Reinforcement Learning-Based Controller Parameter Optimization for Photovoltaic Inverters

With the increasing integration of new energy generation, the study of control technologies for photovoltaic (PV) inverters has gained increasing attention, as they have a significant impact on the voltage stability of the entire power grid. Traditional methods for designing inverter control parameters suffer from the drawbacks of cumbersome optimization processes and suboptimal control performance. To address these challenges, this paper proposes a novel reinforcement learning-based algorithm for PV inverter parameter optimization. The algorithm incorporates dynamic voltage performance metrics as rewards and leverages deep neural network functions to learn from empirical data, enabling online self-tuning and parameter optimization. The aim is to enhance the voltage stability of inverters at grid connection points. To demonstrate the effectiveness of the proposed approach, we present a case study on a virtual synchronous generator, optimizing the integral coefficient in the control system using the proposed algorithm. Experimental results reveal that, compared to traditional parameter tuning methods, the proposed algorithm is able to eliminate the need for laborious manual tuning, effectively optimizes controller parameters, and thus enhances the dynamic response performance of the controller.