Research On Driving Behavior Of New Energy Buses Based On Machine Learning

In the past decades,there has been a significant increase in extreme weather worldwide.In response to the challenges posed by global climate change,energy conservation,emission reduction and green travel have become the consensus of all sectors.New energy vehicles are gradually occupying the traditional motor vehicle market with their advantages of energy saving and environmental protection.In particular,the city bus system also gradually tends to use more new energy vehicles.After the completion of the bus system upgrade iteration,to ensure the safety and comfort of the new energy bus operation process has become an urgent issue.This thesis uses machine learning algorithms to build a model based on the operational data of Beijing 51 new energy buses to identify abnormal driving behaviors that pose safety risks or affect passenger feelings experience and vehicle conditions.The specific work is as follows:Firstly,this thesis analyzes the characteristics of new energy bus operation data and carries out data pre-processing according to its characteristics,mainly including distance segmentation based on GeoHash algorithm,identification of longitude and latitude outliers based on isolated forest algorithm,and filling of missing values based on SG filtering method.Meanwhile,this thesis constructs an evaluation index system for driving behavior of new energy buses,including speeding tendency factor,variable speed tendency factor and vehicle state factor.Secondly,this thesis constructs aggregation indicators to evaluate the driving behavior of new energy buses based on all the instantaneous data on each road section.Five different models,including evaluation models and clustering models,are constructed based on the aggregated indicators.The results of each model are analyzed in conjunction with the definition of driving behavior in this thesis and found that all five types of models can classify the driving behavior of new energy buses into abnormal driving behavior and excellent driving behavior.Because these five models meet the conditions of similar function and small homogeneity,this thesis uses the voting method to construct the integrated model.The experimental results show that the classification effect of the integrated model is better than each single model.Finally,in order to improve the real-time performance of the above integrated model,this thesis constructs a classification model for the driving behavior of new energy buses based on the instantaneous data before aggregation with the help of the results of the integrated model.The algorithms used in the model mainly include random forest algorithm,XGBoost algorithm,LightGBM algorithm,logistic regression model,and plain Bayesian algorithm.Comparing the prediction results of each model,it can be seen that the LightGBM model has the best overall performance and can quickly identify abnormal driving behavior based on real-time data,which increases the practicality of the prediction model.In summary,this thesis uses unsupervised machine learning algorithms to classify the driving behavior of new energy buses.The supervised machine learning algorithm is further used to improve the timeliness and applicability of the model,so that abnormal driving behavior can be identified quickly.It provides technical support to realize real-time warning and ensure the safety and comfort during operation.