MOGNN-EC: A Multi-Objective Optimization Framework for Efficient Energy Management and Charging Coordination in Electric Vehicle Networks

Abstract

The fast growth of electric vehicles (EVs) raises imminent challenges in energy management, charging scheduling, and grid resilience. Traditional rule-based scheduling, reinforcement learning, and metaheuristic techniques suffer from scalability, flexibility, and real-time decision-making issues. This paper introduces MOGNN-EC, a graph neural network (GNN) approach to multi-objective optimization for intelligent EV energy management and scheduling. MOGNN-EC applies EVs, charging points, and power supplies in the form of a dynamic graph to exploit spatiotemporal interdependences and achieve real-time optimization of energy distribution, load balancing, and renewable usage. Experimental evaluations indicate 92% energy efficiency, 75% use of renewables, and 0.4 s computation time for each round of optimization, along with a minimized mean absolute error (MAE) of 2.8 for demand forecasts. MOGNN-EC also optimizes battery thermal management via waste heat recovery in low-temperature climates. Scalar issues in scaling up to voluminous EV networks, e.g., computational intensity and availability of data, could be addressed using federated learning, uncertainty-aware multi-objective optimization, and vehicle-to-grid (V2G) operations. Further work will expand MOGNN-EC in multi-agent reinforcement learning (MARL) for cooperative sharing of energies and fault-tolerant real-time forecast operations. The evaluations prove MOGNN-EC to form an extensible, adjustable, and sustainable remedy for future EV energy infrastructures and intelligent grids.