A Novel Method for Extracting Near-Infrared Spectral Features based on a Random Frog Algorithm for the Performance Analysis of XLPE Cables

Near-infrared spectroscopy (NIR) shows great potential for condition assessment of cross-linked polyethylene (XLPE) cables. In order to better extract the features of near-infrared spectroscopy for XLPE samples, the paper uses a random frog (RF) algorithm for feature extraction of NIR. First, the paper slices samples from 110 kV XLPE cables and conducts accelerated thermal aging tests at 105°C, 135°C, 155°C, and 180°C, obtaining aged samples. Then, NIR tests and elongation at break (EAB) tests are performed on the samples to acquire the near-infrared spectroscopy and EAB curves. Finally, using variance as the evaluation indicator, RF calculations are performed on the near-infrared spectroscopy data, resulting in six sets of feature data: 135°C-168h, 135°C-840h, 180°C-14h, 180°C-26h, 180°C-30h, and 180°C-32h. Further analysis of these six sets of feature data reveals that the NIR peak values and EAB change curves are highly consistent with aging time. The slope of the curve at 135°C is very close, and the trend of the curve at 180°C is consistent. The results indicated that the aging time curve characteristics of the near-infrared spectroscopy and EAB after RF selection are highly consistent.