Прогнозирование электропотребления на горнопромышленных предприятиях с использованием статистических методов

Abstract

Forecasting of electric power consumption with due account of assessed impact of various factors helps to make efficient technical and managerial decisions to optimize the electric power consumption processes, including preparation of bids for the wholesale electric power and capacity market. The article uses multivariate methods of statistical analysis and econometric methods based on time series analysis for model designing. The paper presents the results of developing the following models: a multifactor model of electrical power consumption using the regression analysis, the Principal Component Method with the assessment of the impact of production factors on electrical power consumption using elasticity coefficients, as well as the energy saving factor based on a variable structure model; trend additive and multiplicative forecast models of electrical consumption that take into account the seasonality factor, models with a change in trends, a linear dynamic model of electrical power consumption that takes into account the production output; a forecast adaptive polynomial model of electrical power consumption as well as the Winters model. The developed forecast models have a sufficiently high accuracy (accuracy of the MAPE was below 7%). The choice of the model type to forecast the electrical power consumption depends on the quantitative and qualitative characteristics of the time series, the structural relation between the series, the purpose and objectives of the modeling. In order to enhance the accuracy of the forecast it is required to regularly refine the model and adjust it to the actual situation with the due account of new factors and production trends while building different versions of scenarios and combined forecast models of electrical power consumption