Application of nonparametric ML on forecasting the production of a large-scale solar power plant: Kom-Ombo case study

Abstract

Renewable Energy grew by 50 % globally in 2023, where three quarters of this energy is coming from solar PVs. The variable production of solar energy needed for PVs makes the prediction of the output power vital for avoiding technological and economic issues. Machine Learning (ML) models were used as a nonparametric approach to predict power output in several studies but only on small-scale solar power plants. This work investigates the implementation of several ML algorithms to predict the PV output power of large-scale solar power plants, where the Kom Ombo 26 MW power plant is taken as a case study. The Liner Regression (LR), Decision Tree (DT), and Random Forest (RF) algorithms were tested, where the LR model showed the lowest RMSE and R² values and was further improved after removing the night hours from the dataset. In addition, the Long Short-Term Memory (LSTM) model showed the highest accuracy when used with the historical records of the Kom Ombo power plant. Finally, the LSTM model was used to predict the PV output power for the Kom Ombo power plant to choose the maintenance day of the plant which resulted in substantial power and profit savings. Similarly to [6] who used Quantile Regression Forests (QRF) on a 2 MW solar power plant, we were able to show that nonparametric ML can be reliable in forecasting power output from a 20 MW solar power plant.