An integrated industrial PV panel cleaning recommendation system for optimal dust removal

Developed from prior research, this paper presents a comprehensive study on the effectiveness of the PV Nexus Cleaning Recommendation System (PNCRS), an intelligent cleaning recommendation system designed for optimizing photovoltaic (PV) panel cleaning schedules under various environmental conditions. Traditional fixed interval and performance degradation strategies for PV panel maintenance often prove inadequate, primarily due to their inability to adapt to fluctuating environmental impacts, particularly under extreme weather conditions. These conventional methods, while straightforward, fail to capture the complex interplay between environmental variability and panel efficiency, leading to suboptimal cleaning schedules and diminished energy output. The intelligent cleaning recommendation system utilizes real-time environmental adaptability, data-driven decision making, and comprehensive profit optimization to significantly determine PV panel cleaning schedule over traditional methods. In this paper, the intelligent cleaning recommendation system PNCRS incorporates cutting-edge data augmentation and machine learning techniques, including Variational Mode Decomposition (VMD) and Conditional Generative Adversarial Networks (CGANs). This integration is essential for enhancing data representation, particularly in scenarios where input data is sparse or unrepresentative, such as during unusual weather patterns. Additionally, the system employs the Wavelet Packet Energy Transmissibility Function (WPETF) to innovatively reduce the model’s dependency on less impactful environmental features under extreme conditions. Furthermore, the system uses profit-based Bayesian Optimization (BO) to dynamically adjust the importance weights of model features when profit curves deviate from expectations. Our evaluation of the PNCRS across two PV farms with unique operational features quantitatively validates its effectiveness, as profit curves indicate a 29% profit increase at Farm 1 and a 34% profit increase at Farm 2.