A Model of Classification of Consumers on the Retail Electricity Market

In this paper, we study the problem of forecasting the energy consumption of real estate objects. The purpose of the work is to increase the efficiency of distribution companies by improving the mechanism for determining planned consumption volumes. The task of the work is to construct a mathematical model based on existing data and further forecast on its basis. To achieve the goal of the work, 2 approaches are used: forecasting energy consumption based on time series (regression task) and forecasting the energy consumption profile for each object (classification task). Various methods of machine learning, as well as clustering, were used in the work. We compared machine learning methods that solve different problems with each other, and selected the best ones. As a result, the ARIMA and Support Vector Machine methods are compared for predicting energy consumption. The choice of a particular algorithm will depend on the goals of the companies that produce and distribute energy.