Designing Hybrid Vehicle Architectures: Utilizing an Automatic Generation and Optimization Approach

Different architectures for hybrid powertrains exist. Traditionally, to compare their efficiency, one energetic model is developed per architecture and is optimized. Today, this heuristic approach can be replaced by a systematic method that explores all the possibilities in a defined search space. This article presents hybrid electric vehicle (HEV) optimization and comparison using an automatic methodology for architecture generation. The components of the series-parallel hybrid architecture, with the exception of planetary gears, are chosen as starting components. All possible architectures are generated and then filtered to determine the most promising ones. The latter are automatically assessed and optimized using a general hybrid model. The process includes a genetic algorithm (GA) for design optimization and dynamic programming (DP) for control optimization. Given two different objectives to minimize, namely, battery size and fuel consumption in our considered case, the Pareto fronts of the most promising architectures are presented and compared to the Pareto fronts of reference series-parallel architectures. Several of the filtered architectures are found to be better than the references (their consumption is roughly 5% lower), and a gain in powertrain sizing and cost is found. The best architecture is explored, and its operation is analyzed to determine the functionalities that lead to its energy efficiency.