A Novel Deep Learning-Assisted Framework for the Assessment of Real-Time Dust Accumulation Data on Solar PV Modules

Soiling, usually caused by dust deposition, is a significant difficulty for photovoltaic (PV) systems because it obstructs light absorption, resulting in reduced power output and the creation of hot spots. Monitoring the soiling ratio can give critical insights into PV module performance under dirty circumstances, allowing for power output prediction. This study gives a complete performance analysis of PV modules impacted by soiling based on real-time data obtained over six months in Badli, New Delhi, India. A soiling monitoring system captured key variables such as soiling ratio, transmission loss, and temperature, which served as the foundation for the development of a mathematical model to estimate PV power production under soiled circumstances. The findings show that dust deposition causes a considerable 17% loss in power production, emphasizing the necessity for effective soiling mitigation techniques. Deep learning models, especially stacked long short-term memory (LSTM) and bidirectional LSTM, were used to forecast power output under soiling circumstances. Stacked LSTM outperformed Bi-LSTM, with a R² score of 0.9913 and a mean squared error (MSE) of 0.0078. Training time was 17.35 s. By precisely estimating dirty power output, this study makes it easier to schedule cleaning cycles, improves PV module performance, and contributes to sustainable solar energy generation in dust-prone areas.