Intelligent Energy Management for Fuel Cell Bus Based on Enhanced Soft Actor-Critic Algorithm

Energy management strategies (EMSs) directly affect the fuel economy of hybrid electric vehicles, and deep reinforcement learning (DRL) has become the mainstream method for energy management in recent years. This paper proposes an intelligent DRL-based EMS for an urban fuel cell bus (FCB) powered by both fuel cell and battery to improve the energy efficiency of the FCB. To begin, an enhanced soft actor-critic (SAC) algorithm integrating the standard SAC algorithm with the prioritized experience replay (PER) mechanism is innovatively formulated. Then, an intelligent EMS based on the enhanced SAC algorithm is developed. After that, the superiority of the proposed EMS is evaluated in detail by comprehensive comparisons. Simulation results indicate that the proposed EMS accelerates the convergence speed by 62.79% while improving the fuel economy by 5.18% in comparison with the EMS based on standard SAC.

I. Introduction

With the increasing concern of energy crisis and traffic pressure, electrification has become the main trend of the automobile industry, especially the urban public transportation sector [1]. The fuel cell bus (FCB) powered by fuel cell and lithium battery enjoys both long-distance traveling and zero-emission, providing an effective solution for the sustainable development of urban public transport [2]. To coordinate the fuel cell and the battery of an FCB to work efficiently, effective energy management strategies (EMSs) are needed to improve the fuel economy of the FCB.