Data-Driven Energy Management for Series Hybrid Electric Tracked Vehicle

This paper proposes a data-driven energy management strategy (EMS) for a series hybrid electric tracked vehicle (SHETV). Firstly, according to the configuration characteristics of the SHETV powertrain, a simulation model for the development of EMSs is built. Secondly, combined with the design requirements, a global optimal EMS based on dynamic programming (DP) is developed. Then, the optimal control sequence is obtained and the NARX deep neural network is employed to extract the global optimal control rules and establish the mapping relationship between characteristic parameters and power allocation. Finally, the IC engine-generator power unit (IGPU) output power prediction model, battery state of charge (SOC) stabilizer, and low-pass filter are designed respectively, and the design of the data-driven EMS is completed. In order to verify the performance of the designed strategy, different driving cycles are used for offline training of the neural network and online verification of the effectiveness of the strategy. The simulation results show that the proposed EMS can effectively maintain the SOC of the battery and the fuel economy is improved by 10.89% compared with the EMS based on frequency domain power allocation.