Science and Technology for Energy Transition 79, 45 (2024)

Ó The Author(s), published by EDP Sciences, 2024

https://doi.org/10.2516/stet/2024034

Decarbonizing Energy Systems: Smart Grid and Renewable Technologies Available online at:
stet-review.org
Fang Rui-ming and Qiangqiang Liao (Guest Editors)
REGULAR ARTICLE

Enhanced Randomized Harris Hawk Optimization of PI controller

for power flow control in the microgrid with the PV-wind-battery
system
Gollapudi Pavan* and A. Ramesh Babu

School of Electrical and Electronics, Department of Electrical and Electronics Engineering, Sathyabama Institute of Science
and Technology, Chennai 600119, India

Received: 18 January 2024 / Accepted: 15 May 2024

Abstract. Microgrids, characterized by their ability to work individually or in combination with the main
power system, play a pivotal role in addressing the growing demand for reliable and sustainable energy
solutions. This work concentrates on the integration of sustainable energy sources, specifically photovoltaic
(PV), and wind generation and a battery storage system within a microgrid framework. Additionally, a power
flow control strategy is implemented to enhance the dynamic behaviour and stability of the microgrid. The
proportional-integral (PI) controller is a fundamental component in regulating the microgrid’s power flow,
ensuring optimal performance under varying operating conditions. However, tuning the PI controller parame-
ters is a difficult task because of the dynamic and nonlinear nature of renewable energy sources. In this work,
the application of the Enhanced Randomized Harris Hawk Optimization (ERHHO) to fine-tune the PI con-
troller is proposed, using the algorithm’s ability to mimic the hunting behaviour of hawks in finding optimal
solutions. The PV-Wind-Battery microgrid system is modelled, and the proposed algorithm is employed to
optimize the PI controller parameters for efficient energy management. The ERHHO algorithm’s
exploration-exploitation balance is harnessed to navigate the complex solution space and converge to optimal
PI controller settings, thereby enhancing the microgrid’s stability and performance. The study evaluates the
effectiveness of the proposed ERHHO-based PI controller tuning through comprehensive simulations. Perfor-
mance metrics such as transient response, overshoot, settling time, and steady-state error are analysed to
validate the robustness and efficiency of the proposed method. Compared to its nearest optimization algorithm,
with the proposed algorithm rise time is reduced by 50%, overshoot is reduced by 75%, settling time is reduced
by 66%, and finally, a percentage of reduction of steady-state error is 45%. The outcomes of this research
contribute to the advancement of microgrid control strategies, offering a novel approach to PI controller tuning
in the context of diverse renewable energy sources. The integration of the Harris Hawk Optimization algorithm
provides a promising avenue for enhancing the operational efficiency and reliability of microgrids, paving the
way for sustainable and resilient energy systems in the aspect of growing energy landscapes.

Keywords: Microgrid, Renewable energy, Photovoltaic (PV), Wind generation, Proportional-integral (PI)
controller, Harris Hawk Optimization (HHO) algorithm, Microgrid stability, Simulation analysis.

1 Introduction deviations and running expenses [1]. The integration of a

power flow control within a Microgrid (MG) holds the
Distributed Generations (DGs) emerge as a viable remedy promise of diminishing working costs and enhancing energy
for the prevalent challenges encountered by global energy utilization efficiency [2]. DGs encompass a diverse array of
systems. The challenges are related to network overloading, energy sources, spanning renewable options like wind tur-
greenhouse gas emissions, system stability and reliability bines (WT), and photovoltaic (PV) systems, and alongside
and elevated utilization costs. The application of power flow non-renewable alternatives such as micro-turbines (MT),
control in microgrids for substantial commercial structures diesel generators (DiG), and gas engines (GE) [3]. Micro-
introduces complexities in curtailing network load grids consist of various integral components, including
flexible load management, DGs, and storage devices [4].
* Corresponding author: [email protected] Equipped with controllers, this microgrid model adeptly
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
2 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

manages the system by regulating loads to optimize solar exploration advantages of the SCA with the capability of
output. Furthermore, it facilitates bidirectional power flow, a steady-state genetic algorithm to avoid premature conver-
enabling the utilization of power from residential homes, gence. The results illustrate that the SSGA method signifi-
hostels, and academic departments [5, 6]. cantly enhances the operation of complex design problems
Power flow control in microgrids can operate in either in comparison to similar optimisation methods.
centralized or decentralized modes [7]. In centralized modes, In [20], a novel approach was introduced that integrates
the microgrid’s power exchange primarily depends on the Sine–Cosine Algorithm (SCA) for learning and initializ-
market prices, whereas decentralized modes allow for ing individuals to solve diverse optimization problems. This
autonomous power exchange without being constrained method is enhanced by incorporating nearly contradictory
by market prices [8, 9]. Microgrids confront significant chal- learning approaches and methods incorporating contrasting
lenges related to stability, efficiency, and energy protection. principles, thereby improving the effectiveness and speed of
These challenges arise from factors such as reverse power overall exploration. The efficacy of the proposed method
flow from power sources voltage instability, transient states was assessed by standard functions, and the results under-
within the microgrid, notable frequency changes in stan- scored its superiority over alternative approaches, showcas-
dalone operating modes, and uncertainties in availability ing significant performance improvements.
management dynamics. Ensuring appropriate load alloca- In [21], an innovative approach was employed by
tion, specifically during system vulnerability, often requires combining the gravitational search method with the firefly
microgrids to employ high levels of angle droop. Power flow optimizing technique. This combined methodology aimed
control also suffers from various challenges like increasing to improve the adjustment of controller settings and was
penetration of renewable energy sources, dynamic load vari- applied within a 2 area hydrothermal power system. In
ations and voltage and frequency regulation. To tackle [22], researchers adopted the arithmetic optimization
these issues, a comprehensive tuning of controller gains is approach (AOA) to meticulously fine-tune a fuzzy-PID
required to improve the performance of microgrids [10, 11]. controller. This fine-tuning process is adopted in the impact
In recent times, researchers have introduced resilient of the high voltage direct current (HVDC) link, strategi-
and innovative strategies inspired by nature to effectively cally addressing the limitations associated with alternating
tackle complex challenges like Proportional Integral con- current (AC) transmission.
troller (PI) tuning across diverse domains. Famous method- In [23], researchers implemented a flower pollination
ologies include Genetic Algorithm (GA) [12], Particle optimization algorithm (FPOA) within Multi-Agent Power
Swarm Optimization (PSO) [13], and various others [14]. Systems (MAPSs) to develop a Proportional-Integral-
In [15], a novel technique called Chaos Cultural Sine– Derivative (PID) controller utilizing spontaneous flower
Cosine Algorithm (CCSCA) was introduced to address pollination techniques. To address Load Frequency Control
premature convergence issues in the Sine–Cosine Algorithm (LFC) in interconnected MAPSs, a bacterial foraging
(SCA). This approach incorporates a social algorithm as approach was incorporated, subject to Particle Swarm
the population structure and utilizes disordered methods Optimization (PSO), with a focus on the Proportional-
for particle search. Performance evaluation on benchmark- Integral (PI) controller. This was applied under various
ing functions demonstrated that the CCSCA surpassed the fitness functions, encompassing both standard and
basic SCA algorithm in global search capability, producing customized criteria, and considered two segments within
superior results. In [16], a binary Sine–Cosine Algorithm non-reheat thermal systems [24].
(SCA) approach was suggested to direct the unit commit- In [25], an advanced technique called the Hybrid Sine–
ment problem. The binary SCA approach utilized enhanced Cosine Algorithm (HSCA) was formulated, using SCA to
sigmoidal features to effectively represent the continuous address optimization problems. The results demonstrated
search space. Results showcased the efficiency of the binary that the proposed HSCA consistently outperforms other
SCA method related to alternative approaches. In [17], relevant approaches, yielding superior scheduling outcomes
introduced a discrete version of the Sine–Cosine Algorithm across multiple cases. Its capacity to produce higher-quality
(SCA) designed for optimizing complicated problems. The results establishes its effectiveness in addressing optimiza-
effectiveness of the proposed discrete SCA was demon- tion challenges.
strated through multiple benchmark functions, emphasizing The increasing integration of renewable energy sources,
its efficiency. within microgrids has spurred the need for effective control
In [18], a hybrid optimization depending on the Sine– strategies to optimize energy management. Proportional-
Cosine Algorithm (SCA) was developed to effectively tackle Integral (PI) controllers play a crucial role in regulating
optimization problems. This hybrid approach integrates system dynamics and ensuring stability. Microgrids serve
concepts and operators from various algorithms, fostering diverse loads with varying power consumption patterns.
a robust interchange of exploration and exploitation The dynamic nature of these load profiles introduces
capabilities. To evaluate its effectiveness, seven benchmark challenges in PI controller tuning, as controllers must be
functions were employed. The outcomes revealed that able to adapt to changes in load demand quickly and
the proposed hybrid method surpassed other similar efficiently to maintain grid stability and ensure reliable
approaches, providing a highly competitive solution. In power supply. Microgrids are designed to operate both
[19], a novel hybrid algorithm named Steady State Genetic connected to the main grid and in islanded mode, where
Algorithm (SSGA) was introduced for tackling they function independently during grid outages. PI con-
complex problems. This hybrid method amalgamates the trollers must be tuned to facilitate a smooth transition
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 3

between grid-connected and islanded modes while main- power measured at the output of the respective DC/DC
taining voltage and frequency stability within acceptable converters. P dc available DC power at DC bus.
limits. The dynamics of microgrid components, such as The total power generated by renewable sources is
inverters and energy storage systems, can exhibit nonlinear determined using equation (1), which incorporates the total
behaviour under certain operating conditions. Tuning PI power flow model. The power balance is computed by
controllers to effectively regulate these nonlinear dynamics considering the output of the hybrid sources and the load
while ensuring stability and performance is a complex task demand at the AC bus.
that requires careful consideration of system dynamics and
control strategies. P g ðt Þ ¼ ½P l ðt Þ  P hres ðt Þ ð2Þ
The focus of this study revolves around utilizing the

Enhanced Randomized Harris Hawk Optimization (HHO) P l ðt Þ ¼ P hres ðt Þ þ P g ðt Þ ð3Þ
algorithm for tuning PI controllers within the power flow
control of a microgrid featuring PV, and wind generation. Q g ðt Þ ¼ ½Q l ðt Þ  Q hres ðt Þ ð4Þ
The complex relationship of these diverse energy sources
necessitates an intelligent optimization approach to fine-tune

the PI controllers, thereby enhancing the overall perfor- Q l ðt Þ ¼ Q hres ðt Þ þ Q g ðt Þ : ð5Þ
mance and efficiency of the microgrid. This research explores
the application of ERHHO in the PI controller tuning, P g is active power provided by the grid, Q g reactive power
aiming to attain an optimal balance among energy genera- by the grid, P l load active power, Q l load reactive power,
tion, utilization, and system stability in a microgrid incorpo- P hres active power generated by HDGS and Q hres reactive
rating a mix of renewable and conventional energy sources. power by HDGS.
The battery power of the storage unit is influenced by
the duration of discharge, where it acts as an energy source,
and the duration of charging, where it functions as a load.
2 Hybrid DG system (HDGS) However, ensuring a consistent power balance proves chal-
lenging due to the nonlinear fluctuations on the consumer
The schematic of the hybrid system is presented in Figure 1. side and the unpredictable nature of sustainable energy
The PV-wind-battery system contains a wind turbine, PV [27]. To tackle this challenge, the system unit requires a
system and battery, working in hybrid to facilitate energy high-performance operation mode that enables efficient
transfer from the DC bus to the grid. The PV generation power regulation. The active and reactive power calculated
system is connected to the DC bus through an MPPT- by direct and quadrature axis voltages and currents are
operated DC-DC converter. The wind generation system given as shown in equations (6) and (7).
is connected to the DC bus through a rectifier and
MPPT-operated DC-DC converter. The battery storage 3
P I ðt Þ ¼ ½v d  i d þ v q  iq  ð6Þ
system is connected to the DC bus through a bidirectional 2
DC-DC converter. The design of power flow control and
management modules for the PV-wind-battery system unit 3
is derived from the operational modes of the load and grid. Q I ðt Þ ¼ ½ v q  i d þ v d  i q  : ð7Þ
2
During the grid-connected mode, the active and reactive
output powers of the PV-wind-battery system unit are v d and v q are direct and quadrature axis load voltages,
adjusted to align with their corresponding reference values. i d and iq are direct and quadrature axis load currents.
Therefore, the voltage source inverter (VSI)-based control
system must select the suitable power control mode to
ensure optimal performance. A sinusoidal pulse width 3 Modelling of sources
modulation (SPWM) is adopted to control the VSC. To
accommodate varying load requirements, the power pro- 3.1 Mathematical formulation of solar PV module
vided by the main grid and PV-wind-battery system needs
to be flexibly adjusted. Maintaining a consistent supply of The simplified circuit representation of a solar panel is
active and reactive power necessitates the efficient control depicted in Figure 2.
of power flow among the PV-wind-battery system and the The PV cell can be defined as an ideal current source
grid and load. Additionally, a reliable connection between denoted as I ph with a series of parallel resistances as shown
the power obtained from the main grid and the DG system in Figure 2. The output current of an ideal solar cell is
is essential for meeting the load demand [26]. described in equation (8)
The power balance equation must be fulfilled by the I ¼ I ph  I d : ð8Þ
DC-link and point of common coupling (PCC). Total DC
Power available at DC bus by energy sources in a microgrid I is PV output current, I d is diode current, I ph photon
is defined as current.
P dc ð t Þ ¼ P WT ðt Þ þ P PV ðt Þ þ P Bat ðt Þ : ð1Þ In the realm of semiconductor theory, the primary
mathematical equation that interprets the I-V characteris-
P WT power generated by wind generation system, P PV tics of the PV cell is known as Shockley’s diode current
power generated by solar generation system, P bat battery equation, as depicted in equation (9)
4 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 1. Management structure of HDGS with the proposed controller.

I s is the saturation current, q is the electron charge, V oc is

the open circuit voltage, N s series connected cells per mod-
ule, K Boltzmann’s constant, T o nominal cell temperature
in kelvin.
Given the sensitivity of the photovoltaic (PV) genera-
tion output to weather conditions such as irradiance and
temperature, the use of a DC-DC converter becomes crucial
for regulating both the output voltage and power. Figure 3
illustrates the importance of regulating the PV voltage to
maximum power point voltage (V mpp ), a critical factor in
harnessing the peak power potential from the PV arrays
Figure 2. Equivalent circuit of real model for PV panel. by the implementation of a DC-DC converter. The maxi-
mum power point voltage of a PV panel can be determined
    from its characteristic curve, often referred to as the IV
qV oc curve (current-voltage curve) or P-V curve (power-voltage
Id ¼ Is exp 1 ; ð9Þ
N s KAT o curve). The P-V curve shows the relationship between the
power output (P) and the voltage (V) across the terminals
then of the PV panel. It is derived from the IV curve and pro-
    vides insights into the maximum power point (MPP) at
qV oc
I d ¼ I ph  I s exp 1 : ð10Þ which the panel operates most efficiently. The peak point
N s KAT o on the P-V curve where the power output (P) is maximized
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 5

By continuously adjusting the operating point based on

the observed changes in power output, the P&O algorithm
converges towards the MPP of the PV system. However,
it’s worth noting that the P&O algorithm may oscillate
around the MPP, especially under rapidly changing environ-
mental conditions or when the system is operating near local
power maxima that are not the global MPP. To mitigate
these issues, advanced MPPT algorithms with additional
features and control strategies may be used in conjunction
with or instead of the P&O algorithm. Nonetheless, the
P&O algorithm remains popular due to its simplicity and
effectiveness in many PV system applications.

3.2 PMSG and wind turbine

Output power from a wind turbine (P x Þ is given as in
equation (11)
1
P x ¼ qAC p ðk; bÞV 3x : ð11Þ
2
q is the air density, A wind turbine swept area in m2, k is
the tip speed ratio, b is the pitch angle. The tip speed ratio
Figure 3. I-V and P-V characteristics of PV array. “ k ” is defined as the ratio between the blade tip speed “x,”
the wind turbine blade radius “r,” and the wind speed “V x ,”
and it can be expressed by equation (2).
represents the maximum power point (MPP) of the PV
panel. As depicted in Figure 3, the control of PV voltage rx
k¼ : ð12Þ
to reach V mpp is essential for maximizing power extraction Vx
from the PV arrays via the DC-DC converter. The P&O
MPPT (Perturb and Observe Maximum Power Point Hence, the mechanical energy generated by wind turbines
Tracking) algorithm, shown in Figure 4, is employed to gen- may be regulated by adjusting both “k” and “b.” For any
erate the duty cycle necessary for the DC-DC converter to given “b,” there exists a respective power coefficient (Cp )
achieve maximum power extraction. Figure 5 presents the vs tip speed ratio (k) plot, where each curve has an optimal
DC-DC converter utilized for the PV array. Cp value, denoted as Cp opt corresponding to an optimal tip
The Perturb and Observe (P&O) algorithm is a widely speed ratio, kopt . By controlling these parameters, the
used MPPT technique in PV systems. It is a simple and effi- mechanical energy of the turbine can be adjusted to match
cient method for continuously adjusting the operating point the available wind speed, with the maximum Cp attained at
of a PV system to ensure it operates at its maximum power the optimized rotational speed when “b” equals zero.
point (MPP), where the power output is maximized. The The mechanical energy generated by wind turbines is
algorithm starts by initializing the operating point of the converted into electrical energy with a PMSG (Permanent
PV system. This can be done by setting the operating volt- Magnet Synchronous Generator). The voltage and current
age or current to a predetermined value, typically close to relation for the PMSG are defined as shown in equations
the expected MPP. The algorithm measures the power out- (13) and (14):
put of the PV system at the current operating point. This v dgen ¼ ðRa þ pLd Þi dgen  xe Lq i qgen ð13Þ
can be done by multiplying the current and voltage read-
ings obtained from the PV system. Based on the power
measurement obtained in the previous step, the algorithm v qgen ¼ ðRa þ pLq Þi qgen þ xe Ld i dgen þ uf xe : ð14Þ
perturbs the operating point by either increasing or decreas-
ing the voltage or current slightly. This perturbation is typ- v dgen and v qgen correspond to the stator dq-axis voltages of
ically small to ensure that the system remains near the the PMSG, while idgen and i qgen represent the stator dq-axis
MPP. After perturbing the operating point, the algorithm currents of the PMSG.
measures the power output of the PV system at the new The electromechanical developed torque T e is deter-
operating point. The algorithm compares the power output mined as provided in equation (15):
at the new operating point with the power output at the 3 p
previous operating point. If the power output has increased, Te ¼ uf i qgen þ ðLd  Lq Þi qgen idgen : ð15Þ
2 2
it continues to perturb the operating point in the same
direction. If the power output has decreased, it reverses In the equation, uf represents the amplitude of the flux
the direction of perturbation. These steps are repeated iter- linkage, and p denotes the pole pairs. Ld direct axis induc-
atively, continuously adjusting the operating point of the tance, Lq quadrature axis inductance,
PV system based on the comparison of power outputs at The relationship between mechanical torque and speed
different points. is defined by equation (16)
6 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 4. MPPT algorithm.

challenge when it comes to selecting the optimal step size

for rotor speed adjustments.
The P&O algorithm for MPPT in wind generation
systems involves continuously perturbing the turbine
parameters (such as rotor speed or blade pitch angle) and
observing the resulting changes in power output. It begins
by detecting variations in wind speed, typically measured
using an anemometer. Once a change in wind speed is
detected, the algorithm perturbs the turbine parameters
in a specific direction (e.g., increasing rotor speed) and
observes the resulting change in power output. If the power
output increases, the algorithm continues perturbing in the
same direction. However, if the power output decreases, it
Figure 5. PV system with boost converter. reverses the perturbation direction. This continuous adjust-
ment process helps the algorithm converge towards the
    Maximum Power Point where the wind turbine operates
2 dxr 2 most efficiently. The P&O algorithm iteratively adjusts
Te ¼ J þ xm þ TL : ð16Þ
p dt p turbine parameters based on the observed changes in power
output, aiming to reach the MPP. It repeats this process
xm rotor speed in rad/s, T L is load torque in N/m. until the algorithm converges to a stable operating point
The power produced by the PMSG can be regulated by where further parameter adjustments do not significantly
a DC-DC converter. This converter is operated by an increase power output. The algorithm continuously
MPPT algorithm. Before the DC-DC boost converter takes monitors wind conditions and adjusts turbine parameters
control of the regulation, the PMSG’s output is connected accordingly to maintain optimal performance, ensuring effi-
to a diode bridge rectifier. cient power extraction from the wind.
To improve the dynamic performance of the wind The step size during the perturbation process can either
generation system, an adjustable step size (P and O) algo- stay the same or vary. Using a constant step size in Conven-
rithm is used to guide and manage the DC-DC converter. tional P&O algorithms can present certain difficulties. To
This approach ensures that the system reaches its peak address these limitations, an adjustable step P&O algo-
power output. Differing from traditional P&O algorithms, rithm is used, with the step size determined by the wind
this adaptive method tackles numerous shortcomings speed. By making use of the wind turbine characteristics,
that can adversely affect the performance of the wind gener- estimate the parameter kopt , which is typically around 8.1.
ation system, especially in systems characterized by inertia. Afterwards, calculate the optimal rotor speed using
Traditional P&O algorithms encounter a substantial equation (17), as shown below:
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 7

Figure 6. MPPT flow chart for wind generation.

kopt V w
xoptm ¼ : ð17Þ
R
kopt optimal tip speed ratio,
xopt m is directly related to wind speed expressed in
equation (18)
tw
xopt mp ¼ xbase m : ð18Þ
tbase
tbase refers to the reference wind speed, xbase m represents
the rotor speed at the reference wind speed, xopt mp indi-
cates the current optimal generator speed, and tw repre-
sents the current wind speed.
The control system will make significant corrections to
the rotor speed when it changes from the required optimal
speed. If the rotor speed gets closer to the required level the
adaptive ratio Radp ; will decrease until the optimal speed is
achieved and maintained as shown in equation (19)
 
xopt mp  xm
Radp ¼ K per : ð19Þ
xopt mp
Figure 7. Turbine speed VSs output power.
The perturbation constant K per determines the degree of
accuracy in achieving maximum power output. Figure 7
presents the variation in output power of wind turbines rotor at its optimal speed (sopt m ). Following this, the input
with respect to turbine speed. current reference (I wref ) for the DC-DC converter can be
The flowchart presented in Figure 6 defines the adaptive evaluated from the torque reference as in equation (20).
step P&O MPPT algorithm. After the algorithm determi-
xm
nes the ideal rotor speed, a PI controller comes into play I wref ¼ sopt m : ð20Þ
to produce a reference torque aimed at keeping the PMSG V dcw
8 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 8. Wind generation system with boost converter.

A DC current regulator with PI control is employed to

maintain the output voltage of the DC-DC converter. This
regulator generates the necessary duty cycle to control the
IGBT switch. Figure 8 depicts the wind system incorporat-
ing the DC-DC converter. The output of the PI 1 controller
is reference torque sopt m which is given as in equation (21)
Z
sopt m ¼ K p1 ðxopt m ðt Þ  xm ðt ÞÞ þ K i1 ðxopt m ðt Þ  xm ðt ÞÞdt :

ð21Þ

K p1 and K i1 are PI controller gains.

Output of the PI 2 controller is the duty cycle (d w Þ Figure 9. Battery storage system with DC-DC converter.
required by the converter which is given as
Z
d w ¼ K p2 I Wref ðt Þ  I W ðt Þ þ K i2 I Wref ðt Þ  I W ðt Þ dt : I batref ¼ K p3 V batref ðt Þ  V dcbat ðt Þ
Z
ð22Þ þ K i3 V batref ðt Þ  V dcbat ðt Þ dt: ð25Þ
K p2 and K i2 are PI controller gains.
K p3 and K i3 are PI controller gains, V batref battery reference
3.3 Battery energy storage system (BESS) voltage, V dcbat actual battery voltage (Figure 10).
Output of the PI 4 controller is the duty cycle of the
The battery’s behaviour can be described by its state of switches in the bidirectional converter which is given as in
charge (SoC) expressed as a percentage, which is linked equation (26)
to the battery’s energy level, denoted as BLðtÞ at a given Z
time expressed in equation (23) d bat ¼ K p4 I batref ðt Þ  I bat ðt Þ þ K i4 I batref ðt Þ  I bat ðt Þ dt :
BLðt Þ ð26Þ
SOC ðt Þ ¼  100% : ð23Þ
BLcap
K p4 and K i4 are PI controller gains, I batref battery reference
subjected to as in equation (24) current, I bat actual battery current.
SOC min < SOC ðt Þ < SOC max : ð24Þ

Here, BLcaps represents the battery’s initial nominal capac- 4 Current control strategy of DC–AC
ity, while SOC min and SOC max denote the upper and lower converter
bounds of the battery’s SoC, respectively. These limits are
established to minimize the impact of battery ageing and, The active and reactive power controller introduces four PI
consequently, to prolong the battery’s lifespan. Figure 9 controllers to enable regulated active and reactive power
depicts the utilization of a DC-DC converter and a VSI delivery to the load. These controllers manage the power
for connecting the battery to the AC bus. flow between the grid and the utility.
The PI controller generates the duty cycle of the Equations (27) and (28) showcase the reference current
two switches, within the bidirectional DC-DC converter. utilized for achieving the control objective. The two PI
According to Figure 9, the output of the PI 3 controller controllers within the power controller establish communi-
is reference battery current I batref which is given as in cation with the external control loop to generate the refer-
equation (25), ence current vectors i dref and i qref .
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 9

Figure 10. Control strategy of the DC–AC converter.

Z
used in control systems to regulate a process variable.
idref ¼ K p5 P ref ðt Þ  P ðt Þ þ K i5 P ref ðt Þ  P ðt Þ dt
Heuristic optimization algorithms can be employed to auto-
ð27Þ mate the tuning process of a PI controller. These algorithms
explore the parameter space, searching for optimal con-
Z troller settings that minimize a performance metric such
i qref ¼ K p6 Q ref ðt Þ  Q ðt Þ þ K i6 Q ref ðt Þ  Q ðt Þ dt : as integral of time-weighted absolute error (ITAE), over-
shoot, or settling time. When applying heuristic optimiza-
ð28Þ tion algorithms to PI controller tuning, it’s essential to
define the objective function based on the specific control
K p5 and K i5 are PI controller gains, P ref reference active system’s performance criteria. Additionally, the algorithm
power, Q ref reference reactive power, P actual active power parameters (e.g., population size, and mutation rate)
and Q actual reactive power. should be carefully selected and tuned for the application.
The controller’s output is crucial to ensure precise track- It’s often beneficial to run multiple optimization trials
ing and minimize inverter ripple. This objective is accom- and analyse the results to ensure robustness and reliability
plished by employing two PI controllers to eliminate in achieving optimal controller settings.
current errors, along with an inverter current feedback The objective of design optimization is to identify the
and grid voltage feedforward loop, which enhances both optimal design by minimizing an objective function, adjust-
steady-state and dynamic performance. The controller ing design variables, and adhering to specified design con-
employs a PWM system to generate voltage vectors with straints. When undertaking design optimization, it is
reduced harmonic distortion. Besides the PWM system, common to encounter scenarios where multiple design crite-
the controller incorporates a current feedback loop and a ria or objective functions must be considered simultane-
grid voltage feedforward loop to improve steady-state and ously. In such instances, the design problem transforms
dynamic performance. These loops are used to remove cur- into a multi-objective optimization challenge.
rent errors, achieve precise tracking, and minimize inverter In contrast to conventional optimization methods tai-
drift as can be seen in equations (29) and (30). lored for scalar objective functions, the approach to multi-
Z objective optimization differs significantly. Traditional
v dref ¼ K p7 i dref ðt Þ  id ðt Þ þ K i7 i dref ðt Þ  i d ðt Þ dt methods that focus on a single objective function are inad-
equate for handling the complexity and nuances associated
ð29Þ with optimizing multiple, potentially conflicting objectives
simultaneously. Therefore, specialized multi-objective opti-
Z
mization techniques become essential in addressing the
v qref ¼ K p8 iqref ðt Þ  i q ðt Þ þ K i8 iqref ðt Þ  i q ðt Þ dt : intricacies of these multifaceted design problems.
Multi-objective optimization can be defined as shown in
ð30Þ equations (31)–(34)
K p7 and K i7 are PI controller gains, i dref reference direct min F ðx; k Þ ¼ ½F 1 F 2 . . . F n 
T
ð31Þ
axis current, i qref reference quadrature axis current, i d
actual direct axis current and iq actual quadrature axis u ðx; k Þ  0 ð32Þ
current.
During grid-connected mode, the HDGS unit controls
both the magnitude and phase angle of the inverter current v ðx; k Þ ¼ 0 ð33Þ
to deliver the predetermined active and reactive power val-
ues to the grid. The control strategy for active and reactive x i;lb  x i  x i;ub ði ¼ 1; 2 . . . d Þ : ð34Þ
power in the HDGS unit relies on frequency and voltage
regulation. Evaluating the performance of the HDGS unit Consider the objective function vector F, which is a func-
involves considering the outputs of the PV irradiance, wind tion of the design vector x and a constant parameter vector
turbine, and battery. p. The inequality and equality constraints are denoted by
Tuning PI controller gains is essential to ensure effective u and v respectively. Additionally, x i;lb and x i;ub represent
and stable control of a system. PI controllers are commonly the lower and upper bounds for the ith design variable.
10 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

The predominant approach for multi-objective optimiza- 5.1 Exploration stage

tion is the weighted sum method. This method involves
integrating multiple objectives into a unified scalar objec- During this stage, hawks select random perching locations,
tive function. It achieves this by assigning a weighting influenced by the positions of other group members or the
factor to each objective function and computing their locations of rabbits. This behaviour is modelled as shown
sum. In essence, the transformed scalar objective function in equations (37) and (38)
is a weighted combination of the individual objective func-
X ð t þ 1Þ
tions, facilitating a consolidated optimization criterion
highlighted as F objective in equation (35). (
X rand ðt Þ  r 1 jX rand ðt Þ  2r 2 X ðt Þj; q  0:5
F objective ¼ a 1 F 1 þ a 2 F 2 þ . . . þ a n F n : ð35Þ ¼
X rabbit ðt Þ  X m ðt Þ  r 3 ðLB þ r 4 ðUB  LB Þ; q < 0:5
Where a 1 ; a 2 . . . ; a n are the weights of the respective objec- ð37Þ
Pn
tive functions. If a i ¼ 1 and 0  a i  1, the weighted
i¼1 1 XN

sum is characterized as a convex combination of objectives. X m ðt Þ ¼ X i ðt Þ : ð38Þ

N i¼1
In this scenario, each optimization focusing on a single
objective identifies a specific optimal solution point along In this context, X ðt þ 1Þsignifies the forthcoming location of
the Pareto front. The weighted sum method subsequently hawks in the subsequent cycles, where X rabbit ðt Þ denotes the
varies weights systematically, resulting in distinct single- prey’s current location, X(t) represents the current posi-
objective optimizations that yield different optimal tions of the hawks. r 1 ; r 2 ; r 3 ; r 4 , and q denote random num-
solutions. bers within the range (0, 1). X rand ðt Þ corresponds to the
In this work, the optimization of PI controller gains is location of a randomly chosen member from the hawk and
achieved by adopting the Integral of Time-weighted Abso- X m ðt Þ indicates the mean value of the location of hawks.
lute Error (ITAE) as the single objective function. The
objective function to be minimized is formulated as the 5.2 Transformation
sum of the integral of the absolute error multiplied by time
for each individual PI controller. The optimization goal is to The prey’s escape energy plays a crucial role in the interme-
minimize this aggregated ITAE, thereby enhancing the diate stage, and it is assessed using equations (39) and (40).
dynamic response and overall performance of the control  
system. The final objective function that should be min- E1 ¼ 2 1 
t
ð39Þ
imised is given in equation (36) T
F objective ¼ a 1 ITAE 1 þ a 2 ITAE 2 þ . . . þ a n ITAE n : ð36Þ
E ¼ E 0E 1 ; ð40Þ
The Enhanced Randomized Harris Hawk Optimization
algorithm is employed as the optimization technique for “t ” represents the present iteration, “E 0 ” denotes the initial
determining the optimal Proportional-Integral (PI) con- prays energy, which varies within the random range of 1
troller gains (proportional gain and integral time constant). to 1, and “T ” represents the maximum number of iterations.
This optimization process aims to enhance the dynamic
response of the system. The goal is to optimize all eight 5.3 Exploitation stage
PI controller gains (Two PI controllers are responsible for
In this phase, the hawks engage in pursuit of the prey,
regulating the boost converter in the wind generation
employing four distinct hunting approaches and considering
system, another two PI controllers govern the bidirectional
the prey’s diverting behaviour. A successful capture hinges
converter in the battery storage system and the remain-
on two factors: the escaping energy (E) and the escape
ing four PI controllers operate within the DC–AC con-
potential (r). The variable escape potential (r) is a random
verter) within the control system, aiming to achieve
number in the range of 0 to 1.
improved control over active and reactive power, as well
When r is equal to or greater than 0.5, and the absolute
as enhanced regulation of DC and AC side voltages. The
value of the escaping energy (|E|) is also equal to or greater
PI controller gains subject to optimization include
than 0.5, the hawks initiate a soft besiege as defined by
K p1 ; K p2 ; . . . K p8 , representing the eight proportional gains,
equations (41) and (42). This signifies that the prey pos-
along with I t1 ; I t2 ; . . . I t8 , which signify the integral time
sesses sufficient energy but their attempt to escape has been
constants.
unsuccessful.
5 Harris Hawks algorithm (HHO) X ðt þ 1Þ ¼ X ðt Þ  E jX rabbit ðt Þ  X ðt Þj ð41Þ

The HHO algorithm draws inspiration from the various X ðt Þ ¼ X rabbit ðt Þ  X ðt Þ : ð42Þ
hunting strategies employed by hawks when searching for
and capturing their hunt. HHO is a particle-based optimiza- Here, DX(t) indicates the difference between the positions
tion technique composed of 3 distinct steps: exploration, of the hawks at iteration t and the present location of the
transformation, and exploitation. prey. Additionally, X rabbit ðt Þ denotes the leap strength,
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 11

which undergoes random variation with each iteration. hunting capabilities of particle-based algorithms. Typically,
When r is greater than or equal to 0.5, and the absolute chaotic maps are incorporated into one or more phases
value of the escaping energy (|E|) is less than 0.5 then, of these algorithms, including the initial population
X ðt Þ is given as in equation (43). generation, exploration, or exploitation stages. The way
locations are initialized has a significant impact on both
X ðt Þ ¼ X rabbit ðt Þ  E jX ðt Þj : ð43Þ the population’s diversity and the algorithm’s stability.
While the ERHHO algorithm can ensure the randomness
When the escape potential (r) is less than 0.5, and the abso- of the s particles position during initialization. Chaotic
lute value of escaping energy (|E|) is equal to or greater sequences possess unpredictability and randomness.
than 0.5, hawks employ a more calculated, gentle encir- Chaotic mappings can initialize random numbers that
clement strategy known as “hard besiege” with gradually follow in the range of 0 and 1. The features and random-
intensifying swift dives. This behaviour is detailed as shown ness of this map can enhance performance by altering
in equations (44) and (45). the starting positions of hawks which is given in equation
Y ¼ X rabbit OV  E jJX rabbit ðt Þ  X ðt Þj ð44Þ (52)
8
> Xi
>
< a if X i < a
Z ¼ Y þ S  LF ðDÞ : ð45Þ
X iþ1 ¼ : ð52Þ
>
Here, “D” represents the dimension of the problem, “S” rep- : 1  X i if X  a
>
i
resents a random matrix with an order of 1  D, “LF” 1a
stands for the Levy flight relation, as given in equations
(46) and (47). In this equation, X iþ1 is the new location of the hawks after
applying chaotic mapping, X i denotes the hawks’ position,
ur and “a” is set to 0.8.
LF ðd Þ ¼ 0:01  1=b ð46Þ
jv j
!b1 6.2 Exploration factor
sð1 þ bÞ  sinpb=2
r¼ : ð47Þ During the exploration step, the ERHHO primarily calcu-
s 1 þ b2  b  2b2
1
lates position updates using equations (37) and (38), where
r1 and r3 represent random values within the limits of 0 and
The process of updating the positions of the hawks can be 1. While these locations introduce randomness into every
represented through the following equation (48), wherein iteration of the overall exploration, they may lack the
“u” and “v” are vectors with an order of 1  d, “b” is a con- desired adaptability. The conventional HHO models a
stant in the range of 0 to 1.5, and s represents the Gamma scenario where hawks may effectively chase and identify
function. their prey. However, there are occasions when the prey is
8 not easily visible, and hawks might take several hours to
< Y if F ðZ Þ < F ðX ðt ÞÞ detect their target.
X ðt þ 1Þ ¼ : ð48Þ
: Z if F ðZ Þ < F ðX ðt ÞÞ The parameters r1 and r3 can be regarded as the
step length, where larger values make hawks move more
quickly, and conversely, smaller values result in slower
In cases where the prey’s energy is completely consumed, a movement. There are two distinct scenarios for hawks
robust besiege is initiated, indicated by the conditions when finding their prey: immediate detection and long-term
“r” is less than 0.5 and the absolute value of “E” is less than search. In the first scenario, it’s essential to consider
0.5. The computation of “Y”, “Z” is illustrated in equations randomness in the length of iteration. In contrast, in the
(49) and (50), next situation, the step length should decrease. Since the
Y ¼ X rabbit ðt Þ  E jJX rabbit ðt Þ  X ðt Þj ð49Þ probability of a hawk spotting prey rises over time, hawks
must initially search in a widespread limit with larger
steps, and then transition to a more precise search in the
Z ¼ Y þ S  LF ðDÞ ð50Þ later iterations. Exploration factor (ef ) is given as in
8 equation (53)
< Y if F ðZ Þ < F ðX ðt ÞÞ    2 !
X ðt þ 1Þ ¼ : ð51Þ b p t
: Z if F ðZ Þ < F ðX ðt ÞÞ ef ¼ b  rand   cos  : ð53Þ
2 2 T

The value of “b” is set to 2, which has been determined to

6 Proposed algorithm yield satisfactory results through experimental testing.
The term ðb  rand  b=2Þ is presented to introduce step
6.1 Tent chaotic map length randomness by producing random numbers within
the range of (b/2, b/2). The cosine relation is employed
Over recent years, several researchers have demonstrated to create a non-linear convergence, transitioning from 1 to
the effectiveness of chaotic maps [28] in enhancing the 0 with each iteration.
12 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Then equation (37) is simplified as in equation (54) is retained and carried forward into the next iteration,
a greedy strategy is employed, which is detailed in
X ð t þ 1Þ equation (56)
( 8
X rand ðt Þ  ef jX rand ðt Þ  2r 2 X ðt Þj; q  0:5 < X m ðt þ 1Þ; f ðX m ðt þ 1Þ < f ðX Þðt þ 1Þ
¼ : X ð t þ 1Þ ¼ :
X rabbit ðt Þ  X m ðt Þ  ef ðLB þ r 4 ðUB  LB Þ; q < 0:5 : X ðt þ 1Þ f ðX ðt þ 1Þ < f ðX Þðt þ 1Þ
m
ð54Þ
ð56Þ
Enhancing Exploration in Optimization Algorithms: The
Role of the Exploration Factor. 6.4 The details of ERHHO
In optimization algorithms, the exploration factor plays
a critical role in determining the search behaviour during The ERHHO algorithm focuses on refining certain aspects
the iterative process. At the outset of the optimization pro- of HHO [36]. While HHO is effective in local exploitation,
cess, the exploration factor widens the permissible range of it tends to lack in total exploration. ERHHO addresses this
step lengths. Initially, the step lengths are confined to the by linking the transition from exploration to exploitation to
interval (0, 1). However, the exploration factor introduces the prey’s energy. In the initial iterations representing the
a shift, expanding the limits to (b/2, b/2). This adjust- exploration phase, the range of particles is typically inade-
ment allows for more extensive exploration during the early quate, leading to a slower convergence rate. As iterations
iterations. As the optimization algorithm progresses, the progress, the energy of the prey decreases, and the algo-
exploration process evolves. Initially, the broader explo- rithm enters a phase of local exploitation.
ration range allows the algorithm to explore a wide variety To adapt to these changing conditions, ERHHO
of solutions. As the number of iterations increases, the employs four distinct hunting techniques depending on
exploration factor gradually narrows the step length limits. the energy of the prey and the probability of escaping.
Consequently, the algorithm transitions from broad explo- Hence, the tent chaotic map is added to augment the par-
ration to finer exploration. Fine exploration focuses on ticle’s diversity. The fine-tuned crucial parameters during
smaller regions of the optimization space, aiming for preci- the exploration phase using the exploration factor. As part
sion and convergence toward optimal solutions. Despite the of the exploitation step, incorporated a random walk strat-
narrowing of step length limits, the exploration factor egy to improve the capability to escape local optima. These
ensures that randomness is preserved. Randomness is essen- combined approaches enhance both the convergence speed
tial for escaping local optima and discovering global optima. and accuracy, effectively elevating the overall optimization
By maintaining some degree of randomness in step lengths, performance of the program. ERHHO comprises three
the algorithm avoids getting stuck in suboptimal regions. fundamental components: initialization, location updating,
and fitness estimation. The generation of locations involves
an algorithmic complexity of O(N*D), with N represent-
6.3 Random walk strategy
ing the particle size and D representing the problem’s
During the exploitation stage of the HHO algorithm, hawks dimensions. Fitness solution evaluation takes O(N) compu-
update their positions using four search strategies. While tational effort. To safeguard against local optima, incorpo-
this boosts exploration, proceeding to the next iteration rated a random walk strategy, increasing the mathematical
without interference can lead the algorithm to be stuck in complexity to O (2  N  D  T), where T represents the
local optima. To mitigate this, common strategies like the number of iterations. Consequently, the total computa-
Gaussian random walk, Levy flight, and Brownian motion tional complexity of the proposed ERHHO algorithm is
[29] are applied. These strategies tend to stabilize within O (2  N  D  T). A detailed flowchart for the ERHHO
certain value ranges with high probability but can experi- optimization is shown in Figure 11.
ence substantial changes with low probability, introducing Steps involved in optimization.
a level of deviation to the methods.
When the fitness value is the same as in the previous  Defining the objective function that quantifies the
iteration, this step can be initiated which induces a performance of the microgrid.
deviation in the location of a hawk based on a chang-  Initialize a population of potential solutions (candidate
ing parameter calculated as ðc  rand ðiÞ  c=2Þ  sets of PI controller gains) using the ERHHO algo-
cos ðp=2  ðt=T Þ and declines as reiterations progress. rithm. This typically involves randomly generating a
This can be defined as in equation (55) set of potential gain values within the specified bounds.
  Evaluate the fitness of each candidate solution by
c
X m ði Þ ¼ X ði Þ þ c  rand  applying the corresponding PI controller gains to the
2 microgrid model and simulating its performance under
 2 ! ! various operating conditions and scenarios. The fit-
p t ness function assesses how well the controller achieves
 cos   ðX ðiÞ  X rabbit : ð55Þ
2 T the desired control objectives.
 The ERHHO algorithm iteratively explores the solu-
X m ðiÞ represents the latest location after implementing the tion space, updating the candidate solutions based
random walk strategy. To ensure that the effective result on their fitness values. It employs a combination of
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 13

Figure 11. Flowchart of the ERHHO optimization.

exploration (random search) and exploitation (local  The optimization process continues for a predefined
search) strategies to converge towards optimal or number of iterations or until a termination criterion
near-optimal solutions. At each iteration, the algo- is met, such as reaching a specified level of conver-
rithm perturbs the solutions, evaluates their fitness, gence or exceeding a maximum number of function
and updates them accordingly. evaluations.
14 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

 Once the optimization process is complete, extract the Table 1. Parameters of the Hybrid DG system.
best solutions found by the ERHHO algorithm, corre-
sponding to the optimal PI controller gains for the Parameter Value
microgrid. PV array
Manufacturer and model Sharp NU-SC350,
349.965 W
7 Simulation results Open circuit voltage ðV oc Þ 46.9 V
This section is dedicated to presenting simulation results Short circuit current ðI sc Þ 9.6 A
that aim to establish the credibility and efficacy of the Maximum power point 38.5 V
proposed control method ERHHO. A key performance voltage ðV mpp Þ
metric involves assessing the method’s ability is to maintain Maximum power point 9.09 A
the DC bus voltage, as well as active and reactive powers current ðI mpp Þ
within specified ranges across various Microgrid (MG) oper- Series-connected modules per 10
ational modes. The efficacy of the proposed control is illus- string
trated through simulation results conducted within the Parallel strings 43
MATLAB/SIMULINK environment. Table 1 presents the
specifications of the system adopted for simulation results. PMSG
To establish the superiority of the developed methodology Stator resistance 0.9585 X
based on Enhanced Randomized Harris Hawk Optimization Stator inductance 0.835 lH
(ERHHO), a comparative analysis is performed against Torque constant 4.29 Nm/A
outcomes obtained using Monte Carlo Artificial Bee Colony Inertia 0.0027 kg.m2
Optimization (MCABC) [30], Hybrid Teaching Learn-
Wind turbine
ing Based Optimization and Equilibrium Optimizer
(TLBO-EO) [31], Enhanced Slime Mold Optimization Algo- Mechanical output power 150 kW
rithm (SMO) [32], Artificial Rabbits Algorithm (ARA) [33], Base wind speed 10 m/s
Coot Bird Metaheuristic Optimizer (CB) [34], and Ziegler- Rotor diameter 24.6 m
Nichols (ZN) [35]. The control strategy outlined in Section 3 Swept rotor area 475 m2
is implemented for the system presented in Figure 1. The Battery
simulation involves the utilization of PI gains that have been
optimized through the proposed algorithm. Optimized gains Nominal voltage 500 V
by the proposed algorithm and various algorithms from the Rated capacity 1000 Ah
literature are presented in Table 2. Three cases are consid- Grid 11 kV, 50 Hz
ered to check the efficacy of the optimized PI gains by the
proposed algorithm: reference active and reactive power
change, irradiance and wind speed variation, and isolation
of PV and wind generation systems. and reactive power supplied to the load by the PV-wind-
Case 1: Reference Active and Reactive Power change. battery system. Figure 17 specifics the active and reactive
In this case, dynamic changes in active and reactive power powers provided by the grid. Notably, between 8 and
references (P ref and Q ref ) are examined over specific time 12 s, any excess power generated by the PV-wind-battery
intervals. These references are adjusted as follows: from system, beyond what is needed by the load, is absorbed
0 to 4 s, P ref is set to 300 kW and Q ref to 250 kVAR; from by the grid. The generated power by the PV and wind
4 to 8 s, then increased to 400 kW and 350 kVAR; sources is presented in Figure 18. With an irradiance of
from 8 to 12 s, a further increase is observed to 600 kW 1000 w/m2 generated power from PV is 143.6 kW and with
and 550 kVAR; from 12 to 16 s, a reduction to 350 kW a wind speed of 12 m/s generated power from the wind gen-
and 300 kVAR occurs, and from 16 to 20 s, stabilize at eration system is 145.25 kW. Simultaneously, Figure 19
300 kW and 200 kVAR. The connected load remains con- illustrates the DC link voltage of the DC bus before the
stant at 500 kW and 400 kVAR. Additionally, the environ- DC–AC converter. The efficiency of DC link voltage track-
mental inputs for the PV system include an irradiance of ing is compared between the proposed algorithm and previ-
1000 W/m2 and a temperature of 35 °C, while the wind ous algorithms in Figures 20 and 21, specifically during the
speed input for the wind generation system is set at time intervals 0–1 s and 3–6 s. To numerically quantify the
12 m/s. The comparison in tracking efficiency of the micro- comparison, Table 3 provides metrics such as rise time,
grid, employing optimized Proportional-Integral (PI) con- peak overshoot, settling time, and steady-state error for
troller gains through Enhanced Randomized Harris Hawk DC bus voltage, active power, and reactive power. Table 4
Optimization (ERHHO) and other algorithms (MCABC, presents performance indices, including Integral Absolute
TLBO-EO, SMO, ARA, CB, and ZN), is depicted in Error (IAE), Integral Square Error (ISE), Integral Time
Figure 12. For a more detailed examination, Figures 13 Absolute Error (ITAE), and Integral Time Squared Error
and 14 offer better visualization of the injected active and (ITSE). IAE is the cumulative absolute error over time,
reactive powers by the microgrid between 0 to 1 s and and it signifies the response speed and oscillations during
11.6 to 13 s, respectively. Figure 16 provides the active stability conditions. IAE is 0.0204, 0.0176 and 0.0179 for
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 15

Table 2. Optimized gains of PI controllers using proposed and various algorithms.

Algorithm Kp1, Ki1 Kp2, Ki2 Kp3, Ki3 Kp4, Ki4 Kp5, Ki5 Kp6, Ki6 Kp7, Ki7 Kp8, Ki8
Proposed 0.3754, 1.4758, 0.1145, 0.8872, 0.0745, 0.0745, 0.4210, 0.4210,
ERHHO 5.162 3.8540 0.6375 1.8776 0.9547 0.9547 1.5247 1.5247
MCABC 0.1454, 1.2074, 0.5470, 0.6345, 0.0154, 0.0154, 0.9055, 0.9055,
7.3152 4.9654 0.7459 2.0457 1.4716 1.4716 1.8749 1.8749
TLBO-EO 0.1198, 1.1925, 0.5578, 0.5973, 0.0073, 0.0073, 1.0061, 1.0061,
6.9450 4.7599 0.7041 2.2761 1.5473 1.5473 2.0965 2.0965
SMO 0.1588, 1.0875, 0.5696, 0.5579, 0.0204, 0.0204, 1.0064, 1.0064,
6.9844 4.0014 0.7542 2.3498 1.4918 1.4918 2.2876 2.2876
ARA 0.1746, 1.2252, 0.6074, 0.6928, 0.0319, 0.0319, 0.8547, 0.8547,
6.2214 3.9475 0.8874 2.7519 1.4473 1.4473 2.8936 2.8936
CB 0.1225, 1.3024, 0.6457, 0.7018, 0.0337, 0.0337, 0.8871, 0.8871,
6.8345 4.1745 0.8832 2.0457 1.4750 1.4750 2.9324 2.9324
Ziegler Nicholas 1.3652, 2.574, 1.085, 1.1850, 0.1574, 0.1574, 1.5479, 1.5479,
10.6547 8.1475 3.745 11.195 5.1275 5.1275 5.7480 5.7480

Figure 12. Active and reactive power provided by PV-WIND and battery system.

DC voltage regulation, active and reactive power respec- aim to offer a comprehensive understanding of the micro-
tively. Compared to its nearest optimization algorithm, grid’s performance under varying conditions and the effi-
the proposed algorithm achieved 40–60% of less IAE indi- cacy of the ERHHO algorithm in optimizing PI controller
cating improved performance. ISE quantifies the cumula- gains (Figure 15).
tive squared error over time, and it provides a balance Figures 13 and 14 provide clearer visualizations of
between overshoot and settling time. ISE is 0.0109, 0.0113 injected active and reactive powers during critical time
and 0.0111 for DC voltage regulation, active and reactive intervals. ERHHO ensures more precise and stable power
power respectively. ITSE integrates the absolute error mul- injection, contributing to enhanced overall microgrid
tiplied by time over time and this value defines the Balance performance. In Figure 16, ERHHO optimally manages
between response speed and error magnitude. ITSE is the active and reactive power supplied to the load by the
0.0042, 0.002, and 0.0024 for DC voltage regulation, active PV-wind-battery system, as well as the interaction with
and reactive power respectively. ITAE integrates the abso- the grid. Notably, between 8 and 12 s, ERHHO effectively
lute error multiplied by time and it emphasizes minimizing regulates excess power, demonstrating improved control
error during transient response. ITAE is 0.000092, 0.00008 over power flow and grid interaction compared to other
and 0.000081 for DC voltage regulation, active and reactive algorithms. Figures 19, 20, and 21 showcase the superior
power respectively. This detailed analysis and comparison performance of ERHHO in maintaining stable DC link
16 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 13. Active and reactive power provided by PV-WIND and battery system from 0 to 1 s.

Figure 14. Active and reactive power provided by PV-WIND and battery system from 11.6 to 13 s.

voltage. Particularly during critical time intervals (0–1 s existing optimization algorithms. It provides enhanced
and 3–6 s), ERHHO ensures precise tracking, minimizing tracking efficiency, more stable power injection, improved
deviations and outperforming alternative algorithms. regulation of load power supply and grid interaction, and
Table 3 provides numerical metrics, including rise time, superior stability in DC link voltage. The quantitative
peak overshoot, settling time, and steady-state error for comparisons further validate ERHHO’s effectiveness in
DC bus voltage, active power, and reactive power. ERHHO optimizing the microgrid’s control system for various
consistently yields better results across these metrics operational scenarios.
compared to the other optimization algorithms. Table 4 Case 2: In this case, the effectiveness of optimized
highlights the performance indices, including Integral Proportional-Integral (PI) gains in ensuring continuous
Absolute Error (IAE), Integral Square Error (ISE), Integral power supply to the load is investigated by considering
Time Absolute Error (ITAE), and Integral Time Squared alterations in irradiance and wind speed. The changes in
Error (ITSE). ERHHO excels in minimizing these perfor- irradiance are detailed as follows: from 0 to 3.5 s, it is
mance indices, indicating superior optimization in terms 400 W/m2; from 3.5–7 s, it increases to 600 W/m2; from
of system response and stability. The proposed ERHHO 7 to 10.5 s, it further rises to 800 W/m2; from 10.5 to
algorithm demonstrates notable advancements in optimiz- 14 s, it peaks at 1000 W/m2; from 14 to 17.5 s, it decreases
ing PI controller gains for the microgrid compared to to 750 W/m2, and from 17.5 s onward, it stabilizes at
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 17

Figure 15. Active and reactive power provided by PV-WIND and battery system with proposed ERHHO.

Figure 16. Load active and reactive power. Figure 17. Grid active and reactive power.

550 W/m2. Simultaneously, the wind speed experiences 600 W/m2, again increased to 120.3 kW when irradiance is
variations: from 0 to 3 s, it is 9 m/s; from 3 to 8 s, it 800 W/m2, and 143.6 kW when irradiance is 1000 W/m2.
increases to 11 m/s; from 8 to 12 s, it further rises to The wind generation system generates the powers of
12 m/s; from 12 to 17 s, it peaks at 13 m/s, and from 109.1 kW, 133.3 kW, 145.4 kW and 157.52 kW when wind
17 s onward, it decreases back to 9 m/s. this variation of speed is 9 m/s, 11 m/s, 12 m/s and 13 m/s respectively.
irradiance and wind speed is presented in Figure 22. Figure 24 provides a comprehensive overview of the refer-
Figure 23 illustrates the corresponding changes in generated ence and actual active and reactive powers injected by
power from the Photovoltaic (PV) and wind sources. For the PV-wind-battery system. Reference active and reactive
irradiance of 400 W/m2 generated power from PV is powers are reduced to 100 kW and 50 kVAR between
59.84 kW, it increased to 90.27 kW when irradiance is 12 and 16 s. Consequently, the injected powers by the
18 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 20. DC bus voltage comparison from 0 to 1 s.

Figure 18. PV and wind generated power.

Figure 19. DC bus voltage.

converter are adjusted to match these reference values.

This reduction in injected power by the converter is effec-
tively managed by the grid, ensuring the maintenance of Figure 21. DC link voltage comparison from 3.9 to 6 s.
constant load active and reactive powers, as depicted in
Figure 25. The active and reactive power profiles of the grid
are showcased in Figure 26. During the period between 3 to 7 s, and second, from 8 to 12 s, as visually represented
12 and 16 s, the surplus power generated by the PV and in Figure 29. Throughout these periods, the remaining
wind sources that are not directed to the grid is efficiently power essential for the DC to AC converter to follow refer-
stored in the battery, as evidenced by Figure 27. As the ence values is supplemented by the battery system, as illus-
battery undergoes charging, the State of Charge (SOC) trated in Figure 30. The initial SOC of the battery is chosen
experiences an increase during this timeframe. Figure 28 as 50% and as the battery is in discharging mode SOC
presents the behaviour of the DC bus voltage throughout starts decreasing from 50%. This approach allows for a
the analysed duration. The detailed examination of these dynamic management of power sources, ensuring that the
figures underscores the intricate interplay between varia- DC to AC-converter receives the necessary power support
tions in irradiance and wind speed, the response of the from the battery during the isolated periods of PV and wind
PV-wind-battery system, and the adaptability of the grid generation.
and battery in maintaining a stable and continuous power The proposed algorithm demonstrates significant
supply to the load. improvements over several other optimization algorithms,
Case 3. In this case, PV and wind generation system including MCABC, TLBO-EO, SMO, ARA, CB, and ZN.
powers are reduced to zero by disconnecting them from The microgrid, equipped with the ERHHO-optimized Pro-
the HDGS to check the efficacy of the proposed optimiza- portional-Integral (PI) controller gains, exhibits superior
tion in providing continuous supply to the load. The PV accuracy in following dynamic changes in active and reac-
and wind generation are disconnected from the overall tive power references compared to the alternative optimiza-
power system during two distinct time intervals: first, from tion algorithms.
Table 3. Comparison of metrics between proposed algorithm and previous algorithms from the literature.
Algorithm DC bus voltage Active power Reactive power
Rise Overshoot Settling Steady state Rise Overshoot Settling Steady state Rise Overshoot Settling Steady state
time(s) (%) time(s) error (%) time(s) (%) time(s) error (%) time(s) (%) time(s) error (%)
Proposed 0.0254 1.0248 0.0408 0.3389 0.0199 1.6702 0.0327 0.1587 0.0199 1.6860 0.0325 0.0355
ERHHO

The Author(s): Science and Technology for Energy Transition 79, 45 (2024)
MCABC 0.0406 5.0580 0.1197 0.6778 0.0406 4.7707 0.1174 0.1679 0.0381 4.0156 0.1066 0.0401
TLBO-EO 0.0425 5.9083 0.1311 0.7117 0.0363 3.0116 0.0942 0.1687 0.0291 7.9938 0.0932 0.0452
SMO 0.0355 11.4332 0.1235 0.6947 0.0638 0.8024 0.1024 0.1696 0.0417 2.1603 0.0964 0.0413
ARA 0.0423 4.7517 0.1277 0.7625 0.0294 8.0050 0.0941 0.1669 0.0281 9.6522 0.0928 0.0479
CB 0.0456 8.7447 0.1482 0.8133 0.0373 4.7931 0.1075 0.1770 0.0334 8.5389 0.1068 0.0485
Ziegler 0.0607 39.9643 0.7925 0.8756 0.1147 31.1323 0.9171 0.1801 0.1148 31.1441 0.9193 0.0623
Nicholas

Table 4. Comparison of performance indices between proposed algorithm and previous algorithms from literature.
Algorithm DC bus voltage Active power Reactive power
IAE ISE ITAE ITSE IAE ISE ITAE ITSE IAE ISE ITAE ITSE
Proposed ERHHO 0.0204 0.0109 0.0042 0.000092 0.0176 0.0113 0.002 0.00008 0.0179 0.0111 0.0024 0.000081
MCABC 0.0363 0.0181 0.0086 0.00028 0.0347 0.0224 0.0026 0.00032 0.0326 0.0209 0.0028 0.00028
TLBO-EO 0.0387 0.0192 0.0092 0.00032 0.03 0.0195 0.0024 0.00024 0.0286 0.0178 0.0027 0.00020
SMO 0.0386 0.0183 0.0091 0.0003 0.0452 0.0296 0.0031 0.00057 0.0332 0.0215 0.0029 0.00029
ARA 0.0381 0.0188 0.0097 0.0003 0.0285 0.0179 0.0024 0.00020 0.0287 0.0176 0.0028 0.00020
CB 0.0450 0.0219 0.0107 0.00042 0.0321 0.0208 0.0025 0.00027 0.0324 0.0202 0.0029 0.00026
Ziegler Nicholas 0.1174 0.0479 0.0259 0.0036 0.1771 0.0903 0.0385 0.0080 0.1771 0.0903 0.0386 0.0085

19
20 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Figure 22. Irradiance and wind speed change in PV and wind

generation system. Figure 25. Load active and reactive power.

Figure 23. PV and wind generated power.

Figure 26. Grid active and reactive power.

Figure 24. Active and reactive power provided by PV-WIND

and battery system with proposed ERHHO. Figure 27. Battery terminal power and SOC.
 The Author(s): Science and Technology for Energy Transition 79, 45 (2024) 21

such as photovoltaic (PV) and wind generation. The

utilization of a power flow control strategy and the propor-
tional-integral (PI) controller within the microgrid frame-
work is crucial for ensuring optimal performance under
varying operating conditions. However, tuning the PI con-
troller parameters poses a challenge due to the dynamic
and nonlinear nature of renewable energy sources. To
address this challenge, proposed the application of the
Enhance Randomized Harris Hawk Optimization
(ERHHO) algorithm for fine-tuning the PI controller.
Utilizing the ERHHO algorithm’s ability to mimic the
hunting behaviour of hawks in finding optimal solutions,
demonstrated its efficacy in navigating the complex solu-
tion space and converging to optimal PI controller settings.
This optimization contributes to the microgrid’s dynamic
Figure 28. DC bus voltage. behaviour, particularly in the context of diverse renewable
energy sources. The comprehensive simulations conducted
in this study evaluate the effectiveness of the proposed
Enhance Randomized Harris Hawk Optimization-based
PI controller tuning. Performance metrics such as transient
response, overshoot, settling time, and steady-state error
are analysed to validate the robustness and efficiency of
the proposed method. Compared to its nearest optimization
algorithm MCABC, with the proposed algorithm rise time
is reduced by 50%, overshoot is reduced by 75%, settling
time is reduced by 66%, and finally, the percentage of
reduction of steady-state error is 45%. Rise time is 0.0254,
0.0199, and 0.0199 s for DC voltage, active and reactive
power respectively. Steady-state error is 0.3389, 0.1587,
and 0.0355 for DC voltage, active and reactive power.
The results of this research mark a significant advancement
in microgrid control strategies, offering a promising avenue
for PI controller tuning in the presence of various renewable
energy sources. The integration of the Enhance Random-
Figure 29. PV and wind generated power.
ized Harris Hawk Optimization algorithm proves to be a
valuable tool in enhancing the operational efficiency and
reliability of microgrids. This research contributes to the
broader goal of establishing sustainable and resilient energy
systems, crucial for addressing the evolving energy land-
scapes and growing demand for reliable and environmen-
tally friendly energy solutions.

Declaration of generative AI and AI-assisted

technologies in the writing process
During the preparation of this work, the author(s) used
ChatGPT in order to proofread the manuscript. After using
this tool/service, the author(s) reviewed and edited the con-
tent as needed and take(s) full responsibility for the content
of the publication.

References
Figure 30. Battery terminal power and SOC.
1 Sen S., Kumar V. (2018) Microgrid control: A comprehensive
survey, Annu. Rev. Control 45, 118–151.
8 Conclusion 2 Chaudhary G., et al. (2021) Review of energy storage and
energy management system control strategies in microgrids,
This paper presents a novel and effective approach to Energies 14, 16, 4929.
enhance the control strategies of microgrids, specifically 3 Muqeet H.A., et al. (2022) Sustainable solutions for
focusing on the integration of renewable energy sources advanced energy management system of campus microgrids:
22 The Author(s): Science and Technology for Energy Transition 79, 45 (2024)

Model opportunities and future challenges, Sensors 22, 6, 20 Chun Y., Hua X. (2023) Improved sine cosine algorithm for
2345. optimization problems based on self-adaptive weight and
4 Hossain Md.A., et al. (2019) Evolution of microgrids with social strategy, IEEE Access.
converter-interfaced generations: Challenges and opportuni- 21 Alsoul M., et al. (2023) A new efficient hybrid approach for
ties, Int. J. Electr. Power Energy Syst. 109, 160–186. machine learning-based firefly optimization, Iraqi J. Sci.
5 Haseeb J., et al. (2021) Optimal energy management of a 4600–4612.
campus microgrid considering financial and economic anal- 22 Patel C., Malakar T., Sreejith S. (2023) Assessment of
ysis with demand response strategies, Energies 2021, 14, converter performance in hybrid AC–DC power system
8501. under optimal power flow with minimum number of DC
6 Bin L., et al. (2022) Scheduling and sizing of campus link control variables, Energies 16, 15, 5800.
microgrid considering demand response and economic anal- 23 Thiruvenkadam S., Kim H.-J., Ra I.-H. (2020) Hybrid fuzzy
ysis, Sensors 22, 16, 6150. and flower pollination optimization algorithm for optimal
7 Elnady A., et al. (2022) A comprehensive review of centralized dispatch of generating units in the existence of electric
current/power control schemes for parallel inverters and AC vehicles, in2020 6th International Conference on Advanced
microgrids, IEEE Access 10, 125061–125085. Computing and Communication Systems (ICACCS), IEEE.
8 Kim H., et al. (2019) Direct energy trading of microgrids in 24 Alweshah M., et al. (2020) Flower pollination algorithm for
distribution energy market, IEEE Trans. Power Syst. 35, 1, solving classification problems, Int. J. Adv. Soft Comput.
639–651. Appl. 12, 1, 15–34.
9 Farzin H., et al. (2017) A market mechanism to quantify 25 Singh N., Kaur J. (2021) Hybridizing sine–cosine algorithm
emergency energy transactions value in a multi-microgrid with harmony search strategy for optimization design
system, IEEE Trans. Sustain. Energy 10, 1, 426–437. problems, Soft Comput. 25, 16, 11053–11075.
10 Jumani T.A., et al. (2019) Optimal power flow controller for 26 Dashtaki A.A., et al. (2023) Optimal management algorithm
grid-connected microgrids using grasshopper optimization of microgrid connected to the distribution network consid-
algorithm, Electronics 8, 1, 111. ering renewable energy system uncertainties, Int. J. Electr.
11 Jumani T.A., et al. (2020) Swarm intelligence-based optimiza- Power Energy Syst. 145, 108633.
tion techniques for dynamic response and power quality 27 Fahim K.E., et al. (2023) A state-of-the-art review on
enhancement of AC microgrids: A comprehensive review, optimization methods and techniques for economic load
IEEE Access 8, 75986–76001. dispatch with photovoltaic systems: Progress, challenges,
12 Vishal V., et al. (2014) Online PI controller tuning for a and recommendations, Sustainability 15, 15, 11837.
nonlinear plant using genetic algorithm. In: 2014 Innovative 28 Wang Y., et al. (2023) Adaptive chimp optimization
Applications of Computational Intelligence on Power, Energy algorithm with chaotic map for global numerical optimiza-
and Controls with their impact on Humanity (CIPECH), tion problems, J. Supercomput. 79, 6, 6507–6537.
IEEE. 29 Lawley S.D. (2023) Extreme statistics of superdiffusive Lévy
13 Abdolrasol Maher G.M., et al. (2022) Optimal PI controller flights and every other Lévy subordinate Brownian motion,
based PSO optimization for PV inverter using SPWM J. Nonlinear Sci. 33, 4, 53.
techniques, Energy Rep. 8, 1003–1011. 30 Zamee M.A., Han D., Won D. (2023) Integrated grid
14 Ramasamy V., et al. (2023) A comprehensive review on forming-grid following inverter fractional order controller
Advanced Process Control of cement kiln process with the based on Monte Carlo artificial Bee Colony optimization,
focus on MPC tuning strategies, J. Process Control 121, 85– Energy Rep. 9, 57–72.
102. 31 Sayed E.A., et al. (2023) Enhancement of PV performance
15 Jia L.Y., et al. (2023) A weighted-sum chaotic sparrow by using hybrid TLBO-EO optimization, Ain Shams Eng. J.
search algorithm for interdisciplinary feature selection and 14, 3, 101892.
data classification, Sci. Rep. 13, 1, 14061. 32 Abid S., et al. (2023) Development of slime mold optimizer
16 Al-Betar M.A., et al. (2023) Economic load dispatch using with application for tuning cascaded PD-PI controller to
memetic sine cosine algorithm, J. Ambient Intell. Human. enhance frequency stability in power systems, Mathematics
Comput. 14, 9, 11685–11713. 11, 8, 1796.
17 Gupta S., Rong S. (2023) Diversity-enhanced modified sine 33 El-Sehiemy R., et al. (2023) Proportional-integral-derivative
cosine algorithm and its application in solving engineering controller based-artificial rabbits algorithm for load frequency
design problems, J. Comput. Sci. 72, 102105. control in multi-area power systems, Fractal Fract. 7, 1, 97.
18 Pham V.H.S., Nguyen Dang N.T., Nguyen V.N. (2023) 34 Hussien A.M., et al. (2022) Coot bird algorithms-based
Hybrid sine cosine algorithm with integrated roulette wheel tuning PI controller for optimal microgrid autonomous
selection and opposition-based learning for engineering operation, IEEE Access 10, 6442–6458.
optimization problems, Int. J. Comput. Intell. Syst. 16, 1, 35 Huba M., et al. (2021) Making the PI and PID controller
171. tuning inspired by Ziegler and Nichols precise and reliable,
19 Mousa A.A., El-Shorbagy M.A., Farag M.A. (2020) Steady- Sensors 21, 18, 6157.
state sine cosine genetic algorithm based chaotic search for 36 Song M., et al. (2022) Modified harris hawks optimization
nonlinear programming and engineering applications, IEEE algorithm with exploration factor and random walk strategy,
Access 8, 212036–212054. Comput. Intell. Neurosci.