Day ahead forecasting for solar and wind electricity production using machine learning techniques

Abstract

This paper explores the forecasting of renewable energy power output, specifically wind power in Bogdanci and solar farms in the Republic of N. Macedonia, with the end goal of achieving a day ahead forecast. Accurate forecasting of renewable energy is critical for reliable and efficient energy gen eration, making this study important for the energy industry and policymakers. The study uses historical data from MEPSO for the last 4 years (since 2020) and employs various statistical and machine learning techniques, including autocorrelation function (ACF), partial autocorrelation function (PACF), periodogram, linear regression, decision tree regressor, random forest regressor, support vector machine, XGBRegressor, Lasso, and Ridge, to predict, or rather forecast power output. Results indicate that accurate forecasting can be achieved using these methods, with potential implications for the adoption of renewable energy sources. The models were evaluated using mean squared error, mean absolute error, and r2 score.