Research On The Prediction And Influencing Factors Of Heavy Truck Fuel Consumption Based On Data Mining

As the main carrier for undertaking road transportation tasks,heavy-duty trucks have played an important role in the development of the transportation industry.Due to its heavy body load and long operating time,it has caused a large amount of fuel consumption and environmental pollution,seriously affecting the environment and economic development.Therefore,relying on data mining technology to establish a fuel consumption prediction model based on the operating parameters of heavy-duty trucks,analyzing the impact of driving behavior on fuel consumption of heavy-duty trucks,is of great significance for energy conservation and emission reduction,and achieving the"dual carbon strategy".This study focuses on the fuel consumption of heavy-duty trucks,and establishes regression prediction models for instantaneous and fragmented fuel consumption.When predicting instantaneous fuel consumption,based on the collection and preprocessing of a large number of real-time operating conditions of heavy-duty trucks,correlation analysis is conducted on each driving data to extract 7 strongly correlated factors to construct an LSTM prediction model.By iteratively adjusting parameters,the model reaches its optimal state.At the same time,compare the performance of LSTM model with that of BP neural network,support vector machine regression and random forest algorithm.When predicting fuel consumption in segments,the collected working condition data is constructed into kinematics segments according to the rule of"ignition start stop",and a series of driving behavior characteristics are derived.The segment fuel consumption prediction model is constructed using Light GBM algorithm,and the fitting ability of the model is improved by combining feature selection and Optuna hyperparameter search,and performance comparison is made with XGBoost,random forest,and support vector machine regression algorithm,To provide a research basis for analyzing the relationship between driving behavior and fuel consumption.Finally,based on the SHAP framework,the prediction results of the optimal fragment fuel consumption model are visually explained,and the important influencing factors selected are analyzed based on global,single factor,and dual factor interactions.The results show that the instantaneous fuel consumption model based on LSTM has the best prediction performance,with MAE,MSE,and R~2 of 1.485,16.437,and 0.978,respectively;For fragment fuel consumption,the Light GBM model has the best prediction performance,with MAE,MSE,and R~2 of 2.036,13.042,and 0.873,respectively;The average speed,average throttle pedal value,average negative acceleration,and average positive acceleration have a significant impact on segment fuel consumption;Ensuring stable and appropriate speeds,reducing bad driving behavior,planning transportation vehicles for staggered trips,and optimizing vehicle reservation systems can appropriately reduce fuel consumption.The relevant conclusions can provide a certain theoretical reference for energy conservation and emission reduction in highway transportation enterprises.