Analysis And Prediction Of Pure Electric Bus Driving Energy Consumption Based On Data

Accurate driving energy consumption prediction can solve the problem of driver ’s " mileage anxiety." It is also the basis for reasonably arranging the charging time of pure electric vehicles and realizing the optimal distribution of charging facilities.It can also avoid the difficulty of grid dispatching caused by the random charging behavior of pure electric vehicles.Based on a large number of driving data of pure electric buses,this paper analyzes and predicts the influencing factors of driving energy consumption.In this dissertation,random forest model and multiple linear regression model are established,which have good prediction accuracy in the extracted features.The effects of vehicle factors,environmental factors,driving behavior and traffic conditions on the energy consumption prediction of pure electric buses are studied.The results show that vehicle-related factors,driving behavior and environmental factors have a greater impact on the prediction of driving energy consumption of pure electric buses,and traffic conditions have less impact.Finally,the running time analysis,k-fold cross validation and robustness analysis of the best model are carried out.Firstly,data mining is carried out on the historical driving data of pure electric buses.Through the combination of bus status information and parking time,the driving data is segmented.Cluster analysis was used to screen 32889 driving segments.According to the extracted velocity and acceleration characteristics,five motion states are divided.Mining weather data on weather websites based on location information and time information.Secondly,the potential relationship between multiple influencing factors and driving energy consumption is explored.The physical model and data model of energy consumption of pure electric bus are analyzed.The influence of vehicle factors,environmental factors,driving behavior and traffic conditions on the driving energy consumption of pure electric vehicles is analyzed.Thirdly,the construction of multiple linear regression model and random forest model is completed.Twenty feature quantities related to the vehicle itself,environment and driving behavior are extracted as input.Multiple correlation coefficient analysis,t test and F test were performed on the multiple linear regression model.The two main parameters of ntree and mtry are debugged for the random forest model.Finally,the influence of influencing factors on the prediction accuracy of the model is explored.The model prediction results show that considering all the influencing factors can significantly improve the prediction performance,but relatively speaking,the vehicle ’s own influencing factors contribute the most to the prediction performance of the model.The study also found that the contribution of environmental factor variables and driving behavior factor variables to the improvement of model prediction accuracy is second only to the vehicle ’s own factor variables,and the traffic information factor variables have less influence.In addition,the five operating conditions show that the average energy consumption per kilometer is the lowest when the bus running mileage is 15 km to 25 km,the speed is greater than 25 km / h,and the temperature is 15 ℃ to 25 ℃.