Bi-objective reinforcement learning for PV array dynamic reconfiguration under moving clouds
File(s)
Accepted version
Author(s)
Liu, Chen
Wang, Yu
Cui, Qiushi
Pal, Bikash
Type
Journal Article
Abstract
During the daily operation of photovoltaic (PV) array, it often encounters partial shading conditions, primarily caused by varying degrees of obstruction from moving clouds. One favored method for alleviating the negative impacts of partial shading conditions (PSC) is the reconfiguration of PV arrays. Nevertheless, the conventional method of PV array reconfiguration primarily focuses on maximizing power output, neglecting the consideration of switching device lifetime and the complexity of control involved. This paper introduces a bi-objective reinforcement learning approach for the dynamic reconfiguration of PV arrays in the presence of moving clouds. The objective is to maximize power output while minimizing the number of switches. To validate the efficacy of this approach, two distinct models of moving cloud scenarios were constructed, and simulations were conducted using a 10 × 10 PV array and a 15 × 15 PV array. Then, we compare it with other candidate algorithms. We found that the maximum power capacity achieved by the BiPPO algorithm is 22.62 % higher than other algorithm. Meanwhile, compared to the other algorithms, BiPPO significantly decrease the frequency of PV array reconfiguration by 1.32 to 1.76 times at different time intervals. Through these comparative studies, we verify the effectiveness of the proposed approach.
Date Issued
2025-08-01
Date Acceptance
2025-02-25
Citation
Electric Power Systems Research, 2025, 245
ISSN
0378-7796
Publisher
Elsevier
Journal / Book Title
Electric Power Systems Research
Volume
245
Copyright Statement
© 2025 Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Subjects
Bi-objective reinforcement learning
Engineering
Engineering, Electrical & Electronic
Pareto optimization
Partial shadings
POWER
PV array reconfiguration
SCHEME
Science & Technology
Technology
Publication Status
Published
Article Number
111579
Date Publish Online
2025-03-07