Driving condition feature driven collaborative energy conversion and management strategy for electric commercial vehicles with the dual-motor coupled powertrain system

A widely adjustable powertrain system and efficient energy conversion strategy are key to expanding the operating conditions and applications of fuel cell commercial vehicles. However, under complex long time operating scenarios, fluctuating power demand and strongly coupled control variables make it difficult for empirical energy management strategy to ensure both efficiency and accuracy. To address this challenge, a driving condition feature driven collaborative energy management strategy is proposed for fuel cell hybrid buses with a dual-motor coupled powertrain, reformulating energy management as a unified hydrogen -electrical -mechanical conversion problem integrating offline optimization knowledge, temporal signal features, and experimentally calibrated mechanism constraints. Firstly, real-world vehicle signals are analyzed to establish a data-driven mechanism model by integrating the motor efficiency map, battery degradation model, fuel cell efficiency model, and mode shift cost matrix obtained. On this basis, the dynamic programming is used to generate optimal control sequences, while a one-dimensional convolutional neural network extracts driving condition features from multisource temporal signals. Then a multi-head strategy network is further trained to achieve real-time collaborative output. Experiments across four practical operating scenarios show that the proposed strategy achieves up to 97% of offline dynamic programming performance and reduces the average total cost by 9.82% compared with a rule-based strategy, while feature embedding improves mechanical-to-electrical and hydrogen-to-electrical conversion efficiencies by over 3% and 6.02%, respectively.