Short Term Load Forecasting Using Machine Learning Algorithms: A Case Study in Turkey

Abstract

In this study, short-term load forecasting of the Gebze region in Turkey was carried out using Machine Learning-based prediction algorithms such as Artificial Neural Networks, Decision Tree, Support Vector Regression and K-Nearest Neighbor. Load demand and weather variables such as temperature, humidity, pressure and wind speed are used as input variables in the forecast models. Error metrics such as Mean Absolute Error, Mean Squared Error, Mean Absolute Percentage Error and R-squared were used to control the prediction success of the proposed algorithms and models. As a result, the predictions made with all the proposed algorithms are within reliable and acceptable ranges, and Support Vector Regression algorithm showed the best performance with an error of 1.1%.