Abstract: Compared with the traditional vehicles, the shorter driving mileage and longer charging time are two technology issues for electric vehicles. Hence, the accurate prediction of electric vehicle energy consumption has important significances for mitigating the driver's "range anxiety". Taking the actual operating of electric vehicle in Tianjin, the vehicle driving data are divided into several kinematic segments to analyze the related factors affecting the energy consumption of electric vehicles during driving period including the influence of driving state and operating condition on energy consumption and the influence of braking energy recovery on driving range. For improving the accuracy of energy consumption model, Markov chain Monte Carlo algorithm is applied to predict the curve of future driving cycle. Meanwhile, the neutral network is employed to identify the categories of driving cycles such as congestion driving cycle, city driving cycle and high-speed driving cycle. Then the significant features are extracted from the three driving cycles, respectively. The features are fed into XGBoost algorithm to construct energy consumption model for realizing accurate energy consumption prediction. The results indicate that the proposed method can effectively improve the accuracy of energy consumption compared with the traditional methods.

Key words: electric vehicle, kinematic segment, Markov chain Monte Carlo, energy consumption prediction