Driving-cycle-recognition-based adaptive ECMS for fuel economy improvement in 48 V P0 mild hybrid electric vehicles

To mitigate the performance limitations imposed by the inherent drive-side electrical energy scarcity in 48 V P0 mild hybrid electric vehicles, this study proposes a Driving-Cycle-Recognition-Based Adaptive ECMS (CR-AECMS). It employs an offline-training/online-control framework. In the offline phase, PCA extracts representative driving features, and SOM clustering partitions driving cycles into four typical patterns; a multi-pattern EF library under global SOC-balance constraints is then built via GWO optimization across complete driving cycles. In the online phase, real-time pattern recognition leverages short-term historical speed data for dynamic EF mapping. A high-fidelity Model-in-the-Loop (MIL) platform for the 48 V P0 MHEV is developed and calibrated against empirical data. Based on this platform, a comprehensive benchmark evaluation compares RB, SOC-Feedback-Based Adaptive ECMS (S-AECMS), and CR-AECMS under both standardized cycles and a highly stochastic, independent RDE route. Compared with RB, CR-AECMS reduces equivalent fuel consumption by 1.7% under WLTC and 1.8% under CLTC-P. Crucially, it further delivers a notable 3.6% fuel-saving improvement under RDE while maintaining robust charge sustainability.