Reinforcement Learning Driven Volt-Var Control with PV Inverters and Solid-State Transformer-Enabled EV Charging Stations

The growing proliferation of renewable energy resources and the rising acquisition of electric vehicles (EVs) have considerably increased the concern about volt-var control (VVC) in active distribution networks. Innovative approaches like photovoltaic (PV) inverters and EV charging stations (EVCS) based on a solid-state transformer (SST), which can provide reactive power adjustment, help stabilize the grid voltage dynamics. However, conventional VVC approaches incorporating slowresponding devices lack adaptability to react efficiently against dynamic environmental conditions, limiting their performance to balance the mitigation of voltage violations and the reduction of power losses. Along with that, conflicting mathematical objectives further complicate the application of deep reinforcement learning (DRL)-based optimization. To address these challenges, an innovative VVC framework is presented that employs SSTenabled EVCS capable of providing bidirectional flow and dynamic reactive power compensation, alongside smart PV inverters to improve real-time adaptability. Furthermore, to optimize the proposed multi-objective problem efficiently, a novel two-critic DRL learning scheme has been used to handle distinct mathematical objectives effectively. The conceptualized approach has been implemented for the IEEE-33 bus system. The simulation results demonstrated that the proposed scheme achieved approximately 32.74 % and 31.19 % reductions in power losses for TC-DDPG and TC-SAC, respectively. Meanwhile, the voltage violation rate (VVR) nearly converges to zero. Hence, contributed to the optimized integration of distributed generations (DGs) and EVs in active distribution networks (ADNs).