Aging Assessment of Oil-Impregnated-Paper Electrical Equipment via Near Infrared Spectroscopy Powered by Improved PCA-RBF-NN: Modelling and Field Practices

Yuan Li,Wenbo Zhang,Han Li,Yaoyu Xu,Guanjun Zhang
DOI: https://doi.org/10.1109/tdei.2021.009813
IF: 2.509
2021-12-01
IEEE Transactions on Dielectrics and Electrical Insulation
Abstract:We report our recent progress in quantitative aging assessment of the oil-impregnated-paper (OIP) equipment (i.e., degree of polymerization, DP) by near infrared spectroscopy (NIRS). The NIRS database is built by incorporating 8 types of insulating paper and total 478 differently aged samples. We propose an improved PCA-RBF-NN model to address the nonlinear correlation between DP of insulating paper and spectra, and hence strengthening the prediction accuracy for field assessment. In the improved model, the principle component analysis (PCA) and the filtering layer are two essential procedures for eliminating the noises and unrelated information from the spectra. The field practices show that the improved PCA-RBF-NN model owns better performance than the classic PLS model and general RBF-NN model on the disassembled bushing (RMSE: 56 vs 109 vs 124) and transformer (RMSE: 50 vs 237 vs 244), respectively. The NIRS powered by the improved algorithm can provide a rapid solution to the aging condition assessment of the OIP power equipment in the field.
engineering, electrical & electronic,physics, applied
What problem does this paper attempt to address?