Feature value extraction method applied to ultrasonic partial discharge detection

Abstract

The invention discloses a feature value extraction method applied to ultrasonic partial discharge detection. According to the invention, a wavelet packet decomposition and mutual information amount-based IMF component screening method is adopted. Firstly, wavelet packet decomposition is performed. Preliminary frequency division is carried out on the signal. The IMF component of the single-frequency component is obtained. The initial bandwidth of the EMD decomposition signal is reduced. The ultrasonic signal is decomposed more accurately, the problems of modal aliasing and false IMF componentsexisting in EMD decomposition are effectively solved by applying a mutual information amount method, the accuracy of time-frequency analysis is improved, waveform characteristics are extracted more effectively, and finally, the ultrasonic partial discharge signal is classified by applying a neural network.

Description

technical field [0001] The invention belongs to the field of ultrasonic technology, in particular to a feature value extraction method applied to ultrasonic partial discharge detection. Background technique [0002] With the continuous development of modern power grids, people have put forward higher standards for the safety and reliability of power grid operation. Partial discharge detection, as an effective live detection method without damaging the equipment itself, has attracted more and more attention. and favored. [0003] When the insulation part of electrical equipment deteriorates due to various reasons, it will cause partial discharge in severe cases, and even insulation breakdown and damage, of which more than 70% are caused by partial discharge. Carrying out partial discharge live detection is a need for the safety of the power system. It is convenient for electric power workers to grasp the insulation status of the substation operating equipment in a timely man...

AI Technical Summary

Problems solved by technology

For example, the Fourier transform analysis method loses the time information during the transformation, and cannot judge when a specific signal occurs; although the wavelet transform can better analyze the time-frequency characteristics of partial discharge signals, it relies too much on the wavelet basis. The selection of the wavelet base is limited, and the energy leakage will occur during processing, so it is difficult to make an accurate time-frequency analysis of the signal. The empirical mode decomposition is decomposed based on the information of the signal itself, which is suitable for dealing with nonlinear and non-stationary signals, but the decomposition In the process, there are problems such as envelope fitting, mode aliasing, end effect and false IMF components, which lead to inaccurate signal analysis results

Images