Toward High-Fidelity Datasets for Hybrid Electric Vehicle Energy Management Incorporating Power Converter Non-idealities

초록

Reliable energy management strategies (EMS) for hybrid electric vehicles (HEVs) increasingly depend on high-quality datasets that accurately represent real-world operating conditions. However, most existing datasets are generated using simplified system models that neglect power converter non-idealities and their interaction with vehicle dynamics, leading to biased training and unreliable performance evaluation of EMS algorithms. This paper presents a high-fidelity dataset generation framework that integrates detailed vehicle longitudinal dynamics, hybrid energy storage system (HESS) models, and comprehensive DC-DC converter loss characteristics. The HEV architecture consists of a battery and a supercapacitor (SC), and datasets are generated through large-scale simulations executed on a high-performance computing server. The framework produces time-series datasets capturing realistic power flows, current stress, and converter efficiency under representative urban driving cycles. Dataset fidelity is assessed through statistical coverage of operating conditions, transient power behavior, and consistency between mechanical demand and electrical response. Results demonstrate that incorporating converter non-idealities significantly alters power distribution and efficiency metrics, establishing a benchmark-quality foundation for data-driven and model-based EMS development.

키워드