Design and Evaluation of PPO-Based Coordinated P–Q Control for a Smart-Inverter BESS in PV–EV Distribution Networks

Phase-asymmetric electric vehicle (EV) charging can aggravate voltage unbalance in radial distribution feeders, while a smart-inverter battery energy storage system (BESS) can provide both energy shifting and fast reactive-power support. This study designs and evaluates a proximal policy optimization (PPO)-based controller for coordinated active- and reactive-power dispatch under asymmetric EV demand. The analysis uses a corrected phase-decoupled three-phase extension of the IEEE 33-bus feeder, with voltage unbalance quantified using the negative-to-positive sequence voltage ratio. Active-power-only (P-only) and coordinated P–Q controllers are evaluated using identical BESS hardware, operating limits, and terminal state-of-charge (SOC) requirements, enabling the incremental contribution of inverter reactive-power capability to be assessed under matched conditions. Across three independent training seeds, P–Q control reduces daily feeder loss from 3874.822 to 3809.751 kWh/day, worst voltage unbalance factor (VUF) from 2.7879% to 2.5940%, and mean hourly maximum VUF from 1.5422% to 1.3628%, corresponding to reductions of 1.679%, 6.954%, and 11.632%, respectively, relative to the no-BESS case. The corresponding P-only reductions are 0.610%, 4.362%, and 8.427%, respectively. All reported learned-policy evaluations return the BESS to 50% terminal SOC, preventing the comparison from being biased by net battery depletion. Without retraining, the P–Q policies retain approximately 5.66–7.86% mean worst-VUF reduction under the tested load, photovoltaic, EV demand, and combined-profile perturbations. The results demonstrate repeatable voltage unbalance mitigation and show that coordinated P–Q operation provides an incremental benefit beyond active-power-only control under the investigated conditions. However, residual undervoltage under severe loading and the phase-decoupled feeder representation limit the scope of the conclusions.