2018 Smart Grid Conference (SGC)

A New Load Frequency Control Strategy for an AC

Micro-grid: PSO-based Fuzzy Logic Controlling
Approach
Ahmadreza Abazari Mehdi Ghazavi Dozein Hassan Monsef
Electrical and Computer Engineering Electrical and Computer Engineering Electrical and Computer Engineering
Department Department Department
College of Engineering, University of Tehran College of Engineering, University of Tehran College of Engineering, University of Tehran
Tehran, Iran Tehran, Iran Tehran, Iran
[email protected] [email protected] [email protected]

Abstract— Lack of sufficient synchronous generators (SG) and which concern control, stability, and the efficiency of MGs
high integration of renewable energy sources (RES) raise several [1, 2]. Also, MG connection mode may worsen system
challenges in micro-grid (MG) operation, especially in terms of operation. MGs can be operated in two different connection-
system stability. In fact, MG suffers from low-inertia condition modes: 1) islanded mode, 2) grid-connected mode. In grid-
which results in high risk of frequency instability following
connected mode, a supportive connected power grid mitigates
disturbances. Therefore, it is mandatory to compensate the lack
of sufficient frequency control ancillary services (FCAS) by the load disturbances; nonetheless it is mandatory to utilize
alternative solutions. This paper presents an optimal Fuzzy-logic the MG with intelligent flexible control strategies in order to
based secondary frequency control in order to improve MG achieve a robust performance and stability in islanded mode
frequency stability through management of system active power. [3].
The proposed Fuzzy-logic controller benefits from robustness Some changes in system state parameters and
and simplicity in order to ameliorate frequency deviations. To operational conditions are fast especially when the MG is in
improve the efficiency of the method, Fuzzy membership the islanded-mode and it has fluctuant RESs. In this case,
function parameters are optimized based on a multi-objective traditional controllers cannot satisfy the permissible system
function using Particle Swarm Optimization (PSO) algorithm.
frequency range while a disturbance enters. Also, the
Finally, the proposed strategy is assessed in various scenarios.
traditional controllers are not optimal for whole set of
Index Terms—Micro-Grid (MG), Load Frequency Control operational conditions and configurations. As a consequence,
(LFC), Particle Swarm Optimization (PSO), Fuzzy Logic on the condition of islanded-mode, the need for a robust
Controller controller arises in order to handle a wide range of system
operational conditions.
I. INTRODUCTION Recently, many literatures have presented new control
Operation of electricity networks in the structure of strategies in order to provide stability services for MGs in the
micro-grids (MGs) can result in the reliability improvement, case of islanded mode. Numerous controllers including, but
as well as economic and environmental advantages. not limited to, intelligent control, adaptive control, robust
Generally, MG is a hybrid power generation/energy storage control, and model predictive control (MPC) have been
system combining all different kinds of available renewable proposed in order to maintain system frequency within
energy sources (RES) with available energy storage units. normal operating band [4].
There are different types of distributed generations (DGs), A hierarchical control structure including local level,
such as wind turbine generators (WTGs), photovoltaic (PV) supplementary level, and tertiary level can be considered for
panels, fuel cells (FCs), diesel engine generators (DEGs), MGs [5]. Centralized and decentralized structures frequently
battery energy storage systems (BESS), and flywheel energy presented as two different types of secondary control methods
storage systems (FESS). in MGs. In centralized topology, a central entity is in
Although DGs in MG framework bring many authority to specify MG situation, send/receive information
technical advantages, the coexistence of manifold energy and commands. In decentralized method, each element of
resources with different dynamic raises several challenges MGs is accommodated by local controllers and consequently

978-1-7281-1138-4/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

feedback signals are necessary to control and modify MG deserves attention that the authors evaluate the robustness of
operation [6]. In this paper, a centralized method is suggested proposed control strategy by exposing it to several
for secondary frequency control scheme which is appropriate disturbances, namely load disturbance and wind unit output
for islanded MGs. In [7], for both islanded and grid- power fluctuation.
connected modes, a potential function for each controllable II. SYSTEM MODELING
unit ameliorates the MG performance. In [8], authors have
introduced an approach such that distributed secondary There are different kinds of dynamic models in
frequency control in comparison with [7] does not depend on previous researches regarding MGs. For instance, in [11,12],
a central controller and DGs can work independently. the authors take advantage of some simplified dynamical
Another set of investigations have focused on models for DGs/storage units. Some distributed sources may
generalized droop controllers (GDC). In [9], as the GDC is have high-order dynamic frequency response models, but
highly sensitive to micro-grid configuration, the authors have low-order dynamic models are accurate in order to assess the
considered an Adaptive Neuro-Fuzzy Inference System frequency control problem. Figure 1 shows an MG dynamic
(ANFIS), thus resulting in independency. The proposed model and, what is more, the relevant system parameters for a
method is trained by a desired input-output data set of GDC, typical MG are represented in the case study.
and it is implemented on control loops of converters existing In “Fig. 1”, energy storage systems efficiently supply
in distributed generations (DGs). Subsequent to [9], Bevrani insufficient energy of MG within a short period of time in
et.al. have employed a fuzzy technique based GDC for a order to protect system stability. It takes time to charge FESS
simultaneous frequency regulation in an islanded MG in [10]. and BESS, thus limiting the storage time constant to several
From the author’s point of view, PI controller has a weak seconds [13]. This work expresses the BESS and FESS
performance on the condition of serious load changes. So, a transfer function as a first-order model. As shown in “Fig. 1”,
proposed fuzzy technique is utilized to improve PI the diesel-engine generator model consists of a turbine and a
coefficients. governor, which are denoted through a typical first-order
This paper investigates a centralized Fuzzy logic-based transfer function respectively. Coupled with time-variant
control strategy in order to improve the secondary frequency wind nature, the wind turbine output power is fluctuant like a
control performance in an islanded MG. In the proposed natural source. So, the wind turbine generator (WTG) is
method, the particle swarm optimization (PSO) algorithm modelled as a power fluctuation source of the isolated MG
determines Fuzzy membership function parameters, thus fine- [14]. Similarly, a random power source is an appropriate
tuning the performance of centralized controller. In fact, this candidate to show the photovoltaic (PV) model. In addition,
paper considers a multi-objective function, including the fuel cell is illustrated with a three-order model which is
frequency response characteristics, and employs PSO with sufficient for frequency analysis [15].
intention of the multi-objective function minimization. In
order to assess how the proposed control approach works, the
simulation results are compared to the conventional PI control
method designed by Ziegler-Nicholas technique. Also, it fully
dt
d

K1
K2

1
R ΔPLoad
u
α′

β′

1 1 PDEG 1 Δf
 1 + sT 1 + sT  
g
t D + 2Hs

PFESS
PWTG 1
1 + sTFEESS
P FC
PBESS
1 1 1
1 + sTFC 1 + sT IN 1 + sT1/c 1
1 + sTBEESS

1 1 PPV
1 + sT IN 1 + sT1/c

Figure 1. MG diagram including the proposed control strategy for supplementary (secondary) frequency control section

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

III. CONTROL MECHANISM

A. Design of Fuzzy Based Controller
This research introduces a fuzzy-based controller
which is optimized by PSO algorithm shown in “Fig. 2(a)”. K1 β′
This algorithm is able to choose the best numerical value for f U
the parameters of fuzzy based controller based to a multi-
objective function [16].
d
For this assumption controller, two various signals dt
K2 α′
can be considered as input signals. Frequency deviation (∆ ) (a)
and the changes of frequency deviation ( ⁄ ) enter the
fuzzy controller and the output signal is exploited to alter the
set point of active power injection by distributed energy
resources in LFC scheme. In order to acquire a well-
organized and quick response from this controller, the
membership functions are considered seven-segment
triangular shape including LN (Large negative), MN
(medium negative), SN (small negative), ZO (Zero), SP (b)
(small positive), MP (medium positive), and LP (large
positive). In the following, the Mamdani-type is considered
as interface system and 49 fuzzy rule based are introduced in
Table I in order to connect the input signals to output signal
of this controller.

TABLE I. FUZZY RULE BASE

The derivative of frequency

(c)
LN MN SN ZO SP MP LP Figure 2. (a) Fuzzy logic controller, the pattern of membership function
LN LP LP LP MP MP SP ZO for (b) input signals (c) output signal
MN LP MP MP MP SP ZO SN
perspective of the designer. Besides, the number of
Frequency

SN LP MP SP SP ZO SN MN
ZO MP MP SP ZO SN MN MN individuals in each population generally is initialized by the
SP MP SP ZO SN SN MN LN number of unknown parameters that is four in this objective
MP SP ZO SN MN MN MN LN function. In the following, the search space can be considered
LP ZO SN MN MN LN LN LN as a matrix with the dimension of × in which each
component depicts a parameter of membership functions. In
According to the Fig. 2(b), for the main numerical values this matrix, determines the number of populations and
of input signals including a, b and c, one constrain should be identifies the number of individuals in each population,
considered so that < < < < < is applied respectively. As a result, the search space matrix can be
within the frequency control plan. Besides, one parameter represented as
must be optimized according to a multi-objective function. = (2)
×
This parameter is shown in Fig. 2(c). Finally, this multi- Throughout the initial process of PSO algorithm, random
objective function can be introduced as follows so that number process generates a value for each component of
is the integral squared error (ISE) of the frequency search space while this inequality must be satisfied ( <
response and ST is called settling time of frequency deviation: < < < < ). In the next step, particles move to
the best position so that objective function represented in (1) is
. .( , , , ) = × ∆ + × (1) at the best minimum point. The velocity vector of the PSO
algorithm can be considered as follows:
B. PSO Algorithm
In order to find the best optimum parameters of = × + . . − +
membership function, the suggested method takes advantage . .( − ) (3)
of particle swarm optimization (PSO) algorithm. In the PSO Where is velocity of individual- at iteration- , is a
algorithm, space search size can be defined based on two weighting parameter, and are weighted factors,
miscellaneous parameters including the number of and are considered as random numbers between 0 and
populations and the number of individuals in each population. 1, is the position of individual-i, is the best
The number of populations is selected according to the position of individual-i until iteration-k, and is

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

considered as the best position of group until iteration-k. In A. Scenario I: The proposed Fuzzy-logic controller on the
this algorithm, the individual moves from the current position diesel generator for the maintenance of stability:
to the subsequent position so that this statement can be In this scenario, it is supposed to assess the frequency
represented for position vector as
deviation, 0.1 per unit load increment is applied. The
= + (4) simulation results for two different control strategies which
include primary and secondary frequency controls are
illustrated in “Fig. 4”. The primary frequency control is
IV. SIMULATION RESULTS AND DISCUSSION
associated with the local control mechanism in order to
A typical micro-grid shown in “Fig. 1” is utilized in order prevent the power system from entering to emergency band.
to evaluate the suggested frequency control mechanism. The secondary frequency control mitigates the frequency
Besides, the MG works in the islanded mode and its excursion relatively and the steady state error of deviation
parameters are presented in Table II and Table III. A will be zero after a short period of time. This paper focuses
noteworthy point is to consider a specific amount of power
on frequency regulation program, specifically secondary
generated by the DEG as spinning reserve for secondary
frequency control. In order to clarify how the proposed
frequency control. The fact is that the diesel engine generator
(DEG) is well known as a fast source for power generation strategy works, it needs to look on the diesel generator output
during load increment with high durability and efficiency. power in the secondary frequency control. The diesel
According to Fig. 3, ∆ indicates the frequency deviation, generator output variation curve is illustrated in “Fig. 5” for
and are considered as governor time constant and turbine two types of control strategies. The “Fig. 5” clearly illustrates
(generator) time constant respectively. In addition, depicts that the DEG output power comes to a constant value (i.e.
speed regulation coefficient of the DEG. Simulations are approximate 0.1 per unit) from 3 second onward. Indeed, the
conducted in three different scenarios as follows: proposed Fuzzy logic controller could decrease the settling
time. This can sequentially reduce the cost of operation. Also,
Δf in comparison to the conventional PI controller, the proposed

Ksup 1
R TABLE III. THE RATED VALUE OF UNITS USED IN MICRO-GRID
s
Pmax Rated MG based load
Generation Type
ΔPDEG Power(kW) (kW)
1 1
 Wind 100
1 + sT g 1 + sT t

PV 30

Figure 3. Diesel engine generator model in load frequency control studies Fuel Cell 70 200
Diesel Generator 160
FESS 45

TABLE II. PARAMETERS VALUES OF THE AC MICRO-GRID BESS 45

value parameter
( / ) 0.015
2 ( ) 0.1667 0.03

0.26 0.02
F re q u e n c y D e v i a ti o n ( H z )

0.08 0.01

0.4 0

0.004 -0.01

0.04 -0.02

( / ) 3 -0.03
Secondary frequency control (Conventional)
0.1 -0.04
Secondary frequency control (Fuzzy)
0.1 -0.05 Primary frequency control (Conventional)
primary frequency control (Fuzzy)
-0.06
0 1 2 3 4 5 6 7 8 9 10
Time (s)
Figure 4. The frequency deviation curve in the case of 0.1 per unit
load disturbance for two different control schemes (primary/secondary
frequency control)

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

specifically membership function parameters. Table IV

illustrates the simulation results with respect to the optimal
0.12 values, when the optimisation process is applied to the micro-
grid for 10 times. Based on data shown in Table IV, patterns
The Output Power Variation of DEG (p.u.)

0.1
of input/output signals can be depicted in “Fig. 6”.
0.08 B. Scenario II: Cooperation in frequency control with DEG
Conventional PI Controller and fuel cell in the case of load disturbance:
0.06
In this scenario, the impact of fuel cell contribution in
Proposed Controller secondary frequency control is assessed. The fuel cell unit
0.04
contributes to frequency regulation scheme as an ancillary
0.02 source of energy in order to mitigate the effect of frequency
excursion on the DEG and consequently the life span of DEG
0 Conventional PI Controller is increasing relatively. From “Fig. 7”, it is clear that the
Proposed Controller
cooperation of FC in secondary frequency control leads to
-0.02
0 2 4 6 8 10 have a proper frequency nadir point, that is -0.028 Hz. It is
Time (s) important to note that in the first scenario the frequency nadir
Figure 5. The DEG incremental output in the secondary frequency point was -0.04 Hz. It can be inferred that the FC unit
control participation in the secondary frequency control can reduce
Fuzzy-based strategy has a powerful performance in order to the frequency initial drop in the case of load disturbances.
damp the DEG output power oscillations, thus enhancing the Moreover, it should be underlined that the FC cooperation
life time of DEG. can ameliorate the frequency oscillation when a load
As previously declared, a clear agreement on the disturbance enters. Another interesting point seen in this
nonlinear behaviour of frequency excursion is introduced scenario, the conventional PI controller has even worse
when there have been various membership function performance than the previous scenario regarding oscillation
parameters. According to the foregoing studies, the damping. It is an observation that the FC cooperation
amelioration of frequency excursion depends on the value of increases the problem complexity and the system inertia.
numerical amounts of membership functions considerably. It Consequently, the conventional PI controller is not
is believed that the inaccurate selection of these parameters compatible with such complexity and has a worse
will have some unfavourable ramifications on the frequency performance compared to the first scenario. “Fig. 8(a)”
behaviour of the MG. The urgent require to optimize the illustrates the power output variation of DEG in this scenario.
parameters of membership function arises when the fuzzy In addition, “Fig. 8(b)” depicts how FC participates in the
based controller wants to cope with the mentioned issue in secondary frequency control. The proposed controller helps
LFC scheme. The suggested PSO-based algorithm regulates FC to have a steep increase in its output power. The FC
the best amounts for unrevealed numerical values, output power comes to 0.08 in an instant. The proposed
controller helps FC to have a steep increase in its output
power. The FC output power comes to 0.08 in an instant.
TABLE IV. BEST VALUE BY PSO ALGORITHM

a b c d O.F.
Answer 0.1495 0.3746 0.6770 0.8099 0.076895
0.03

0.02
F r e q u e n c y D e v ia ti o n ( H z )

0.01

-0.01

-0.02

-0.03
Secondary frequency control(Conventional)
-0.04
Secondary frequency control(Fuzzy)
-0.05 primary frequency control(Conventional)
primary frequency control(Fuzzy)
-0.06
0 1 2 3 4 5 6 7 8 9 10
Time (s)
Figure 7. Frequency deviation in the presence of Fuel cell and DEG
concurrent participation for two different control schemes
Figure 6. The membership function patterns for input/output (primary/secondary frequency control)
signals when PSO parameters are based on the first case

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

0.1
0.1
0.08
Conventional PI Controller
The Output Power Variation of DEG (p.u.)

Wind Power Fluctuation (p.u)

0.08 0.06

0.04

0.06 0.02

0
0.04 -0.02
Proposed Controller -0.04
0.02
-0.06

-0.08
0 Conventional PI Controller
Proposed Controller -0.1
0 20 40 60 80 100
Time (s)
-0.02
0 2 4 6 8 10 (a)
Time (s)
(a) -3
x 10
8
0.08
The Output Power Variation of Fuel Cell (p.u.)

6
0.07

Frequency Deviation (Hz)

0.06 Proposed Controller 4

0.05 2

0.04
0
0.03
Conventional PI Controller
-2
0.02

0.01 -4 Conventional PI controller

Proposed Controller
Conventional PI Controller
0 Proposed Controller -6
0 20 40 60 80 100
-0.01 Time (s)
0 2 4 6 8 10
Time (s) (b)
(b) Figure 9. (a) The wind power fluctuation (b) The frequency deviations
Figure 8. (a) The output power variation of DEG on the condition of 0.1 per of the MG in presence of wind output power variations
unit load disturbance (b) The output power variation of FC on the condition
of 0.1 per unit load disturbance

V. CONCLUSION
C. Scenario III: Cooperation in frequency control with DEG
and fuel cell in the case wind speed fluctuation.
This paper presents a novel strategy by which diesel
In this scenario, changes in weather condition make and fuel cell units participate into FCAS. As the Fuzzy
wind power variation so that the wind power changes is controller is dependent on its membership function
shown in “Fig. 9(a)”. In “Fig. 9(b)” the simulations have parameters, the PSO algorithm is employed to find the best
been depicted for two different cases. These results affirm value of membership function parameters. In this paper three
that the suggested fuzzy controller not only mitigates the various scenarios are conducted in order to analyse the
drastic influence of wind fluctuations effectively and restrains proposed frequency control strategy performance. The
the frequency excursion from entering the limited range, but simulation results confirm that the proposed Fuzzy-logic
also outperforms the conventional controller so far as its based controller not only arrest the frequency following
damping effect is concerned. disturbances, but also outperforms the PI controller in various
operating conditions.

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.
2018 Smart Grid Conference (SGC)

REFERENCES
[1] J. A. P. Lopes, C. L. Moreira, A. G. Madureira, “Defining
control strategies for microgrids islanded operation,” IEEE
Trans. Power Syst.,vol. 21, pp. 916-924, May. 2006.
[2] J. M. Hossein, H. R. Pota, M. A. Mahmud, M. Aldeen,
“Robust control for power sharing in microgrids with low-
inertia wind and PV generators,” IEEE Trans. Sustain.
Energy, vol. 6, pp. 1067-1077, July. 2015.
[3] H. Bevrani, M. R. Feizi, S. Ataee, “Robust Frequency control
in an islanded microgrid: H∞ and µ- synthesis approaches,”
IEEE Trans. Smart Grid, vol. 7, pp. 717, March. 2016.
[4] J. Yang, Z. Zhili, T. Yufei, Y. Jun, H. Habio, W. Yunliang,
“Load frequency control in isolated micro-grids with electrical
vehicles based on multivariable generalized predictive
theory,” Energies, Vol. 8, pp. 2145-2164, March. 2015.
[5] J. M. Guerrero, P. C. Loh, T. L. Lee, “Advanced control
architectures for intelligent microgrids-Part II: Power quality,
energy storage, and AC/DC microgrids,” IEEE Trans. Ind.
Electron., vol. 60, pp. 1263-1270, April. 2013.
[6] E. Backlund, N. Pogaku, M. Prodanovic, C. Hernandez-
Aramburo, T. C. C. Green, “Energy management in
autonomous microgrid using stability-constrained droop
control of inverters,” IEEE Trans. Power Electon., vol. 23, pp.
2346-2352, Sept. 2008
[7] A. Mehrizi, R. Iravani, “Potential-function based control based
control of microgrid in islanded and grid-connected modes,”
IEEE Trans. Power Syst., vol. 25, pp. 1883-1891, Nov. 2010.
[8] Q. Shafiee, J. M. Guerrero, J. C. Vasquez, “Distributed
secondary control for islanded microgrid- A novel approach,”
IEEE Trans. Power Electron.,vol. 29, pp. 1081-1091, Feb.
2014.
[9] H. Bevrani, Sh. Shokoohi, “An intelligent droop control for
simultaneous voltage and frequency regulation in islanded
microgrids,” IEEE Trans. Smart Grid, vol.4, pp. 1505-1531,
Sept. 2013.
[10] S. Ahmadi, Sh. Shokoohi, H. Bevrani, “A fuzzy logic-based
droop control for simultaneous voltage and frequency
regulation in an AC microgrid,” Electrical Power & Energy
Syst.,vol. 64, pp. 147-155, Jan. 2015.
[11] D. J. Lee, L. Wang, “Small-signal stability analysis of an
autonomous hybrid renewable energy power generation/
energy storage system part I: Time-domain simulations,” IEEE
Trans. Energy Convers., vol. 23, pp. 311-320, March. 2008.
[12] H. Bevrani, F. Habibi, P. Babahajyani, M. Watanabe,
“Intelligent frequency control in an AC microgrid: Online
PSO-based fuzzy tuning approach,” IEEE Trans. Smart
Grid,vol. 3, pp. 1935-1944, Dec. 2012.
[13] K. Agbossou, M. Kolhe, J. Hamelin, T. K. Bose,
“Performance of a stand-alone renewable energy system based
on energy storage as hydrogen,” IEEE Trans. Energy
Convers., vol. 19, pp. 633-640, Sept. 2004.
[14] A. Abazari, M. G. Dozein, H. Monsef, “An optimal fuzzy-
logic based frequency control strategy in a high wind
penetrated power system,” Journal of the Franklin Institute,
vol. 355, pp. 6262-6285, Sept. 2018.
[15] S. Obara, “Analysis of a fuel cell micro-grid with small-scale
wind turbine generator,” Int. J. Hydrogen Energy,vol. 32, pp.
323-336, March. 2007.
[16] A. Abazari, H. Monsef, B. Wu, “Load frequency control by
de-loaded wind farm using the fuzzy-based PID droop
controller,” IET Renew. Power Generat., 2018, in press.

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on January 25,2022 at 22:46:12 UTC from IEEE Xplore. Restrictions apply.