Dust impact on photovoltaic technologies: a comparative analysis using deep recurrent neural networks

Photovoltaic (PV) behavior and productivity are affected by environmental parameters such as temperature, humidity, dust, etc. This study was conducted in Sohar, Oman, six standalone PV modules, each rated at 100 W, were meticulously tested outdoors for 35 days, specifically focusing on the detrimental effects of dust. What sets this research apart is the pioneering use of a deep recurrent neural network (DRNN) to comprehensively analyze and simulate the adverse impact of dust on PV power generation. The study employed various PV module technologies, including monocrystalline, polycrystalline, and flexible monocrystalline modules. The experimental findings, which encompassed a comparative analysis of both clean and dusty PV modules over the course of a month, revealed substantial performance degradation: 30.24%, 28.94%, and 36.21% for the respective PV module technologies. Furthermore, the presence of dust not only reduced power output but also led to lower module temperatures. The results obtained through practical experiments underscore the heightened negative influence of dust on solar panels, as explained by the DRNN network, resulting in a significant decrease in energy production across all panels. In the cross-validation phase at epoch number 500, the MSE for the DRNN module was 0.0145, while in the training phase at epoch number 827, it was 0.0178. A set of performance factors has been applied to test the accuracy of the results predicted from the proposed model DRNN. The sensitivity analysis showed that the flexible monocrystalline panel (Clean) is less uncertain based on its sensitivity rate of 0.0092. It is followed by the flexible monocrystalline panel (Dusty) with a rate of 0.0176.