Hierarchic Multi-Agent Energy Management for Extended Driving Range Through Battery Cell Balancing

Battery Electric Vehicles (BEVs) are being researched as a viable method to mitigate transportation-related pollution. However, a principal obstacle hindering their extensive acceptance is the driving range, in which Internal Combustion Engine (ICE) vehicles maintain a competitive edge. This study presents a Hierarchical Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (HMATD3)-based Energy Management Strategy (EMS) to improve EMS efficiency in BEVs by utilizing the unique benefits of the TD3 algorithm to extend driving range in BEV. Reinforcement Learning (RL), a branch of Machine Learning (ML), has gained substantial attention due to its adaptability and self-learning capabilities. As a subset of RL, the TD3 algorithm has demonstrated remarkable effectiveness in optimizing various aspects of electric vehicle (EV) operations. Its ability to mitigate overestimation bias and enhance stability makes it a promising approach for energy management strategies. The suggested approach leverages the flexibility of RL-based EMS through ongoing interaction with the fluctuating driving environment, facilitating real-time optimization. The method employs a hierarchical multi-agent architecture to successfully tackle multi-objective optimization difficulties, enhancing energy distribution and battery management. The effectiveness of the suggested method is assessed in comparison to other RL-based approaches, Particle Swarm Optimization (PSO), and an Enhanced Fuzzy Logic-Based EMS. Simulation results indicate that the HMATD3-based EMS surpasses current techniques, significantly enhancing driving range and energy efficiency in BEVs.