Forecasting fuel consumption in means of transport with the use of machine learning

Abstract

Transport is a key factor influencing greenhouse gas emissions. In relation to this, the issues and challenges facing the transport industry were presented. The issues of challenges for the transport industry related to the European Green Deal were discussed. It discussed how the transport system is critical for European companies and global supply chains. The issues related to the exposure of society to costs are presented: greenhouse gas emissions and pollution. The article deals with the issues of managing transport processes in an enterprise. It was decided to raise the topic of fuel consumption in means of transport. Based on a review of the scientific literature, 3 categories of features are indicated: the vehicle characteristics, the driver's characteristics, and the route's impact on fuel consumption. The study is based on actual data from the archives of the GPS vehicle monitoring system. Data was collected on 1890 routes operated between May 30, 2020, and May 31, 2021. The routes were performed by twenty-nine drivers and 8 vehicles. The vehicles are 40-ton road sets consisting of a tractor unit and a semi-trailer. The analysis of factors influencing fuel consumption is presented. The methodology for conducting feature engineering is described. The benefits of using the method of reducing fuel consumption are presented. The possibilities of using the methods of forecasting electricity and hydrogen consumption in various means of transport, including public transport, where indicated. The data is processed using the Pandas library. The models are compared according to the MAE success measure. The application of methods of working with large data sets is presented. The calculations are made with the help of the NumPy library. Data visualization is done with Matplotlib and Seaborn. Scikit-Learn models are used.