A two-term energy management strategy of hybrid electric vehicles for power distribution and gear selection with intelligent state-of-charge reference

This paper presents a two-term energy management strategy (EMS) to obtain optimal power distribution with proper gear selection under intelligent state of charge (SOC) reference for parallel hybrid electric vehicles (HEV). A long-term SOC planning level utilizes dynamic programming (DP) to calculate SOC trajectories under many repetitive routes. Then, the characteristics of driving route and relatively SOC value from DP algorithm are respectively as input and output data to train artificial neural network to generate intelligent SOC reference planning model. In the short-term online optimization level, deep neural network model is built to forecast velocity sequence over each predictive horizon. According to route characteristics, this SOC reference model could be real-time gained for model predictive control (MPC) scheme as terminal SOC value in each prediction horizon. Moreover, based on the SOC constraint and predictive velocity, MPC is employed to achieve energy management online by DP optimization solver in combination with adjacently searching gear skill. Numerical simulations show that MPC with intelligent SOC reference planning and adjacently searching gear methods has yielded the desirable performance of the fuel economy compared with the fixed SOC constraint MPC. More importantly, inaccurately short-term speed prediction in real cycles indicating the favorable robustness of the proposed methods, which the adaptability is urgent for practical application.

Automated summary

This paper introduces a novel two-term energy management strategy and SOC reference model, achieving energy management and speed prediction optimization through deep neural networks and dynamic programming, resulting in desirable fuel economy performance.