Study on the Pollution Level of Silicone Rubber Composite Insulators Based on BPNN

Pollution is one of the key factors leading to flashovers in power transmission lines, and accurately detecting the degree of pollution is crucial for assessing the insulation state of insulators. To this end, this paper utilizes hyperspectral imaging technology, collecting hyperspectral data from the surface of insulators and improving data quality through preprocessing methods such as black and white calibration, and standard normal variate transformation. Subsequently, the hyperspectral data are dimensionally reduced using the successive projection algorithm (SPA), and the insulator surface pollution levels are categorized with the backpropagation neural network (BPNN) algorithm. Experimental results show that the SPA-BPNN model, after feature extraction, achieves an overall accuracy of 95.56% in classifying the pollution levels of insulators, which is superior to the all band-BPNN model without feature extraction. This research not only enhances the accuracy of pollution level classification but also provides a scientific basis for the reliability of grid equipment and the reliability of grid operations.