the case in any real-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. istic MG. In [18], the application of flexible multi-state switches
Digital Object Identifier 10.1109/JPHOTOV.2019.2897579 in a three-phase multi-level PFC rectifier has been explored. In
2156-3381 © 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.

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 MATHEW et al.: MULTI-LEVEL CONTROL AND OPTIMIZATION SCHEME FOR ISLANDED PV BASED MICROGRIDS 823

Fig. 1. Analyzed LVdc microgrid.

[8], a supervisory scheme of control with a central controller 1) a complete multi-level control architecture (system level
that receives grid characteristic inputs from remote terminal and node level) that proposes to combine the merits of
units and sends control outputs to local controllers (LCs) is several previously proposed schemes;
discussed. The analyzed scheme, however, has not discussed 2) an optimization scheme for controlling power flow for
results during system transients due to load changes. Moreover, charge and discharge of batteries and super-capacitors
the supervisory controller is responsible for computation of ad- based on characteristics, such as load demand and SOC;
mittance (Ybus ) matrix, which increases the burden on the central 3) application of battery-supercapacitor hybrid energy stor-
controller significantly for larger systems. Phurailatpam et al. age (HES) systems and their related control for the ab-
proposed a MG scheme optimized for rural applications specif- sorption of sharp transients in the system;
ically for a village in Chattisgarh, India [9]. The MG is highly 4) most importantly, this paper discusses how the proposed
simplified, consisting of two generation systems (solar and wind scheme can be easily expanded and modified for larger
turbine), a battery and a few loads, and focuses more on the eco- systems.
nomic aspects of implementing an MG. In [10], a distributed
control scheme wherein current sharing is achieved through dy- D. Organization
namic consensus between neighboring nodes is discussed. The
The remainder of this paper is organized as follows: Section II
model includes four source agents inter-connected through tie
analyses the proposed system and explains, in detail, the control
lines and loads connected to each node. Although the control
and optimization functions of the proposed scheme. Section III
scheme is dynamic and responsive, it has not been tested for
develops and analyses mathematical models of the various sub-
more complex MGs with more constraining parameters. With
systems present in the grid. Section IV shows the simulations of
the recent developments in PV-based MGs, the literature on the
the proposed scheme in MATLAB/Simulink Environment and
control and stability for the same has grown as well. In [20], the
conducts a stability analysis of the same for different scenarios.
“Generation-Storage Coordination” of a radial dc MG with high
Section V discusses the experimental results of the proposed
penetration of PV generators is discussed. The control algorithm
system. The paper finally concludes in Section VI with the dis-
aims to balance the state of charge (SOCs) of storage devices
cussion of de/merits and the future scope for improvements in
to prevent their overuse. The stability of PV generators has also
the proposed scheme.
been explored, such as in [17], in which a non-linear describing
function have been used to conduct the accurate stability anal-
ysis of PV generators with consideration of P&O based power II. MG SYSTEM AND CONTROL SCHEME
control. The studied ring 48-V dc MG is shown in Fig. 1. Since the
MG is in islanded mode, power balancing is carried out solely
C. Key Contributions
by control of distributed energy resources and compensation
The salient features of this paper are as follows: systems. The MG consists of three dc buses connected to their

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 824 IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 9, NO. 3, MAY 2019

corresponding subsystems. Loads are distributed throughout the The order of compensation is such that the VCS and CCS
system and connected to each bus. Maximum possible load on compensate power demand of the MG and supercapacitors com-
the system is 20 kW. pensate for these systems during sharp transients so as to prevent
any damage to them. The main purpose of energy storage in MG
A. Roof-Top Solar Photovoltaic (SPV) System with renewable energy resources is to maintain the quality of
power supply by compensating power generation and consump-
A 20-kW SPV System (constructed out of eighty 250-W tion mismatch. Battery systems are commonly used in order to
modules) under variable irradiance and temperature is connected regulate the oscillating output of SPV power plants [16]. How-
to Bus 1 along with its battery storage. The dc/dc boost converter ever, recent research has shown that even though batteries with
performs maximum power point (MPP) tracking (MPPT) to high specific power are available, the combination of batteries
extract maximum power from the SPV array using incremental and SCcap s can achieve a better performance. By handling the
conductance (InC) algorithm with an added integral regulator sharp power surges using supercapacitors, we can improve the
robustness. This can be replaced with more advanced MPPT life span of the battery by reducing the current stresses on it.
techniques if needed. The InC algorithm works on the principle
that the derivative of the SPV module power is zero at MPP,
positive at left of MPP, and negative at right of MPP [11]. C. Proposed Control Hierarchy
The InC loop continuously changes the value of Vref based The proposed control architecture is shown in Fig. 2(a). Lo-
on present and past values of voltage (Vk , Vk −1 ) and current cal bus controllers are responsible for bus voltage regulation and
(ik , ik −1 ). Vref determines the output voltage of the converter carry out subsystem-specific functions, such as MPPT in solar
and stops being updated once the InC of the system equals photovoltaic system, and keep a check on the present subsystem
the instantaneous conductance of the system, as given in the status. This includes SOC values, bus voltage, converter cur-
following: rents, and power generation in the case of generation nodes. The
supervisory control (SC) workflow algorithm has been shown in
dP d (I.V ) dI I
=0 ⇒ =0 ⇒ =− (at Pm pp ) . (1) Fig. 2(b). This information is relayed to the SC, which maintains
dV dV dV V an overall model of the system. The SC, using this information,
The battery storage allows the SPV to operate under MPPT determines the power demand and power generation in the MG.
continuously as it can store extra energy being generated, if The SC workflow algorithm has been shown in Fig. 2(b).
any, through a bidirectional buck/boost converter. The battery is LCs are shown as L1, L2, . . . , Ln. To streamline the working
also responsible for regulating the bus voltage by operating its of the entire system, the MG compensation operation has been
converter in constant voltage (CV) mode, allowing the SPV’s divided into two modes, viz., battery compensation mode and
converter to operate in constant compensation (CC) mode and high compensation mode (HCM). These two modes induce cor-
inject maximum power into the MG. The operation of bidirec- responding bus modes through messages sent from the SC to
tional buck/boost converter is studied further in Section III. Ln. If in surplus, the SC determines which compensation sys-
tem to turn into a load to feed this extra power. This is decided
by using a charge priority list, which states the order in which
B. Compensation Systems
energy-storage devices of the system must be charged. During
This paper divides the grid-compensation system into two deficit generation conditions, the compensation systems go into
parts, variable compensation (VC) and CC. VC systems (con- discharge mode, decided by the discharge priority list, to supply
nected at Bus 2) include electric vehicles (EVs) and are variable this deficit, while the SPV system runs in MPPT mode. Once this
in the sense that at any given point of time, the number of decision has been done, the SC simply sends a charge signal to
EVs could be either none or several. CC systems, on the other the corresponding Ln. In HCM, the mismatch between genera-
hand, are permanently-installed batteries (connected at Bus 3) tion and demand cannot be compensated by the battery systems.
that are always available to be used for grid compensation (if In a typical MG, this would be done through load shedding (for
their present SOC allows it). The batteries considered for this PG  PL ) or derated power generation (for PG  PL ), where
paper are of the Li-ion type as they have proven to have higher PL , PG are load power and generated power, respectively.
energy density and efficiency as compared with other battery A new feature of the proposed scheme is the ability to change
types (such as lead-acid, NiCd, or NiMH) [12]. This paper also the voltage level of the entire MG with just a single command
focuses on the use of supercapacitors (SCcap ) to aid batteries to the SC. The SC sends this reference voltage value to all LCs,
connected to the MG during sharp current transients that might hence making them raise/lower their bus voltage levels accord-
harm these sensitive devices. The advantages of using SCcap s in ingly. This feature is used in HCM, in which the voltage level
various aspects of MGs have been discussed in [13]–[15]. The (Vbus ref ) of the MG can be brought down/up by an accept-
elements used in grid-compensation system and their ratings are able value, so as to try to accommodate more load/generation
listed below. in excess conditions. This prevents unnecessary load shedding
VCS: Battery (12 V, 42 Ah, Li-ion) or generation waste. Being a centralized system, inaccuracy
Supercapacitor (500 F, 32 V). due to data collision between neighboring nodes is prevented.
CCS: Battery (24 V, 42 Ah, Lead-Acid) On the other hand, most centralized systems produce signifi-
Supercapacitor (500 F, 32 V). cant burden on the central controller which, in turn, increase

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 MATHEW et al.: MULTI-LEVEL CONTROL AND OPTIMIZATION SCHEME FOR ISLANDED PV BASED MICROGRIDS 825

Fig. 3. bidirectional dc–dc converter.

this case, which is another advantage of the proposed scheme

over distributed control schemes.

III. ANALYSIS OF MG CONTROL LOOPS

The control systems used in the proposed scheme have been
discussed in this section. While the initial parts discuss each
node and their subsystems, the last part discusses how each
individual part has been woven together to obtain the entire
multi-level control hierarchy scheme.

A. Bidirectional Buck–Boost Converter

The circuit of the converter is shown in Fig. 3. The converter
circuit also includes parasitic resistances for inductors and ca-
pacitors, denoted by rL and rc , respectively. Switches S1 and S2
represent power switches. L, C, u, iL , Vc , ic , R are the inductor,
capacitor, input voltage, inductor current, voltage across capac-
itor, capacitor current, and load, respectively. The state-space
model (SSM) for the converter can be represented by
x = Ax + Bu (2)
y = Cx + Du (3)
where A, B, C, D are the system matrices, x is the state-variable
matrix, x is the derivative state-variable matrix, y is the output
matrix, and u is the input matrix.
The SSM of such a converter is similar to that of a simple
boost converter, owing to the fact that the two circuits are similar
except for the replacement of the free-wheeling diode with a
Fig. 2. (a) Multi-level control hierarchy. (b) SC workflow algorithm. power switch. The SSM for the converter can be derived from
averaging the individual SSMs for when 1) S1 is ON, S2 is OFF
for d duty cycle [shown in (4)]; and 2) S1 is OFF, S2 is ON
the chances of failure. The proposed control scheme, however, for d = 1 − d duty cycle [shown in (5)]. The state variables
shares the burden of computation and regulation between LCs selected are x1 = iL (inductor current) and x2 = Vc (capacitor
and SC to maximize efficiency and reduce chances of failure. voltage).
The SC does not have to deal with handling the inner workings  −r L  1
L 0
of each node, like the central controller does.  L
xON = −1
x+ u (4)
The nature of the proposed scheme allows it to be extended 0 C (R +r c ) 0
for larger systems. If another node has to be added to the MG, ⎛    ⎞ 1
−1
another entry must be made into the SC lists. This entry would 1
rL + R .r c
R +r c 1 + rc
R +r c ⎠

xOFF = ⎝ L L
x+
L
u.
include the nature of the node (generation, load, or compensa- R −1 0
tion) and any other information required by the SC during its C (R +r ) c C (R +r ) c

workflow algorithm. No changes need to be made for LCs, in (5)

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 826 IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 9, NO. 3, MAY 2019

The averaged SSM model can be derived using

A = d . AON + d . AOFF (6)
B = d . BON + d . BOFF (7)
to produce
⎛     ⎞
−1
1
rL (1 − 2d) + R .r c
.d 1+ rc
.d
x = ⎝L R +r c L R +r c ⎠x
R .(1−d) −1
C (R +r c ) C (R +r c )
 
1
+ L
u. (8)
0
The C and D matrices can be derived in a similar fashion to
give the following:
1
10
y= x+ L u. (9)
01 0
This final averaged SSM represents the converter through one
complete switching cycle of period Ts . By controlling the gating
pulses to switches S1 and S2, the converter can operate in boost
(forward) or in buck (reverse) modes. This control is decided by
the SC, as shown in Section III-B.
Fig. 4. Fully active HES system.
B. Battery-Supercapactior (HES) System
HES systems form an integral part of the proposed scheme TABLE I
PI CONTROL PARAMETERS
as 1) the supercapacitor absorbs sharp transients during peak
demand in the microgrid; and 2) the battery helps regulate the
bus voltage at each node. In the absence of these systems, power
regulation and balancing in the MG must be achieved by con-
trolling the outputs of generation units, such as the SPV unit.
However in the case of peak demand, generation units must pro-
duce the maximum power possible. Hence, the application of
such HES systems allows us to shift the burden of control from
generation units. The battery-supercapacitor model is used in
the VCS and CCS nodes. The subsystem and its control scheme
are shown in Fig. 4.
The chosen topology for the HES system is that of fully active
HES, which consists of a battery and a supercapacitor, connected
to the same dc bus through bidirectional dc/dc converters. The
proportional integral (PI) controller has been tuned accordingly
with gain values as shown in Table I. The PI control for the
battery includes a simple integral regulator, which takes the
error (Vbus − Vbus ref ) as its input signal. The supercapacitor
control is a cascaded PI scheme in which the Iscref value is
controlled using and is brought to zero in steady state. The
reference value is also controlled depending on characteristics,
such as the discharge current of the battery, SOC of the battery,
and the left-over charge on the supercapacitor. All parameters
shown in Fig. 4 are self-explanatory, where Dx represents the
diode, and Qx represents the switch of converter.

C. SPV Battery Storage System

A detailed circuit figure for the subsystem is shown in Fig. 5.
At this node, the converter (unidirectional) of the SPV panel is
operated under MPPT so as to extract maximum solar power. Fig. 5. SPV-battery hybrid system.

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 MATHEW et al.: MULTI-LEVEL CONTROL AND OPTIMIZATION SCHEME FOR ISLANDED PV BASED MICROGRIDS 827

Fig. 7. (a) Conventional transient detection method in HESS. (b) Simulation

results for SPV and battery system X axis: Time (s) with intervals of 0.2 s.

LC works on SOCs of various storage devices at that node) and

the corresponding “Local Signals” from the LC that, in turn,
Fig. 6. Condensed control scheme. affect the local subsystem. Hence, there exist closed loops in
the control system of the proposed scheme at various levels.
The converter (bidirectional) of the battery works in CV mode,
regulating the bus voltage with a PI controller. The battery IV. SIMULATION RESULTS AND ANALYSIS
storage is charged from the microgrid in excess power con-
ditions and, in this scenario, the charging current of the battery The simulation results for different nodes and the entire MG
is controlled so as to regulate bus voltage. All parameters shown have been presented and discussed in this section.
in Fig. 5 are self-explanatory, where Dx represents the diode,
and Qx represents the switch of converter. The switching be- A. CCS/VCS Node Simulation
tween charge and discharge modes of the battery is controlled
The CCS and VCS nodes consist of battery-supercapacitor
by the inductor and capacitor (LC), which determines whether
HES systems that are controlled by their respective LCs on the
Vbus ≥ Vbus ref and controls the buck/boost switch accordingly.
basis of signals received from the supervisory control. The bat-
The PI control scheme for the battery is similar to that in the HES
tery is responsible for maintaining a regulated bus voltage at the
system and consists of a single integral regulator. The mode of
node, while the supercapacitor provides power during transients
the battery also depends on a control signal mode_pv, which is
so as to limit the effect of these sharp transients on the battery.
sent from the supervisory control, in case the SPV battery needs
The transient detection has been done using a conventional low-
to be used for power compensation in the grid.
pass filter (LPF) approach as shown in Fig. 7(a). However, the
proposed HES system is slightly modified by adding a ramp
D. Condensed MG Control Scheme signal to the battery converter with the supercapacitor’s SOC
A condensed control diagram (see Fig. 6) is given to explain as another sensing parameter as shown in Fig. 4. Whenever a
the various control loops in the proposed system. At the MG sudden load change occurs, at that particular instant, LPF atten-
level, the closed loop consists of continuous feedback from uates the signal passing through battery converter. Hence, the
LCs and corresponding outputs from the SC. It can be seen signal will pass through the supercapacitor. This processed sig-
that the SC gives outputs called “Mode Signals” that affect the nal is decreased, once the battery reaches to its steady state. The
microgrid. The decisions taken by the SC are based on data steady-state current of the battery is obtained by controlling the
sent from the LCs through what is called as “Status Signals.” ramp signal. The simulation results of a step-load response have
At the node level, the closed loop is formed by the “Feedback been shown in Fig. 7(b). It can be seen that the supercapacitor
Signals” from the local subsystem (for example, the VCS node delivers power only while battery current (ib ) is in transients

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 828 IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 9, NO. 3, MAY 2019

Fig. 8. Simulation results for SPV and battery system X axis: Time (s) with
intervals of 1 s.

and gradually decreases to zero as the battery reaches the load

demand value.
This entire process takes after a load change that has occurred
in the system. The time window this process takes can be further
increased by maintaining the SOC in supercapacitor, i.e., exter-
nal charge circuit. The response time of the battery PI controller
can be made slower or faster depending upon required output.
The maximum variation observed in bus voltage is ±5% and
this can be mitigated using larger supercapacitors.

B. SPV Batter Storage Node Simulation

The simulation of the SPV-battery node has been carried
out with various conditions, such as sudden increment in ir- Fig. 9. (a) Step-load change power measurements X axis: Time (s) with inter-
radiance (250–1000 W/m2 ), sudden decrement in irradiance vals of 1 s. (b) Bus measurements for VCS voltage variation X axis: Time (s)
with 0.2s intervals.
(1000–250 W/m2 ), and sudden increment in load and sudden
decrement. Further, the system has been tested with variable
temperature as well. The above-specified conditions are simu- is working with the three tier principle, i.e., identity tier, data
lated and are shown in Fig. 8. tier, and app tier. Where application (app) can act as a user inter-
During step-load and irradiation/temperature changes, the face with user name and password with data stored in server as
scheme allows only a ±0.5% variation in bus voltage. From shown in Fig. 8. In this paper, MODBUS communication proto-
the simulation results, it is clear that as the solar irradiation col (in hardware) has been used for all converters. The proposed
reduces, the charging current of the battery reduces, and vice system has been tested with various conditions as follows:
⎫
versa. The priority has been given for SPV before the battery. In Vbus1 ref = Vbus1 + ((Vbus1 − Vbus3 ) + (Vbus1 − Vbus2 ))⎪ ⎬
case the battery is fully charged, the SPV operates in de-rated
Vbus2 ref = Vbus2 + ((Vbus2 − Vbus3 ) + (Vbus2 − Vbus1 )) .
mode. The maximum power from SPV is extracted by using InC ⎪
⎭
algorithm, explained in Section III. Please consider the X-axis Vbus3 ref = Vbus3 + ((Vbus3 − Vbus1 ) + (Vbus3 − Vbus2 ))
as time in seconds unless specifically stated otherwise. (10)
1) Step-Load Variation: During a step-load change, the LCs
C. LVDC MG Simulation try to compensate for the changes in bus voltage through their
The proposed scheme is tested for its stability and accuracy respective PI control schemes. In the event of less power be-
for different system variations, such as 1) irradiance and temper- ing generated, the SC sends discharge signals to the CCS and
ature variation of SPV cell; 2) step-load changes in the system; VCS, hence increasing the power they inject into the MG. The
and 3) variation in VCS (change in number of connected EVs). simulation results for step-load variations are shown in Fig. 9(a).
Further, as initially stated, the voltage across each converter is During the simulation, the irradiance and temperature values
maintained by its LC and power is managed by the dynamic of the SPV panel have been simultaneously changed as well.
consensus average voltage technique with the help of central- 2) VCS Voltage Variation: In this simulation, the voltage of
ized command from supervisory control. In addition, the system the VCS is changed in a step manner. The VCS voltage depends

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 MATHEW et al.: MULTI-LEVEL CONTROL AND OPTIMIZATION SCHEME FOR ISLANDED PV BASED MICROGRIDS 829

on the total number of EVs plugged in and can change at any

instant of time. This change can be recreated by changing the
terminal voltage of the VCS system and observe the corre-
sponding changes in the MG. For analysis, the voltage has been
changed by a step of 12 V (increase and subsequent decrease).
The simulation results, shown in Fig. 9(b), confirm that the sys-
tem can adequately compensate for such changes by allowing a
change in bus voltage no more than ±6.25%. A vital reason for
this compensation ability is the presence of the supercapacitor
which can work extremely well, as is shown, during transients.

V. EXPERIMENTAL ANALYSIS
The prototype of the proposed system is developed and tested
for various conditions. The experimental system mainly consists
of three blocks, which are roof top solar system, VCS and PCS.
For the purpose of experimentation, solar array simulator (SAS)
(100 V, 20 A) is used to emulate the characteristics of roof top
solar system. A 48-V, 20-Ah Li-ion battery is used in conjunc-
tion with SAS either to charge the excess energy or to support
the grid. The PCS and VCS block consist of dc–dc bidirectional
converters (50 V, 20 A). The PCS is equipped with a super
capacitor of 36 V, 500 F. The entire dc MG system is imple-
mented on SILOV dc grid research test bed [19]. The proposed
control logic is implemented on Labview-SP1-2017 platform
with 64-bit, Intel i5 processor based Windows 10 operating
system.
Fig. 10. (a) Battery-supercapacitor hybrid energy storage system with its
A. Case 1: Battery-Supercapacitor System for Sudden implemented control scheme [X-axis—Time (s), Y-axis—Current (5 A/Div)].
(b) Battery-supercapacitor hybrid energy storage system with its change in SOC
Changes in Load [X-axis—Time (s), Y-axis—Current (5 A/Div)].
The importance of combined operation of battery su-
percapacitor is well accepted by the community due to its
inherent energy management capabilities. For sudden load
changes/surges, supercapacitor will come into the picture, while
battery slowly ramps up its current to the desired level. Initially,
experiments have been conducted for sudden load increments
and decrements, and the results are shown in Fig. 10(a). Further,
some experiments have been conducted with various levels of
SOCs and the results are explored in Fig. 10(b). At the node
of HESS, when the load is suddenly increased from 0 to 5 A,
the initial transient has been absorbed by the supercapacitor
while the battery current ramps up slowly. The system has been
investigated with sudden load loss from 5 to 0 A; in this case,
supercapacitor tries to energize. These two cases have been
shown in Fig. 10(a). Some other cases have been tested and Fig. 11. PV emulator I–V and P–V characteristics with load line and MPP.
denoted in Fig. 10(b); when the supercapacitor is fully charged X axis: Voltage (V) with intervals of 2 V. Y axis: Current (A) with intervals of 1
A. Y axis: Power (W) with intervals of 10 W.
(100% SOC) and battery is at 80% SOC, a sudden load change
occurs. Because of this, supercapacitor comes into the picture to
done with a separate power supply as well. If the supercapacitor
handle the transient. In the meantime, battery starts to ramp up
charges with separate power supply, the ramp time of battery
to the required voltage and takes almost 5 s of time as shown in
extends to almost 10–15 s with the same capacity. This type of
Fig. 10(b). This time gap is because its SOC was 100%. When
scenario may particularly be suited for EV node, i.e., VCS node.
another sudden surge takes place, the supercapacitor was not
fully charged, i.e., SOCs is 90% and can handle the transient
only for 2 s, which is shown in Fig. 10(b). From this, one can B. Case 2: Step Decrease/Increase in Load Current
conclude that if sufficient time is allowed for supercapacitor to The roof-top SPV system (using SAS) with battery en-
get fully charged, the ramp time of battery will be extended, ergy storage is set to run at variable irradiance from 100 to
which subsequently improves the life of battery. This may be 1000 W/m2 , with open-circuit voltage of 44 V, short-circuit

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 830 IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 9, NO. 3, MAY 2019

current drawn from the PCS node also increases from 4 to 5.5 A
to maintain the MPP. Even though load is suddenly increased,
the dc-bus voltage remains within the limits of 48 V as shown in
Fig. 12(a). Another case has been considered for load changes
at different time intervals as shown in Fig. 12(b). At time t =
1 s, load on Bus 1 is suddenly reduced from 6 to 4 A, and at
that particular instant, battery current at Bus 2 tries to increase
from 3 to 3.5 A, and the remaining current is delivered to su-
per capacitor. From this experiment, it is found that the dc-bus
voltage is still maintained as shown in Fig. 12(b). In the same
case, at t = 20 s, when the overall load suddenly increases, all
the three buses start to respond, while Bus 1 from 3 to 5 A, Bus
2 from 3 to 6 A, and Bus 3 from 3 to 5 A, respectively. In both
the cases, the bus voltage is constant and is maintained at 48 V.
The above-mentioned analysis is shown in Fig. 12(b). Hence,
from the above-mentioned analysis one can conclude that, as
the load increases, the power given by battery increases, and
vice versa. However, in this case MPPT algorithm slightly devi-
ates (around 2%) from MPP by drawing more current, which is
quite acceptable in practical cases. Further, the proposed system
is tested for sudden change in irradiance, i.e., sudden decrement
to 800 W/m2 , and then to 500 W/m2 , and then to 700 W/m2 ,
and then to 900 W/m2 . For those sudden changes in SPV pen-
etration, the deficit power is either supplied by battery storage
system or by other connected buses.
Another important feature of the proposed scheme is that it
has the inherent capability for accurate load sharing as well.
When the SC issues the command to the LCs for current, all
the controllers can respond at a time and can operate as per the
requirement as shown in Fig. 12(c). At t = 2 s, when SC passes
a command of decrement in load for all buses from 3 to 0 A, all
systems are responded with in 10 ms by maintaining its dc-bus
voltage, while the same is true for vice versa as well. Hence,
from this analysis, it can be concluded that the MG can be
controlled with a single command from SC, while maintaining
its dc bus.

VI. CONCLUSION
The scheme proposed in this paper has been analyzed and its
control and optimization capabilities have been explored. The
proposed scheme is a possible improvement over purely central-
ized and decentralized schemes that suffer from central failures
and data collision/inaccuracy, respectively. An important feature
Fig. 12. (a) Gradual load increment on three node scenario. (b) Sudden load
of the proposed scheme is the ability to expand itself to larger
increment and decrement at VCS node. (c) Implementation of load sharing MGs with simple entries to the supervisory control’s work-
phenomenon with proposed scheme. flow. The scheme has been applied to a PV-based MG and an-
alyzed through simulation in MATLAB/Simulink environment
current of 5.5 A at nominal temperature 298 K, and with zero- for bus voltage regulation and load sharing between generation-
series resistance as shown in Fig. 11. With the implementation of compensation systems. The application of HES systems in the
maximum power tracking algorithm, the peak power obtained MG for absorbing system transients and aiding battery systems
at 1000 W/m2 is 200 W with its MPP current at 5.2 A. The has been experimentally analyzed. The experimental analysis
obtained results with its parameters are shown in Fig. 11. The checks the stability, response time, accuracy of the scheme in
performance of the proposed converter is initially tested with hardware, and the results have been discussed. Future research
sudden increment in load and then with sudden decrement in may include applying the proposed control scheme in a grid-
load. When the load is gradually increased from 2 to 5 A, the connected dc MG where power can be supplied to/from the main
current supplied to the other two buses are also increased and grid to achieve compensation.

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.
 MATHEW et al.: MULTI-LEVEL CONTROL AND OPTIMIZATION SCHEME FOR ISLANDED PV BASED MICROGRIDS 831

REFERENCES Philip Mathew is currently working toward the

B.Tech. degree at electrical and electronics engi-
[1] The World Bank, “The state of electricity access report (SEAR) 2017,” neering at the National Institute of Technology,
The World Bank, Rep. no. 114841. Washington, DC, USA, vol. 2, 2017. Karnataka, India.
[2] P. Savage, R. R. Nordhaus, and S. P. Jamieson, “DC microgrids: Benefits Throughout his undergraduate studies, he has
and barriers,” Yale School of Forestry and Environmental Studies - From been leading development teams working on several
Silos to Systems: Issues in Clean Energy and Climate Change. Manhattan, environment-friendly products, such as electric grass
KS, USA: Kansas State University, 2013, pp. 51–66. cutters and electric vehicles, which aim to ease the
[3] S. Peyghami, H. Mokhtari, and F. Blaabjerg, “Decentralized load sharing transition of the community into an eco-friendly fu-
in an LVDC microgrid with an adaptive droop approach based on a super- ture. At present, he is responsible for conducting an
imposed frequency,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 5, Energy Audit of his campus to optimize distribution
no. 3, pp. 1205–1215, Sep. 2017. lines, identify faulty loads, and transformers so as to minimize wastage, reduce
[4] T. Dragičević, X. Lu, J. C. Vasquez, and J. M. Guerrero, “DC microgrids— power cuts, and enable the integration of solar energy into the campus grid.
Part I: A review of control strategies and stabilization techniques,” IEEE His research interests include smart grid control and optimization technologies,
Trans. Power Electron., vol. 31, no. 7, pp. 4876–4891, Jul. 2016. integration of renewables into modern power systems, and application of power
[5] T. Dragičević, X. Lu, J. C. Vasquez, and J. M. Guerrero, “DC Microgrids- electronics in power systems.
Part II: A review of power architectures, applications, and standardization
issues,” IEEE Trans. Power Electron., vol. 31, no. 5, pp. 3528–3549,
May 2016.
[6] X. Lu, J. M. Guerrero, K. Sun, and J. C. Vasquez, “An improved droop con-
trol method for DC microgrids based on low bandwidth communication Sreedhar Madichetty (M’16) received the B.Tech.
with DC Bus voltage restoration and enhanced current sharing accuracy,” degree from the Jawaharlal Nehru Technologi-
IEEE Trans. Power Electron., vol. 29, no. 4, pp. 1800–1812, Apr. 2014. cal University, Anantapur, Anantapuramu, India, in
[7] M. A. Setiawan, A. Abu-Siada, and F. Shahnia, “A new technique for si- 2010, and the M.Tech (Topper and Gold Medal)
multaneous load current sharing and voltage regulation in DC microgrids,” and the Ph.D. degrees from the KIIT University,
IEEE Trans. Ind. Inform., vol. 14, no. 4, pp. 1403–1414, Apr. 2018. Bhubaneswar, India, in 2012 and 2015, respectively.
[8] S. R. Mudaliyar and S. Mishra “Coordinated voltage control of a grid In 2014, he joined the Department of Electrical
connected ring DC microgrid with energy hub,” IEEE Trans. Smart Grid. and Electronics Engineering, Birla Institute of Tech-
[9] C. Phurailatpam, B. S. Rajpurohit, and L. Wang, “Optimization of DC nology and Science, Pilani, India, as a Lecturer. He
microgrid for rural applications in India,” in Proc. IEEE Region 10 Conf. is currently a Postdoctoral Fellow with the Depart-
(TENCON), Singapore, 2016, pp. 3610–3613. ment of Electrical Engineering, Indian Institute of
[10] S. Sahoo and S. Mishra, “A distributed finite-time secondary average Technology, Delhi, India. He has authored more than 30 research articles (in-
voltage regulation and current sharing controller for DC microgrids,” cluding papers in international journals, conferences, and book chapters). His re-
IEEE Trans. Smart Grid, vol. 10, no. 1, pp. 282–292, Jan. 2019, doi: search interests include power electronics, power system studies, and renewable
10.1109/TSG.2017.2737938. energy.
[11] A. K. Gupta and R. Saxena, “Review on widely-used MPPT techniques
for PV applications,” in Proc. Int. Conf. Innov Challenges Cyber Secur.,
Noida, India, 2016, pp. 270–273.
[12] P. Garcia, J. P. Torreglosa, L. M. Fernandez, and F. Jurado, “Viability
study of a FC-battery-SC tramway controlled by equivalent consumption Sukumar Mishra (M’97–SM’04) received the B.E.
minimization strategy,” Int. J. Hydrogen Energy, vol. 37, no. 11, pp. 9368– degree from the University College of Engineering,
9382, Jun. 2012. Burla, India, in 1990, and the M.E. and Ph.D. degrees
[13] M. Uzunoglu and M. S. Alam, “Dynamic modeling, design, and simulation from the Regional Engineering College, Rourkela, In-
of a combined PEM fuel cell and ultracapacitor system for stand-alone dia, in 1992 and 2000, respectively.
residential applications,” IEEE Trans. Energy Convers., vol. 21, no. 3, In 1992, he joined the Department of Electrical
pp. 767–775, Sep. 2006. Engineering, University College of Engineering, as
[14] S. Pay and Y. Baghzouz, “Effectiveness of battery-supercapacitor com- a Lecturer, and subsequently, became a Reader in
bination in electric vehicles,” in Proc. IEEE Bologna Power Tech Conf., 2001. He is currently a Professor with the Depart-
vol. 3, 2003, p. 6. ment of Electrical Engineering, Indian Institute of
[15] Y. Zhang, Z. Jiang, and X. Yu, “Small-signal modeling and analy- Technology, New Delhi, India. He is working as a
sis of battery-supercapacitor hybrid energy storage systems,” in Proc. National Thermal Power Corporation (NTPC) Chair Professor and has previ-
IEEE Power Energy Soc. General Meet., Calgary, AB, Canada, 2009, ously worked as the Power Grid Chair Professor, New Delhi. He is also serving
pp. 1–8. as an Independent Director with the Cross Border Power Transmission Com-
[16] G. Wang, M. Ciobotaru, and V. G. Agelidis, “Power smoothing of large pany Ltd, Gurugram, India, and an Industry Academic Distinguish Professor.
solar PV plant using hybrid energy storage,” IEEE Trans. Sustain. Energy, His research interests include power systems, power quality studies, and renew-
vol. 5, no. 3, pp. 834–842, Jul. 2014. able energy. He has authored more than 100 research articles (including papers
[17] Y. Xia, W. Wei, M. Yu, and P. Wang, “Stability analysis of PV genera- in international journals, conferences, and book chapters). He has handled many
tors with consideration of P&O based power control,” IEEE Trans. Ind. research projects and industrial consultancies.
Electron. Dr. Mishra is a Fellow of the IET (U.K.), National Academy of Sciences,
[18] D. F. Cortez and I. Barbi, “A three-phase multilevel hybrid switched- India (NASI) (India), INAE (India), Institution of Engineers, India (IE) (In-
capacitor PWM PFC rectifier for high-voltage-gain applications,” IEEE dia), and Institution of Electronics and Telecommunication Engineers (IETE)
Trans. Power Electron., vol. 31, no. 5, pp. 3495–3505, May 2016, (India). He received the Indian National Science Academy (INSA) Medal for
doi: 10.1109/TPEL.2015.2467210. Young Scientists, the Indian National Academy of Engineering (INAE) Young
[19] S. Madichetty, DC Grid Research Test Bed by Silov Solutions Pvt. Lim- Engineer Award, and the INAE Silver Jubilee Young Engineer Award. He is
ited. [Online]. Available: http://silovsolutions.com/ currently the Vice-Chair of the IEEE Intelligent System subcommittee of the
[20] Y. Xia, M. Yu, P. Yang, Y. Peng, and W. Wei, “Generation-storage coor- Power and Energy Society and also an Editor for the IEEE TRANSACTIONS ON
dination for islanded DC microgrids dominated by PV generators,” IEEE SMART GRID and an Associate Editor for IET Generation, Transmission, and
Trans. Energy Convers. Distribution.

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on November 04,2024 at 06:25:57 UTC from IEEE Xplore. Restrictions apply.