FEDERATED REINFORCEMENT LEARNING FOR INTELLIGENT ELECTRIC VEHICLE CHARGING MANAGEMENT

Abstract

The Intelligent Electric Vehicle (EV) Charging Management System using Federated Reinforcement Learning (FRL) is a privacy-preserving and AI-based solution for optimizing electric vehicle charging. The system combines Reinforcement Learning (RL) to make intelligent charging and Vehicle-to-Grid (V2G) decisions with Federated Learning (FL), enabling multiple EVs to collaboratively improve their charging models without sharing sensitive user data. Each EV learns from its local charging history based on factors such as battery State of Charge (SoC), electricity prices, departure time, and user preferences. A central server aggregates only the trained model parameters using the Federated Averaging (FedAvg) algorithm to create an improved global model. Developed using Flask, PyTorch, MySQL, and modern web technologies, the system provides smart charging recommendations, realtime battery monitoring, and grid management features. The proposed approach reduces charging costs, protects user privacy, improves grid stability, and supports efficient energy management for future smart transportation systems