GIS mechanical fault diagnosis method and device

Patent metadata

Field	Value
Publication number	US-12368288-B2
Application number	US-202017117106-A
Country	US
Kind code	B2
Filing date	Dec 9, 2020
Priority date	Nov 29, 2019
Publication date	Jul 22, 2025
Grant date	Jul 22, 2025

Abstract

A GIS mechanical fault diagnosis method and the device are disclosed. The method includes: collecting vibration signals to be measured of various excitation sources of GIS in mechanical operation; performing wavelet packet-feature entropy vector extraction on the vibration signals to be measured, when it is determined that the vibration signals to be measured are abnormal according to standard vibration signals in the normal state; inputting the extracted wavelet packet-feature entropy vectors into the pre-trained BP neural network for GIS mechanical fault identification, and outputting the corresponding fault. The disclosure integrates the vibration signals under the action of various excitation sources, extracts the feature entropy vectors according to the entropy theory, and constructs and trains a BP neural network that can classify and recognize various GIS mechanical faults, so as to perform comprehensive and effective GIS mechanical faults diagnose.

First claim

The inventionn claimed is:

1. A gas insulated switchgear (GIS) mechanical fault diagnosis method, comprising:
collecting vibration signals to be measured of various excitation sources of GIS in mechanical operation;
performing wavelet packet-feature entropy vector extraction on the vibration signals to be measured, in response to determining that the vibration signals to be measured are abnormal according to standard vibration signals in normal state; and
inputting extracted wavelet packet-feature entropy vectors into a pre-trained back propagation (BP) neural network for GIS mechanical fault identification, and outputting corresponding fault;
wherein the vibration signals to be measured comprise vibration signal excited by an operating mechanism and vibration signal excited by electromagnetic force, and the operating mechanism comprises a circuit breaker, an isolating switch, and a grounding switch;
wherein determining that the vibration signals to be measured are abnormal according to standard vibration signals in normal state comprises:
extracting an envelope of the vibration signal excited by the operating mechanism to obtain envelope area by using Hilbert method; at the same time, extracting vibration energy of the vibration signal excited by the electromagnetic force at set frequency by using fast fourier transform (FFT) method;
determining that the vibration signals to be measured are abnormal in response to that the envelope area corresponding to the vibration signal excited by at least one operating mechanism satisfies a set first formula, and/or the vibration energy satisfies a set second formula; in which,
the first formula is:
| E S −E S0 |/E S0 ×100%≥5%

E S is the envelope area corresponding to the vibration signal excited by a certain operating mechanism, and E S0 the envelope area corresponding to the standard vibration signal excited by the operating mechanism; and
the second formula is:
| E Q −E Q0 |/E Q0 ×100%≥5%

E Q is the vibration energy of the vibration signal excited by electromagnetic force at the set frequency, and E Q0 is the vibration energy of the standard vibration signal excited by electromagnetic force at the set frequency;
wherein the set frequency is 100 Hz;
a collection time of the vibration signal excited by the electromagnetic force is 2 seconds, and a sampling frequency is 20 KHz; and
a collection period of the vibration signal excited by the operating mechanism is from a start time to an end time of an operation of the operating mechanism and 0.5 second extended.

2. The GIS mechanical fault diagnosis method of claim 1 , wherein determining that the vibration signals to be measured are abnormal in response to that the envelope area corresponding to the vibration signal excited by at least one operating mechanism satisfies the set first formula, and/or the vibration energy satisfies the set second formula comprises:
determining that the vibration signals to be measured are abnormal in response to that the envelope areas corresponding to the vibration signals excited by at least two operating mechanisms satisfy the set first formula, and the vibration energy of the vibration signal excited by the electromagnetic force satisfies the set second formula.

3. The GIS mechanical fault diagnosis method of claim 1 , wherein after collecting vibration signals to be measured of various excitation sources of GIS in mechanical operation, the method further comprises:
performing wavelet soft threshold denoising processing on the vibration signals to be measured; and
for the vibration signals to be measured after denoising processing, performing wavelet packet decomposition by using db10 of Daubechies wavelet series as wavelet base function.

4. The GIS mechanical fault diagnosis method of claim 3 , wherein performing wavelet packet-feature entropy vector extraction on the vibration signals to be measured comprises:
performing k-layer wavelet packet decomposition on the vibration signals to be measured, and reconstructing the vibration signals at 2 k -th node of k-th layer; and
extracting an envelope of each reconstructed vibration signal, and dividing each envelope into N segments according to principle of equal integral energy; and extracting the wavelet packet-feature entropy by normalized value process, thereby obtaining the wavelet packet-feature entropy vectors.

5. The GIS mechanical fault diagnosis method of claim 4 , wherein the wavelet soft threshold is 2.1, and k=3; N=15.

6. The GIS mechanical fault diagnosis method of claim 1 , wherein training method of the BP neural network comprises:
setting 8 neurons in an input layer and 3 neurons in an output layer, with transfer functions all being tan-sigmoid, training function being traingd, and expected error being 0.01;
using the standard vibration signals in the normal state and the wavelet packet-feature entropy vectors of the vibration signals corresponding to various GIS mechanical faults as input to perform training of the BP neural network; and
determining that a number of hidden layer neurons is 10 and a step size is 0.31 according to training results.

Assignees

• Electric Power Science & Res Institute Of State Grid Tianjin Electric Power Company
• State Grid Tianjin Electric Power Co
• State Grid Corp China

Inventors

• Qu Bin
• Zhang Li
• Chen Rong
• Zhou Liansheng
• Gan Zhiyong
• Zhang Chi
• Li Guohao
• He Jin
• Wang Kun
• Wang Ziyue
• Wang Jian
• Fan Wei

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/048

  Activation functions · CPC title

• G06F2218/12

  Classification; Matching · CPC title

• G06F2218/06

  by applying a scale-space analysis, e.g. using wavelet analysis · CPC title