Известия Томского политехнического университета: Инжиниринг георесурсов (Jul 2022)
FORECASTING OF ELECTRIC ENERGY CONSUMPTION BY AN INDUSTRIAL ENTERPRISE USING MACHINE LEARNING METHODS
Abstract
The relevance of the research is caused by the need to develop energy-saving approaches through the use of data mining tools to improve the efficiency of the management decision-making process and, as a result, more optimal use of energy resources. In particular, forecasting the consumption of electric energy of an industrial facility will simplify the process of making managerial decisions and will help minimize the cost of electricity for the production of a unit of production. The availability of an accurate forecast will enable the use of reserve capacities during peak load hours for the electric power complex. In practice, the existing methods of calculating the load on the power grid are not always suitable for forecasting, so the study is interdisciplinary in nature, combining important practical significance and the development of new recommendations regarding the use of machine learning algorithms. The main aim of the research is to analyze scientific papers containing proposals to improve the accuracy of determining energy loads using data mining, as well as to develop a machine learning model that allows you to create a reliable forecast of electricity consumption for an industrial enterprise. Objects of the research is an industrial enterprise that is characterized by the complexity of determining the energy characteristics of technological equipment. Methods: analytical method, methods of mathematical statistics, methods of machine learning, complex generalization of scientific achievements and practical experience in the use of data processing tools in the tasks of predicting the load on the power grid. Results. A review of literature sources covering the application of data mining in energy consumption management is carried out, and the main results of forecasting total electricity consumption according to industrial facility data are presented. The methods of data mining used to solve energy saving problems for various objects are considered. A machine learning model is built based on the gradient boosting algorithm of the CatBoost library, which allows obtaining a forecast of electricity consumption by months with a reliability level of 92 %. The results of the study are relevant for decision-making at the tactical and strategic levels of enterprise management for medium-term (monthly) and long-term (from a year to several years) forecasting of electrical loads, respectively.
Keywords