Authors: S.M. Karpenko, E.A. Ematin, N.V. Karpenko, G.M. Lebedev

Title of the article: Forecasting of an Industrial Enterprise's Electricity Consumption Based on Statistical Models

Year: 2026, Issue: 4, Pages: 3-10

Branch of knowledge: 2.4.2. Electrotechnical complexes and systems (engineering)

Index UDK: 621.313:676.026

DOI: 10.26730/1816-4528-2026-4-3-10

Abstract: To improve the efficiency of an industrial enterprise's electricity consumption management, it is necessary to increase the accuracy of forecasting, including daily forecasting. The article provides a rationale for the relevance of daily forecasts and the selection of statistical models for their construction. A research methodology based on the Box-Jenkins methodology is proposed, which includes studying the structure of electricity consumption and production volumes based on autocorrelation and partial autocorrelation functions, conducting cross-correlation analysis, testing for cointegration and causality, selecting the type of model and determining its parameters, calculating the accuracy of the MARE, and obtaining linear convolution (aggregation) of forecasts to estimate the electricity consumption of the company's departments (wings). The studies were based on archival data on daily electricity consumption and production volumes at a mining and metallurgical enterprise. Forecast models of daily electricity consumption of the ADL type have been constructed. The Engel-Granger test was used to identify the presence of a "false regression" effect in the levels of time series. The model coefficients were estimated using the maximum likelihood method. To improve the accuracy of the forecast, a combined approach was proposed for constructing an industrial enterprise's electricity consumption forecast as an aggregation of individual forecasts from departments (shops) with a piecewise linear (segmented) trend, taking into account the weight coefficients that reflect the contribution of individual forecasts to the overall electricity consumption. The nonlinear electricity consumption trend was represented as a segmented trend, where the local linear trends correspond to the periods of technological tasks that significantly influence the nature of electricity consumption and the type of the forecast model. This approach improves the accuracy of the MARE prediction on the training and test sets by 2.2 and 1.5 times, respectively.

Key words: Electricity consumption industrial enterprise daily forecasting statistical models segmented trend aggregation of forecasts

Receiving date: 08.06.2026

Approval date: 01.09.2026

Publication date: 27.08.2026

CITE DOWNLOAD