Experimental evaluation of industrial pollution impact on photovoltaic system performance and machine learning application

Nowadays, the rapid increase in energy requirements has scaled up the demand for photovoltaic (PV) panels in parallel with technological improvements and their cost benefits. In Solar Power Plants (SPP), maintenance and cleaning processes carried out during the operation cycles play critical roles in the long-term performance of photovoltaic (PV) panels, as much as system design and feasibility analyses. Dust and industrial pollutant accumulation on the PV panel surface over time prevent solar radiation from reaching the cells, resulting in significant losses in electrical efficiency. This study presents experimentally the effect of industrial pollution that has occurred over the years on the performance of PV panels of a power plant installed in an industrial region. Two of the fully polluted 250 W polycrystalline solar panels were selected randomly. One of the panels was thoroughly cleaned, whereas the latter remained dirty, with pollution levels of 75%, 60%, 45%, 30%, and 15% sequentially. In the experiments, PV solar panels were oriented at a 25° angle using the same mechanical structure, and the effects of various pollution levels on electrical performance were evaluated comparatively. As a result, the electrical efficiency losses were observed as 45%, 39%, 38%, and 34.8% with respect to the pollution levels of 75%, 60%, 45%, 30%, and 15%, respectively. Additionally, the back temperature varied between 44 °C and 48 °C for the fully cleaned panel, whereas it ranged from 42 °C to 45 °C for panels with various pollution levels. In terms of pollutant analysis, it was determined that iron oxide (Fe2O3) was the most dominant contaminant, accounting for 37.37%, followed by calcium oxide (CaO) at 29.30%. Furthermore, the Gradient Boosting and Multi-Layer Perceptron models demonstrated superior prediction performance, achieving R2 scores above 90% and the lowest RMSE, MAE, and MSE scores.