An Optimal Control Approach for Plug-In Electric Vehicles in Active Distribution Systems Using Deep Reinforcement Learning

The penetration of plug-in electric vehicles (PEVs) and distributed energy resources (DERs) is increasing in distribution systems, potentially leading to significant technical and economic challenges. To tackle these challenges, this paper introduces a novel framework for effectively managing DERs and EVs within active distribution systems (ADSs), incorporating time-varying ZIP load models. A deep reinforcement learning (DRL)-based control approach is developed that simultaneously optimises both technical and economic objective functions for the efficient operation of ADSs. For this purpose, the PEVs are integrated with different nodes of the ADS through solid-state transformers (SSTs). Based on available generation, load demand and EV charging profiles, the control algorithm regulates reactive power flow using SSTs and minimises the operational cost as well as power loss of the ADS. The proposed framework is successfully applied and evaluated on standard IEEE systems, demonstrating its efficacy in solving the problem of integrating PEVs and DERs using solid-state transformers.