Physics-informed data-driven modeling approach for commuting-oriented hybrid powertrain optimization

Abstract

The physics-informed data-driven approach is instructive in powertrain modeling with high-fidelity, which can be adopted for powertrain optimization to further improve vehicle performances. In this paper, a plug-in hybrid electric vehicle (PHEV) equipped with a novel powertrain is considered as the research object, and a series of vehicle bench experiments are carried out for operative data acquisition. With the collected data, the physics-informed data-driven models for engine fuel consumption and battery state of health (SOH) loss are proposed. The accuracy and superiority of the proposed model are demonstrated compared with other approaches. In order to realize the commuting-oriented powertrain optimization, the driving cycle representing the commuting scenario is reconstructed based on the collected traffic data in Beijing. The synthesized driving cycle generated by the proposed method can describe the practical application scenario in Beijing within the allowable error range. In addition, the non-dominated sorting genetic algorithm-II (NSGA-II) is adopted to further optimize the parameters for the hybrid powertrain according to vehicle performances. The multi-objective optimization is completed with the engine fuel consumption and battery state of health (SOH) loss as targets. The results indicate that the optimized configuration parameters can reduce fuel consumption by 4.85% and battery capacity loss by 3.7% compared with the original vehicle.