Transformer fault classification and identification method based on vibration analysis method

Abstract

The invention discloses a transformer fault classification and identification method based on a vibration analysis method. The method comprises the following steps: S1) selecting a transformer test object, and collecting transformer vibration signals under different states as sample data; S2) calculating the sample data by utilizing a Hilbert-Huang transform EEMD (ensemble empirical mode decomposition) method to obtain an intrinsic mode function; S3) extracting characteristic vector V in each intrinsic mode function component; S4) carrying out dimensionality reduction on the characteristic vector through a principal component analysis method, coordinates being projected to a two-dimensional image; S5) carrying out classification on the sample data through a K neighbor method; S6) calculating the distance between a test sample and an original sample through a distance formula; S7) carrying out mode recognition; and S8) outputting corresponding transformer fault type in mode recognition. Operation state of a transformer can be judged visually and effectively.

Description

technical field [0001] The invention relates to the field of electric equipment detection, in particular to a transformer fault classification and identification method based on a vibration analysis method. Background technique [0002] Among the various equipment in the power system, the transformer is an expensive and important equipment, and its reliability is of great significance to ensure the safe operation of the power grid. The fault statistics of transformers over the years show that transformer windings and iron cores are the most faulty components, and transformer accidents caused by windings and iron cores account for 70% of the total transformer accidents. [0003] At present, there are more and more methods for detecting the operation status of transformers. The short-circuit impedance method judges the winding deformation by measuring the reactance of the winding off-line and observing the change of its impedance value, but this method has low sensitivity and...

AI Technical Summary

Problems solved by technology

However, according to EEMD, the feature vector of the transformer vibration signal is extracted, the feature vector is multiple sets of data, and when there are many test samples, the test sample will become a multidimensional array

Multidimensional arrays are not conducive to calculations, and are not conducive to observation in reality; at the same time, Fisher's discriminant method can only achieve linear classification, but cannot achieve nonlinear classification. In the experimental data, due to the uncertainty of the experimental measurement data, it is inevitable that there will be wrong data. These erroneous data will affect the effect of linear classification

Images