IEEE SYSTEMS JOURNAL, VOL. 16, NO.

2, JUNE 2022 2765

Switching Matrix Enabled Optimal Topology

Reconfiguration for Maximizing Power Generation in
Series–Parallel Organized Photovoltaic Systems
Xiaolun Fang , Qiang Yang , Senior Member, IEEE, and Wenjun Yan, Member, IEEE

Abstract—Mismatch losses (ML) problem can directly reduce

the power generation of a photovoltaic (PV) array as well as
complicate maximum power point tracking (MPPT) algorithms
by causing multiple peaks to appear in the P-V characteristic.
Hence, PV array reconfiguration techniques have become highly
popular to minimize the ML of PV arrays. In this article, an optimal
dynamic topology reconfiguration method based on a switching
matrix is presented for series–parallel (SP) interconnected PV
array to mitigate the ML, thereby maximizing the power generation
of the PV array. The design of the switching matrix is presented that
enables each string of the PV array to be divided into two sections
to carry out the efficient dynamic topology reconfiguration with
improved flexibility and scalability. The proposed reconfiguration
method is assessed and compared with the SP scheme and the
recently developed cross diagonal view method. The effectiveness of
the proposed solution for the TCT scheme is also confirmed through
simulation experiments. The numerical results indicate that the Fig. 1. Deployment of distributed PV systems with a complex environment.
multiple peaks on P-V curves can be reduced and MPPT can be
implemented with a simple control algorithm. Additionally, the
maximum power generation improved by 20.13% and 14.39% on development of PV renewable energy systems imposes an urgent
average compared to the SP scheme and TCT scheme, respectively, technical challenge to ensure the reliability and safe operation
for different mismatch conditions of partial shading and random
PV module failures. of PV systems as well as the power generation efficiency [4].
The power generation efficiency of PV arrays can be affected by
Index Terms—Mismatch losses (ML) problem, power various internal and external factors, but one of the most crucial
generation, photovoltaic (PV) array reconfiguration, switching
matrix.
issues is array mismatch losses (ML) due to partial shading, dust,
and module defects or failures [5]. Especially, the mismatch
problem is more serious in urban distribution PV systems due
I. INTRODUCTION to complex environments, as illustrated in Fig. 1. The ML
problem in PV arrays causes the affected modules to dissipate
N RECENT years, photovoltaic (PV) energy, one of the most
I important renewable energy sources, attracted extensive at-
tention due to the pursuit of a low-carbon economy [1]. Though
the power generated by other modules leading to the reduced
maximum power point (MPP). Moreover, the hotpots formed
due to prolonged partial shading deteriorate the power genera-
large-scale PV systems have been a major part of PV’s growth
tion efficiency and system reliability [4]. The bypass diodes are
in recent years, rooftop solar is the most dramatic part of the
required to bypass the shaded or damaged modules to protect the
solar revolution [2]. The statistics in [3] reported that the total
PV modules from the effects of mismatch conditions (MCs) [6].
installed PV capacity in China was 216 GW (67.07 GW of
However, such bypassing may introduce multiple peaks in the
distributed PV systems) in the first half of 2020. Such rapid
P-V characteristic, and hence can impose additional complexity
on the MPP tracking algorithms or lead to the deviation from
Manuscript received November 11, 2020; revised February 2, 2021; accepted the MPP.
March 6, 2021. Date of publication April 5, 2021; date of current version June
13, 2022. This work was supported in part by the National Key Research and In literature, various forms of interconnection schemes of PV
Development Program of China under Grant 2018YFB1500900, in part by array have been reported, e.g., simple series (SS), series–parallel
the Natural Science Foundation of China under Grant 51777183, and in part (SP), total cross ted (TCT), and bridge linked (BL) [7]. Various
by the Major Scientific Project of Zhejiang Lab. under Grant 2018FD0ZX01.
(Corresponding author: Qiang Yang.) interconnection schemes of PV array are shown in Fig. 2. In
The authors are with the College of Electrical Engineering, Zhe- an SS interconnection scheme, all PV modules are connected
jiang University, Hangzhou 310027, China, and also with the Zhejiang in series and the array output current is limited by the smallest
Lab., Hangzhou 310027, China (e-mail: [email protected]; [email protected];
[email protected]). module current. In an SP interconnection scheme, PV modules
Digital Object Identifier 10.1109/JSYST.2021.3065131 are connected in series in a string, and PV strings are connected
1937-9234 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 2766 IEEE SYSTEMS JOURNAL, VOL. 16, NO. 2, JUNE 2022

PV modules in the column and the row need to be equal. A novel

reconfiguration method named cross diagonal view (CDV) was
proposed in [13] to allocate the module numbers logically and
spread the shading effect throughout the entire array for 9 × 9
SP and 9 × 9 TCT interconnection schemes. The CDV method
can allocate the module numbers logically for any number of PV
Fig. 2. Different PV array interconnection topologies: (a) simple series (SS),
(b) series-parallel (SP), (c) total cross ties (TCT); and (d) bridged link (BL). modules. Later in [14] and [15], the novel dominance square and
competence square based on the mathematical puzzle were pro-
posed. These methods adopt the logic-based number placement
that locates the number or alphabet to distribute shading effects
throughout the array. In reality, the implementation of static PV
array reconfiguration solutions requires much effort to carry out
a physical relocation reconfiguration. Also, this technique can
perform shade dispersion for concentrated shadow, but it does
not always maximize the power generation under all random
partial shadings.
Fig. 3. Equivalent circuit models of a PV cell.
On the other hand, the dynamic reconfiguration methods carry
out reconfiguration by altering the electrical connections of
the PV array. In literature, a collection of optimization algo-
rithms [e.g., Genetic Algorithm (GA), Particle Swarm Optimiza-
tion (PSO), Artificial Ecosystem-Based Optimization (AEO),
and Multiobjective Grey-Wolf Optimizer (MOGWO)] were ex-
ploited to optimally reconfigure a 9 × 9 TCT interconnected
PV array [16]–[19]. In these methods, each PV module can
be flexibly connected with any other module in the adjacent
column through changing the electrical connections, thereby
maximizing the power generation under partial shading pat-
terns. However, such solutions require the installation of a large
number of switches and may be computationally prohibitive for
large-scale distributed PV systems in the practical deployment.
Also, a dynamic reconfiguration solution using a switching
matrix arranged between fixed and adaptive PV modules to
Fig. 4. 10 × 8 SP interconnected PV array. reconfigure a PV array was proposed in [20]. In this article, an
intelligent reconfiguration approach based on fuzzy partitioning
in parallel. In TCT, PV modules are connected in parallel in a for a 9 × 4 TCT interconnected PV array was proposed and eval-
row, and each row is connected in series. The BL interconnection uated. Similarly, a method utilizing a switching matrix-based on
scheme is similar to the TCT but half of its connected to reduce the standard deviation for 3 × 4 SP connected PV array was
cable losses [8]. For the improvement of power generation from developed in [21]. However, such a solution cannot work well
the PV system, various reconfiguration solutions were developed in certain partial shading patterns, e.g., the shadows that only
for different types of interconnection schemes. The work in appear on certain modules.
[9] summarized the state-of-the-art of reconfiguration strategies In summary, the aforementioned reconfiguration solutions
for various types of interconnection schemes to enhance the have some obvious limitations. The static reconfiguration so-
maximum power generation under MCs. The reconfiguration lutions cannot guarantee the maximum power generation under
strategies mainly fall into two classes: static reconfiguration and all random shadow and MCs due to the random module defects
dynamic reconfiguration. and failures. The dynamic reconfiguration solutions proposed
For the static reconfiguration techniques, the PV modules are in [16] and [17] generally require a large number of switches
relocated without altering electrical connections, and one of the that can significantly increase the investment and maintenance
most common methods is based on the Sudoku puzzle. In [10] cost as well as the operational complexity for large-scale dis-
and [11], the authors proposed the sudoku and improved sudoku tributed PV systems. Besides, the performance of the dynamic
arrangement, respectively, for the 9 × 9 TCT interconnected PV reconfiguration solutions using a switching matrix (e.g., [20]
array to reduce the impact of partial shading of modules in any and [21]) can degrade under some shadow patterns in practice.
row to improve the overall power generation. In [12], an arrow Further, the pros and cons of existing static and dynamic PV
sudoku puzzle pattern was studied to enhance the maximum array reconfiguration approaches are summarized in Tables I
power generation for a 6 × 6 TCT interconnected PV array. and II, respectively.
However, the sudoku puzzle-based reconfiguration schemes can To this end, this article proposed a novel switching matrix-
only relocate the PV modules in a column, but cannot change the based reconfiguration method to realize the PV array recon-
position in a row. Also, it is required that the number of allocated figuration to enhance power output for SP interconnected PV

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 FANG et al.: SWITCHING MATRIX ENABLED OPTIMAL TOPOLOGY RECONFIGURATION FOR MAXIMIZING POWER GENERATION 2767

TABLE I
SUMMARY OF STATIC TOPOLOGY RECONFIGURATION APPROACHES

TABLE II
SUMMARY OF DYNAMICAL TOPOLOGY RECONFIGURATION APPROACHES

array under different MCs. The main technical contributions

made in this article can be summarized as follows. 1) An
SP interconnected PV array can be reconfigured by the pro-
posed switching matrix-based reconfiguration method combined
with the optimization algorithm. The proposed reconfiguration
method allows flexible alteration of PV module arrangements
in the context of distributed PV systems by the use of switch-
ing matrix, and hence outperforms the existing reconfiguration
solutions with significantly improved power generation under
various MCs. 2) The design of the switching matrix is presented
that enables each string of the PV array to be divided into two
sections to carry out the efficient dynamic reconfiguration. As a
result, the scalability of the reconfiguration solution is improved
as the number of required switches will not have to increase
along with the number of installed PV modules in a string. The
performance of the proposed solution is extensively evaluated
by using the CDV method [13] and SP without reconfiguration
as the comparison benchmarks. The numerical result for a range
of MCs considering the partial shading and random module fail-
ures confirmed the effectiveness and efficiency of the proposed
method.
The remainder of this article is organized as follows: Sec-
tion II presents the details of a PV array using the 10 × 8 Fig. 5. Switching matrix-based reconfiguration.
SP interconnection as an example. Section III described the
proposed switching matrix-based reconfiguration method and an
optimal reconfiguration solution. The performance of the pro- comparison studies and numerical results analysis. Finally, the
posed reconfiguration method is evaluated in Section IV through conclusive remarks and future work are given in Section V.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 2768 IEEE SYSTEMS JOURNAL, VOL. 16, NO. 2, JUNE 2022

Fig. 6. Designed switching matrix for 10 × 4 SP interconnected PV array.

TABLE III
PV MODULE SPECIFICATIONS AT STC

II. MODEL OF A PV MODULE AND ARRAY

A. Modeling of a PV Cell
Fig. 7. Inputs and outputs of the optimization engine.
The PV array is composed of a set of PV modules that consist
of a large number of cells. The equivalent circuit of a PV cell is
string. Similarly, the PV module labeled “210” indicates that
shown in Fig. 3. The series resistance of the PV cells in the path
the PV module is located at the 10th row of the second string.
of the current flow is represented by Rs [Ω]. The shunt resistance
The current produced by the PV module labeled “ij” is calcu-
offered to the leakage current is denoted by Rsh [Ω] [25]. The
lated as
mathematic modeling of a PV cell can be given by the following:
    Gij
V + IRs V + IRs Iij = kij Im = Im (2)
I = Iph − I0 exp −1 − (1) G0
ηVT Rsh
where i and j represent the string number and the row number in
where I[A] and V [V ] represent the generated current and voltage a string. G0 and Gij represent the standard irradiance (G0 = 1000
of a PV cell. Iph [A] and I0 [A] represent the photo current W/m2 ) and the irradiance received by the module ij, respectively.
and the reverse saturation current, respectively. ηis the diode Im [A] is the current generated by a module at standard irradiance
ideality factor (usually between 1 and 2 for a single cell). VT [V ] conditions.
is the thermal voltage and can be represented as VT = kT /q, Since PV modules in a string are in series, the available current
where k is the Boltzmann’s constant (1.38 × 10−23 J/K), q is of a string is restricted by the smallest current generation PV
the elementary charge (1.6 × 10−19 C), and T [K] is the absolute module that is not bypassed, and the voltage of the string #i is
temperature. The specifications of the PV module under standard equal to the sum of voltage produced of all modules in a string.
test conditions (STC, G = 1000 W/m2 , and T = 25 ◦ C) are Hence, the voltage generation of string #i can be written as
presented in Table III. 10

Vi = Vij . (3)
B. PV Array With SP Interconnection Topology j=1
In an SP interconnected scheme, the PV modules in a string Since eight strings are connected in parallel, applying Kirch-
are connected in series and a set of such strings are connected hoff’s circuit laws, the PV array current is equal to the sum of
in parallel [26]. In this article, a 10 × 8 SP interconnected PV the string current. The current produced by the array (Ia [A]) can
array is considered. As shown in Fig. 4, this PV array comprises be written as
80 PV modules, and each PV module is labeled to indicate its 8

location. For example, the PV module labeled “307” indicates Ia = Is,i . (4)
that the PV module is located at the seventh row of the third i=1

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 FANG et al.: SWITCHING MATRIX ENABLED OPTIMAL TOPOLOGY RECONFIGURATION FOR MAXIMIZING POWER GENERATION 2769

Fig. 9. Arrangement of PV array using the CDV method [12]: (a) before
reconfiguration, (b) during the process, and (c) after reconfiguration.

Fig. 8. Flowchart of the proposed optimal solution for PV array reconfigura-

tion.

The overall output voltage of the array (Va [V ]) is limited by

the minimum voltage generation string theoretically [27], and
the power produced by the array (Pa [W ]) can be written as

Pa = Ia Va = Ia min{Vi , 1 ≤ i ≤ 8}. (5) Fig. 10. PV module arrangement and mismatch patterns for case 1, case 2,
and case 3: (a), (d), (g) SP without reconfiguration; (b), (e), (h) SP with the CDV
method [13]; and (c), (f), (i) SP with the proposed reconfiguration method.
III. PROPOSED SWITCHING MATRIX-BASED
RECONFIGURATION SOLUTION
In this article, a switching matrix-based reconfiguration divide the whole array into 16 parts, i.e., each string is divided
method with optimal reconfiguration algorithmic solution is pro- into two parts, and each part consisting of five PV modules.
posed to maximize the power generation of SP interconnected In the proposed method, two ends of each part are connected to
PV array under MCs. the switching matrix. A simplified switching matrix for a 10 × 4
SP interconnected PV array and the positions of sensors (current
and voltage) of each module are shown in Fig. 6. The switching
A. Switching Matrix-Based PV Reconfiguration Method matrix is controlled by the control block with the optimization
To solve the mismatch problem, a switching matrix-based engine. The inputs and the outputs of the optimization engine
reconfiguration method is proposed. This proposed method is to are illustrated in Fig. 7. Through controlling such a switching
achieve the reconfiguration of an SP interconnected PV array. matrix, in total 16 parts can be combined in pairs, i.e., any two
In the proposed reconfiguration method, for a PV array, each parts can be connected in series in a string.
column is divided into two parts according to the principle of To carry out the topology reconfiguration in a timely manner,
minimizing the difference in the number of PV modules between two strategies, i.e., event-triggered or time-triggered mecha-
two parts. In this article, a 10 × 8 SP interconnected PV array nisms, can be adopted in the control block for decision-making.
is considered. As illustrated in Fig. 5, the proposed method is to The former can immediately respond to the changes of external

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 2770 IEEE SYSTEMS JOURNAL, VOL. 16, NO. 2, JUNE 2022

can also use to address the optimization problem with different

computational resources requirements and complexities.
The proposed optimal reconfiguration solution-based opti-
mization technique focuses on enhancing the maximum power
output of SP interconnected PV array under different MCs and
the procedure of optimal reconfiguration solution is presented
as follows:
Step 1. Irradiance Estimation and Initialization: Even though
the irradiance on each module can be known by using the
irradiance sensors, it can alternatively be derived from the
module-level voltage and current measurements with lower cost
using
Gij = α[Iij + I0 (eVij /ηVT − 1)] (6)
where Iij [A] and Vij [V ] are, respectively, the voltage and current
measurements of the module ij. α, I0 , and ηVT can be estimated
from the manufacturer’s datasheet values [16].
In this article, considering an M×N SP interconnected PV
array is initialized with a high temperature T0 , a 10 × 8 PV
irradiance matrix calculated by (6) is used as the initial solution.
The irradiance matrix is denoted as “G,” and the elements in each
column of “G” are arranged from the largest to the smallest to
obtain a matrix “S.”
Step 2. Partition of the Irradiance Matrix: According to Fig. 5,
the irradiance matrix “G” is divided into 16 submatrices, and the
size of each submatrix is 5 × 1 (five rows and one column).
Step 3: Obtainment of a New Solution: A perturbation is given
to the current solution to obtain a new solution. In this article,
Fig. 11. P-V and I-V characteristics: (a) PV-curve (case 1), (b) I-V curve (case
1), (c) PV-curve (case 2), (d) I-V curve (case 2), (e) PV-curve (case 3), and (f)
“reproduction,” “crossover,” and “mutation” in the GA are used
I-V curve (case 3). to generate the perturbation. This method can obtain the new
solution randomly and flexibly, and the obtained new solution
changes for random scenarios, e.g., a passing cloud. The recon- is represented by “Gnew .”
figuration can be carried out once the predefined criterion is met, The elements in each column of the irradiance matrix “Gnew ”
e.g., 10% improvement of MPP can be obtained after reconfig- are arranged from largest to smallest to get a matrix “Snew .”
uration, but it may cause frequent switching matrix operations. Step 4. Definition of the Objective Function: According to
The latter respond to the external changes periodically, e.g., the (2)–(5), the maximum power output of the SP interconnected PV
control block can make a decision at the time slot of 10 min. In array corresponding to the current solution “S” can be written
practice, the selection of time- or event-triggered mechanisms as

8
i=1 min(Sij )
can be determined by operators of the PV infrastructures.
C(S) = Pmax = max × I m × Va ,
G0
B. Optimal Algorithmic Solution
(j = 1, 2, · · · 10) (7)
The switching matrix-based reconfiguration needs to be im-
plemented through an optimization process. It is known that the where C corresponds to the objective function. Pmax represents
simulated annealing (SA) algorithm is efficient to address many the maximum power output of the PV array and Sij is the element
combinatorial optimization problems with low implementation of the ith column and the jth row of the solution “S.”
complexity. The GA [16] is a widely adopted optimization Step 5. Selection of the Solution: The value of the objective
technique that mimics the process of natural evolution to search function corresponding to the current solution and the new
for the optimal solution through “reproduction,” “crossover,” solution is compared to determine whether the new solution is
and “mutation” operations. The combination of SA and GA has accepted. The difference between the two objective function
been confirmed efficient to pursue the global optimal solutions values can be calculated as
[28]. In this article, due to the low implementation complexity
ΔD = C (Snew ) − C (S) . (8)
and excellent optimization performance of the SA and GA
algorithms, the SA algorithm with the genetic operator, i.e., The objective function value corresponding to the current
SA–GA joint algorithm, is implemented to identify the optimal solution is represented by C(S), and the new objective function
reconfiguration solution. It is worth noting that, other optimiza- value is represented by C(Snew ). If the value of the objective
tion algorithms, e.g., PSO [17], AEO [18], and MOGWO [19], function corresponding to the new solution is better than the

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 FANG et al.: SWITCHING MATRIX ENABLED OPTIMAL TOPOLOGY RECONFIGURATION FOR MAXIMIZING POWER GENERATION 2771

Fig. 12. Performance comparison of three methods for case 1–3: (a) GMPP, (b) ML, and (c) ER.

Step 6. Temperature Update: Based on (9), the simulated

temperature is updated to a lower value and then go back to
step 4

Tk = Tk−1 × α (10)

where Tk and Tk−1 are the current and the previous iteration SA
temperature, receptively, and α represents the rate of tempera-
ture drops.
Step 7. Termination Criterion: Repeat steps 3–6 until the
simulated temperature is lower than the cooling temperature Tn
and the optimal solution can be obtained.
For the proposed algorithmic solution, the parameter settings
are critical to the computational complexity and the precision of
the obtained solution, and the trade-off needs to be made to select
the appropriate parameter [28]. In this article, the parameters
of SA+GA are defined and set as follows: initial temperature
T0 = 1000, rate of temperature drop α = 0.98, and cooling
temperature Tn = 0.01. The flowchart of the proposed optimal
algorithmic solution implemented for PV array reconfiguration
is illustrated in Fig. 8.

IV. SIMULATIONS AND NUMERICAL RESULTS

To evaluate the effectiveness of the proposed switching
matrix-based reconfiguration method, a 10 × 8 SP intercon-
Fig. 13. PV module arrangement and mismatch patterns for case 4, case 5, nected PV array is used. In simulations, the specifications and
and case 6: (a), (d), (g) SP without reconfiguration; (b), (e), (h) SP with the CDV parameters of PV modules are given in Table III. The PV array
method [13]; and (c), (f), (i) SP with the proposed reconfiguration method. comprises in total of 80 modules and the total power generation
without any shadows under the STC is 20.008 kW. In total six
previous one, i.e., ΔD > 0, the new solution is accepted and cases (i.e., case 1–case 6) representing different MCs are consid-
this new solution is the basis for the next iteration. If ΔD < 0, ered in this article. Specifically, case 1–case 3 consider different
then the decision to keep or discard the new solution is based on partial shading scenarios, i.e., short-wide shading, long-narrow
probability. The probability of the current solution replaced by shading, and two-part shading, respectively, and case 4–case 6
the new solution is written as consider the scenarios with both partial shading and random
failures on modules. Further, a 10 × 8 TCT interconnected
1 ΔD > 0
P = (9) PV array is used to verify the effectiveness of the proposed
1/exp( − ΔD/Tk ) ΔD < 0
reconfiguration method for the TCT scheme. Three additional
where Tk and k are the current SA temperature and the num- cases (i.e., case 7–case 9) with the same mismatch scenarios
ber of iterations, respectively. This formula corresponds to the as case 4–case 6 are implemented for performance evaluation.
Metropolis–Hastings algorithm. While SA temperature is high, In this article, the 3.00 GHz Core i7-9700 CPU and a 16.0 GB
the algorithm is more likely to accept the solution worse than RAM are used for computation hardware and the simulations are
the current solution. The Metropolis criterion [29] allows the carried out using MATLAB/Simulink (version 2018a). The time
algorithm to jump out of the local optimums found in the early for decision-making of the proposed reconfiguration solution for
stages of execution. all tested scenarios is less than 0.4 s.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 2772 IEEE SYSTEMS JOURNAL, VOL. 16, NO. 2, JUNE 2022

where n and m represent the number of strings and the number

of modules in each string, respectively.
ER: It refers to the ratio of the GMPP under MCs and GMPP
under the STC. The ER can be determined as
GMPPMCs
ER (%) = × 100%. (12)
GMPPSTC

A. Performance Evaluation Under Partial Siding Patterns

The simulated partial shading patterns for case 1 (short-wide
shading), case 2 (long-narrow shading), and case 3 (two-part
shading) are shown in Fig. 10(a), (d), and (g), respectively. Four
inconsistent irradiation levels, i.e., 900, 700, 400, and 200 W/m2
are adopted in these cases. The shade dispersion results with the
CDV-based reconfiguration for three different cases (i.e., case
1–case 3) are presented in Fig. 10(b), (e), and (h), respectively.
It can be observed that the shaded areas in the modules are
distributed in the array. The PV module arrangements after
the reconfiguration using the proposed reconfiguration method
for three different cases are presented in Fig. 10(c), (f), and
(i), respectively. The result shows that the number of strings
with shaded modules in the arrangement after reconfiguration
is reduced compared with the SP without reconfiguration for all
evaluated cases. It also can be observed that the modules that
are most seriously shaded are connected to the same string as
many as possible when the proposed reconfiguration solution is
adopted. Such pheromone can effectively improve the overall
power generation of the distributed PV systems.
Fig. 14. P-V and I-V characteristics: (a) PV-curve (case 4), (b) I-V curve (case In Fig. 11, the P-V and I-V characteristics of the proposed
4), (c) PV-curve (case 5), (d) I-V curve (case 5), (e) PV-curve (case 6), and (f) solution are presented against the scenarios of SP without re-
I-V curve (case 6).
configuration and CDV method, where the MPP of the PV
array is highlighted. The I-V curves of the proposed solution
demonstrate more graceful degradation under MCs compared
For all simulated cases of MCs, the performance of the
with the other two comparative benchmarks. This indicates that
proposed solution is evaluated against the SP scheme without
the proposed solution is more capable to minimize the ML effect
reconfiguration and reconfiguration based on the CDV method
than the other two methods. Also, the proposed reconfiguration
[13]. In the existing reconfiguration methods, the CDV method
solution method produces smoother P-V curves with negligible
is considered suitable for the arbitrary size of the PV array and
local maximum power points. It can be seen that the proposed
is efficient to mitigate the ML. Fig. 9(a) illustrates a 10 × 8
solution can outperform the other two solutions for all evaluated
PV array, i.e., in total 80 modules (labeled from 1 to 80) that
cases with more produced power generation. For example, in
can be reconfigured using the CDV method through two steps
case 1, the proposed solution can produce additional power
as suggested in [13]: 1) the 10 × 8 array is first converted to a
generation of 2.541 and 2.192 kW compared with the SP with-
10 × 10 PV array and the CDV method is carried out, as shown
out reconfiguration and the CDV method, respectively. Similar
in Fig. 9(b); and 2) the 10 × 10 array is further reduced to 10 × 8
results can also be observed for evaluation case 2 and case 3 that
array by removing the last two columns (i.e., gray area) and the
confirm the superiority of the proposed solution.
modules are renumbered consecutively from 1 to 80, as shown
Here, the performance of the proposed solution is further as-
in Fig. 9(c).
sessed using three performance metrics, i.e., GMPP, ML, and ER
To quantify the performance of reconfiguration solutions
through a comparative study, and the numerical results are shown
for different simulation cases, three performance metrics are
in Fig. 12(a)–(c), respectively. It is shown that the proposed
adopted and defined as follows: 1) global maximum power point
solution is with the highest GMPP for all evaluated cases. In
(GMPP); 2) ML [30]; and 3) execution ratio (ER) [31].
detail, the ML of the proposed solution is 17.059%, 13.501%,
ML: It refers to the difference between the sum of the max-
and 14.831% less than the case without reconfiguration and
imum power of each module and the GMPP of the PV array
14.716%, 25.351%, and 12.0175% less than the CDV-based
under MCs (GMPPMCs ). The ML can be determined as
method for case 1–3, respectively. Also, the ER of the proposed
n m solution is 14.0979%, 11.1573%, and 12.2559% higher than
i=1 Pij − GMPPMCs
ML (%) = j=1
n m × 100% (11) the case without reconfiguration and 12.162%, 20.950%, and
i=1 j=1 Pij 9.931% higher than the CDV method.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 FANG et al.: SWITCHING MATRIX ENABLED OPTIMAL TOPOLOGY RECONFIGURATION FOR MAXIMIZING POWER GENERATION 2773

Fig. 15. Performance comparison of three methods for case 4–6: (a) GMPP, (b) ML, and (c) ER.

Fig. 16. PV module arrangement and mismatch patterns for case 7, case 8, and
case 9: (a), (d), (g) TCT without reconfiguration; (b), (e), (h) TCT with the CDV
method [13]; and (c), (f), (i) TCT with the proposed reconfiguration method. Fig. 17. P-V and I-V characteristics: (a) PV-curve (case 7), (b) I-V curve (case
7), (c) PV-curve (case 8), (d) I-V curve (case 8), (e) PV-curve (case 9), and (f)
I-V curve (case 9).
B. Performance Evaluation Under Partial Shading With
Random Failures Patterns
For the MC with both partial shading and random failures, curves. Also, the proposed solution can obtain the highest MPP
three different cases (i.e., case 4–case 6) are evaluated, as shown of 12.835, 13.380, and 12.178 kW, respectively, for case 4–6.
in Fig. 13(a), (d), and (g), respectively. Apart from partial Again, the performance in terms of GMPP, ML, and ER are
shadings, two failed PV modules (labeled as “210” and “803”) further evaluated and the results are shown in Fig. 15(a)–(c),
without power generation are set for all evaluated cases. The ar- receptively. It shows that the proposed solution can consistently
rangements of PV modules for three cases using the CDV-based produce the highest GMPP for the evaluated cases. The ML of
method are presented in Fig. 13(b), (e), and (h), respectively. The the proposed solution is 17.469%, 12.350%, and 14.227% less
module arrangements after reconfiguration using the proposed than the SP without reconfiguration, and 12.419%, 20.367%,
solution are shown in Fig. 13(c), (f), and (i), respectively. and 11.557% less than the CDV method. Also, the ER of
The P-V and I-V curves of the three evaluated solutions the proposed method is 14.04 79%, 9.9312%, and 11.4403%
are presented in Fig. 14. It can be observed that the proposed higher than the condition without reconfiguration and 9.987%,
reconfiguration solution can produce more smooth P-V and I-V 16.378%, and 9.293% higher than the CDV method.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 2774 IEEE SYSTEMS JOURNAL, VOL. 16, NO. 2, JUNE 2022

Fig. 18. Performance comparison of three methods for case 7–9: (a) GMPP, (b) ML, and (c) ER.

C. Performance Evaluation for TCT Topology Implemented on average compared with the SP scheme and the CDV method,
With the Proposed Solution respectively. Further, the effectiveness of the proposed method
In this article, the proposed reconfiguration solution is mainly for the TCT scheme is also verified confirmed with the improve-
ment of MPP by 14.39% and 16.33% on average compared with
investigated for the PV array with the SP interconnected scheme.
the TCT scheme without reconfiguration and the CDV recon-
It is noteworthy to mention that the solution can be adopted
for other interconnection schemes, e.g., the TCT scheme. The figuration method. Also, the proposed solution can improve the
scalability of deployment as the number of required switches
system model of the TCT interconnected scheme and problem
formulation can be found in [32]. For a 10 × 8 TCT scheme, the will not have to increase along with the number of installed PV
proposed method is to divide the overall array into twenty parts, modules in a string.
On the other hand, once the reconfiguration solution is
i.e., each row is divided into two parts, and each part consisting
of four PV modules. adopted, the installation and maintenance cost of the switching
Three cases (i.e., case 7–case 9) with both partial shading and matrix, electric measurement sensors, and the electrical cables
need to be included in the PV system investment. In reality, the
random failures are used to validate the proposed reconfiguration
method for the TCT scheme, as shown in Fig. 16(a), (d), and (g), proposed solution can bring out long-term economic benefit,
respectively. The proposed solution is evaluated by comparing particularly for the distributed PV systems that often suffer from
the module shadings and failures. In future work, several direc-
the CDV method, and the arrangements after reconfiguration
are shown in Fig. 16(b), (e), and (h), and Fig. 16(c), (f), and (i), tions are considered worth further research effort. An extensive
cost-benefit analysis of the proposed reconfiguration method is
respectively.
required before the practical deployment in real PV systems.
The P-V and I-V curves of these three evaluated cases are
presented in Fig. 17 and the results of GMPP, ML, and ER The anomaly detection of field sensors needs to be included in
consideration of system maintenance to enhance the reliability
are shown in Fig. 18(a)–(c), receptively. It can be observed
of PV arrays. Also, the optimal design of the switching matrix
that, similar to the experiment results of the SP interconnected
scheme, the proposed solution can produce more graceful P-V needs to be further exploited to reduce the deployment and
maintenance complexity, and hence minimize the deployment
and I-V curves and outperform the other two solutions for all
evaluated cases with more power generation. Also, The ML of hurdles in practice. Finally, the field deployment and test of the
the proposed solution is 16.739%, 7.307%, and 10.064% less proposed solution will be carried out to verify its effectiveness.
than the TCT without reconfiguration, and 14.979%, 9.803%,
and 14.010% less than the CDV method. Also, the ER of the
proposed method is 12.175%, 5.315%, and 7.320% higher than REFERENCES
the condition without reconfiguration and 10.895%, 7.130%, [1] J. D. Bastidas-Rodríguez, J. M. Cruz-Duarte, and R. Correa, “Mis-
and 10.190% higher than the CDV method. matched series–parallel photovoltaic generator modeling: An implicit
current–voltage approach,” IEEE J. Photovolt., vol. 9, no. 3, pp. 768–774,
May 2019.
V. CONCLUSION [2] H. S. Sahu, S. K. Nayak, and S. Mishra, “Maximizing the power gen-
eration of a partially shaded PV array,” IEEE J. Emerg. Sel. Topics
This article presented a novel flexible switching matrix-based Power Electron., vol. 4, no. 2, pp. 626–637, Jun. 2016.
reconfiguration method and adopted the SA–GA joint optimal [3] “2020 Q2 PV installations utility and distributed by province,” China
Energy Portal, 2020. [Online]. Available: https://chinaenergyportal.org/
algorithmic solution for a 10 × 8 SP interconnected PV array for en/2020-q2-pv-installations-utility-and-distributed-by-province/
power generation maximum under MCs. The proposed solution [4] D. S. Pillaib, J. P. Rama, M. S. S. Nihanthb, and N. Rajaseka, “A simple,
is extensively evaluated through comparison with the existing sensorless and fixed reconfiguration scheme for maximum power enhance-
ment in PV systems,” Energy Convers. Manage., vol. 172, pp. 402–417,
solutions for a range of MCs. The P-V and I-V characteristics are 2018.
analyzed and the performance is assessed for three performance [5] Y. Mahmoud and E. F. El-Saadany, “Enhanced reconfiguration method
metrics, i.e., GMPP, ML, and ER. The comparative studies for reducing mismatch losses in PV systems,” IEEE J. Photovolt., vol. 7,
no. 6, pp. 1746–1754, Nov. 2017.
indicate that the proposed solution consistently provides the [6] S. Vemuru, P. Singh, and M. Niamat, “Analysis of photovoltaic array
maximum power generation, minimum ML, and highest ER. with reconfigurable modules under partial shading,” in Proc. 38th IEEE
More precisely, the MPP is increased by 20.13% and 22.04% Photovolt. Specialists Conf., 2012, pp. 1437–1441.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.
 FANG et al.: SWITCHING MATRIX ENABLED OPTIMAL TOPOLOGY RECONFIGURATION FOR MAXIMIZING POWER GENERATION 2775

[7] B. Dhanalakshmi and N. Rajasekar, “Dominance square based array [29] C. Jiang and C. Wang, “Improved evolutionary programming with dynamic
reconfiguration scheme for power loss reduction in solar photovoltaic (PV) mutation and metropolis criteria for multi-objective reactive power opti-
systems,” Energy Convers. Manage., vol. 156, pp. 84–102, 2018. mization.” IEE Gener., Transmiss. Distrib., vol. 152, pp. 291–294, 2005.
[8] P. R. Satpathy, S. Jena, and R. Sharma, “Power enhancement from partially [30] S. G. Krishna and T. Moger, “Optimal sudoku reconfiguration technique
shaded modules of solar PV arrays through various interconnections for total-cross-tied PV array to increase power output under non-uniform
among modules,” Energy, vol. 144, pp. 839–850, 2018. irradiance,” IEEE Trans. Energy Convers., vol. 34, no. 4, pp. 1973–1984,
[9] G. S. Krishna and T. Moger, “Reconfiguration strategies for reducing Dec. 2019.
partial shading effects in photovoltaic arrays: State of the art, ” Sol. Energy, [31] A. S. Yadav and V. Mukherjee, “Line losses reduction techniques in
vol. 182, pp. 429–452, 2019. puzzled PV array configuration under different shading conditions,” Sol.
[10] B. I. Rani, G. S. Ilango, and C. Nagamani, “Enhanced power generation Energy, vol. 171, pp. 774–783, 2018.
from PV array under partial shading conditions by shade dispersion using [32] X. Fang, W. Yan, and Q. Yang, “Optimal reconfiguration of PV array under
Su Do Ku configuration,” IEEE Trans. Sustain. Energy, vol. 4, no. 3, mismatch conditions using simulated annealing algorithm,” in Proc. IEEE
pp. 594–601, Jul. 2013. 3rd Conf. Energy Internet Energy Syst. Integr., 2019, pp. 981–986.
[11] G. S. Krishna and T. Moger, “Improved sudoku reconfiguration tech-
nique for total-cross-tied PV array to enhance maximum power under
partial shading conditions,” Renewable Sustain. Energy Rev., vol. 109,
pp. 333–348, 2019.
[12] V. M. R. Tatabhatla, A. Agarwal, and T. Kanumuri, “Performance en-
hancement by shade dispersion of solar photo-voltaic array under con-
tinuous dynamic partial shading conditions,” J. Cleaner Prod., vol. 213,
pp. 462–479, 2019. Xiaolun Fang received the B.S. degree in electrical
[13] M. J. Bosco and M. C. Mabel, “A novel cross diagonal view configuration engineering and automation from Zhejiang Univer-
of a PV system under partial shading condition,” Sol. Energy, vol. 158, sity of Technology, Hangzhou, China, in 2018. She
pp. 760–773, 2017. is currently working toward the Ph.D. degree in elec-
[14] B. Dhanalakshmi and N. Rajasekar, “Dominance square based array trical engineering at Zhejiang University, Hangzhou,
reconfiguration scheme for power loss reduction in solar PhotoVoltaic China.
(PV) systems,” Energy Convers. Manage., vol. 156, pp. 84–102, 2018. Her research interests include renewable energy
[15] B. Dhanalakshmi and N. Rajasekar, “A novel competence square based PV analysis, energy efficiency, and optimization.
array reconfiguration technique for solar PV maximum power extraction,”
Energy Convers. Manage., vol. 174, pp. 897–912, 2018.
[16] T. S. Babu, J. P. Ram, T. Dragičević, M. Miyatake, F. Blaabjerg, and N.
Rajasekar, “Particle swarm optimization based solar PV array reconfigura-
tion of the maximum power extraction under partial shading conditions,”
IEEE Trans. Sustain. Energy, vol. 9, no. 1, pp. 74–85, Jan. 2018.
[17] S. N. Deshkar, S. B. Dhale, J. S. Mukherjee, T. S. Babu, and N. Rajasekar,
“Solar PV array reconfiguration under partial shading conditions for
maximum power extraction using genetic algorithm,” Renewable Sustain. Qiang Yang (Senior Member, IEEE) received the
Energy Rev., vol. 43, pp. 102–110, 2015. B.S. degree (first class Hons.) in electrical engineer-
[18] D. Yousri, T. S. Babu, S. Mirjalili, N. Rajasekar, and M. A. Elaziz, “A ing and the M.Sc. (with distinction) and Ph.D. degrees
novel objective function with artificial ecosystem-based optimization for in electronic engineering and computer science from
relieving the mismatching power loss of large-scale photovoltaic array,” Queen Mary College, University of London, London,
Energy Convers. Manage., vol. 225, 2020, Art. no. 113385. U.K., in 2001, 2003, and 2007, respectively.
[19] D. Yousri, S. B. Thanikanti, K. Balasubramanian, A. Osama, and A. Fathy, He was a Postdoctoral Research Associate with the
“Multi-objective grey wolf optimizer for optimal design of switching Department of Electrical and Electronic Engineer-
matrix for shaded PV array dynamic reconfiguration,” IEEE Access, vol. 8, ing, Imperial College London, London, U.K., from
pp. 159931–159946, 2020. 2007 to 2010 and has been involved in a number of
[20] M. Karakose, M. Baygin, N. Baygin, K. Murat, and E. Akin, “An intelligent high-profile U.K. EPSRC and European IST research
reconfiguration approach based on fuzzy partitioning in PV arrays,” in projects. He visited University of British Columbia and University of Victoria
Proc. IEEE Int. Symp. Innov. Intell. Syst. Appl., 2014, pp. 356–360. Canada as a Visiting Scholar in 2015 and 2016. He is currently a Full Professor
[21] K. Ş. Parlak and M. Karaköse, “An efficient reconfiguration method based with the College of Electrical Engineering, Zhejiang University, Hangzhou,
on standard deviation for series and parallel connected PV arrays, ” in China, and has authored/coauthored more than 200 technical papers, applied
Proc. Int. Conf. Renewable Energy Res. Appl., 2014, pp. 457–461. 60 national patents, coauthored two books and several book chapters. His
[22] R. Venkateswari and N. Rajasekar, “Power enhancement of PV system via research interests include communication networks, smart energy systems, and
physical array reconfiguration based Lo Shu technique,” Energy Convers. large-scale complex network modeling, control, and optimization.
Manage., vol. 215, 2020, Art. no. 112885. Prof. Yang is the Fellow of British Computer Society as well as the Senior
[23] A. Srinivasan, S. Devakirubakaran, and B. M. Sundaram, “Mitigation of Member of China Computer Federation.
mismatch losses in solar PV system—Two-step reconfiguration approach,”
Sol. Energy, vol. 206, pp. 640–654, 2020.
[24] A. M. Ajmal, V. K. Ramachandaramurthy, A. Naderipour, and J. B.
Ekanayake, “Comparative analysis of two-step GA-based PV array re-
configuration technique and other reconfiguration techniques,” Energy
Convers. Manage., vol. 230, 2021, Art. no. 113806.
[25] P. Srinivasa Rao, G. Saravana Ilango, and C. Nagamani, “Maximum
power from PV arrays using a fixed configuration under different shading
conditions,” IEEE J. Photovolt., vol. 4, no. 2, pp. 679–686, Mar. 2014.
[26] X. Qing, H. Sun, X. Feng, and C. Y. Chung, “Submodule-based modeling Wenjun Yan (Member, IEEE) received the M.Sc. and
and simulation of a series-parallel photovoltaic array under mismatch Ph.D. degrees from Zhejiang University, Hangzhou,
conditions,” IEEE J. Photovolt., vol. 7, no. 6, pp. 1731–1739, Nov. 2017. China, in 1991 and 1994, respectively.
[27] A. A. Mansur, M. R. Amin, and K. K. Islam, “Determination of module He is currently a Full Professor with the Col-
rearrangement techniques for non-uniformly aged PV arrays with SP, TCT, lege of Electrical Engineering, Zhejiang University.
BL and HC configurations for maximum power output,” in Proc. Int. Conf. His research interests include robust control, fault
Elect., Comput. Commun. Eng., 2019, pp. 1–5. diagnosis of hybrid systems, home service robotics,
[28] J. Wang, G. Jin, Y. Wang, and X. Chen, “Genetic simulated annealing and coordinated dispatch for electricity and heat in
algorithm used for PID parameters optimization,” in Proc. Int. Conf. Artif. microgrid.
Intell. Comput. Intell., 2009, pp. 397–401.

Authorized licensed use limited to: Shakuntla Educational & Welfare Society Galgotias. Downloaded on July 24,2022 at 06:09:02 UTC from IEEE Xplore. Restrictions apply.