Electrical Engineering

https://doi.org/10.1007/s00202-023-02206-6

ORIGINAL PAPER

Load frequency control of an isolated microgrid using optimized

model predictive control by GA
Ayan Kumar Tudu1 · Nageswarappa Naguru1 · Sunita Halder Nee Dey1 · Subrata Paul1

Received: 20 June 2023 / Accepted: 10 December 2023

© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024

Abstract
A novel method of frequency of control of isolated microgrid by optimization of model predictive controller (MPC) is
proposed in this study. The suggested controller is made for a microgrid that employs renewable energy sources as well
as storage systems. The proposed control scheme makes use of MPC to continuously optimize and modify the controller
coefficients. The MPC parameters specifically the input control rate weight parameter, prediction horizon, and control horizon
are optimized using genetic algorithm. The frequency perturbations that occur after the power fluctuations in the microgrid
due to the power imbalance are minimized by the proposed controller by sending a control signal to the sources. This power
imbalance is caused by the sporadic nature of power generation by renewable energy resources like wind and solar units
and load disruption in an isolated microgrid. The suggested control method efficacy is assessed through simulation work on
the islanded microgrid in MATLAB environment wherein the proposed controller performs better than conventional MPC,
fuzzy-based MPC, and particle swarm optimization-based MPC. In comparison with other approaches, the suggested control
method performs better when parameters are changed and has been able to successfully reduce the number of oscillations and
amplitude of frequency fluctuations. It is also more resilient to the indeterminacy of microgrid specifications when compared
to other approaches.

Keywords Microgrid · Model predictive control · Load frequency control · Genetic algorithm · Controllable power-generating
unit

1 Introduction or isolated mode [2]. The microgrid’s capacity to function as

an island and on its own is one of its most significant advan-
With the growing demand for electrical power, unconven- tages. Islanding increases the quality of power and reliability
tional sources have been incorporated into power systems, of the RESs associated with the microgrid. The microgrid
which has led to an increase in system complexity and uncer- frequency is the same as that of the main grid in the case
tainty. In modern power systems, renewable energy sources of the grid-connected mode of operation of the microgrid.
(RESs) are primarily used as alternative generation units. But, in the islanded operation the microgrid must regulate
While the rising usage of RESs offers numerous benefits, both frequency and voltage individually [3–6]. Due to the
it also poses new challenges, such as their ability to oper- imbalance between the electrical power generated and the
ate stably alongside conventional generation units [1]. The electrical demand of the load, the supply frequency moves
presence of these abundant RESs in distribution systems has away from its specified value. This imbalance between power
given rise to a new microgrid setup. By definition, a micro- generated and load in an isolated microgrid is mainly due to
grid is a low-voltage network that is connected to a small the presence of intermittent nature of the power generation
distributed energy-generating system, a number of loads, and of RESs.
a storage device. It can be operated in either grid-connected The issue of controlling the real power of generating
units in response to a change in system frequency and the
exchange of patch line power within predetermined limita-
B Nageswarappa Naguru
[email protected] tions is referred to as load frequency control (LFC). The
load frequency fluctuates with the variation in active power
1 Electrical Engineering Department, Jadavpur University,
Kolkata 700032, India

123
 Electrical Engineering

generated by the RESs and with load demand. The load fre- of RES. In order to preserve the stability of frequency
quency must be kept within specified limits for the safety in low-carbon grids, [16] examines the necessity of inte-
and acceptable operation of electrical equipment, as well as grating synthetic inertia and primary frequency response
the microgrid’s stability. Numerous control strategies can be (PFR) from renewable energy (RE) facilities. It offers a
found in the literature in the domain of regulation of the framework for the stochastic security-constrained unit com-
frequency of a microgrid that is isolated. To regulate the mitment that takes into consideration the intermittent nature
frequency of island microgrids, droop control is frequently of the generation of energy from RES and forecasts and eval-
utilized. This droop control is dependent on the property of uates the system-wide frequency response requirement using
the active power balance of distributed generator (DG) con- machine learning techniques. For low-inertia microgrids, the
verters. To increase the stability, dependability, and power researchers in [17] suggest an intelligent synthetic inertia
quality of microgrids, [7] combines direct current vector control (SIC) model with an adaptive fuzzy logic controller.
management with droop control. The article also gives a In order to meet the system’s demands for fast Frequency
general description of the three hierarchical levels of con- response (FFR) [17], the model optimizes the output of RESs.
trol that are used to control the microgrid, as well as typical [18, 19] discuss the possibilities of providing fast frequency
distributed generating and energy storage units. In order response to low-inertia microgrids.
to achieve optimal power sharing across dispersed gener- Metaheuristic algorithms also find application in the fre-
ators, [8] suggests a plug-and-play optimization technique quency management of microgrids. In order to enhance the
for islanded microgrid operations. The suggested method application of automatic generation control for load fre-
compares each DG to a set of user-defined fictitious DGs quency management of interconnected power systems, the
the droop relations and operation cost functions of which authors in [20] present a novel control structure that makes
have been predetermined and are utilized as a benchmark. A use of a fractional-order proportional–derivative cascade
droop-based demand response scheme is proposed in [9] for controller. The power system’s nonlinearities and uncertain-
smartly managing loads in integrated microgrid systems. [10] ties are intended to be handled by the suggested controller.
covers the implementation of coordinated dynamic droop The parameters of the controller were optimized by the
control for fuel cells and wind turbines in a microgrid. It wild horse optimizer algorithm. [21] addresses the usage
compares the effectiveness of dynamic and steady droop con- of a fractional-order PID controller in a two-area reheat
trol techniques, and the outcomes are examined in terms of thermal type interconnected power system for automatic
frequency regulation and power production. [11] suggests generation control (AGC) where the metaheuristic method
a secondary frequency control strategy for islanded micro- of artificial gorilla troops optimizer (GTO) was applied to
grids that greatly enhances frequency regulation, using an find the controller parameters. This work also takes into
adaptive time-variable droop feature and distribution-level account nonlinearities like generation dead band and gen-
phasor measurement units (D-PMUs). The frequency regu- eration rate constraints. [22] proposes the application of one
lation of a microgrid using a proportional–integral controller plus proportional–derivative with filter-fractional-order pro-
is covered in [12]. Here, Ziegler–Nichols (Z-N) approach is portional–integral ((1 + PDF)-FOPI) controller as a new
used to manually tune up the parameters of the controller. control scheme for the microgrid load frequency control. This
[13] proposes the application of a fractional-order propor- controller is assisted by the marine predator algorithm. [23]
tional–integral–derivative (FOPID) controller optimized by demonstrates the use of a social-spider optimizer (SSO) to
genetic algorithm (GA) to regulate the frequency and power tune up the controllers of the PID controller for an effective
deviations in a microgrid system. It shows that the pro- LFC of an isolated two-area hybrid microgrid system. [24]
posed GA-FOPID controller effectively reduces frequency suggests using GTO to improve the LFC in a wind energy-
and power changes in the microgrid. Different LFC archi- based microgrid system that uses wind power. [25] examines
tectures in different power system setups are thoroughly a novel load frequency controller with a quasi-oppositional
analyzed in [14]. A thorough understanding of the LFC harmony search algorithm (QOHSA) for an autonomous
evaluation of different power system designs is offered in hybrid power system model (HPSM) made up of energy stor-
this work by the comprehensive tabular study of frequency age and numerous power-generating units.
control architectures. It also highlights the significance of In recent years, predictive control has emerged as one
improved controllers for LFC planning in complex power of the suitable controllers for taking into account numer-
system domains. ous uncertainties in a system. The model predictive control
[15] proposes a new robust virtual inertia control-based (MPC) technique is an advanced method that is extensively
coefficient diagram (CDM) method. The proposed controller used in the chemical process industry, automobile engineer-
improves the robustness and the validity of the system in ing, and electrical engineering [26]. In terms of addressing
the presence of unpredictability of the parameters of an limitations, forecasting future behavior, and optimizing con-
islanded microgrid where there is a very high penetration trol actions, the adoption of MPC has proved to be beneficial.

123
Electrical Engineering

The MPC method is an effective feedback control optimiza- (i) Introducing a new predictive control-based method to
tion technique that makes use of a model of a system to dampen frequency fluctuations in the islanded mode of
forecast the future behavior of the variable to be controlled. operation of a microgrid.
The future behavior is dependent on the best choice of actu- (ii) The MPC control parameters, namely prediction and
ating action by taking into account the minimization of control horizon, are also optimized using GA while
the cost function. This actuating action is the result of for- simultaneously optimizing the control input rate weight
ward forecast outcomes within a predetermined time horizon which was absent in earlier works.
[27]. MPC has evolved into a capable method for microgrid
control. [28] delves into the application of optimal model The paper is organized as follows: a brief overview of
predictive control (MPC) to regulate virtual inertia in self- the model predictive controller is provided in Section II,
sufficient microgrids. The study highlights the efficacy of followed by a detailed discussion on the usage of conven-
MPC in controlling microgrid operations and maintaining tional MPC for regulation of load frequency of a standalone
stable grid frequency. The optimal design of MPC is achieved microgrid in Section III. Section IV discusses the GA-based
through the utilization of the African vultures optimization MPC for load frequency regulation in a standalone microgrid.
algorithm (AVOA). An innovative approach was suggested The proposed algorithm’s efficiency is assessed through var-
in [29] to improve frequency regulation in AC microgrids by ious simulations, and the results and discussions on them are
combining a fuzzy control system with a model predictive exhibited in Section V. Finally, in Section VI the conclusions
controller (MPC). The proposed method involves adjusting are drawn.
the damping coefficient and the virtual inertia of the virtual
synchronous generator (VSG) and modifying the optimal
rated power of the VSG by the application of the MPC 2 Outline of model predictive control (MPC)
method. In [30], a novel methodology for load frequency
regulation in isolated microgrids has been introduced and In process industries, especially in the oil and petro-chemical
verified. It was shown that the fuzzy adaptive model predic- sectors, model predictive control is a popular optimization
tive control system outperforms traditional systems with a strategy [34]. It is presented as a heuristic method in [35],
swifter and more flexible response. [31] suggests a frame- where its effectiveness in industrial process applications is
work for the coordinated operation of a residential microgrid verified. The current measurements and projected future
that is connected to a grid. The framework uses the MPC output values are linked to the MPC calculations. These com-
method, in which the system is optimized at each deci- putations must comply with a set of control actions in order
sion time based on short-term forecasts of the generation to best forecast the response to follow the particular points
of load demand, renewable energy resources, and electric- [36] (Fig. 1).
ity prices. [32] implements an MPC method that employed The discrete time model of a plant taking into account the
mixed-integer linear programming to optimize the microgrid plant noise and disturbance is given by
operations. Satisfactory utilization of renewable energy and
storage capabilities in varying grid connection situations is x1 (k + 1)  A1 x1 (k) + B1 u(k) + Bd w(k) (1)
made possible by the proposed hybrid microgrid model in
[33] that employs switched model predictive control. Most
load frequency control MPC systems have shifted from a cen- y(k)  C1 x1 (k) + D1 u(k) (2)
tralized approach to a decentralized or distributed approach.
However, there has been limited research on enhancing the where u is the control input and y is the plant output, x is the
flexibility of MPC. state variable vector with dimensions of n 1 × n 1 , and w(k)
This paper proposes a predictive control approach for is the input disturbance vector. In a moving horizon control,
frequency regulation wherein a genetic algorithm (GA) is the current control input does not affect the present output;
employed to optimize the control parameters of the MPC, it affects directly the output at (k + 1) instant. So, D1  0 in
namely the prediction horizon, control horizon along with Eq. (2).
the parameter Rw in the cost function of the MPC. A control Taking the difference on both sides of Eqs. (1) and (2),
signal is generated by the proposed method which is sub-       
sequently applied to the RESs within the microgrid. Under x1 (k + 1) A1 O1T x1 (k) B1
 + u(k)
normal conditions and in cases of uncertainty and system y(k + 1) C 1 A1 1 y(k) C1 B1
parameters, the proposed controller displays superior perfor- (3)
mance in damping microgrid frequency changes compared to  
other controllers such as conventional MPC and fuzzy MPC. x1 (k)
y(k)  [O1 1] (4)
The major contributions of this article are as follows: y(k)

123
 Electrical Engineering

Fig. 1 Structure of model

predictive controller [36]

 
Here, O1  0 0 ... . The size of O1 is 1 × n 1 . where Y is set of predicted output, Rs is reference vector, and
    R is control input rate weight matrix.
A1 O1T B1 ⎡ ⎤
Assume here, Am  , Bm  , 1
 
C 1 A1 1 C1 B1 ⎢1⎥
⎢ ⎥
C m  O1 1 . ⎢1⎥
⎢ ⎥
⎢ ⎥
The matrices (Am , Bm , C m ) represent the enhanced dis- R S  ⎢ . ⎥ ∗ r , size of R S is N P , r , is reference value.
⎢ ⎥
crete state space model of the given plant. N P is the prediction ⎢.⎥
⎢ ⎥
horizon, and NC is the control horizon and is less than N P . ⎣.⎦
The disturbance matrix in Eq. (1) is left out as the disruption 1
in the current step has no effect on the future. R  Rw I N C×N C , Rw is any real positive number, (Rw ≥
At sampling time instant k, k > 0, the sequence of control 0), I is a unit identity matrix of the dimension NC × NC .
input is given by, To determine sequence of control inputs U , ∂U ∂O
0

u(k), u(k + 1), u(k + 2), u(k + 3)......u(k + NC − 1)

−1
U  ϕ T ϕ + R ϕ T (R S − F x(k)) (7)
Here, though N C number of control inputs are calculated,
only the first control input is implemented to the plant at the
time instant k, moving to next sampling time k + 1, again
the sequence of control inputs is calculated to obtain reced-
ing horizon control. The future state variable vector will be
estimated using the information about the current state vari- 3 Load frequency control of an isolated
able. Using the current and future state variable vector along microgrid using conventional MPC
with the sequence of control input, the output vector can be
obtained by applying the following equations, The block diagram representation of the microgrid under con-
sideration is shown in Fig. 2. This microgrid is isolated from
the grid and is in its standalone mode of operation. It is made
Y  F x(k) + ϕU (5)
up of a diesel generator unit, a fuel cell unit, a wind power-
⎡ ⎤ generating unit, a solar power-generating unit, and a battery
CA storage unit [37–39]. The feedback control system of the
⎢ ⎥ diesel generator and the impact of MPC on the best possible
⎢ C A2 ⎥
Here F  ⎢ ⎥ϕ  control parameter adjustment is displayed. The difference
⎣ ... ⎦
C ANP between the reference frequency and system frequency is
⎡ ⎤ applied as the input to the controller. The controller output
CB 0 ... 0
⎢ ⎥ is the sequence of input control to the fuel cell and diesel
⎢ C AB CB ... 0 ⎥
⎢ ⎥. generators present in the microgrid. It is taken into consid-
⎣ ... ... ... ... ⎦ eration that the renewable energy generators are running at
CA N P −1 B CA N P −2 B ... C A N P −N C B full capacity. So, only the fuel cell and diesel generators units
The objective function to determine sequence of control are considered to be controllable power-producing units. The
input U , discrete time state space model of the microgrid under con-
sideration is used to determine the control inputs to the plant.
O  (Y − R S )T (Y − R S ) + U T RU (6) The dynamic state space model of the microgrid is shown in

123
Electrical Engineering

Fig. 2 Isolated microgrid model for the LFC problem

Eqs. (8)-(16): PL —Load disturbance.

PS —Disturbance in solar power.
• 1
f  (Psfilt + Pmd + Pw − PL Pmd —Power fluctuation in mechanical power output of
(2H ) turbine in Diesel unit.
+Pffilt − Pbat − D ×  f ) (8) Pgd —Power fluctuation in governor of Diesel unit.
Pffilt —Power fluctuation in filter unit of Fuel cell.
• 1 Pfinv —Power fluctuation in invertor unit of Fuel cell.
 P  (PS − PSinv ) (9)
Sinv (Tinv ) Pfc —Fuel cell unit power output fluctuation.
Pbat —Battery unit power output fluctuation.
• 1
 P  (PSinv − PSfilt ) (10) Pcd —Input to the diesel unit from MPC.
Sfilt(Tfilt ) Pcf —Input to the Fuel cell unit from MPC.
 
• 1 f Tt —Time constant of turbine.
 P    Pcd − − Pgd (11) Tg —Time constant of governor.
gd Tg (R)
Tfilt —Time constant of filter unit.
• 1   Tinv —Time constant of invertor unit.
P  Pgd − Pmd (12)
md (Tt ) Tfc —Time constant of fuel unit.
  Tb —Time constant of battery energy storage system.
• 1 f
P  Pcf − − Pfc (13) The dynamic equations shown above are used to formulate
fc (Tfc ) (R)
state space model of the plant given in Eq. (17) and Eq. (18).
• 1 In Eq. 17, A, B, C, and D are calculated as shown in Eq. (20)
 P  (Pfc − Pfinv ) (14)
finv (Tinv )
• 1 dx(t)
 P  (Pfinv − Pffilt ) (15)  Ax(t) + Bu(t) + Bd u(t) (17)
Sfilt (Tfilt ) dt
• 1
P  ( f − Pbat ) (16)
bat (Tb ) y  C x(t) (18)

where 
 f —Frequency deviation, where x   f PS_ inv PS_filt Pgd Pmd Pfc
T
R—Droop in frequency. Pf_inv Pf_filt Pbat .
D—Damping coefficient of the microgrid, The load frequency is regulated by controlling the con-
H—Inertia constant of rotating parts in microgrid. trol inputs to the diesel and fuel cells. These optimal control

123
 Electrical Engineering

Table 1 Microgrid parameter values 4 Load frequency control of an isolated

microgrid using genetic algorithm-based
Parameter Value Parameter Value
MPC (GAMPC)
D (p.u. MW/Hz) 0.015 Tg (s) 0.08
2H(s) 0.1667 Tt (s) 0.4
A better performance from an MPC will be achieved when
the parameters of MPC are chosen optimally. The parameters
Tfc (s) 0.26 Tb (s) 0.1
that influence the behavior of MPC are control horizon N C ,
Tinv (s) 0.04 R (Hz/p. u MW) 3
prediction horizon N P , input control rate weight parameter
Tfilt (s) 0.004
Rw , and sampling time. Here, genetic algorithm (GA) is used
to tune the parameter values N C , N P , and Rw .
Optimal tuning of the mentioned parameters gives bet-
inputs are calculated by using MPC, such that the cost func- ter performance of MPC compared with conventional MPC,
tion O as shown in Eq. (19) is minimized. where these parameters N C , N P , and Rw are selected ran-
domly. The objective function to be minimized is given in
O  ( f ref −  f o )T ( f ref −  f o ) + U T RU (19) Eq. (6). The upper bound parameters of the parameters N C ,
N P, and Rw are 50, 50, and 10, and the lower bound for the
Here, in the objective function shown above the first term same parameters was chosen to be 2, 1, and 1. The choice of
is related to minimizing of error between the reference and the these bounds for the parameters of the MPC has been arrived
predicted output and the second term is related to optimal pre- after careful literature survey.
dicted controlinput that optimizes the cost function U 
T
Genetic algorithm uses natural process of biological evo-
Pcd Pcf , R  Rw I NC ×NC , Rw is a real positive num- lution [40]. It involves natural process of selecting fittest
ber, Rw ≥ 0, I is a unit identity matrix of the dimensions individual for reproduction in the next generation. Genetic
NC × NC . algorithm solves an optimization problem in five steps. They
For the conventional MPC method, the prediction horizon are:
is taken as 10 sampling times and control horizon is taken
as 2 sampling times with sampling time of 0.01 s [30]. The
parameter values of the microgrid shown in Fig. 2 are given
in Table 1. 1. Initial population: the genetic algorithm process starts
⎡ ⎤ with selecting the size of initial population depending on
− D 0 1
0 2H 1
0 0 2H − 2H
1 1
the optimization problem. Typically, hundreds or thou-
⎢ 2H 2H
⎥
⎢ 0 − Tinv 01
0 0 0 0 0 0 ⎥ sands of initial populations are selected.
⎢ ⎥
⎢ 0 0 ⎥
Tfilt − Tfilt
1 1
⎢ 0 0 0 0 0 ⎥ 2. Fitness function: it helps in determining the best individ-
⎢ ⎥
⎢ − RT1 0 0 − 1
0 0 0 0 0 ⎥ ual with best chromosome among the selected population
⎢ g Tg ⎥
⎢ 1
− 1 ⎥
A⎢ 0 0 0 0 0 0 0 ⎥ initially. These best chromosomes are being used for
⎢ Tt Tt
⎥
⎢− 1 0 0 0 0 − T1 0 0 0 ⎥ reproduction in the next generation.
⎢ R fC fc ⎥
⎢ ⎥
⎢ 0 0 0 0 0 T1 − T1 0 0 ⎥ 3. Selecting chromosomes: in this the best individual with
⎢ inv inv ⎥
⎢ 0 0 ⎥ best chromosomes is selected for reproduction in the next
Tfilt − Tfilt
1 1
⎣ 0 0 0 0 0 ⎦
1
0 0 0 0 0 0 0 − T1 generation. The individuals are then paired to get the
T b b
⎡ ⎤ ⎡  ⎤ reproduction done. The objective of this step is to get
0 0 −1 2H 1 2H 0 the best individual for better chances of giving optimal
⎢ 0 
⎢ 0 ⎥ ⎥
⎢
⎢ 0 0 1 Tinv ⎥
⎥ solution to the problem.
⎢ ⎥ ⎢ ⎥
⎢ 0 0 ⎥ ⎢ 0 0 0 ⎥ 4. Cross over of chromosomes: in this step the genetic infor-
⎢  ⎥ ⎢ ⎥
⎢ 1 Tg 0 ⎥ ⎢ 0 0 0 ⎥ mation of two parents selected in step 3 is exchanged to
⎢ ⎥ ⎢ ⎥
B⎢⎢ 0 0 ⎥ ⎥ E ⎢⎢ 0 0 0 ⎥ ⎥
get new generation population of equal size that of parent
⎢ 0 1 T ⎥ ⎢ 0 ⎥ population.
⎢ fc ⎥ ⎢ 0 0 ⎥
⎢ ⎥ ⎢ ⎥ 5. Mutation: in this step new genetic information is added
⎢ 0 0 ⎥ ⎢ 0 0 0 ⎥
⎢ ⎥ ⎢ ⎥ to the new generation. Mutation is done by switching few
⎣ 0 0 ⎦ ⎣ 0 0 0 ⎦
bits in the chromosome.
0 0 0 0 0
 
C  100000000 ; D0
 T  T (20)
U  Pcd Pcf ; W  PL PW PS The flowchart of the proposed method is given in Fig. 3.

123
Electrical Engineering

diesel unit and fuel cell unit are described in Fig. 4d. It can
be seen from Fig. 4a and b that the GAMPC is aptly able to
bring the frequency deviation down at the load changes at a
much faster rate compared to the other adopted methods. The
optimally found parameters are Rw  1, N P  50, and N C 
11. Moreover, there is significant reduction in the variation
of the cost function over the simulation period as compared
to the other MPCs used.

Case II: Wind power, solar power as well as the load

disturbance are varying with time In this instance, it is
presumed that the RESs like solar power and wind power, that
are present in the microgrid and the load is also undergoing a
change. Figure 5a shows this variation of power. The micro-
grid frequency response in this case is shown in Fig. 5b where
again the proposed GAMPC outperforms the conventional
MPC and fuzzy MPC and the PSO-based MPC. Figure 5c
shows the control inputs for the fuel cell unit and diesel unit
generated by the proposed GAMPC. The responses of the
cost function of MPC during the simulation period are shown
in Fig. 5d.
In this case, there are variations in wind power, solar power
as well as load variation; still it can be seen from Fig. 5b that in
Fig. 3 Flowchart of the proposed method this case also the GAMPC outperforms the other mentioned
methods. In this case, also the optimally found parameters
5 Results and discussion are Rw  1, N P  50, and N C  11. The variations of cost
functions of the MPC as shown in Fig. 5d also certify that
The performance of a GAMPC for LFC of a microgrid the GAMPC is performing better than the conventional MPC,
is verified using MATLAB simulations. In MATLAB, the fuzzy-based MPC, and the PSO-based MPC.
microgrid block diagram model illustrated in Fig. 2 is sim-
ulated. Figure 3 gives the flowchart of the proposed method Case III: Variation of the microgrid parameters The pro-
wherein the GA-based MPC is used to find the control input posed genetic algorithm-based model predictive control is
rate (Rw ) of the MPC as well as the optimal prediction hori- used to study the system response upon introduction of vary-
zon (N P ) and control horizon (N C ) for the MPC. To test the ing parameters into the microgrid model. The variations in
operation of GAMPC, various load disturbances and power the parameters are as follows: R  + 30 percent, D  − 40
generation changes in wind and solar power generators are percent, H  + 50 percent, T t  − 50 percent, T g  + 50
introduced into the microgrid. In all the cases, the perfor- percent, T b  − 45 percent. The load disruption and variabil-
mance of GAMPC is compared to that of conventional MPC ity in wind and solar power generation that were addressed
having fixed tuning parameters Rw  8, N P  10, and N C in Case II are both present in this case. Figure 6a shows the
 2 and with the fuzzy MPC as described in [30] as well as frequency deviation for case III. Figure 6b shows the con-
with PSO-based MPC [41]. The various test cases are: trol input generated by in Case III, and Fig. 6c shows the
cost functions of MPC during the simulation period. The
Case 1: With only load disturbance In this case, load results obtained again verify the superior performance of the
frequency regulation is investigated when there is no wind GAMPC. The optimally tuned parameter for this case is Rw
generation or solar power and the microgrid system is sub-  1; N P  100; N C  12.
jected only to a step load disturbance, as depicted in Fig. 4a. The performance index parameter integral square error
Figure 4b depicts the microgrid frequency response in the (ISE) was calculated and compared for all the three sim-
presence of only load disturbance where it can be observed ulation cases. The comparison of the proposed genetic
that the proposed GAMPC ensures a faster and better change algorithm-based MPC with conventional MPC controller,
in frequency deviation when compared with the conventional fuzzy MPC, and PSO MPC is evaluated and summarized
MPC, the fuzzy MPC, and the PSO-based MPC. The varia- in Table 2 wherein the obtained results indicate that the
tion of the cost function over the simulation period is shown proposed GAMPC has an improved performance over the
in Fig. 4c. The control inputs generated by GAMPC for the conventional MPC controller, fuzzy MPC, and PSO MPC.

123
 Electrical Engineering

Fig. 4 a Load disturbance for

case I. b microgrid frequency
deviation response for case I.
c cost functions responses of
MPC for case I. d control input
generated by GAMPC for case I

123
Electrical Engineering

Fig. 5 a Load disturbance, wind

power variation, and solar power
variation for case II. b microgrid
frequency deviation response for
case II. c control Input Generated
by GAMPC for case II. d cost
functions responses of MPC for
case II

123
 Electrical Engineering

Fig. 6 a Microgrid frequency deviation response for case III. b control input generated by GAMPC for case III. c cost functions responses of MPC
for case III

123
Electrical Engineering

Table 2 Performance index (ISE)

obtained in different cases Conventional MPC Fuzzy MPC PSO MPC GAMPC

Case I 0.000537 0.000490 .000483 0.000082

Case II 0.007959 0.001995 0.007144 0.001028
Case III 0.007959 0.003485 0.010849 0.001896

The effectiveness of the proposed GAMPC is highlighted in bold

Table 3 Settling time (in secs) of

 f in different case studies for Conventional MPC Fuzzy MPC PSO MPC GAMPC
different MPCs
Case I 21.458477 21.996166 21.557886 20.517875
Case II 20.744237 20.476483 21.124644 20.308087
Case III 20.744237 21.054887 23.695576 20.433017

The effectiveness of the proposed GAMPC is highlighted in bold

The settling times of all the three simulation cases are pre- algorithm-based MPC can be used for frequency regulation
sented in Table 3 wherein the effectiveness of the proposed in further microgrid uses. However, the proposed study is
GAMPC is again confirmed. It can be observed from Table 3 limited to a microgrid whose storage system gives constant
that GAMPC takes less time to reach steady state compared output. Future work will further exploit the performance of
to the conventional MPC controller, fuzzy MPC and PSO proposed controller for different power system with storage
MPC. system having variable output, variation in sampling time of
the MPC, and the effect of state of charge of battery on the
frequency management of microgrids.
6 Conclusion Author contributions AKT, NN, SHND, and SP wrote the main
manuscript text and prepared the figures. All authors reviewed the
In order to adjust load frequency effectively and quickly manuscript.
for an isolated microgrid, a genetic algorithm-based MPC
Funding This declaration is “Not Applicable”.
was proposed in this article. The studied isolated microgrid
included a wind power-generating unit, a solar power- Availability of data and materials All data supporting the findings of
generating unit, a diesel generator unit, a fuel cell unit, and this study are available within the paper.
a battery storage unit. The dynamic model of the energy-
producing units and storage unit was modeled in MATLAB. Declarations
Only the power produced by the fuel cell units and diesel
generator is considered to be controllable as it is presumed Conflict of interests This declaration is “Not Applicable”.
that all the renewable energy generators are producing their
maximum capacity output. The input to the MPC is the differ- Ethical approval This declaration is “Not Applicable”.
ence between the reference frequency and system frequency.
The reference frequency is taken to be zero in this study. The
system frequency is affected by the presence of each renew- References
able energy source as well as the storage unit. In the study
1. Chowdhury C, Chowdhury SP, Crossley P (2009) The institution of
for load frequency control, it is shown that the tuning param- engineering and technology. In: Microgrids and Active Distribution
eters Rw , N P , and N C of the MPC affect the effectiveness Networks. London, U.K.: Inst Eng Technol
of the controller. The robustness of the suggested algorithm 2. Lasseter RH, Akhil A, Marnay C, Stephens J, Dagle J, Guttrom-
was checked in three different scenarios wherein the pro- son R, Meliopoilous AS, Yinger R, Eto J (2002) Integration of
Distributed Energy Resources–The CERTS MicroGrid Concept,
posed GAMPC performed better compared to conventional CERTS white paper
MPC, fuzzy MPC, and PSO MPC where only the control 3. Dashtdar M, Dashtdar M (2020) Voltage-frequency control (vf)
input rate (Rw ) of the MPC is optimized and the parameters of islanded microgrid based on battery and MPPT control. Am J
N P and N C are kept constant. The projected optimal tun- Electr Comput Eng 4(2):35–48
4. Hosseinimoghadam SMS, Roghanian H, Dashtdar M, Razavi SM
ing method offers the best possible performance in a variety
(2020) Power-sharing control in an islanded microgrid using virtual
of operating situations, particularly when certain parame- impedance. In: 2020 8th International Conference on Smart Grid
ters are intermittent in nature. Hence, the proposed genetic (icSmartGrid), pp 73–77

123
 Electrical Engineering

5. Dashtdar M, Dashtdar M (2020) Power control in an islanded 21. Sah SV, Prakash V, Pathak PK, Yadav AK (2022) Fractional Order
microgrid using virtual impedance. Sci Bull Electr Eng Faculty AGC Design for Power Systems via Artificial Gorilla Troops
20(1):44–49 Optimizer. In: 2022 IEEE International Conference on Power Elec-
6. Hosseinimoghadam SMS, Dashtdar M, Dashtdar M, Roghanian H tronics, Drives and Energy Systems (PEDES), Jaipur, India, pp 1–6.
(2020) Security control of islanded micro-grid based on adaptive https://doi.org/10.1109/PEDES56012.2022.10079975
neuro-fuzzy inference system. Sci Bull Ser C 1:189–204 22. Pathak PK, Yadav AK, Padmanaban S, Kamwa I (2022) Frac-
7. Ramezani M, Li S (2016) Voltage and frequency control of islanded tional cascade LFC for distributed energy sources via advanced
microgrid based on combined direct-current vector control and optimization technique under high renewable shares. IEEE
droop control. In: 2016 IEEE Power and Energy Society General Access 10:92828–92842. https://doi.org/10.1109/ACCESS.2022.
Meeting (PESGM), pp 1–5. Boston, MA, USA 3202907
8. Cingoz F, Elrayyah A, Sozer Y (2016) Plug-and-play nonlinear 23. El-Fergany AA, El-Hameed MA (2017) Efficient frequency con-
droop construction scheme to optimise islanded microgrid opera- trollers for autonomous two-area hybrid microgrid system using
tions. IEEE Trans Power Electron 32(4):2743–2756 social-spider optimiser. IET Gener Transm Distrib 11(3):637–648
9. Elrayyah A, Cingoz F, Sozer Y (2017) Smart loads management 24. Ramesh M, Yadav AK, Pathak PK (2023) Artificial gorilla troops
using droop-based control in integrated microgrid systems. IEEE optimizer for frequency regulation of wind contributed microgrid
J Emerg Sel Top Power Electron 5(3):1142–1153 system. J Comput Nonlinear Dynam 18(1):011005. https://doi.org/
10. Roozbehani S, Hagh MT, Zadeh SG (2019) Frequency con- 10.1115/1.4056135
trol of islanded wind powered microgrid based on coordinated 25. Shankar G, Mukherjee V (2016) Load frequency control of an
robust dynamic droop power sharing. IET Gener Transm Distrib autonomous hybrid power system by quasi-oppositional harmony
13:4968–4977. https://doi.org/10.1049/ietgtd.2019.0410 search algorithm. Int J Electr Power Energy Syst 78:715–734
11. Rodrigues Y, Abdelaziz M, Wang L (2020) D-PMU based 26. Bordons C, Garcia-Torres F, Ridao MA (2020) Model predictive
secondary frequency control for islanded microgrids. IEEE control of microgrids. Springer, Cham
Trans Smart Grid 11:857–872. https://doi.org/10.1109/TSG.2019. 27. Joshal KS, Gupta N (2023) Microgrids with model predictive con-
2919123 trol: a critical review. Energies 16:4851. https://doi.org/10.3390/en
12. Maneesh (2015) Frequency Control of a Microgrid by Using PI 16134851
Controller In: IEEE Trans 28. Saleh A, Hasanien HM, Turky AR, Turdybek B, Alharbi M, Jurado
13. Regad M, Helaimi M, Taleb R, Gabbar HA, Othman AM (2019) F, Omran WA (2023) Optimal model predictive control for virtual
Fractional order PID control of hybrid power system with renew- inertia control of autonomous microgrids. Sustainability 15:5009.
able generation using genetic algorithm. In: 2019 IEEE 7th Inter- https://doi.org/10.3390/su15065009
national Conference on Smart Energy Grid Engineering (SEGE), 29. Long B, Liao Y, Chong K, Rodríguez J, Guerrero J (2021) Enhance-
pp 139–144 ment of frequency regulation in AC microgrid: a fuzzy-MPC
14. Kumari S, Pathak PK (2023) A state-of-the-art review on recent controlled virtual synchronous generator. IEEE Trans Smart Grid
load frequency control architectures of various power system con- 12:3138–3149. https://doi.org/10.1109/TSG.2021.3060780
figurations. Electric Power Compon Syst. https://doi.org/10.1080/ 30. Kayalvizhi S, Kumar D (2017) Load frequency control of an iso-
15325008.2023.2234373 lated micro grid using fuzzy adaptive model predictive control.
15. Ali H, Magdy G, Li B, Shabib G, Elbaset A, Xu D, Mitani Y IEEE Access 5:16241–16251. https://doi.org/10.1109/ACCESS.
(2019) A new frequency control strategy in an islanded micro- 2017.2735545
grid using virtual inertia control-based coefficient diagram method. 31. Zhang Y, Zhang T, Wang R, Liu Y, Guo B (2015) Optimal
IEEE Access 7:16979–16990. https://doi.org/10.1109/ACCESS. operation of a smart residential microgrid based on model pre-
2019.2894840 dictive control by considering uncertainties and storage impacts.
16. Kushwaha P, Prakash V, Bhakar R, Yaragatti UR (2022) Synthetic Sol Energy 122:1052–1065. https://doi.org/10.1016/J.SOLENER.
inertia and frequency support assessment from renewable plants in 2015.10.027
low carbon grids. Electric Power Syst Res 209:107977. https://doi. 32. Parisio A, Rikos E, Glielmo L (2014) A model predictive control
org/10.1016/j.epsr.2022.107977 approach to microgrid operation optimization. IEEE Trans Control
17. Nema S, Prakash V, Pandžić H (2022) Adaptive synthetic inertia Syst Technol 22:1813–1827. https://doi.org/10.1109/TCST.2013.
control framework for distributed energy resources in low-inertia 2295737
microgrid. IEEE Access 10:54969–54979. https://doi.org/10.1109/ 33. Maślak G, Orłowski P (2022) Microgrid operation optimization
ACCESS.2022.3177661 using hybrid system modeling and switched model predictive con-
18. Prakash V, Pandzic H (2023) Fast frequency control service pro- trol. Energies. https://doi.org/10.3390/en15030833
vision from active neighborhoods: opportunities and challenges. 34. Mohamed F, Koivo H (2012) Online management genetic
Electric Power Syst Res 217:109161. https://doi.org/10.1016/j. algorithms of microgrid for residential application. Energy
epsr.2023.109161 Convers Manage 64:562–568. https://doi.org/10.1016/J.ENCO
19. Kushwaha P, Prakash V, Yamujala S, Bhakar R (2023) Fast fre- NMAN.2012.06.010
quency response constrained electric vehicle scheduling for low 35. Garcia CE, Prett DM, Morari M (1989) Model predictive control:
inertia power systems. J Energy Storage 62:106944. https://doi. Theory and practice—a survey. Automatica 25:335–348
org/10.1016/j.est.2023.106944
20. Pathak PK, Yadav AK (2023) Design of optimal cascade control
approach for LFM of interconnected power system. ISA Trans
137:506–518. https://doi.org/10.1016/j.isatra.2023.01.029

123
Electrical Engineering

36. Richalet J, Rault A, Testud J, Papon J (1978) Model predictive 41. Dashtdar M, Flah A, El-Bayeh CZ, Tostado-Véliz M, Durra AA,
heuristic control: applications to industrial processes. Automatica Abdel Aleem HE, Ali ZM (2022) Frequency control of the islanded
14:413–428 microgrid based on optimised model predictive control by PSO.
37. Seborg DE, Edgar TF, Mellichamp DA, Doyle FJ III (2016) Process IET Renew Power Gener 16(10):2088–2100. https://doi.org/10.
dynamics and control. Wiley, New York 1049/rpg2.12492
38. Lee D-J, Wang L (2008) Small-signal stability analysis of an
autonomous hybrid renewable energy power generation/energy
storage system part i: time-domain simulations. IEEE Trans
Publisher’s Note Springer Nature remains neutral with regard to juris-
Energy Convers 23(1):311–320. https://doi.org/10.1109/TEC.
dictional claims in published maps and institutional affiliations.
2007.914309
39. Bevrani H, Habibi F, Babahajyani P, Watanabe M, Mitani Y (2012)
Springer Nature or its licensor (e.g. a society or other partner) holds
Intelligent frequency control in an AC microgrid: online PSO-based
exclusive rights to this article under a publishing agreement with the
fuzzy tuning approach. IEEE Trans Smart Grid 3(4):1935–1944.
author(s) or other rightsholder(s); author self-archiving of the accepted
https://doi.org/10.1109/TSG.2012.2196806
manuscript version of this article is solely governed by the terms of such
40. Goldberg DE (1989) Genetic algorithms in search, optimization
publishing agreement and applicable law.
and machine learning, reading. Addison-Wesley Publishing Com-
pany, New York

123