Braking energy management strategy for electric vehicles based on working condition prediction

To improve the mileage capacity of electric vehicles (EVs), a dual-motor front-wheel-drive EV is considered as the research object. Through experiments with actual vehicles, data from four typical working conditions are collected; a C4.5 decision tree algorithm is developed to train a working condition recognition model. The long short termmemory neural network is used to train four deep-learning working condition prediction models, and the particleswarm algorithm is used to optimize their structural parameters. The braking strength, demand torque, and demand speed are determined based on the predicted working conditions. Based on four common braking energy recovery control strategies, front- and rear-wheel braking force distribution strategies are formulated according to the changes in braking strength. The maximum regenerative braking torque and remaining mechanical braking torque provided by the front wheels are optimized. The Seagull Optimization Algorithm is used to optimize the torque distribution of the dual motors on the front wheels and improve the working efficiency of the motors. In the experimental conditions, the recovered energy at 100 km is 2.6 kWh; the energy recovery rate is 19.1%, and the power consumption ratio is reduced by 15.8%, improving the EV cruising range.