Research on Intelligent Voltage Regulation Method Within a Station Based on Multi-Agent Reinforcement Learning

In response to the issues that existing voltage regulation methods cannot finely perceive the load data within the station area, have low regulation precision, and often use active power regulation leading to high electricity costs for users, a novel intelligent voltage regulation method within the station area based on multi-agent reinforcement learning is proposed. Firstly, a photovoltaic (PV) output prediction technology based on a hybrid neural network is proposed to predict the load for the previous day. Secondly, a strategy generation mechanism for voltage regulation within the station area that integrates active and reactive power control is designed, and based on this, an intelligent voltage regulation method within the station area using multi-agent reinforcement learning is proposed. Finally, the feasibility of this method is verified through the voltage regulation situation of the same station area at different times of the same day and the same times on different days. The experimental results show that the regulation algorithm can effectively achieve voltage regulation at the PV connection point and prevent voltage limit violations at the PV connection point.