Abstract:

This paper aims to study the time-varying configuration energy management strategy for hybrid distributed drive heavy vehicles based on real vehicle test data. Firstly， based on the real vehicle test data and by using the dynamic programming algorithm， the chassis configuration with optimal energy consumption is found， and the long short-term memory （LSTM） neural network is trained to complete the configuration optimization in three typical scenes. Then， the RULE_LSTM algorithm is proposed based on rule judgment. Its matching accuracy is 11.76% higher than that using LSTM network configuration with optimal energy consumption， and its frequency of configuration switching is reduced by 33.3%. Next， based on traffic flow the prediction on long-scale operating condition information is completed and the optimal chassis configuration matching and reference SOC trajectory generation are fulfilled. Based on radial basis function neural network， a short-scale operating condition prediction sequence is generated as the input of the subsequent algorithm. Finally， time-varying configuration is adopted to optimize the control variables， meanwhile the strong speed change rate constraint and SOC reference trajectory guidance are also introduced to implement the guided multi-APU predictive energy management strategy. The results show that with above-mentioned measures taken， the fuel consumption is reduced by 10.60%， 3.95%， and 2.06% respectively.

Key words: hybrid distributed drive heavy vehicles, energy consumption optimization, chassis configuration, energy management strategy