Application of machine learning in electric load forecast

Abstract

The forecast of the daily load diagram is an integral part of the planning and management of the power system. The system operator continuously forecasts the load and, in accordance with the forecasts, ensures the operating reserve and checks the security of the electricity transmission network. Therefore, the quality of the forecast of the daily load diagram has a direct impact on the costs and safety of the power system operation. The development of machine learning algorithms and the corresponding computer infrastructure has enabled the application of machine learning in a wide range of problems. A suitable problem for such algorithms is the problem of forecasting the daily load diagram. In this report, a brief overview of the algorithms used to forecast the daily load diagram is given. Also, the report presents the results of applying a simple machine learning algorithm to the forecast of the daily load diagram in the Republic of Croatia.

Josip Vasilj

Associate Professor | Department of Power Grids and Substations

Researcher and Associate Professor at the Faculty of Electrical Engineering, Mechanical Engineering, and Naval Architecture in Split, where he teaches courses related to engineering economics, power system analysis, power grids and machine learning. His research focus is the application of advanced numerical methods to problems in the analysis and planning of power system operations.

Damir Jakus

Full Professor | Department of Power Grids and Substations

Researcher and a full professor at the Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture in Split. His research interests include power system optimization and planning, RES integration, electricity market modeling.