A voltage and reactive power control method is used in substations to control transformer taps and shunt capacitors/reactors in order to maintain the voltage of a power system. It becomes difficult with the conventional method to ensure the voltage within stipulated band for a power system with a massive penetration of photovoltaic (PV) generation systems. In order to address this issue, the authors have first used machine learning to estimate the voltages at the consumer end and then proposed a voltage and reactive power control method using deep reinforcement learning that considers the voltage at all consumer ends. The effectiveness of the proposed method is studied through simulations performed on an example system with a massive PV and the results are found to be quite promising.
J-STAGEがリニューアルされました！https://www.jstage.jst.go.jp/browse/-char/ja/
