Research On Fuel Consumption Prediction Of Heavy Trucks Based On Road Driving Conditions

Heavy trucks,as an important means of transportation,have a significant impact on the environment due to their unique usage scenarios and body structure,resulting in a high proportion of exhaust emissions.To address the issues of environmental pollution and energy scarcity,the energy-saving and emission reduction of heavy trucks has gradually become a research hotspot.This article analyzes and explores the influencing factors of fuel consumption under different road driving conditions of heavy-duty trucks and establishes a fuel consumption prediction model for heavy-duty trucks.This model is of great significance for reducing vehicle energy consumption.The main tasks are as follows:（1）Construct different road driving conditions for heavy-duty trucks.Firstly,preprocess the original data,including methods such as duplicate data processing,abnormal data processing,missing data filling,and data denoising.Then divide the processed data into multiple short stroke segments,define and calculate parameters that can reflect the characteristics of short stroke segments,and construct a dataset of working condition characteristics.Next,use principal component analysis and K-means clustering methods to cluster short distance segments,and construct three types of road driving conditions based on their characteristics and actual road traffic conditions:congested,suburban smooth,and high-speed.Finally,analyze the characteristics of various driving conditions and differences in fuel consumption,explore the driving behaviors that affect fuel consumption in various road driving conditions,and construct a dataset of heavy truck fuel consumption prediction features.（2）Construct a heavy truck fuel consumption prediction model based on least squares support vector machine（LSSVM）.Using feature datasets of three types of road driving conditions as experimental inputs,construct a heavy-duty truck fuel consumption prediction model based on LSSVM and RBF neural network.Evaluate the prediction effect of the model using Root-mean-square deviation（RMSE）,average absolute error（MAE）,determination coefficient（²）,and relative error（RE）by comparing with the recurrent neural network（RNN）and support vector machine（SVM）models.The experimental results show that the LSSVM prediction model performs relatively well under three types of road driving conditions.（3）Construct a least squares support vector machine（APSO-LSSVM）heavy truck fuel consumption prediction model based on adaptive particle swarm optimization.To solve the problem of large local error of the LSSVM model,use the particle swarm optimization algorithm to optimize the regularization parameterγand nuclear parameterσof the LSSVM model and establish a PSO-LSSVM model.To address the issues of the particle swarm optimization algorithm being prone to falling into local optima and premature convergence in the later stage,improve the inertia weight value,learning factor(8₁ and(8₂,and introduce cross-mutation operation to construct a heavy truck fuel consumption prediction model based on APSO-LSSVM.The experimental results show that under three types of road driving conditions,the APSO-LSSVM model has higher prediction accuracy and better adaptability compared to other models,providing a feasible method for predicting fuel consumption of heavy trucks.