Energy Demand Forecasting for Hybrid Microgrid Systems Using Machine Learning Models

This study aims to design energy demand forecasting models for energy management in hybrid microgrid systems using optimized machine learning techniques. By incorporating temperature, humidity, season, hour of the day, and irradiance, the complex relationship between these input parameters and the yield of photovoltaics, generator, and grid energy sources is examined. Five different machine learning models including linear regression, random forest (RF), support vector regression, artificial neural network, and extreme gradient boosting models are adopted in this study. Evaluation of model performance shows that the RF model is the best candidate for the dataset, with a mean-squared error of 0.2023, mean absolute error of 0.0831, root-mean-squared error of 0.4498, and R² score of 0.9992. Shapley additive explanations analysis identified key predictors such as hour, irradiation, and season while highlighting the negative impact of humidity and day of the week on energy demand.