Using machine learning prediction to design an optimized renewable energy system for a remote area in Italy

Abstract

Due to the lack of fossil fuels, there is a significant demand to employ renewable energy systems (RES) worldwide. This paper proposes designing an optimized RES for a remote microgrid that relies solely on solar and wind sources. The proposed RES aims to provide reliable and efficient energy to the microgrid by using machine learning algorithms to forecast the power output of the solar and wind sources. This forecasting will help the system to anticipate and adjust to changes in the weather patterns that may affect the availability of solar and wind. In addition, the system advisor model (SAM) software is used to design the hybrid solar/wind system, considering factors such as the size of the microgrid and the available resources. The system comprises a 60-kW wind system of ten turbines and a 100-kW PV system spread out over the houses. The results show that random forest regression (RFR) models achieved a high level of accuracy in predicting solar power generation, as evidenced by their low mean squared error (MSE) and high R² values. Additionally, a proposed hybrid system can generate enough energy to meet the area's needs.