Abstract

Natural dust accumulation on the installed solar photovoltaic (PV) modules reduces the transmittance, thereby degrading the conversion efficiency of solar PV plants. Therefore, proper monitoring and cleaning of solar modules are necessary to ensure the efficient performance of solar PV modules. For new installations, the prediction of performance is needed based on the already installed sight-specific PV plant data. In this work, Machine Learning (ML) algorithms have been used to predict the transmittance of solar PV module surface under the impact of natural dust accumulation over seasonal variations. Annual experimental dataset from solar PV modules surface (low-iron glass) has been utilized to predict the transmittance using ML algorithms such as K-Nearest Neighbour (KNN), Decision Tree, Random Forest, and Ensemble learning (EL) - Gradient Boost algorithm, on three different tilt positions: horizontal, inclined and vertical positions. A detailed comparative analysis of different ML models for different glass positions using performance evaluation metrics has shown that the ensemble learning-based Gradient boost algorithm is the most accurate in predicting the optical performance of solar modules with R2, MAE, RMSE, and MAPE values of 0.99, 0.736, 0.55 and 0.7% respectively. The prediction results obtained in this work claim to be useful for optimal scheduling of cleaning the solar modules to improve the overall performance of site-specific solar power plants.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call