Pollution level detection of silicone rubber insulators based on deep extreme learning machine

Silicone rubber insulators are widely used in power transmission lines due to their excellent hydrophobicity and strong resistance to pollution and flashover, but the insulators are subject to pollution and other factors during operation, which can destroy many of their excellent properties and cause very serious impacts on the power system. It is an effective strategy to prevent accidents and ensure the safe operation of the power grid to obtain the insulator pollution level quickly and accurately. In this paper, the hyperspectral test platform is used to obtain the hyperspectral spectral information of silicone rubber samples with different pollution levels, and the obtained spectral data are pre-processed with black and white correction, multiple scattering correction and SG smoothing to classify the hyperspectral data of silicone rubber samples with different pollution levels into 1-4. The training data and the test data are the hyperspectral data after pre-processing, divided in the ratio of 7:3, and the accuracy of the final test results reaches 95%. Finally, the training results of the deep extreme learning machine were compared with the support vector machine and BP neural network, and it was found that the DELM algorithm can combine the advantages of accuracy and rapidity, and has a greater application value in the classification of the pollution level of silicone rubber composite insulators. It provides a technical and theoretical reference for the non-contact detection of insulator pollution level.