FORECASTING THE LOAD ON THE ELECTRIC GRID OF A MINING AND PROCESSING PLANT

The relevance of the work is due to the need for a comprehensive study of the power supply system of the efficiency of managing the consumption of electrical energy based on predicting the load on the power grid of concentrating plants. The accuracy of its forecast will significantly simplify the process of making managerial decisions in strategic and operational planning and minimize the cost of electricity for the production of a unit of output (ore concentrate). Purpose: substantiation of the efficiency of electricity consumption in the enrichment and processing of ores based on the management and forecasting of the load on the electrical network of concentrating plants. This will make it possible to use industrial energy storage devices at the right time to reduce the consumption from the network during peak hours (when using a two-rate tariff) or in the peak zone of the day (when using a differentiated tariff for electrical energy). Objects: enrichment plants, which are characterized by the complexity of determining the energy characteristics of process equipment and effective management of electricity consumption based on predicting the load on the power grid. Methods: methods of complex generalization, analysis and evaluation of practical experience and scientific achievements in the field of managerial decision-making in strategic and operational planning and minimization of electricity costs for the production of a unit of output were used. In particular, the analysis of literary sources was carried out; used the method of theoretical generalizations with the use of mathematical statistics, physical and mathematical modeling; calculations and feasibility studies, laboratory and full-scale experimental studies, as well as industrial tests in the conditions of operating enterprises according to standard and new methods with the participation of the authors. Results. On the basis of experimental studies, the factors influencing the consumption of electrical energy and to be included in the developed machine learning model were determined, which will significantly simplify the process of making managerial decisions in the strategic and operational planning of electrical energy costs and minimize the cost of electricity for the production of a unit of output. It has been established that the important features are the ambient temperature, the correlation coefficient of which with the target variable - the total uncontrolled load - is 0.17, humidity (correlation coefficient was -0.22), weekend/working day sign (correlation coefficient -0.33), day of the week (correlation coefficient -0.38), time data (correlation coefficient for time 0.19, for month and year < 0.068). A machine learning model based on the extreme gradient boosting regression model (XGB Regressor) is proposed, which makes it possible to obtain the most reliable forecast of the load on the concentrator's power grid. The average absolute prediction error of these models was 3.51, and the coefficient of determination was 0.84 and 0.87 for the training and test samples, respectively. The results of the study can be adapted for other objects. It was noted that the modeling of the work of fuzzy intelligent control system (ISC) showed potential opportunities for reducing the total energy consumption by 15-35% in the modes of additional hydroaccumulating EE generation (two-rate tariff was studied) in the conditions of various enterprises involved in the extraction and beneficiation of iron ore raw materials.