Driving Style Analysis And Energy Consumption Prediction Of New Energy Buses Based On Driving Data

In recent years,pure electric vehicles have been supported by national policies,so the market size has been increasing.The main problem that restricts the development of pure electric vehicles at this stage is the lack of range.Therefore,reducing the energy consumption of pure electric vehicles can improve the range.Studies have shown that driver habits and behaviors have a significant impact on vehicle energy consumption.However,the previous research methods of driver behavior analysis have the defects of overly ideal data sets,limited by data sampling frequency,and insufficient experimental validation.In summary,this paper proposes a theoretical framework and implementation method for driver behavior analysis and energy consumption prediction.The main work and innovation points of this paper are as follows.1.This paper proposes a method for bus driver energy consumption prediction.First,this paper establishes the data set for performing energy consumption prediction by means of data processing such as outlier processing,data travel cut,signal correlation analysis,and feature construction.Then,this paper uses random forest and XGBoost methods to establish a prediction model of driving behavior energy consumption through grid tuning reference,index evaluation,and model comparison.At the same time,this paper uses variable importance scores to select driving behavior features in the model that have a large impact on energy consumption,and analyzes the relationship between behavior feature parameters and energy consumption.The experimental results verified in 20 buses show that the energy consumption prediction model has a good prediction effect on vehicle energy consumption with a root mean square error of 0.042.The relationship between different driving behavior features and energy consumption varies,and the accelerator pedal opening degree has a linear relationship with motor energy consumption within a certain range.2.This paper proposes a method for bus driver behavior analysis.First,this paper establishes a dataset for performing driving style classification through feature extraction,data normalization processing,and clustering.Then,this paper proposes a driving style classification model based on K-means clustering,spectral clustering,and hierarchical clustering methods.This paper combines the results of driving style classification with indepth analysis of the differences in driving behavior characteristics of different driving styles and energy consumption differences,and constructs a direct relationship between driving behavior characteristics and energy consumption.The experimental results show that the driving style classification model effectively distinguishes drivers with different driving behavior tendencies,and the maximum contour coefficient of clustering can reach0.4886.The distribution of driving behavior under different styles has large differences,and the driving behavior of rapid acceleration and deceleration is the driving behavior factor that causes high energy consumption of electric vehicles.3.The innovation of this study is the proposed method of driver behavior analysis and energy consumption prediction under the condition of large sampling interval.Through feature construction,feature normalization and other data processing methods,this paper mines the driving behavior information in the driving data records with very large sampling intervals to achieve effective energy consumption prediction and driving behavior analysis.Moreover,this paper combines energy consumption prediction model and driving style classification model to deeply analyze the relationship between driving behavior characteristics and energy consumption,which is exploratory and practical.The significance of this research is to provide a theoretical basis for bus manufacturers to provide personalized driving assistance to drivers and a scientific basis for bus operators to evaluate different drivers.