With the rapid development of urbanization, a large number of power cables have entered the city's distribution network, which required to pay more attention to the inspection of power cables’ quality. However, the current testing methods for power cable insulation present long testing period, which are unable to ensure full coverage sampling. In order to address the problem, in this paper, non-destructive testing method based on near-infrared (NIR) spectroscopy is studied and applied to the prediction of elongation at break of cross-linked polyethylene (XLPE) cable insulation. 28 kinds of power cables with different service time were studied and the NIR spectra were obtained with the preprocess of smoothing, standard normal variable transformation (SNV), and second-order derivation. In addition, random frog (RF) algorithm was employed to select the wavelengths that are closely related to the structures which are responsible to the properties. 10 wavelengths were further selected for the subsequent modeling by partial least squares regression (PLSR) to predict the elongation at break. The great precision is presented in the results which showed only 1.1% error in prediction. The study provides a convenient method to measure the elongation at break of XLPE insulation without any destruction.

Cable insulation,cross-linked polyethylene,Near-infrared,Elongation at break