Route-Controlled Identification of Hybrid-Vehicle Battery Energy-Flow Regimes Using a Hidden Semi-Markov Model

This study proposes a Hidden Semi-Markov Model (HSMM)-based framework for identifying and comparing traction-battery energy-flow regimes in a hybrid electric vehicle under route-controlled real-world driving conditions. Vehicle operating data were acquired through the OBD-II interface during repeated ECO, NORMAL, and SPORT trips performed along the same road route. Instantaneous battery power was reconstructed from measured voltage and current, while empirical charge–discharge episode durations were first analyzed to assess the validity of the memoryless assumption inherent in conventional Hidden Markov Models. The resulting dwell-time distributions supported the use of an explicit-duration semi-Markov formulation. A six-state HSMM identified physically interpretable regimes corresponding to mild charging/coasting, EV propulsion, hybrid propulsion with electric assistance, stationary/low-power operation, strong regenerative braking, and ICE-supported battery charging. State occupancy, transition structure, dwell time, and episode-level energy reward were subsequently combined into a battery dynamic signature. Repeated NORMAL-mode trips exhibited highly reproducible transition structures despite substantial variability in their vehicle-speed profiles, while ECO and SPORT operation produced distinct redistributions of the same underlying regimes. The proposed framework demonstrates that hybrid-battery operation can be characterized more comprehensively through its temporal and energetic organization than by conventional trip-level indicators such as SOC change or total energy balance alone.