Combined algorithm improves power generation by mitigation partial shading conditions
Combined algorithm improves power generation by mitigation partial shading conditions lead image
The uneven irradiation that occurs when photovoltaic solar cells operate under partial shading conditions (PSCs) can curb their effectiveness due to multiple peaks in the power-voltage curve and resulting power loss. New approaches have shown promise in addressing this power mismatch loss and enhancing overall solar energy utilization by applying reinforcement learning algorithms to combined hybrid photovoltaic-thermoelectric generation (PV-TEG) systems.
Zhou et al. proposed a new reinforcement learning approach that provides a highly adaptable, dynamic optimization process that uses the strengths of photovoltaics and thermoelectric generation to mitigate the effects of PSCs. Combining a value-based approach known as Q-learning with a self-checking algorithm called advantage actor-critic, the technique, called cooperative Q-learning and advantage actor-critic (C-Q-A2C), can enhance solar energy utilization with the inherent temperature of the photovoltaic system and optimize the electrical connections between the two modules of PV-TEG systems.
“The innovation of C-Q-A2C lies in its collaborative framework. Unlike traditional loosely coupled algorithms, it deeply integrates Q-learning and A2C, where Q-learning’s value estimation directly guides A2C’s policy gradient updates,” said author Lei Zhou. “Heuristic exploration principles are embedded in the reinforcement learning agent’s action selection to dynamically guide optimization and escape local optima, forming a unified adaptive framework.”
The group tested the hybrid algorithm on simulations of 9x9 and 15x9 arrangements of cells, as well as a 4x4 experimental system, demonstrating increases of maximum power output and outperforming other algorithms in the paper.
Over 20 trials, deviations from optimal output were below 3%, confirming stability across dynamic shading scenarios.
The group hopes their work stokes further interest in developing dynamic multi-objective optimization frameworks for renewable energy.
Source: “Optimal hybrid PV-TEG systems reconfigurations for effective mitigation of partial shading conditions via cooperative Q-learning and advantage actor-critic algorithm,” by Lei Zhou, Bo Yang, Chuanyun Tang, Zijian Zhang, Jiale Li, Zhenning Pan, Hai Lu, Hongbiao Li, Dengke Gao, and Lin Jiang, Journal of Renewable and Sustainable Energy (2025). The article can be accessed at https://doi.org/10.1063/5.0278532 .
