Research On Fuel Consumption Prediction Model Of Heavy Truck Based On Driving Style

The drive of heavy trucks is inseparable from fuel.With the rapid development of the domestic logistics industry,the number of heavy trucks is increasing,so the demand for fuel imports is also increasing.However,the rising international fuel prices have caused the transportation costs of logistics enterprises to continue to rise.Therefore,how to improve the fuel economy of heavy trucks has become an urgent problem for logistics enterprises.Fuel economy is related to road conditions,vehicle equipment and driving behavior.Compared with objective factors such as road conditions and vehicle equipment,drivers can better regulate driving behavior.Different driving behaviors can be extracted by driving data mining.Driving behavior can reflect the driver ’s driving style.Fuel consumption analysis based on different driving styles will provide reference for driver ’s decision-making.In view of this,the work of this article is as follows :Firstly,the original data is preprocessed.The original data is preprocessed by data merging,removing duplicate values,GPS outlier screening and data standardization,and four categories of indicators are extracted based on the preprocessed data.Subsequently,the complex multi-dimensional data was reduced to low-dimensional data with less data volume by principal component analysis,and 6principal components were extracted from 28 indicators,which effectively reduced the time and calculation required for subsequent experiments.Secondly,based on the preprocessed data,a driving style classification and recognition model based on clustering and support vector machine is proposed.Three clusters are obtained by extracting six principal component clusters,and the driving behavior data related to the three clusters are analyzed.Three types of driving style labels are marked,namely,slam on the accelerator,high-speed driving and frequent variable speed;on this basis,a three-class support vector machine classification model is constructed.The classification model can identify the driving style for the newly added data.After repeated verification,the accuracy of the model is 96.5%.Finally,a fuel consumption prediction model is proposed based on driving style classification.Based on different driving styles,two fuel consumption prediction models are constructed based on random forest RF algorithm and extreme gradient enhancement XGBoost algorithm respectively.The data sets of different driving styles are input into the two models respectively,and the predicted values of fuel consumption are obtained by fitting.The prediction accuracy of each algorithm is calculated by regression evaluation index.By comparison,the accuracy of XGBoost algorithm is 5 % higher than that of RF algorithm,so XGBoost algorithm with better performance is selected for experimental simulation.Considering that the XGBoost algorithm conforms to the characteristics of the black box theory,the Shapley value algorithm is selected to analyze the influence mechanism of the characteristic variables on the fuel consumption prediction value,and then the energy-saving driving strategy suggestions and management enlightenment are put forward.