Improving the Energy Efficiency of Electricity Distribution in the Mining Industry Using Distributed Generation by Forecasting Energy Consumption Using Machine Learning

The article describes the analysis of the principles of reducing the level of power in the energy system, as well as factors influencing the change in the balance (meteorological, social, etc.). The basic principles of price regulation in the wholesale electricity and capacity market are considered. The article describes analysis of methods of power consumption forecasting, as well as factors that are taken into account. This research substantiates the choice of the methodology of forecasting electricity consumption based on data on the wholesale electricity and capacity market. One of the research results is data on energy consumption for further regression analysis based on the methods of moving average and weighted average. The research presents a regression analysis of the time series of electricity consumption for the 2st price zone (Siberian zone) of the wholesale electricity and capacity market for an 8-day period (period length is due to highlight the seasonality in electricity consumption). Using the cross-validation method on the obtained data, we get as the result the choice of smoothing parameters, characterizing the weight of the equalized period. Final result of the research is power consumption forecast based on the Holt-Winters method, determined the forecasting confidence interval, data anomalies of the model, and the model's behavior during training. Funding information. The publication has been prepared with support of the RUDN University Program 5-100.