Smart Charging of EV Using DRL for Peak Load Minimization in Microgrids

Abstract

As the use of Electric Vehicles (EVs) is on the rise, the issue of high-density charging loads impacting microgrid stability is on the horizon. The “Peak-on-Peak” issue, where the evening peak demand meets the decrease in solar generation, is addressed in this research work by developing a Deep Reinforcement Learning (DRL) management system for a 9x9 solar-powered microgrid. The unmanaged EV charging pattern in the 81-node microgrid leads to severe frequency instability and voltage drops of up to 12%. The AI-based system reschedules EV charging patterns to align with peak solar generation. Simulation results demonstrate that the intelligent management system reduces peak grid demand by 32% and lowers consumer electricity costs by 21%, with grid voltage stabilized within a very small margin of ±2.5%.