Application of Linear Regression and Artificial Neural Networks for Load Forecasting the Consumption of Fuel and Energy Resources Train Traction

Processing "big data" of existing information systems is currently possible using modern data mining methods. In relation to forecasting energy consumption for a trip, data mining involves preliminary research and selection of factors that have a significant impact on the amount of fuel and energy resource consumption with subsequent forecasting. The study is aimed at selecting the optimal method for creating a mathematical model for predicting the specific energy consumption for trains' traction. The article contains the results of the factors impact nature such as technical speed, mass of the train, axle load and the number of axles, wind speed and direction, temperature on the amount of energy consumption. The Student's method of statistical data processing, used to determine the homogeneity of the samples under study, and the Pearson correlation analysis method for determining the correlation coefficients between the specific energy consumption for trains traction and the factors supposedly influencing the value of this consumption are applied in the work. The construction of mathematical models for predicting specific energy consumption is based on the methods of multiple linear regression and neural network modeling. The article contains the results of specific electricity consumption studied samples homogeneity checking, the description of the calculated correlation coefficients for each studied group of freight trains. The conducted research argues for the need to develop a methodology for rationing the consumption of fuel and energy resources for train traction using a neural network modeling method.