Time Series Forecasting of Dust Deposition on PV module for Efficiency Planning

Dust deposition on surfaces, particularly in open environments, is a critical issue in tropical and arid regions. For solar energy systems, accumulated dust can significantly reduce the efficiency of photovoltaic (PV) panels, leading to energy loss and increased maintenance costs. Understanding the pattern and behavior of dust accumulation over time is therefore essential not only for energy performance optimization, but also for environmental and operational planning. This study focuses on analyzing dust accumulation data collected during 2018-2022, and forecasting future trends to support better decision-making in solar farm operations and module clening. Dust accumulation in this study area follows a consistent seasonal pattern, beginning in October and ending in September of the following year. The peak typically occurs between January and February, with the dust area fraction ranging from approximately 12.5% to 20%. This recurring trend allows for the application of time series forecasting techniques to predict future dust accumulation. Four models namely SARIMA, Prophet, XGBoost, and LSTM were compared to find the best forecasting method. Among the models, LSTM achieved the best accuracy, followed by XGBoost, while SARIMA and Prophet showed lower performance. Including seasonal features improved the prediction of yearly dust cycles and produced smoother results. This supports the utility of such models in planning solar panel maintenance, improving system efficiency, and understanding long-term environmental impacts.