利用可见光RGB图像监测光伏板上的沙尘积聚水平

Accurate monitoring and assessment of sand–dust accumulation levels are essential for optimizing cleaning schedules of photovoltaic systems in dusty regions. This article proposes an intelligent monitoring method using visible spectrum RGB images of photovoltaic panels. First, a revenue function based on the photovoltaic efficiency loss model is developed, and then a grading scale for sand–dust accumulation levels is proposed. Second, a sand–dust accumulation monitoring framework is designed, comprising image preprocessing operations and a hybrid convolution transformer-encoder network. The preprocessing steps, involving correction and background filtering, aim to eliminate potential interference factors. In addition, an atrous stem module with an extended receptive field and novel relation-aware global attention modules are integrated into the hybrid convolution transformer-encoder network to identify global and local differences between panel images with varying levels of sand–dust accumulation. Experiments conducted on two types of photovoltaic panels confirm the feasibility and effectiveness of the proposed method. The results show that this method can accurately classify sand–dust accumulation levels on photovoltaic panels and exhibits robust generalization performance.