Solar Energy 269 (2024) 112361

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Design and HIL validation of improved prairie dog optimization based

dynamic unitary reconfiguration for partially shaded PV arrays
Bo Yang a, Yuanweiji Hu a, Haoyin Ye b, Jie Zhang c, Xianlong Cheng c, Zilin Li d, Yaxing Ren e,
Yunfeng Yan f, *
a
Faculty of Electric Power Engineering, Kunming University of Science and Technology, 650500 Kunming, China
b
State Grid Lanxi Power Supply Company, 321100 Jinhua, China
c
China Southern Power Grid Honghe Power Supply Bureau, 661100 Mengzi, China
d
Department of Electrical Engineering, Hong Kong Polytechnic University, Kowloon 999077, Hong Kong SAR, China
e
School of Engineering, University of Lincoln, LN6 7TS Lincoln, United Kingdom
f
College of Electrical Engineering, Zhejiang University, 310027 Hangzhou, China

A R T I C L E I N F O A B S T R A C T

Keywords: This work designs an advanced dynamic unitary PV array reconfiguration scheme based on a novel improved
Photovoltaic array prairie dog optimization (IPDO) to solve the power loss issue due to partial shading of PV arrays. This method
Dynamic unitary reconfiguration takes the PV cell as the smallest entity of reconfiguration, and is suitable for large-scale PV arrays. To avoid
Improved prairie dog optimization
invalid and redundant relocation of PV modules, an innovative self-rectify strategy is employed and embedded in
Partial shading mitigation
Self-rectify strategy
the reconfiguration algorithm, which can reduce invalid actions of switching matrix by up to 80 %. Comparative
simulation tests among nine algorithms (IPDO, genetic algorithm, particle swarm optimization, simulated
annealing, four-square sudoku, arithmetic sequence pattern, modified harris hawks optimizer, artificial
ecosystem-based optimization, Runge Kutta optimizer) under eight typical shading types verify the superiority of
IPDO from four metrics, i.e., power loss, power enhancement, fill factor and execution ratio. Particularly, IPDO
reduces power loss by 7.33 % to 30.63 %, enhances the harvested power by 6.1 % to 31.59 %, enhances fill factor
(FF) by 0.579 to 0.773, and enhances execution ratio (ER) by 60.37 % to 92.67 % for asymmetric (20 × 30) PV
array. For symmetric (30 × 30) PV array, IPDO reduces power loss by 9.01 % to 27.86 %, enhances the harvested
power by 8.33 % to 43.69 %, enhances FF by 0.602 to 0.759, and enhances ER by 72.14 % to 90.99 %. Moreover,
a new switching matrix with double-pole multi-throw switches is designed for the required switches number
reduction. Lastly, hardware-in-the-loop (HIL) experiments based on RTLAB are carried out which validate the
implementation feasibility of the proposed approach.

study on it [8–9].
In practice, large-scale PV arrays are usually affected by partial
1. Introduction shading condition (PSC), which is resulted from surrounding buildings,
cloud shadows, dusts, etc. PSC can cause many troubles for PV systems,
With the ever-growing deterioration of ecological environment and such as reduced power generation efficiency, hotspot effect that may
the increasing demands for pollution-free living environment, clean damage the structure of PV cells and multiple-peaks phenomenon to P-V
energy sources such as solar energy [1], wind energy [2–3], tidal energy and I-V curves [10]. Furthermore, it makes maximum power point
[4], and thermoelectric energy [5–6] have become alternatives to tracking (MPPT) algorithm more difficult to track the optimal operation
traditional fossil fuels due to their environmentally friendly advantages point of unsmooth P-V characteristics.
and huge reserves [7]. Among them, photovoltaic (PV) power genera­ To mitigate the multimodality issue caused by PSC, PV array
tion has become a vital approach to obtain renewable energy sources. In reconfiguration techniques are developed, which have three advantages
the context of carbon neutral and net zero strategy, PV power industry compared to other methods: firstly, PV array reconfiguration can
will face new development opportunities. Therefore, PV power genera­ improve and smooth the output characteristics, which is beneficial to
tion has attracted a large number of researchers to conduct in depth

* Corresponding author.
E-mail address: [email protected] (Y. Yan).

https://doi.org/10.1016/j.solener.2024.112361
Received 21 December 2022; Received in revised form 6 January 2024; Accepted 17 January 2024
Available online 26 January 2024
0038-092X/© 2024 International Solar Energy Society. Published by Elsevier Ltd. All rights reserved.
B. Yang et al. Solar Energy 269 (2024) 112361

Nomenclature σ a random value between 0 and 1

Variables Abbreviations
DPV dimension of PV array ASP arithmetic sequence pattern
Dunit dimension of PV unit AEO artificial ecosystem-based optimization
ER execution ratio AI artificial intelligence
FF fill factor FSS four-square sudoku
Io output voltage of PV array, A GA genetic algorithm
Ipn current of PV module in the pth row and nth column, A HIL hardware-in-the-loop
Nswitch number of required switches IPDO improved prairie dog optimization
Nunit number of PV units in a PV array LN long-narrow
Pen (%) power enhancement LW long-wide
Ploss (%) power loss MHHO modified harris hawks optimizer
Vo output voltage of PV array, V MPPT maximum power point tracking
V om row voltage of the mth row, V PSC partial shading condition
PSO particle swarm optimization
IPDO parameters PV photovoltaic
F objective function of IPDO SA simulated annealing
itermax the maximum number of iterations SN short-narrow
LB lower bound of solution in IPDO SP series–parallel
UB upper bound of solution in IPDO SRS self-rectify strategy
xrandom a random solution of IPDO STD standard deviation
ρ inherent parameter of IPDO SW short-wide
ε individual difference between various solutions TCT total-crossed-tied

MPPT algorithm; secondly, it can increase the output power at the

maximum power point under PSC; furthermore, PV array reconfigura­
tion has favorable universal applicability that can be applied to any size
of PV arrays. In addition, distributed photovoltaic power stations
broaden the application scenarios of such framework and reduce overall
implementation cost. Therefore, this work focuses on the algorithm
design of PV array reconfiguration.
Previous studies have classified PV array reconfiguration techniques
into static and dynamic methods. Pillai et al. proposed a column index
static method that obtains good results in shadow dispersion and reduce
the power loss on wires, but was deficient in the flexibility of algorithm
[11]. Sugumar et al. proposed a two-step staircase static reconfiguration
method and applied it to (5 × 5) photovoltaic arrays [12]. Madhusu­
danan et al. designed a magic sudoku method that simplified MPPT
process since maximum power point always appears on the rightmost
zone of P-V characteristic [13]. However, the magic sudoku method can
only provide a limited and fixed topology scheme, as well as other
sudoku-based variant methods, so they lack of the capability of real time
processing [14–16]. Mutual shadows due to the mutual shelter of
adjacent PV panels, are considered and solved in reference [17–19].
Loshu scheme which is inspired by an ancient Chinese venation pattern
named ‘Luo Shu’, was applied to PV array reconfiguration [20]. This
method has the significant advantage of extreme simple execution steps,
but it often suffers from the same drawback as sudoku-based ap­
proaches. Ram et al. innovatively employed a chaotic baker map tech­
nique to randomly relocate the shaded PV panels [21]. The skyscraper
scheme used in a (6 × 6) array acquired lower cable loss than Sudoku
[22]. Static reconfiguration methods have common advantages such as
low implementation cost and simple structure. However, the static
methods only shift the physical position of PV panels that invariably
increase wire losses.
Fig. 1. TCT connected (r × c) PV array.
In the research of energy systems, artificial intelligence algorithms
are widely applied, and Su et al. optimized the battery charging strategy

2
B. Yang et al. Solar Energy 269 (2024) 112361

However, most of the studies mentioned above focus on (9 × 9) PV

arrays, as well as some smaller PV arrays, such as (4 × 2) PV arrays
reconfiguration for rooftops [34]. In addition, a switching matrix is used
to divide (8 × 8) PV arrays into (4 × 4) matrices to increase the
maximum output power [35], and a L-shaped propagated array
configuration method with a new dynamic reconfiguration algorithm
have proposed for (4 × 4) PV arrays [36]. A few works provide solutions
for large-scale PV array reconfiguration on account of the enormous
algorithmic complexity. As a prominent method of renewable energy
generation technology, PV power generation has been increasing glob­
ally year by year. With more installed capacity, the array of PV power
generation becomes larger, which necessitates the study of large-scale
PV array reconfiguration to enhance the efficiency of PV power gener­
ation. Prairie dog optimization (PDO) is a novel nature-inspired meta­
heuristic proposed in 2022, which owns favorable balance between
global exploration and local exploitation [37]. To address this thorny
problem, this work proposes an innovative dynamic unitary reconfigu­
ration technique that focuses on large scale PV arrays. The main con­
tributions and novelties of this work can be clarified as follows:

• A novel dynamic unitary PV array reconfiguration method based on

improved prairie dog optimization (IPDO) is designed to deal with
the power loss of large-scale PV arrays under PSC;
• Two case studies, e.g., asymmetric (20 × 30) and symmetric (30 ×
30) PV arrays under eight typical shading types, are tested with nine
typical algorithms, i.e., IPDO, GA, PSO, simulated annealing (SA)
[38]), four-square sudoku (FSS) [39], arithmetic sequence pattern
(ACP) [40], MHHO, artificial ecosystem-based optimization (AEO)
[41] and Runge Kutta optimizer (Run) [42] to verify the effective­
ness of IPDO;
• Power loss, power enhancement, fill factor [11]), and execution ratio
(ER) [39], are used to quantitatively evaluate various algorithms and
the output efficiency of PV system;
• Information exchange and self-rectify strategies are introduced to
improve the exploitation ability of IPDO and to reduce the invalid
and redundant movements of PV module during reconfiguration
process;
• In order to simplify the switching matrix structure and decrease the
number of switches, a new switching matrix with double-pole multi-
throw switches is designed, which reduces the switching cost of
reconfiguration;
• To further validate the practical operational feasibility of unitary PV
Fig. 2. Schematic diagram of PV unit (example of a (6 × 6) PV array). array reconfiguration method, RTLAB based hardware-in-the-loop
(HIL) experiments are carried out.
using artificial neural network method, and evaluated the battery health
The structure of this work is organized as follows: the modeling of
using deep learning [23,24]. In the study of PV array reconfiguration,
total-crossed-tied (TCT) connected PV array and unitary reconfiguration
dynamic PV array reconfiguration methods based on artificial intelli­
is given in Section 2. Section 2 presents objective function and evalua­
gence (AI) algorithms that only alter electrical connections of PV panels
tion metrics. Then, mechanism of IPDO is elaborated in Section 3.
by a switching matrix have also been largely investigated, such as ge­
Moreover, comprehensive case studies are executed in Section 4 while
netic algorithm (GA) [25], particle swarm optimization (PSO) [26],
thorough discussion and analysis are provided in Section 5. Besides, HIL
modified harris hawks optimizer (MHHO) [27], democratic political
experiment is presented in Section 6. Lastly, Section 7 concludes the
algorithm (DPA) [28], multi-objective grey wolf optimizer (MOGWO)
whole work.
[29], fuzzy logic (FL) [30], machine learning approach (MLA) [31],
jellyfish search algorithm(JSA) [32], and a combination of algorithms
2. Problem formulation for PV array reconfiguration
called dynamic leader-based collective intelligence is adopted [33].
Mirjalili et al. simultaneously optimizes two objectives of maximizing
Series-parallel (SP), bridge-linked and TCT are the most common and
output power and minimizing current difference between adjacent PV
widespread basic configurations of PV array. When illumination con­
rows by MOGWO [29]. Yousri et al have performed simulation tests
dition is ideal, the power output of PV array with all three connections
which show that MHHO obtains better overall performance than that of
are the same. However, the output characteristic of SP is most sensitive
HHO, PSO, GA. In terms of AI algorithms, the effect of AI algorithms on
to irradiance distribution i.e., power loss of SP connected PV array is the
power enhancement is relatively inferior.

3
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 3. Overall implementation structure based on the proposed switching matrix.

∑
c
Table 1 Io = Ipn , pisconstant and p ∈ [1, r] (2)
Comparison of required number of switches using different switching matrix n=1
configurations.
PV array TCT Switching Switching Proposed
where Vom is the row voltage of the mth row; Ipn is the generated current
size matrix [45] matrix [47] switching matrix of PV module in the pth row and nth column; r and c represent the row
and column dimensions of PV array, respectively.
1,908 299 200 90
35,218 1,809 1,200 580
53,128 2,699 1,800 870 2.1.2. Modeling of unitized PV array
A unitized PV array is firstly designed in this study which aims to
address the challenges of large-scale PV array reconfiguration: the
highest under PSC. Comparative tests of power loss of three PV array
complexity of algorithm increases rapidly due to the large number of PV
configurations under PSC was conducted by Nayak et al., and the results
modules and its cost increases significantly due to the use of complicated
indicate that PSC can seriously affect the power derived from PV systems
switching matrix. The entire PV array is divided into PV units in which
[43]. Meanwhile, under each test condition, PV array in TCT connection
one unit contains multiple PV modules. PV units are treated as the
has less power loss and better output characteristics. Therefore, this
smallest units for reconfiguration. In addition, the switching matrix is
work carries out a study on PV array reconfiguration based on TCT
modified to alter the topology of PV array faster and to reduce the
structure.
number of switches, such that the unitized PV array can be established.
To clearly explain the proposed unitary reconfiguration method, a (30 ×
2.1. Modeling of unitized PV array 30) PV array is taken as an example.
An appropriate size of PV units is determined. To balance the
2.1.1. Modeling of TCT connected PV array complexity of algorithm and the size of feasible region of solutions, such
PV array consists of PV cells, whose detailed model can be referred to (30 × 30) PV can be divided into (10 × 10) units, i.e., the size of PV unit
previous work [44]. The output voltage of PV array is superposed by is (3 × 3), which yields:
voltages of r rows, and the output current is the summation of currents of
c branches. The structure of TCT connected PV array is depicted in Fig. 1. Nunit =
DPV
(3)
Here, output current Io and voltage Vo of PV array can be described as Dunit
below:
where DPV and Dunit represent the dimension of PV array and PV unit,
∑
r
respectively; Nunit denotes the number of PV units.
Vo = Vom (1)
m=1
Note that PV modules in a PV unit undergo the same topology
alteration. Moreover, the schematic diagram of PV units is clearly shown

4
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 4. Procedure schematic diagram of SRS.

Fig. 5. Flowchart of IPDO based unitary PV reconfiguration.

5
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 6. Partial shading patterns under asymmetric (2030) PV array.

in Fig. 2. the central control unit. Based on those data, central control unit deduce
Fig. 3 gives the overall implementation process of unitary PV array shading distribution of PV array and provides optimal configuration via
reconfiguration based on IPDO. Information monitoring system moni­ IPDO. According the control signal, switch matrix controls the closure
tors and collects environment and electrical variables such as tempera­ and disconnection of each switch. So that the unitary reconfiguration is
ture, irradiance, current and voltage in real time and transmits them to accomplished. The switching matrix is the hardware infrastructure of

6
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 7. Partial shading patterns under symmetric (3030) PV array.

7
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 8. Irradiance distribution of asymmetric PV array reconfigured by IPDO.

dynamic reconfiguration. In previously published paper [45], the throw switches so that the complexity and number of switches can be
switching matrix is composed of double-pole multi-throw switches and reduced. In addition, the proposed switching matrix is depicted in Fig. 3.
double-pole multi-throw switches. On the basis of this, this work pro­ Note that double-pole switches can only be connected to adjacent switch
poses a modified switching matrix employing only double-pole multi- contacts. The required number of switches can be calculated by:

8
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 9. Characteristic curves of asymmetric PV array before and after reconfiguration by IPDO.

9
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 9. (continued).

10
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 10. The output characteristic curves of asymmetric PV array under SW shading before and after reconfiguration by nine algorithms.

Table 2
Output power obtained by nine algorithms under asymmetric (20 × 30) PV array.
Shadow TCT IPDO PSO GA SA

Pout Pmax Pavg STD Pmax Pavg STD Pmax Pavg STD (W) Pmax Pavg STD
(kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW)

SW 87.74 116.09 115.46 0.31 113.84 111.59 0.95 113.39 111.79 0.64 113.39 112.08 0.44
SN 113.84 125.09 125.09 0 124.64 122.43 0.94 124.64 122.90 0.61 122.84 121.35 0.51
LW 72.36 94.04 93.64 0.41 92.69 91.32 0.73 93.59 91.28 0.82 89.54 87.29 1.00
LN 110.69 118.34 118.34 0 118.34 117.84 0.20 118.34 118.34 0 118.34 116.97 0.14
Diagonal 110.69 117.44 117.44 0 117.44 115.07 1.26 117.44 114.16 1.20 116.09 114.26 0.79
Outer 88.82 107.99 107.38 0.49 107.54 105.76 0.64 106.19 104.73 0.72 104.39 102.04 1.04
Center 91.34 113.84 113.15 0.31 112.49 111.16 0.63 112.04 110.24 0.80 110.24 108.72 0.61
Random 100.88 116.99 116.57 0.27 115.19 113.34 0.89 116.09 114.09 0.72 114.74 115.37 0.44

Shadow FSS ACP MHHO AEO Run

Pout Pout (kW) Pmax Pavg STD Pmax Pavg STD Pmax Pavg STD
(kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW)

SW 104.84 109.34 112.04 110.28 0.93 112.94 111.61 0.45 113.84 113.15 0.49
SN 117.89 117.89 121.94 119.38 1.38 125.09 122.92 0.90 123.29 121.67 0.67
LW 79.19 91.34 88.64 85.16 1.57 90.89 88.75 0.88 91.34 90.47 0.60
LN 112.49 114.29 117.44 116.59 0.37 118.34 118.13 0.14 118.34 118.34 0
Diagonal 107.99 113.39 113.39 111.27 1.02 117.44 115.28 0.93 117.44 115.26 0.79
Outer 95.39 106.19 102.14 100.64 0.73 105.74 102.83 0.95 106.19 105.86 0.57
Center 103.94 111.14 111.14 110.03 0.57 112.49 111.87 0.26 111.14 110.53 0.43
Random 101.74 104.39 111.59 110.13 0.87 114.74 113.81 0.41 114.29 113.58 0.42

11
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 11. Irradiance distribution of symmetric (30 × 30) PV array reconfigured by IPDO.

12
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 12. Characteristic curves of symmetric (30 × 30) PV array before and after reconfiguration by IPDO.

13
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 12. (continued).

14
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 13. The output characteristic curves of symmetric PV array under diagonal shading before and after reconfiguration by nine algorithms.

Nswitch = r × (c − 1) (4) the number of switches and thus decreases the overall implementation
cost. The required number of switches for different sizes of PV arrays are
where r means the number of PV modules in a string; c denotes the tabulated in Table 1.
number of strings in a PV array.
On the other hand, the equipped switch number of normal TCT
configuration is r × (r +1) − 2 +2 × c × (r × c − r) [46], while the 2.2. Optimization objective
configuration is 2 × r × c [47], and 3 × r × c +c − r − 1 [45]. It can be
seen that the proposed modified switching matrix remarkably reduces The objective function of PV array unitary reconfiguration problem
is defined as

15
B. Yang et al. Solar Energy 269 (2024) 112361

Table 3
Output power obtained by nine algorithms of symmetric (30 × 30) PV array.
Shadow TCT IPDO PSO GA SA

Pout Pmax Pavg STD Pmax Pavg STD Pmax Pavg STD (W) Pmax Pavg STD
(kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW)

SW 112.45 161.99 161.58 0.83 159.96 158.36 1.79 159.96 158.02 1.14 156.91 155.43 1.33
SN 170.09 184.26 184.26 0 184.26 182.86 0.90 184.26 182.43 0.16 182.91 182.91 0
LW 120.77 155.24 154.30 0.45 152.54 151.63 1.52 153.21 151.82 1.08 151.19 150.23 0.94
LN 153.89 174.14 173.87 0.34 172.11 170.48 1.05 171.43 169.87 0.79 172.11 170.34 0.53
Diagonal 153.08 174.81 174.07 0.43 171.43 169.58 1.39 170.85 169.28 1.26 170.08 168.41 1.04
Outer 118.79 146.47 146.09 0.51 145.79 144.92 1.55 145.79 144.03 1.60 138.36 135.93 1.28
Center 147.14 178.19 178.19 0 176.83 175.24 1.29 175.48 174.02 0.92 174.81 172.17 0.84
Random 144.26 169.41 168.67 0.63 166.03 164.38 1.17 166.71 164.54 1.14 164.68 163.26 1.07

Shadow FSS ACP MHHO AEO Run

Pout Pout (kW) Pmax Pavg STD Pmax Pavg STD Pmax Pavg STD
(kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW) (kW)

SW 147.82 153.89 157.93 155.75 1.77 159.96 157.83 1.03 157.94 156.25 0.83
SN 179.54 178.86 184.26 183.28 1.08 184.26 184.26 0 184.26 183.82 0.50
LW 137.33 147.14 148.49 146.34 1.20 152.54 151.81 0.72 152.54 151.34 0.54
LN 164.01 167.39 172.11 170.93 0.83 172.11 171.04 0.58 172.11 170.73 0.55
Diagonal 160.64 159.29 168.06 166.51 1.42 170.76 169.15 0.80 170.08 168.67 0.94
Outer 123.52 134.31 136.34 133.82 1.60 140.39 138.54 0.86 143.77 141.21 1.19
Center 164.01 178.19 174.13 172.61 1.20 174.81 173.49 0.53 174.81 173.19 0.41
Random 152.54 147.81 161.31 159.46 1.23 165.36 163.58 0.68 164.68 162.67 0.60

F = max(Po ) = max(Io × Vo ) (5) foraging and burrowing tactics of prairie dogs [37]. Due to its promising
capabilities and robustness in engineering design problems, PDO is
where Po is the total output power of PV array, F is the fitness function adopted to address PV array reconfiguration.
for maximum optimization problem of PV array.
3.1.1. Initialization
PDO categorizes the overall solution domain into isometric zones
2.3. Evaluation metrics and each zone is allocated into a coterie, which contains numerous
searching agents. In other words, all agents are divided into clusters,
In order to accurately evaluate and quantitatively compare various while each cluster seeks for the optimal solution in parallel within its
algorithms, four typical criteria are used [48,49], as follows own scope. Therefore, the global exploration and local exploitation are
Pmax max
STC − Pre
executed repeatedly, which ensures that more searching directions can
Ploss (%) = × 100% (6) be realized. This comprehensively enhances the exploration capability.
Pmax
STC
Similar to other swarm intelligence algorithms, PDO adopts random
Pmax max
re − PPSC initialization of agents as follows:
Pen (%) = max
× 100% (7)
PPSC
Cti = σ × (UB − LB), i = 1, 2, ⋯, m (10)
Pmax
FF = PSC
(8) xj = σ × (ub − lb), j = 1, 2, ⋯, n (11)
VOC × ISC
{
ub = UB/m
Execution ratio (ER) =
Pmax
PSC
× 100% (9) (12)
lb = LB/m
Pmax
STC

where Pmax where Ct i is the ith coterie and m is the number of coteries; xj is the jth
STC and PPSC represent the maximum power point values under
max
agent (prairie dog) and n is the number of agents in one coterie; UB and
standard test conditions (1000 W/m2 and 25 ◦ C) and PSC, respectively;
LB represent the upper and lower bounds of solution domain, respec­
Pmax
re denotes the output power of shaded PV array after reconfiguration;
tively; and σ means a random value between 0 and 1. Moreover, m ≥ n,
VOC and ISC are the open circuit voltage and short circuit current of PV
respectively.
array. In addition, Ploss (%) is the pow execution ratio loss, which reflects
the influence of PSC on PV array; Pen (%) represents the power
3.1.2. Local exploitation
enhancement acquired by PV reconfiguration, which can be used as the
In PDO, the maximum number of iterations is divided equally into
principal criterion for the performance of the reconfiguration algorithm;
four periods to imitate various activities of prairie dogs. During the first
and a larger fill factor (FF) indicates better power exploitation effi­
two periods, PDO follows the exploratory process, which is inspired by
ciency, respectively.
foraging and burrowing behavior. When PDO is in the third and last
quarters, it performs exploitation strategy that mimic the response of
3. IPDO based unitary PV reconfiguration technique prairie dogs to predators and food. Accordingly, the solution update
equations for the first and second periods can be expressed by
3.1. Principal of prairie dog optimization
xbest − xrandom itermax
xi+1 = xbest − Ei × ρ − Levy(n) × , ∀i < (13)
PDO is a recently developed meta-heuristic optimizer that mimics the xbest + ε 4

16
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 14. Comparison of four metrics between different reconfiguration algorithms for asymmetric (20 × 30) PV array.

17
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 14. (continued).

itermax itermax different coteries still deserves further improvement. Therefore, this
xi+1 = xbest × xrandom × DS × Levy(n), ∀ ≤i< (14) work introduces an information exchange strategy to enhance in­
4 2
teractions among coteries and thus to improve the global exploration
Ei = xbest × ε +
xi × xave
(15) capability of PDO. In addition, a self-rectify strategy is embedded in the
xbest × (UB − LB) + ε reconfiguration of unitary PV arrays to eliminate redundant switch
actions.
i
DS = 1.5 × ω × (1 − )2*i/itermax (16)
itermax 3.2.1. Information exchange strategy
Each prairie dog memorizes food sources, burrows, and risks from
where xi+1 represents the solution in the ith iteration; xbest is the best
predators, such information determines solution quality. The searching
agent obtained so far; ρ is an inherent parameter of PDO assigned as 0.1;
space will be more efficiently explored if the information communica­
Levy(n) mimics the pace of prairie dogs to improve the global searching
tion is constructed between different coteries. To obtain better prairie
ability; xrandom denotes a random agent in the current population; ε
dog coteries, which are the foundation of iterative optimization, infor­
represents the individual difference between various agents which is
mation exchange strategy is introduced. Specifically, the worst agent in
usually a small number; xave is the mean of the current population; ω
the first coterie is replaced by the best agent in the second coterie, and
denotes a constant of 1 or − 1, respectively.
the worst agent in the second coterie is similarly replaced by the best one
in the third coterie, and so on. Given the above, the best individual is
3.1.3. Global exploration
migrated to other coteries, thus consequently realizing information ex­
When the iteration number exceeds half, PDO is focused on global
change between each coterie, yields
exploration. As mentioned earlier, the reactions to food can propels
{
prairie dogs to a promising position for further exploitation and better j
xworst = xj+1
solutions. The reaction to danger alerts can help agents keep away from
best
j = 1, 2, ⋯, m − 1 (20)
xworst = x1best
m
bad solutions. Moreover, two tactics can be described as:
j
xi+1 = xbest − θ × Ei − r ×
xbest − xrandom itermax
,∀ ≤i<3
itermax
(17) where xworst represents the worst solution of the jth coterie; x1best denotes
xbest + ε 2 4 the best solution of the first coterie; and m is the number of coteries,
respectively.
xi+1 = xbest × r × PE (18)
3.2.2. Self-rectify strategy
i itermax
PE = 1.5 × (1 − )2*i/itermax , ∀3 ≤ i < itermax (19) The essence of PV array reconfiguration is to exchange two PV
itermax 4
modules with diverse irradiance. In contrast, meta-heuristic algorithms
where θ is signal of food; i means the iteration number; r represents a used in previous studies are all stochastic optimization, which does not
random value from 0 to 1; PE denotes predator effect, respectively. take into account the irradiance of each PV module. That is, there must
exist redundant position swapping for completely equivalent PV mod­
ules. However, existing researches employing meta-heuristic algorithms
3.2. Improved prairie dog optimization PV reconfiguration design almost acquiesce in these superfluous reconfiguration actions. To
address this gap, this work designs a self-rectify strategy (SRS) to avoid
During PDO optimization, various coteries (group of agents) explore useless relocating of PV modules, i.e., reduce the ineffective actions of
the searching space in parallel, but the communication between

18
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 15. Comparison of four metrics between different reconfiguration algorithms for symmetric (30 × 30) PV array.

19
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 15. (continued).

the switching matrix and enhance the efficiency of the optimizer. In 6. Remained PV units are checked in a similar way.
particular, PV units are examined one by one: if the irradiance of relo­
cated PV unit is equal to that of the one originally placed in the position, Lastly, the rectified solution and the optimized irradiance distribu­
the corresponding reconfiguration action is invalid since it does not tion are determined. Obviously, invalid actions of switching matrix are
contribute to the enhancement of PV power generation, and this relo­ significantly reduced by SRS.
cated PV unit should be deported to its last position. In addition, ac­ Combining the above two strategies, IPDO technique is proposed.
cording to Kirchhoff’s voltage and current laws, repositioning of the The steps of IPDO based unitary PV reconfiguration are shown in Fig. 5.
entire row of PV array makes no difference to the output power of PV
array. Therefore, SRS rearranges the rows to remain the first column of 4. Case studies
PV array unchanged, i.e., the vector of [1, 2, 3, …, n].
To better illustrate the procedure of SRS, an example of Case 1 under Two case studies of asymmetric and symmetric arrays are performed
SW shading is depicted in Fig. 4. Firstly, the initial irradiance distribu­ among proposed IPDO and other algorithms (GA, PSO, SA, FSS, ACP,
tion of asymmetric (20 × 30) PV array is equivalently transformed into MHHO, AEO, and Run). Moreover, simulation tests are executed under
PV units with configuration of (10 × 10) and a transitional solution 1 is eight typical shades. Besides, all simulations are accomplished by
generated that is computed by IPDO. After rearranging PV rows, tran­ MATLAB 2019b through a personal computer equipped with an Intel
sitional solution 2 is obtained. Note that all these small squares with CoreTMi5 CPU at 3.0 GHz and a 16 GB of RAM. Meanwhile, ODE45
gray background represent invalid switching actions according to the solver and variable step size are used. PV module type is A10 Green
initial irradiance distribution. The details will be elaborated on PV unit Technology A10J-M60-225 with the maximum power of 224.9856 W,
numbered 5 in the last column: voltage at maximum power point as 30.24 V, current at maximum power
point as 7.44 A, open circuit voltage as 36.24 V and short circuit current
1. In transitional solution 2, PV unit numbered #5 is in the second row, as 8.04 A. In order to guarantee a fair comparison between all adopted
but with invalid action (both 1000 W/m2 before and after algorithms, the population number and terminating iterations are cho­
relocation); sen as 50 and 500, respectively. While each case is repeated 20 times. In
2. PV unit numbered #2 should be sent back to the second row. And PV addition, the assumptions made during the experiment are as follows:
unit numbered #5 should be deported to the last position, i.e., the
position of PV unit numbered #2; therefore, PV unit numbered #5 is • The ambient temperature was kept at 25 ◦ C;
moved to the third row; • The wire loss between the PV panels was not considered during the
3. PV unit numbered #2 is moved back to the second row in the calculation of the output power;
meantime; • The PV panels are not affected by the heat caused by power
4. PV unit numbered #5 in the third row still represents an invalid generation;
switch action (both 1000 W/m2 before and after relocation) and • The switch matrix can act correctly and quickly when receiving the
therefore PV unit numbered #3 should be sent back to the third row; action signal.
5. PV unit numbered #5 is moved to the tenth row (valid action for 900
W/m2 changing to 1000 W/m2); The eight shading patterns are short-wide (SW), short-narrow (SN),

20
B. Yang et al. Solar Energy 269 (2024) 112361

long-wide (LW), long-narrow (LN), diagonal, outer, center, and random,

Case 2
as presented in Fig. 6 and Fig. 7.

5
3
5
4
5
4
6
6
4.1. Asymmetric (20 × 30) PV array

4.1875
Case 1
Run
According to the principal of unitary PV reconfiguration established

5
3
3
5
5
2
5
5
1
in Section 2.1.2, the asymmetric (20 × 30) PV array is divided into 10 ×
10 PV units. Each PV unit contains three parallel PV strings consisting of
Case 2 two series-connected PV modules. These six modules are considered as

4
4
1
3
2
4
5
5
one unit to participate relocation of PV array. Fig. 8 demonstrates the

3.6875
irradiance distribution of the reconfigured asymmetric PV array by
Case 1
AEO

IPDO. Meanwhile, I-V and P-V curves under eight types of shadings

2
5
2
6
4
2
6
2
4
reconfigured by IPDO are illustrated in Fig. 9.
Fig. 8 demonstrates that the original concentrated shading is effec­
Case 2

tively dispersed throughout the whole PV array. Fig. 9 implies that SW,
7
6
4
8
3
7
8
7
LW and outer shadings have a serious impact on the output character­
istic of PV array with jagged multi-peak P-V curve and large power
6.8125
MHHO

Case 1

losses. Therefore, the wider the partial shading, the greater the influence
8
7
7
8
7
8
8
7
7

on the output power characteristics of PV array. The wide shading

should be avoided at all cost. IPDO effectively increases the output
Case 2

power and smooth the characteristic curves of each shading. Apart from
9
8
9
7
8
9
7
1

the slight local peak in P-V curve under outer shading, all other P-V
curves after reconfiguration by IPDO are unimodal curve, which greatly
6.5625
Case 1
ACP

facilitates the precision tracking of MPPT algorithm.

7
8
8
2
8
7
2
4
8

Besides, the comparative characteristic curves of the initial PV array

(TCT configuration) and relocated PV array by nine algorithms under
Case 2

SW partial shading are exhibited in Fig. 10.

8
9
8
9
9
8
9
9

The initial P-V curve without reconfiguration have five peaks, among
nine algorithms, FSS reduces the number of peaks to three but with
8.8125
Case 1

relatively unsatisfactory enhancement of power. The other eight algo­

FSS

10

9
9
8
9
9

9
9
9

rithms all obtained smooth curves. Meanwhile, IPDO acquires the

highest maximum power, while PSO, Run, SA, GA and AEO achieve
Case 2

similar peak values, followed by MHHO and ACP. Output power opti­
5
7
5
6
6
6
6
8

mized by nine algorithms are provided in Table 2, in which Pmax means

maximum power and Pavg is average power in 20 independent runs,
5.8125
Case 1

respectively.
SA

6
3
6
7
6
5
7
8
2

Combined with TCT curves shown in Fig. 9, it can be found that the
available margin for power improvement under LN and diagonal is
Case 2

relatively small. Although PSO, GA, SA, AEO and Run also obtain
2
3
7
2
7
3
3
4

maximum power and zero standard deviation (STD) in LN shading, only

STD of IPDO is zero in both LN and diagonal scenario, which indicates
3.9375
Case 1

that IPDO owns a better stability than that of other alternatives. In

GA

6
3

4
4
3
4
1
6
5

addition, IPDO obtains the highest value in both Pmax and Pavg for all
PSCs, which verifies the effectiveness and advantages of IPDO. Besides,
Case 2

STD of IPDO is the smallest among the nine algorithms in all shadings
2
3
3
2
6
4
5
2

except the center shadow, which validates its robustness and stability.
Case 1

3.875

4.2. Symmetric (30 × 30) PV array

PSO

3
6

3
6
4
3
5
4
4

Similarly, a symmetric (30 × 30) PV array is divided into 100 PV

Case 2

units, each of which becomes a (3 × 3) PV array. The shading pattern

Ranking of all reconfiguration algorithms based on.

1
1
1
1
1
1
1
1

reconfigured by IPDO is exhibited in Fig. 11.

Because of the “first column fixed” rule of SRS, the first column of PV
Case 1
IPDO

unit, i.e., the first three PV modules, remains unchanged, as demon­

strated in Fig. 11. Here, IPDO effectively scatters shadings over the
1
1
1
1
1
1
1
1
1
1
Ranks (case 1/case 2)

whole PV array. But there still exists several concentrated (3 × 3)

Case 2

shadings, which is an inherent drawback of the unitary reconfiguration

10
10
10
10
10
10
10
10

and a major constraint to power optimization. Fig. 12 provides the

characteristic curves of IPDO under eight shadings. Apparently, the
9.9375
Case 1

jagged P-V and I-V curves under each shading are effectively smoothed.
TCT

10
10
10

10
10
10
10
10

However, zigzag pattern still remains in the left corner of I-V curves
9

under SW and LW, together with a dramatically increased power under

Average rank

each shading. Therefore, the outstanding capacity of power enhance­

Final rank
Diagonal

ment under various PSCs fully demonstrate the practicality and uni­
Random
Shadow
Table 4

Center
Outer

versal application of IPDO based unitary PV reconfiguration.

LW
SW

LN
SN

Meanwhile, diagonal shading is selected to compare the effect of nine

21
B. Yang et al. Solar Energy 269 (2024) 112361

Table 5
Switch actions of IPDO based unitary PV array reconfiguration obtained under two cases.
Shadow SW SN LW LN Diagonal Outer Center Random

Case Case Case Case Case Case Case Case Case Case Case Case Case Case Case Case
1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2

Switch Without SRS 85 90 87 89 59 89 87 91 83 93 88 89 92 86 93 90

actions With SRS 55 45 24 28 56 55 18 24 32 49 51 53 37 36 57 68
Decreasing 30 45 63 61 3 34 69 67 51 44 37 36 55 50 36 22
actions

algorithm may be unsuitable for different shadings.

In Case 2, the advantage of IPDO over other methods is more obvious
in diagonal, with a power increase of 2.93 % to 9.65 % compared to that
of other schemes. Pen (%) of IPDO over TCT is 40.69 %, 8.33 %, 27.76 %,
12.98 %, 13.71 %, 22.98 %, 21.1 % and 16.92 % in eight shadings.
Moreover, IPDO achieves the best results in four criteria in all shadings,
as indicated in Table 4.
Besides, switching actions of IPDO based PV array unitary reconfi­
guration with/without SRS are summarized in Table 5, which verifies
the effectiveness of SRS for avoiding unnecessary switching actions.
Obviously, SRS can noticeably reduce mechanical and time costs due to
redundant switching actions. In particular, switching actions is
decreased by up to 80 % with a mean value of 50 %.

6. RTLAB based hardware-in-the-loop experiment

A RTLAB based HIL experiment is carried out to verify the imple­

mentation feasibility of IPDO with real-time processor. The RTLAB
Fig. 16. RTLAB based HIL experiment platform. platform is equipped with an Intel Core i7-8650U 2.11 GHz CPU and 8
GB memory as well as an OP5700 simulator, as illustrated in Fig. 16.
reconfiguration methods, as illustrated in Fig. 13. Here, black dashed P- ODE4 is chosen as the applied solver with the sampling time of 0.003 s.
V curve without reconfiguration (TCT) has four peaks that might lead Other experimental parameters and input data are identical to
MPPT algorithm into a local optimum. The power characteristic curves simulation.
optimized by ACP and FSS still have small sawtooth. Other P-V curves Fig. 17 and Fig. 18 provide the comparison of output characteristic
are smooth, among which SA, GA, PSO, Run and AEO obtain very close curves between simulation and HIL experiment results obtained by IPDO
peak values. Meanwhile, the maximum power obtained by MHHO is under two cases. One can readily observe that their curves match quite
slightly smaller. It is noteworthy that the maximum power acquired by well, which therefore validate the implementation feasibility.
IPDO is obviously larger than that of others under diagonal case. In
particular, IPDO achieves the best results for all shadings as tabulated in 7. Conclusions
Table 3.
A novel IPDO is designed to achieve PV arrays reconfiguration under
5. Discussions various PSC conditions, major findings and conclusions of this work can
be summarized into the following four aspects:
To thoroughly investigate IPDO performance, comprehensive anal­
ysis are conducted on Pavg presented in Table 2 and Table 3. Therefore, • The novel dynamic unitary PV array reconfiguration method signif­
the metrics of power loss, power enhancement, FF and ER are clearly icantly mitigates the computational burden, thus expanding the size
illustrated in Fig. 14 and Fig. 15. of PV arrays available for the research due to its well trade-off be­
For better visualization, bars of IPDO are bordered with a dashed red tween computational speed and solution quality. This indicates that,
line, and the optimum values of each shading are highlighted in red. when the irradiance changes, the rapid calculation can make the
Under SW scenario in Case 1, IPDO increases power by 31.59 %, 4.41 %, switching matrix act more effectively, to achieve the optimal power
4.18 %, 3.85 %, 12.10 %, 6.98 %, 5.90 %, 4.39 % and 2.63 % compared generation efficiency;
to that of TCT, PSO, GA, SA, FSS, ACP, MHHO, AEO and Run, respec­ • The modified switching matrix dramatically reduces the required
tively. Furthermore, IPDO reduces power loss under SW to 14.47 %, switch actions and the implementation cost. This demonstrates that,
which is significantly better than that of others. FF and ER are increased the proposed method can effectively improve the switch lifetime, and
to 0.714 and 85.53 % compared to 0.542 and 65 % for TCT. For other effectively control the power loss caused by switching action in
shadings, IPDO also achieves remarkable enhancements compared to actual production and reduce the cost caused by this reason;
that of all alternatives. However, four metrics scores of GA and Run • IPDO achieves the highest maximum power point as well as the best
under LN are the same as those of IPDO. It is noteworthy that, FSS ob­ ranking of the four metrics under all shadings. Particularly, it in­
tains a negative growth of Pen (%) under diagonal (power decreased after creases output power by 6.1 % to 31.59 % for asymmetric (20 × 30)
reconfiguration), as well as other three criteria. This indicates that static PV array and by 8.33 % to 43.69 % for symmetric (30 × 30) PV array
compared to that of TCT. Besides, the minimum power STD of IPDO

22
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 17. Comparison of asymmetric (20 × 30) PV array obtained by HIL experiment and simulation.

validates its robustness and stability. This shows that, the proposed the grid, it has caused a certain impact on the grid. Through the
algorithm can be effectively applied to the dynamic reconfiguration reconfiguration of PV array described in this paper, the output curve of
of large-scale PV arrays, significantly enhancing the power genera­ PV array can be smoothed and the stability of power grid can be
tion efficiency of PV power station, and reducing the impact of improved.
irradiance change; Future studies will not only focus on the hardware implementation of
• RTLAB based HIL experiments effectively validate the implementa­ IPDO for PV systems under real atmospheric conditions, but also study
tion feasibility of IPDO. the reconfiguration results of IPDO in hybrid systems with PV and other
reconfigurable systems.
With a large number of PV power generation equipment connected to

23
B. Yang et al. Solar Energy 269 (2024) 112361

Fig. 18. Comparison of symmetric (30 × 30) PV array obtained by HIL experiment and simulation.

Declaration of Competing Interest Science Foundation of China (62263014); Yunnan Provincial Basic
Research Project (202201AT070857); Key R&D Project of Zhejiang
The authors declare that they have no known competing financial Province (LQ21F030017); China Scholarship Council.
interests or personal relationships that could have appeared to influence
the work reported in this paper. References

[1] K. Anjaiah, P.K. Dash, M. Sahani, A new protection scheme for PV-wind based DC-
Acknowledgments
ring microgrid by using modified multifractal detrended fluctuation analysis,
Protect. Control Modern Power Syst. 7 (1) (2022) 100–123.
The authors gratefully acknowledge the support of National Natural

24
B. Yang et al. Solar Energy 269 (2024) 112361

[2] H. Laghridat, A. Essadki, T. Nasser, Coordinated control by ADRC strategy for a power extraction under partial shading conditions, IEEE Trans. Sustainable Energy
wind farm based on SCIG considering low voltage ride-through capability, Protect. 9 (1) (2018) 74–85.
Control Modern Power Syst. 7 (1) (2022) 82–99. [27] D. Yousri, D. Allam, M.B. Eteiba, Optimal photovoltaic array reconfiguration for
[3] B. Yang, B. Liu, H. Zhou, J. Wang, W. Yao, S. Wu, H. Shu, Y. Ren, A critical survey alleviating the partial shading influence based on a modified harris hawks
of technologies of large offshore wind farm integration: summarization, advances, optimizer, Energ. Conver. Manage. 206 (2020) 112470.
and perspectives, Protect. Control Modern Power Syst. 7 (17) (2022) 1–32. [28] B. Yang, R. Shao, M. Zhang, H. Ye, B. Liu, T. Bao, J. Wang, H. Shu, Y. Ren, H. Ye,
[4] Z.Y. Ren, Z.X. Yang, H. Li, Y. Xu, W.Y. Li, Y.Y. Wen, A 3-dimensional planning Socio-inspired democratic political algorithm for optimal PV array reconfiguration
method for tidal current power generation farms considering complex submarine to mitigate partial shading, Sustainable Energy Technol. Assess. 48 (2021) (2021)
terrains, IEEE Trans. Sustainable Energy 13 (1) (2022) 220–230. 101627.
[5] Y. Chen, B. Yang, Z. Guo, J. Wang, M. Zhu, Z. Li, T. Yu, Dynamic reconfiguration [29] S. Mirjalili, S. Saremi, S.M. Mirjalili, L.D.S. Coelho, Multi-objective grey wolf
for TEG systems under heterogeneous temperature distribution via adaptive optimizer: a novel algorithm for multi-criterion optimization, Expert Syst. Appl. 47
coordinated seeker, Protect. Control Modern Power Syst. 7 (2022) 38. (2015) 106–119.
[6] Z. Guo, B. Yang, Y. Chen, Z. Li, Q. Li, J. Deng, C. Guo, X. Zhang, B. Tang, M. Zhu, [30] A. Tabanjat, M. Becherif, D. Hissel, Reconfiguration solution for shaded PV panels
S. Qu, Modular Thermoelectric generation arrays reconfiguration under using switching control, Renew. Energy 82 (2015) 4–13.
heterogeneous temperature distribution via improved cooperation search [31] V.S. Narayanaswamy, R. Ayyanar, A. Spanias, C. Tepedelenlioglu, D. Srinivasan,
algorithm: modelling, design and HIL validation, Appl. Therm. Eng. 219 (2023) 2019. Connection topology optimization in photovoltaic arrays using neural
119323, https://doi.org/10.1016/j.applthermaleng.2022.119323. networks. In: 2019 IEEE International Conference on Industrial Cyber Physical
[7] B. Yang, Y. Li, J. Li, H. Shu, X. Zhao, Y. Ren, Q. Li, Comprehensive summary of Systems (ICPS). 6-9 May 2019, Taipei, Taiwan, pp. 167-172.
solid oxide fuel cell control: A state-of-the-art review, Protect. Control Modern [32] B. Yang, M. Zhang, Z. Guo, P. Cao, J. Yang, G. He, J. Yang, R. Su, X. Huang, M. Zhu,
Power Syst. 7 (2022) 36. H. Lu, D. Zhu, Adaptive evolutionary jellyfish search algorithm based optimal
[8] X. Fu, Statistical machine learning model for capacitor planning considering photovoltaic array reconfiguration under partial shading condition for maximum
uncertainties in photovoltaic power, Protect. Control Modern Power Syst. 7 (1) power extraction, Expert Syst. Appl. 215 (2023) 119325.
(2022) 51–63. [33] Y. Wang, B. Yang, Optimal PV array reconfiguration under partial shading
[9] V. Hosseinkhani, M. Sarvi, A new five-level inverter with reduced leakage current condition through dynamic leader based collective intelligence, Protect. Control
for photovoltaic system applications, Protect. Control Modern Power Syst. 7 (2) Modern Power Syst. 8 (2023) 40.
(2022) 287–299. [34] B. Praveen Kumar, D.P. Winston, S. Cynthia Christabel, S. Venkatanarayanan,
[10] B.o. Yang, T. Zhu, J. Wang, H. Shu, T. Yu, X. Zhang, W. Yao, L. Sun, Comprehensive Implementation of a switched PV technique for rooftop 2 kW solar PV to enhance
overview of maximum power point tracking algorithms of PV systems under partial power during unavoidable partial shading conditions, J. Power Electron. 17 (6)
shading condition, J. Clean. Prod. 268 (2020) 121983. (2017) 1600–1610.
[11] D.S. Pillai, R.J. Prasanth, S.S.M. Nihanth, N. Rajasekar, A simple, sensorless and [35] S. Shakthivel, D.P. Winston, S. Periya Backiyam, M. Murugan, Square dynamic
fixed reconfiguration scheme for maximum power enhancement in PV systems, reconfiguration for the partial shaded photovoltaic system–simulation and
Energy Convers. Manage. 172 (2018) 402–417. experimental analysis, Energy Sources Part A 44 (3) (2022) 6868–6885.
[12] S. Sugumar, D.P. Winston, K. Ganesan, M. Pravin, Comparative analysis of Hybrid, [36] A. Srinivasan, S. Devakirubakaran, B. Sundaram, P. Balachandran, S. Cherukuri, D.
conventional and staircase static reconfiguration methods to mitigate partial P. Winston, T. Babu, H. Alhelou, L-Shape propagated array configuration with
shading effects: Unveiling the superiority of two-step staircase (2SS) dynamic reconfiguration algorithm for enhancing energy conversion rate of partial
reconfiguration, Sol. Energy 264 (2023) 112029. shaded photovoltaic systems, IEEE Access 9 (2021) 97661–97674.
[13] G. Madhusudanan, N. Rakesh, S. Senthil Kumar, S. Sarojini Mary, Solar [37] A.E. Ezugwu, J.O. Agushaka, L. Abualigah, S. Mirjalili, A.H. Gandomi, Prairie dog
photovoltaic array reconfiguration using magic su-do-ku algorithm for maximum optimization algorithm, Neural Comput. Appl. 34 (22) (2022) 20017–20065,
power production under partial shading conditions, Int. J. Ambient Energy 43 (1) https://doi.org/10.1007/s00521-022-07530-9.
(2022) 1204–1215. [38] G.S. Krishna, T. Moger, Enhanced power generation for total-cross-tied PV array
[14] S.R. Potnuru, D. Pattabiraman, S.I. Ganesan, N. Chilakapati, Positioning of PV using simulated annealing under partial shading conditions, IET Renew. Power
panels for reduction in line losses and mismatch losses in PV array, Renew. Energy Gener. 14 (17) (2020) 3457–3468.
78 (2015) 264–275. [39] D. Yousri, A. Fathy, E.F. El-Saadany, Four square sudoku approach for alleviating
[15] B.I. Rani, G.S. Ilango, C. Nagamani, Enhanced power generation from PV array shading effect on total-cross-tied PV array, Energy Convers. Manage. 269 (2022)
under partial shading conditions by shade dispersion using sudoku configuration, 116105.
IEEE Trans. Sustainable Energy 4 (3) (2013) 594–601. [40] S. Anjum, V. Mukherjee, A novel arithmetic sequence pattern reconfiguration
[16] V.M. Ram Tatabhatla, A. Agarwal, T. Kanumuri, Performance enhancement by technique for line loss reduction of photovoltaic array under non-uniform
shade dispersion of solar photovoltaic array under continuous dynamic partial irradiance, J. Clean. Prod. 331 (2022) 129822.
shading conditions, J. Clean. Prod. 213 (2018) 462–479. [41] D. Yousri, T.S. Babu, S. Mirjalili, N. Rajasekar, M.A. Elaziz, A novel objective
[17] M. Horoufiany, R. Ghandehari, Optimal fixed reconfiguration scheme for PV arrays function with artificial ecosystem-based optimization for relieving the
power enhancement under mutual shading conditions, IET Renew. Power Gener. mismatching power loss of large-scale photovoltaic array, Energy Convers.
11 (11) (2017) 1456–1463. Manage. 225 (2020) 113385.
[18] M. Horoufiany, R. Ghandehari, A new photovoltaic arrays fixed reconfiguration [42] A.M. Nassef, E.H. Houssein, B.E. Helmy, A. Fathy, M.L. Alghaythi, H. Rezk,
method for reducing effects of one- and two-sided mutual shading, J. Sol. Energy Optimal reconfiguration strategy based on modified Runge Kutta optimizer to
Eng. 141 (3) (2018) 031013. mitigate partial shading condition in photovoltaic systems, Energy Rep. 8 (2022)
[19] M. Horoufiany, R. Ghandehari, Optimization of the sudoku based reconfiguration 7242–7262.
technique for PV arrays power enhancement under mutual shading conditions, Sol. [43] B. Nayak, A. Mohapatra, P. Das, 2017. Optimal hybrid array configuration scheme
Energy 159 (2018) 1037–1046. to reduce mismatch losses of photovoltaic system. In: IEEE 2017 Second
[20] R. Venkateswari, N. Rajasekar, Power enhancement of PV system via physical array International Conference on Electrical, Computer and Communication
reconfiguration based lo shu technique, Energ. Conver. Manage. 215 (2020) Technologies (ICECCT). 22-24 Feb. 2017, Coimbatore, India, pp. 1-7.
112885. [44] B. Dhanalakshmi, N. Rajasekar, A novel competence square based PV array
[21] V.M.R. Tatabhatla, A. Agarwal, T. Kanumuri, Improved power generation by reconfiguration technique for solar PV maximum power extraction, Energy
dispersing the uniform and non-uniform partial shades in solar photovoltaic array, Convers. Manage. 174 (2018) 897–912.
Energy Convers. Manage. 197 (2019) 111825. [45] R. Shao, B. Yang, H. Shu, K. Zeng, H. Zhang, Y. Chen, Optimal reconfiguration
[22] G. Meerimatha, B.L. Rao, Novel reconfiguration approach to reduce line losses of method for photovoltaic arrays based on improved mayfly algorithm, Automat.
the photovoltaic array under various shading conditions, Energy 196 (2020) Electric Power Syst. 46 (11) (2022) 142–150.
117120. [46] G. Velasco-Quesada, F. Guinjoan-Gispert, R. Piqué-López, M. Román-Lumbreras,
[23] S. Su, W. Li, A. Garg, L. Gao, An adaptive boosting charging strategy optimization A. Conesa-Roca, Electrical PV array reconfiguration strategy for energy extraction
based on thermoelectric-aging model, surrogates and multi-objective optimization, improvement in grid-connected PV systems, IEEE Trans. Ind. Electron. 56 (11)
Appl. Energy 312 (2022) 118795. (2009) 4319–4331.
[24] S. Su, W. Li, J. Mou, A. Garg, L. Gao, J. Liu, A hybrid battery equivalent circuit [47] A. Fathy, Recent meta-heuristic grasshopper optimization algorithm for optimal
model, deep learning, and transfer learning for battery state monitoring, IEEE reconfiguration of partially shaded PV array, Sol. Energy 171 (2018) 638–651.
Trans. Transp. Electrif. 9 (1) (2023) 1113–1127. [48] B.o. Yang, H. Ye, J. Wang, J. Li, S. Wu, Y. Li, H. Shu, Y. Ren, H. Ye, PV arrays
[25] S.N. Deshkar, S.B. Dhale, J.S. Mukherjee, T.S. Babu, N. Rajasekar, Solar PV array reconfiguration for partial shading mitigation: Recent advances, challenges and
reconfiguration under partial shading conditions for maximum power extraction perspectives, Energy Convers. Manage. 247 (2021) 114738.
using genetic algorithm, Renew. Sustain. Energy Rev. 43 (2015) 102–110. [49] B. Yang, C. Zeng, D. Li, Z. Guo, Y. Chen, H. Shu, P. Cao, Z. Li, Improved immune
[26] T.S. Babu, J.P. Ram, T. Dragicevic, M. Miyatake, F. Blaabjerg, N. Rajasekar, genetic algorithm based TEG system reconfiguration under non-uniform
Particle swarm optimization based solar PV array reconfiguration of the maximum temperature distribution, Appl. Energy 325 (2022) 119691.

25