燃料电池汽车动力系统及能量管理策略研究进展

Research progress on powertrain and energy management strategy of fuel cell vehicle

动力系统是燃料电池汽车(FCV)的核心,可分为单一式系统与混合动力系统两大类,其中,将燃料电池与辅助电源相结合组成“电-电”混合动力系统,已成为业界主流。本文根据辅助电源类型的不同,提出3类FCV混动系统的构建方案,分别为燃料电池+动力电池方案、燃料电池+超级电容方案、燃料电池+动力电池+超级电容方案,并对各方案的优势和劣势进行比较。同时,本文综述了近年来国内外学者提出的面向FCV的代表性能量管理策略,从理论基础与求解方法的差异出发,将现有燃料电池汽车的能量管理策略分为3类:基于规则定义的策略、基于最优化方法的策略以及基于机器学习的策略,并总结了各类策略在最优性与实时性等方面的优势和劣势。其中,基于规则定义的策略最易实现,在工程应用中最为普遍,但无法实现性能最优;基于最优化方法的策略能够接近甚至达到理论最优,但存在计算量过大、计算耗时过长、实时性差等问题;以强化学习为代表的基于机器学习的策略有望在最优性与实时性之间实现理想的平衡,但目前还存在模型训练耗时长、试错代价高等困难,在实车应用层面还存在一定挑战。基于文献研究与分析,本文提出以下观点:1)以大功率燃料电池为核心的功率混合型系统是FCV混动系统的未来发展方向;2)必须进一步提升智能化程度,根据实际使用场景开发具有个性化的能量管理策略;3)亟需建立关于燃料电池汽车能量管理策略的综合评价体系。

The powertrain system is the core of fuel cell vehicle(FCV),which can be divided into two categories:the single power source system and hybrid powertrain system.Among them,the combination of fuel cell and auxiliary power source to form"electric-electric"hybrid system has become the mainstream.Three types of FCV hybrid systems were summarized according to different types of auxiliary power sources,namely fuel cell+battery,fuel cell+supercapacitor,and fuel cell+battery+supercapacitor,and the advantages and disadvantages of each type were compared.The representative energy management strategies for FCVs proposed in recent years were reviewed.Based on the differences in theoretical foundations and solution methods,the existing energy management strategies for fuel cell vehicles were divided into three categories:rule-based,optimization-based,and machine-learning-based strategies,and the pros and cons of each strategy in terms of optimality and real-time performance were summarized.Among them,the rule-based strategies are the easiest to implement and the most commonly used in engineering applications,but they cannot achieve optimal performance.The optimization-based strategies can approach or even reach the theoretical optimum,but there are problems such as excessive computation,long computation time,and poor real-time performance.The machine-learning based strategies,represented by reinforcement learning,are expected to achieve the ideal balance between optimality and real-time performance,but they still suffer from time-consuming model training and high trial-and-error costs.Therefore,there are still some challenges in practical vehicle applications.Based on literature research and analysis,this paper proposes the following perspectives.1)The systems with high-power fuel cells as the core are the future development direction of FCV hybrid systems.2)It is necessary to further improve the degree of intelligence and develop personalized energy management strategies based on actual usage scenarios.3)There is an urgent need to develop a comprehensive evaluation system for FCV energy management strategies.