CN-116994606-B - Sound diagnosis method for power equipment

CN116994606BCN 116994606 BCN116994606 BCN 116994606BCN-116994606-B

Abstract

The invention discloses a power equipment sound diagnosis method, which belongs to the technical field of power equipment fault diagnosis and comprises the steps of collecting a first signal, preprocessing the first signal to obtain a second signal, decomposing the second signal to obtain a third signal, screening the third signal based on the collection time, and judging the third signal. When the transformer sound is collected, the detection points are all the same in advance calibrated on the transformer each time, the sound sensor is installed at the detection points, the sound sensor of the same model is adopted, the influence of collected transformer sound change caused by the influence of the model, the erection angle, the azimuth and the far and near of the sound sensor is eliminated, the recognition accuracy is improved, the purpose of denoising is achieved by preprocessing the first signal, then the second signal is decomposed into a single third signal, the interference of sound change to fault judgment caused by various abnormal sounds is eliminated, and the recognition accuracy is improved.

Inventors

QIN ZHENG
LU XIN
LIN SONG
FENG YANWEI
HUANG ZHIRONG
Zeng Zuowei
LU FEI
LI ZHE
LI JIANING

Assignees

中国能源建设集团广西电力设计研究院有限公司
上海交通大学

Dates

Publication Date: 20260512
Application Date: 20230714

Claims (1)

1. A method of sound diagnosis of an electrical device, comprising: Collecting a first signal; preprocessing the first signal to obtain a second signal; decomposing the second signal to obtain a third signal; screening the third signal based on acquisition time; Judging the third signal; calibrating detection points on a transformer in advance; Mounting a sound sensor to the detection point; The method comprises the steps of controlling a sound sensor to synchronously acquire original sound signals of a transformer, and taking the original sound signals of the transformer as first signals; recording the corresponding acquisition time of the first signal meter; Preprocessing the first signal to obtain a second signal, wherein the step of obtaining the second signal comprises the following steps: Performing multi-scale wavelet decomposition on the first signal based on a self-adaptive hierarchical threshold wavelet algorithm, analyzing short-time high-frequency components and low-frequency components, denoising through a set threshold processing detail component, and defining a mathematical expression of the first signal as follows: Wherein, the Representing the first signal, i.e. the noisy signal, The original signal is represented by a representation of the original signal, Representing a noise signal; For a pair of Performing discrete wavelet transformation, wherein the mathematical expression is as follows: Wherein, the Representing the discrete wavelet(s) and, Representation of Is used for the transformation of the wavelet coefficients of (a), Representation of Wavelet transform coefficients of (a); For a pair of Wavelet decomposition is performed by a threshold function Filtering the noise-containing signal, wherein the mathematical expression is as follows: Wherein, the The threshold value is indicated and the threshold value, The upper threshold value is represented as such, A lower threshold value is indicated and, , To obtain Changes in Is changing the value of (2) At the same time as the lower threshold value The critical area where the upper threshold value and the lower threshold value are located is changed, the critical area is shrunk by adjusting the m value, the removal proportion of the noise coefficient is further controlled, the noise-containing signal is removed to the maximum extent, and a second signal is obtained; decomposing the second signals to obtain a plurality of third signals, wherein the mathematical expression is as follows: Wherein k represents the number of decompositions, Which represents the center frequency of the wave, Representation of Is used to determine the instantaneous amplitude of (a), Is of instantaneous frequency of In the following Within the interval range of (2), Can be regarded as an amplitude value of At a frequency of Is a harmonic signal of (a); screening the third signal based on the acquisition time includes: comparing the start time and the stop time of the third signal which are different; if the time interval between the start time and the stop time of the two third signals is smaller than 0.1s, the two third signals are independently existed; If the time interval between the start time and the stop time of the two third signals is smaller than 0.1s, the two third signals are represented to be in a concurrency relation; the determining of the third signal includes: collecting audio information of defects of each part of the transformer; Performing short-time spectrum analysis on the audio information, and converting the audio signal into a first spectrogram recorded with joint distribution information of a time domain and a frequency domain by using fast Fourier change; sorting the first spectrograms corresponding to the defects of all parts of the transformer, and establishing an audio sample library; performing short-time spectrum analysis on the third signal, and converting the audio signal into a second spectrogram by using fast Fourier change; comparing the second spectrogram with the first spectrogram in the audio sample library; if the joint distribution information of the time domain and the frequency domain of the second spectrogram is the same as that of the first spectrogram, the defects corresponding to the first spectrogram exist on the parts on the transformer.

Description

Sound diagnosis method for power equipment Technical Field The invention relates to the technical field of power equipment fault diagnosis, in particular to a power equipment sound diagnosis method. Background With the rapid development of the electric power technology, the scale of the power grid is continuously enlarged, and especially the ultra-high voltage, global energy internet and the like of China are proposed to indicate the direction for the prospect of the electric power industry. The large power grid, the extra-high voltage system puts higher demands on the reliability of the high-voltage power equipment, which means that the detection and diagnosis of the high-voltage power equipment put higher demands. In addition to the signals conventionally widely used for detection, the concept of sound for detection of electrical equipment has been proposed by some scholars as the capabilities of sound collection and processing techniques are improved. The sound is an important index for analyzing the running state of equipment, and the sound signal contains a large amount of vibration information. In addition, the equipment does not need to stop running or contact the equipment, the device is simple, the signal acquisition is convenient, the installation mode is flexible, and the normal running of the equipment is not interfered The fault data of the actual running transformer can be caused by various factors causing abnormal running and fault states of the transformer, and the fault data of the actual running transformer can be caused by structural deformation, breakage and abnormal electrothermal stress. The generation of abnormal sounds of the transformer is probably not a single abnormal point, in practice, the sounds are often changed due to the fact that a plurality of positions and a plurality of conditions are abnormal, the analysis is difficult due to the fact that the possible fault factors are various, the arrangement and combination of different numbers of abnormal points are difficult, the background of the running environment of the transformer is complex, a large amount of noise exists to influence the analysis, and the collected sounds of the transformer are changed due to the influences of the model number, the erection angle, the azimuth, the distance and the like of the sound sensor, so that the recognition accuracy is reduced. Disclosure of Invention The invention solves the technical problems that abnormal sounds of the transformer are not generated as single abnormal points, in practice, a plurality of positions and conditions are often abnormal at the same time to cause sound change, the analysis is difficult due to various possible fault factors, the arrangement and combination of different numbers of abnormal points are difficult, the background of the running environment of the transformer is complex, a large amount of noise exists to influence the analysis, and the collected sounds of the transformer are changed due to the influences of the model number, the erection angle, the azimuth, the distance and the like of the sound sensor, so that the recognition accuracy is reduced. The technical scheme includes that the power equipment sound diagnosis method comprises the steps of collecting a first signal, preprocessing the first signal to obtain a second signal, decomposing the second signal to obtain a third signal, screening the third signal based on the collection time, and judging the third signal. The method for diagnosing the sound of the power equipment comprises the steps of calibrating a detection point on a transformer in advance, installing a sound sensor at the detection point, controlling the sound sensor to synchronously collect original sound signals of the transformer, taking the original sound signals of the transformer as first signals, and recording corresponding collection time of a signal meter of the first signals. The power equipment sound diagnosis method according to the present invention is a preferable scheme, wherein preprocessing the first signal to obtain a second signal includes: Performing multi-scale wavelet decomposition on the first signal based on a self-adaptive hierarchical threshold wavelet algorithm, analyzing short-time high-frequency components and low-frequency components, denoising through a set threshold processing detail component, and defining a mathematical expression of the first signal as follows: Wherein, the Representing the first signal, i.e. the noisy signal,The original signal is represented by a representation of the original signal,Representing a noise signal. As a preferable mode of the sound diagnosis method of the electric power equipment, the invention comprises the following steps ofPerforming discrete wavelet transformation, wherein the mathematical expression is as follows: Wherein, the Representing the discrete wavelet(s) and,Representation ofIs used for the transformation of the wavelet coefficients of (a),Represe