Predicting the energy-supply parameters for a transportation process based on multifactor models

Abstract

Results are presented for the construction of a model of an energy-supply parameter characteristic for the transportation process, “electric-energy consumption for traction (plan),” based on the identification thereof with actual production and economic parameters by the example of one of the branches of JSC Russian Railways. The research was carried out based on a database created for monthly actual values over a long period, from 2007 to 2016. A methodology and results are presented for statistical studies of the parameter “electric-energy consumption for traction (plan)” in conjunction with 20 parameters, including “sectoral speed,” “average weight of the train,” “transportation costs.” The methodology involves choosing the factors for a multifactorial model and elimination of the consequences of multicollinearity based on the results of cluster analysis and constructed dendrograms. Variants are presented for the construction of multifactorial models to predict and manage the parameter of “electric-energy consumption for traction (plan)” for control and risk-management tasks. A variant of construction of a multifactorial model for predicting and control of the parameter “electric-energy consumption for traction (plan)” for control tasks and risk-management is given. As a mathematical tool for constructing the multifactorial models, a technique of linear regression equations has been used. The construction of dendrograms and obtaining the regression equation coefficients have been carried out using the Statistica software package. Using the multifactorial model results in improvement in the predictive accuracy of the “electric-energy consumption for traction (plan)” parameter measured with a MAPE estimate, from a 15% level to a level better than 6%.