The Analysis For Driving Energy Consumption And Driving Style Classification Based On Vehicle Operation Data

Compared with traditional fuel vehicles,battery electric vehicles(BEV)have prominent advantages such as low energy consumption and low pollution.In recent years,the scale of China's new energy automobile industry is constantly expanding,and the sales volume of automobiles is also constantly climbing.Therefore,the standards and restrictions,technical indicators and test methods of electric automobiles are also being updated with the continuous maturity of electric automobile industry technology.At the same time,the state subsidy scheme for electric vehicles is also increasingly stringent,so as to promote the development of electric vehicles to the market direction.The main problem affecting the development of BEV is the battery's endurance.By regulating driving behavior,the vehicle energy consumption can be reduced and the driving range can be improved.Meanwhile,the driving behavior analysis can verify the driving characteristics of different driving styles and the running characteristics of vehicles,providing a theoretical basis for the control strategy optimization of BEV.This paper took the daily operation data of a pure electric vehicle as the research object,and defined a total of 33 driving behavior characteristic parameters including the class of speed,speed change,pedal control and time proportion after data preprocessing.The data according to vehicle running state could be divided into starting,parking and operation phase,through the coefficient of maximum mutual information algorithm to calculate the relation between the characteristic parameters of 33.The characteristic parameters which had significant influence on the two parameters of total energy consumption and 100km energy consumption under different conditions were defined.And the results showed that driving behavior that significantly affects energy consumption varied at different stages,so driving behavior which could cause high energy consumption should be avoided.Took the maximum speed and average speed of operating data as the characteristic parameters of distinguishing driving condition,three driving conditions named low speed,medium speed and high speed were clustered by K-Means algorithm.The linear functions of the classification boundary of three kinds of driving conditions were fitted and the driving condition identification model was established.Before classification,the dimension reduction clustering algorithm accuracy of K-Means+t-SNE and t-SNE+K-Means algorithms were compared,and the results showed that t-SNE+K-Means was superior in the classification of driving behavior.The result of driving style classification indicated that the average characteristic parameters under different operation conditions were variant,the average of characteristic parameters of high speed was minimum while that of low speed was maximum.It suggested that the intensity of driving under low speed condition was greater than that under high speed condition,and it also proved the necessity of driving condition identification before driving style classification.Finally,the distribution characteristics of the accelerator pedal opening,brake pedal opening,steering wheel Angle,speed,motor speed,motor torque and 100km energy consumption of different driving styles were analyzed.Results showed that the operation characteristics under different style had a bigger difference,and provided theoretical basis for related technical personnel and the vehicle control strategy to upgrade.