Safety-Integrated Online Deep Reinforcement Learning for Mobile Energy Storage System Scheduling and Volt/VAR Control in Power Distribution Networks

In coupled power distribution and transportation (CPT) system, a joint scheduling framework for mobile energy storage systems (MESSs) Volt/VAR control (VVC) ensures reliable grid operations while supporting electric vehicle loads at charging stations (EVCSs). However, conventional model-based optimization methods MESS VVC may yield suboptimal solutions greater computation times because of operation in uncertain environment CPT systems. To resolve this issue, study proposes model-free deep reinforcement learning (DRL) framework. framework, smart inverters MESSs solar photovoltaic (PV) cooperate to minimize the real loss mitigate violations both MESSs’ state charge (SOC) voltage network, travel via network satisfy EV EVCSs. A routing algorithm based on Dijkstra’s is developed determine optimal destinations MESSs. addition, two safety modules are ensure that neither SOC nor occur by adjusting and/or reactive PV during training process. The integrated into proposed DRL wherein agent performs desired through safe exploration procedure. approach tested IEEE 33-bus 15-node 57-bus 42-node Numerical examples demonstrate advantages terms convergence, loss, SOC/voltage violation.