Machine Learning-Based Optimization of Photovoltaic Panel Cleaning Schedules in Arid Environments: A Case Study of Ryadh, Saudi Arabia

Photovoltaic solar power systems in arid regions offer tremendous energy potential due to high solar irradiance. This study proposes a machine learning framework for predicting the effects of pollution on photovoltaic panels. Environmental data were collected from NASA, the Solar Photovoltaic Information System (PVGIS), and local air quality stations. The XGBoost, Random Forest, and LSTM machine learning models were used in the analysis. The results show that the XGBoost model outperformed the other models with a coefficient of determination (R²) of 0.93, thanks to its ability to accurately predict energy loss due to pollution. The cost-benefit analysis demonstrated that biweekly cleaning maximizes energy recovery and reduces operating costs by up to 15% compared to fixed schedules. The proposed framework integrates real-time dust prediction and dynamic scheduling, providing a scalable solution for solar farms in arid environments. This approach improves the efficiency of photovoltaic solar energy systems, reduces water and labor waste, and supports the deployment of sustainable solar energy in dust-prone areas.