Research on fault diagnosis of transformer based on laser induced fluorescence technology

Abstract

Transformer fault diagnosis is an indispensable part of normal operation of power equipment. Aiming at the problem that traditional gas chromatography and other methods take a long time to detect fault types, this paper proposes a laser-induced fluorescence spectroscopy (LIF) technology, combined with differential mutation brainstorm optimization algorithm (DBSO) to optimize the ELM model to identify transformer fault diagnosis types. Four different transformer oils were selected as experimental samples, including thermal fault oil, electrical fault oil, local damp oil and crude oil. LIF technology was used to obtain different spectral images of four oil samples. Firstly, the obtained fluorescence spectra were pre-treated by MSC and Normalize, and the dimensionality was reduced by PLS. Then, the data after dimension reduction are input into the ELM model for training, and the BSO algorithm and DBSO algorithm are used to optimize the parameters of ELM. Finally, the experiments show that the DBSO-ELM model pre-processed by MSC has the highest recognition accuracy, the goodness of fit (R2) is 1, and the mean square error (MSE) reaches 3.205e-31, which is higher than that of ELM model and BSO-ELM model. In the case of the same pre-processing, the fitness values based on the mean square error rate of the training set are lower than those of the other types of recognition algorithm models. Therefore, the MSC-PLS-DBSO-ELM model has the best recognition effect, which can be extended to transformer fault diagnosis and improve the accuracy and safety of power equipment detection.