Research On Prediction Of Truck Travel Time In Open Pit Mine Based On Machine Learning

Open pit truck scheduling system is generally composed of three parts:the best path selection,vehicle flow planning and real-time scheduling.The best path is determined to minimize the running time,and the vehicle flow planning also takes the running time as the main parameter.The real-time scheduling needs to predict the travel time of trucks on the selected route more accurately.Therefore,the travel time of trucks on the path is an important basis for the implementation of optimal scheduling,and the accuracy of its prediction will directly affect the real-time and reliability of traffic scheduling decision.In order to build an accurate prediction model of truck travel time,taking the truck intelligent scheduling system of a large open pit mine in Henan Province as the background,through the analysis of influencing factors of truck travel time,this paper uses machine learning technology to model and evaluate the research problem,and finally puts forward a stacking model,which can achieve higher prediction accuracy of truck travel time.The specific research contents include:(1)This paper analyzes the research status of vehicle travel time prediction and open-pit Truck travel time prediction at home and abroad by using the methods of literature analysis and induction,and expounds the concept and application fields of machine learning as well as the principles of six common data-driven models,so as to lay a theoretical foundation for the follow-up research.(2)Analysis of influencing factors of truck travel time in open pit mine.Combined with literature and research problems,truck characteristics,road characteristics,meteorological characteristics and time characteristics are identified as the influencing factors of truck travel time.On this basis,truck scheduling data and GPS trajectory data are extracted from truck scheduling platform to generate truck travel time data,Data exploratory analysis(EDA)and Pearson correlation coefficient were used to analyze the causality and correlation between these factors and truck travel time.(3)Construction of truck travel time prediction model in open pit mine.Based on the analysis of the influencing factors of truck travel time,the problems of abnormal value,missing value and non-uniform feature dimension in the truck scheduling data and GPS trajectory data are dealt with,and then the modeling process of truck travel time prediction is given from the perspective of machine learning practice,including the selection of model evaluation index,the selection of model evaluation index,the selection of feature dimension and so on Finally,a stacking model which can accurately predict truck travel time is proposed.(4)Application of truck travel time prediction model in open pit mine.Taking the intelligent dispatching system of a large open pit mine in Henan Province as an example,the single model KNN,SVM,MLP,RF and XGBoost of truck travel time prediction are applied to carry out cross validation and super parameter optimization on the travel time training set respectively,and then the SVM,MLP,RF and XGBoost after super parameter optimization are selected for model stacking according to their validation results,These optimized single model and stack model are used to evaluate the travel time test set,and the prediction performance of RMSE,MAE,MAPE and R~2are compared and analyzed.The results show that the stack model is better than the single model optimized by super parameters in all evaluation indexes,and it is more suitable for truck travel time prediction.