A FPPT Scheme Based on Reinforcement Learning in Photovoltaic Power Systems

For Photovoltaic (PV) power systems to capture the maximum power from solar energy, the maximum power point tracking (MPPT) is necessary. Since the intermittent nature of solar energy and the ever-changing environment, MPPT may cause power and voltage fluctuations which may cause problems for grid stability. Furthermore, the fluctuation may increase the PV converter's heat load and reduce its service life. To limit power fluctuations, a flexible power point tracking (FPPT) scheme is proposed. As a result, the system is more dependable. This article proposes a FPPT scheme based on reinforcement learning (RL) to increase the FPPT performance in terms of high speed and precision. RL makes the system hardware structure simpler and enables faster tracking speed and higher accuracy than traditional methods. The open source OpenAI Gym platform is applied to construct the PV environment. The constructed simulation environment is applied to conduct experiments to evaluate the proposed control scheme's robustness in various environmental scenarios. The simulations results show the feasibility of the scheme with fast response and stable performance.