Distribution Network Expansion-Friendly Adaptive Deep Reinforcement Learning for Inverter-Based Volt-Var Control

In recent years, in response to the challenges posed by the volatility and uncertainty of increasing renewable energy sources (RESs), numerous studies have emerged on inverter-based Volt-Var control (VVC) using data-driven deep reinforcement learning (DRL) methods. However, most of these methods assume fixed distribution network (DN) scales, which may not be applicable under potential DN expansion involving newly connected nodes and controllable devices. To address this issue, this letter proposes a novel DRL framework for inverter-based VVC capable of adapting to evolving DN environments. Specifically, state embedding for expanding state space (SE-ESS) and action branching for expanding action space (AB-EAS) are designed to facilitate model fine-tuning for the expansion of DN scales. Case studies on a modified IEEE 33-bus system validate the proposed method’s strong adaptability to DN expansion.