Short-Term Electricity Consumption Forecasting for a Steel Enterprise

Forecasting of power consumption is one of the ways to improve the validity of managerial decisions in the process of planning costs for electrical energy. The functioning of wholesale electricity markets creates economically advantageous conditions for its participants. At the same time, one of the key requirements is the availability of a forecast of power consumption with a minimum error. Industrial enterprises often act as subjects of wholesale electricity markets, since the purchase of electricity in the wholesale electricity market is a particularly effective and expedient solution for large electricity consumers. However, when drawing up an application for a day under the terms of participation in the "day ahead market" wholesale electricity market, it is not enough to carry out engineering calculations based on data on output volumes, production plans, etc. A more flexible approach is needed, taking into account a number of other factors: seasonal, random, etc. Therefore, the validity of management decisions in the planning process can be enhanced by the development and implementation of a management decision support system integrated into the automated information system of the employee-operator responsible for power consumption planning. It is assumed that the main functionality of this system will be the ability to obtain a reliable forecast through a human-machine interface. This paper presents an approach to creating a predictive models of power consumption based on the use of machine learning algorithms. Several mathematical models were also created, the best of which turned out to be the CatBoost gradient boosting model.