Research on Electric Vehicles Shift Strategy Adapted to Driving Habits

In the field of AMT shift strategy research, adapting the shift strategy to driving habits is a problem that has not been effectively addressed. This paper conducts a complementary study, which is as follows. A longitudinal motion model and a control model of electric vehicles are established by using Simulink. The economic shifting is taken as a prior driving behavior. The parameters such as vehicle velocity and accelerator pedal opening are collected from the driver's active shifting during the driving process. The parameters are used as 8-parameter input and 2-parameter input, respectively. Three neural networks, RNN, LSTM, and Bi-LSTM, are trained to derive six shift strategies and their performance is compared. The outcomes indicate that the Bi-LSTM neural network has significantly better performance than the other two neural networks in terms of the accuracy and loss function of the neural network. The performance of the shift strategy derived from the 8-parameter input is better than that of the shift strategy derived from the 2-parameter input, and the Bi-LSTM-8 shift strategy is the best performer. The Bi-LSTM-8 strategy can be well adapted to the driver's driving habits. In terms of power consumption, the Bi-LSTM-8 shifting strategy consumes an average of 2.13 kW·h more power per 100km than the actual driving, which is basically adapted to the driver's driving habits.