Abstract: In order to evaluate the energy consumption level of plug-in hybrid vehicles (PHEV) in actual driving process and study the essential factors affecting the energy consumption level of PHEV, this paper explores the relationship between the vehicle driving status and driving behavior and the overall energy consumption of PHEV based on the driving record data of 180 drivers. Firstly, principal component analysis (PCA) method is employed to decouple the characteristics of parameter matrix of driving conditions, and the first 5 PCs contributing accumulatively 84% to the overall information are extracted. The principle components are also defined according to the numerical distribution of the influence coefficient matrix. Then, based on kinematics fragments, the K-means algorithm is used to constrain the principal components in turn to form target real vehicle road conditions, which is put into vehicle power flow model to calculate the comprehensive energy consumption level. Finally, Pearson correlation coefficient and covariance value between each principle component and comprehensive energy consumption level are calculated and clustering and sensitivity definitions are given. The results show that the three significant factors representing human-vehicle-road extracted in this paper have a strong sensitivity relationship with energy consumption. The conclusions of this paper have important guiding significance for the selection of PHEV design and control parameters.

Key words: plug-in hybrid electric vehicle, sensitivity analysis, principle components analysis, cluster algorithm