The speed of a hybrid electric vehicle is a critical factor that affects its energy management performance. In this study, we focus on the importance of solving the problem of inaccurate speed prediction in the energy management strategy (EMS) and application of dynamic programming (DP) needs to know the entire driving cycle. A gated recurrent unit neural network (GRU-NN) speed predictive model based on machine learning is developed by using the model predictive control (MPC) framework and solved in the prediction domain by employing DP. The neural network is trained on the training set, which is a collection of standard driving cycles. The results are compared with other two types of speed predictive models to verify the effects of different parameters of different speed predictive models on the state of charge and fuel consumption under Urban Dynamometer Driving Schedule driving cycle. Simulation shows that MPC based on the GRU-NN speed predictive model can effectively improve the fuel economy of hybrid electric vehicles, with a 94.14% fuel economy, which proves its application potential. Finally, the GRU-NN speed predictive model is applied under the Real-World Driving Cycle, whose fuel consumption has a fuel economy of 91.95% compared with that of the original rule-based EMS.

Accidents caused by the failure of high-voltage power battery systems are rising with the increase of pure electric commercial vehicles. The fault tree analysis method based on traditional reliability is no longer suitable for quantitative evaluation of polymorphic systems. In this paper, the polymorphic fuzzy fault tree of the high-voltage power battery system for pure electric commercial vehicles is established and analyzed qualitatively and quantitatively based on a combined theory of the polymorphic theory, fuzzy mathematical theory, group decision theory, and fault tree analysis theory. The results showed that the multistate reliability-analysis method of the fuzzy fault tree could describe the various fault states of the high-voltage power battery system. Through quantitative evaluation of the reliability of system, the low-temperature environment and CAN high and low reverse connection were the weakest links of the system, and the problem of the occurrence probability of each state of the unknown polymorphic bottom event in the sub-fault tree of the deteriorated-state mode was solved quickly using group decision-making to deal with fuzzy probability. It provides theoretical reference for system design and detection process, which has important practical significance for the improvement of high-voltage power battery system.