Optimization of stack size, operating temperature, and battery capacity in fuel cell hybrid electric vehicles for durability and fuel economy under use-level conditions

• Fuel cell/battery hybrid systems are optimized for FCHEV durability and fuel economy. • Optimization is performed to find the optimal stack size and battery capacity. • Total cost analysis shows a distinct minimum at each stack operating temperature. • Total cost decreases at low stack temperature due to longer stack lifetime. In this work, a novel optimization framework for fuel cell hybrid electric vehicles is developed to facilitate custom design of optimal fuel cell stack size and battery capacity by accounting for fuel cell and battery degradation and hydrogen consumption. Validated durability models for catalyst and membrane degradation are used to simulate fuel cell lifetime under real-world transit bus operation in Victoria, B.C., Canada. Two energy management strategies (EMSs) are evaluated for fuel cell–battery power allocation: a rule-based approach and a deep reinforcement learning (DRL) strategy. The stack nominal power and battery capacity are then optimized separately at different operating temperatures using a multi-objective genetic algorithm considering fuel cell durability and fuel economy. The resulting Pareto fronts demonstrate a trade-off between fuel cell lifetime and hydrogen consumption. A total cost analysis is performed to identify the optimal design among all Pareto-front scenarios, yielding the minimum total annual cost. This analysis accounts for the capital and replacement costs of the fuel cell stack, battery pack, and thermal management system, and the hydrogen operating costs. Operating at lower temperatures results in an overall lower total cost due to significantly longer stack lifetimes and only minor compromises in fuel economy and heat generation. The optimal scenario at a fuel cell temperature of 70°C demonstrates a 4% reduction in total cost compared with the 80°C case. Although both EMSs achieve similar performance during optimization, the DRL-based approach demonstrates greater flexibility in adapting to changes in the initial state of charge and altered drive cycles.