مهرداد احمدي

Timely an‎d effective cleaning of photovoltaic (PV) panels is one of the key requirements for the sustainable operation of solar power plants, as dust accumulation—particularly in desert regions—causes a significant decrease in system efficiency. In this study, aimed at optimizing the maintenance an‎d cleaning process of PV panels, various cleaning methods an‎d surface contamination detection systems were first reviewed. Finally, a brush-an‎d-water cleaning mechanism combined with an optical dust sensor was selec‎ted as the most suitable option. Subsequently, using the XGBoost an‎d Ran‎dom Forest machine learning algorithms an‎d based on climatic an‎d environmental pollution data, predictive models were developed for optimal cleaning scheduling in a solar power plant located in Semnan Province, Iran. The results showed that the proposed approach not only reduces unnecessary cleaning operations but also increases power generation efficiency by about 1.2% an‎d improves the cost–benefit ratio (BCR) to 3.49 for robotic cleaning in solar plants—particularly in dusty desert regions of Iran. Accordingly, the optimal cleaning interval was adjusted in proportion to the dust concentration of the target area, resulting in an optimal cycle of 32–33 days an‎d a 49% reduction in annual cleaning costs. Furthermore, the payback period for implementing the robotic cleaning system was calculated to be less than 12 months, demonstrating the economic feasibility of the proposed method. This achievement can play an important role in industrializing the intelligent maintenance of solar power plants an‎d reducing efficiency losses caused by surface contamination.

عملكرد فتوولتائيك , يادگيري ماشين , بهينه سازي , گرد و غبار , آلودگي سطح پنل فتوولتائيك , دوره تميزكاري , نسبت فايده به هزينه , بازده

photovoltaic performance , machine learning , optimization , dust , soiling , cleaning cycle , cost–benefit ratio , efficiency

Mehrdad Ahmadi

Rouhollah Ahmadi