Research On The Detection Method Of Nsdd On Insulator Surface Based On Hyperspectral Technology

With the rapid development of industry and agriculture in China,contamination on the transmission line insulators becomes more and more serious.on the insulator surface increases quickly with frequent dust and sand weather.Frequent dust and sand weather causes a rapid increase in the Non Soluble Deposit Density(NSDD)on the insulator surface.And the problem of contamination flashover needs to be solved.NSDD has an important influence on the occurrence and development of flashover and the NSDD detection is of great guiding significance to the external insulation configuration and contamination flashover prevention.However,the traditional method is complicated and prone to errors.Therefore,it is necessary to research an on-line and non-contact detection methods of NSDD on insulator surface,which could quantitatively measures NSDD insulator surface without manual intervention.In this paper,the detection of NSDD on insulator surface under normal circumstances is studied firstly,and the special condition of sand on insulator surface in sandstorm area is considered,and the detection method is improved.Then the special case that the sand attached to the insulator surface in the sandstorm area is considered and the detection method is improved.In the case of NSDD detection under common conditions,the detection principle based on hyperspectral technology is expounded and the specific procedure is briefly described.According to the introduced procedure,the detection model using different regression algorithms is established based on the hyperspectral data of insulating sheet sample with different pollution levels.The performance of the model under different pre-processing schemes is compared.The results show that the model based on KLEM algorithm using BW+WD+MSC is the best,which R~2 and RMSE are 0.975 and 0.339,respectively.The particularity of the sand attached to the contaminated insulator surface in the sandstorm area is considered and the influence of dust on the hyperspectral line characteristics of the dirty insulating sheet surface is analyzed firstly.Then the characteristic wavelength is extracted used SPA algorithm and the location and characteristics of the characteristic wavelength are analyzed.Based on the original data,first-order derivative data and its characteristic wavelength data,the NSDD detection model is established based on KLEM algorithm.The influence of pre-processing scheme and wavelength data number on the detection effect is compared.The results show that the characteristic wavelength data is beneficial to improve the prediction speed and accuracy of the model.R~2 and RMSE of the optimal model are 0.979 and 2.068,respectively.The model with the best performance in the two cases is applied to the detection of NSDD on the insulator surface,and the factors affecting the accuracy of the model are analyzed.In addition,the visualization of NSDD distribution on the surface is realized,which provides technical support for pollution detection and is of great engineering application value.