New Correlation of PV Modules Soiling and Outdoor Conditions Using Artificial Neural Networks

In the present study a methodology for correlation between soiling rate of solar photovoltaic (PV) glass and weather conditions using Artificial Neural Network (ANN) is presented. To perform the study, a solar PV glazing was exposed outdoor at home Solar Energy Platform in Rabat, Morocco. Regular measurements from April 20, to December 31, 2016, were performed in the platform of High College for Education and Research ENS - Rabat to monitor the change of soiling rate over time. Meteorological data composed by solar irradiance G, ambient temperature T_amb, wind speed W_s, wind direction W_d, relative humidity RH, and rainfall RF were used as input variables for ANN modelling. The performance of the models was evaluated on the basis of statistical indices Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). Results show that the employ of Levenberg-Marquardt back-propagation algorithm, and the active functions Tansig (hidden layer) and Purline (output layer) in an ANN architecture 6-35-1 achieve the best estimations (RMSE = 0.4541, MAPE = 0.348, and R² = 0.9848). Therefore, this method is a promising framework for its application in smart sensors on cleaning systems for PV modules to improve their operational efficiency. Based on Garson Algorithm, the sensitivity analysis has shown that rainfall, followed by wind direction respectively impacts the soiling rate more than the other meteorological parameters.