Net Load Prediction Using Machine Learning Methods: A Case Study of Residential Households

As residential solar photovoltaic (PV) systems are increasingly integrated into distribution networks globally, there is a growing necessity to shift from traditional load forecasting to net load forecasting (i.e., predicting the combination of PV and load collectively). Understanding and predicting the daily net energy of a typical household is crucial for designing and sizing appropriate renewable energy systems and storage solutions. This paper compares the performance of different machine learning methods in predicting the net load patterns of a typical household. The database contained PV production and energy load records from 68 actual households in Cyprus, over a period of one year. The results showed that the eXtreme Gradient Boosting (XGBoost) model outperformed all other models yielding a normalized root mean square error of 5.52%. Despite the inherent unpredictability of net load patterns, the findings indicate that the proposed XGBoost model can achieve high prediction accuracies for household electricity net load profiles.