CN106840637B - Time-Frequency Analysis Method of GIS Mechanical Vibration Signal Based on Improved HHT Algorithm - Google Patents

Abstract

The invention discloses a kind of based on the GIS mechanical oscillation signal Time-Frequency Analysis Method for improving HHT algorithm, it measures GIS and operates normally GIS vibration signal under vibration signal and failure, the characteristic quantity of vibration signal is extracted using overall set empirical mode decomposition, then the signal pre-processed by EEMD is subjected to Hilbert transform pairs GIS mechanical oscillation signal and carries out time frequency analysis.GIS vibration signal can effectively be handled by carrying out time frequency analysis to GIS mechanical oscillation signal using the method for carrying out HT transformation again after the pretreatment of EEMD algorithm, to establish GIS mechanical fault diagnosis database, to realize that live live detection GIS mechanical breakdown provides theoretical foundation.

Description

Time-Frequency Analysis Method of GIS Mechanical Vibration Signal Based on Improved HHT Algorithm

technical field

The invention relates to a time-frequency analysis method for GIS mechanical vibration signals based on an improved HHT algorithm.

Background technique

Gas-insulated metal-enclosed switchgear (Gas Insulated Switchgear, GIS) will produce a variety of mechanical defects, such as loose screws, failure of the circuit breaker operating mechanism, and vibration of the transformer. Therefore, it is possible to improve the operational stability of GIS by diagnosing early defects in GIS by monitoring mechanical faults. The vibration signal generated by the internal failure of GIS is transmitted to the GIS cylinder through the medium. For the GIS equipment in operation, sensors are often placed on the surface of the cylinder to receive the transmitted vibration signal, so as to detect whether the GIS is operating abnormally, that is, a failure occurs. Because the vibration signal detection method has strong anti-interference ability and can complement the ultra-high frequency detection method widely used in the power grid, it is convenient and reliable to detect medium and low frequency signals.

However, at this stage, the fault research on GIS still focuses on the direction of partial discharge, and the collected vibration signals are mostly electromagnetic wave signals with high frequency. There are few researches on the wider and lower frequency mechanical fault signals. Due to the complexity of the GIS structure, there is less research on the vibration characteristics of the GIS in the field. Although there are literatures in the prior art that have measured the vibration signals of the GIS in operation, the data acquisition is difficult and the data volume is limited. Analyze the specific failure problems of GIS. Or perform multiple detection and processing on normal and abnormal HGIS vibration signals, and obtain a correlation diagram of the corresponding vibration amplitude and frequency with respect to frequency from a statistical point of view, but does not indicate the fault problem.

The commonly used method for processing the time-frequency of vibration signals is the HHT algorithm, which is to perform HT processing on the signal obtained after the EMD algorithm preprocessing. However, as a mature method for analyzing time-frequency signals, empirical mode decomposition still has many problems. The more serious ones are the phenomenon of false components and modal aliasing, which are manifested in:

1) A single IMF contains disparate scales;

2) The same scale appears in different IMFs.

Contents of the invention

In order to solve the above-mentioned problems, the present invention proposes a time-frequency analysis method for GIS mechanical vibration signals based on the improved HHT algorithm. A common GIS mechanical fault, the GIS vibration signal under three working conditions (normal signal, screw looseness, transformer oscillation) has been detected multiple times, and the improved HHT method is introduced to process the GIS mechanical vibration signal. frequency analysis, so that this processing method for nonlinear non-stationary signals can be more widely and deeply applied in the field of vibration signal processing.

First of all, in order to avoid ambiguity, the unified interpretation of terms is as follows:

GIS: Gas Insulated Switchgear, gas insulated metal-enclosed switchgear;

HT: Hilbert Transformation, Hilbert transformation;

IMF: Intrinsic Mode Function, Intrinsic Mode Function/Intrinsic Mode Function;

EEMD: Ensemble Empirical Mode Decomposition, overall set empirical mode decomposition;

EMD: Empirical Mode Decomposition, empirical mode decomposition.

In order to achieve the above object, the present invention adopts the following technical solutions:

A time-frequency analysis method for GIS mechanical vibration signals based on the improved HHT algorithm, which measures GIS vibration signals in normal operation and GIS vibration signals under fault conditions. The Hilbert transform of the signal is used for time-frequency analysis of the GIS mechanical vibration signal.

Set the abnormal vibration source at the connecting screw and transformer of GIS circuit breaker operating mechanism and disconnector, including loosening of screws and vibration of transformer based on winding deformation.

The sensor is used to detect the vibration signals under the three conditions of normal, loose screw and transformer oscillation on the lower side of the GIS peephole.

The original signal is sieved into intrinsic mode functions, and a Gaussian white noise signal is added to the signal during the sieving process to provide a consistent reference structure for the time-domain distribution of the remaining components after each EMD decomposition.

The specific process of using the overall set empirical mode decomposition method includes:

(1) Obtain an overall signal by superimposing a group of Gaussian white noise signals on the original signal;

(2) EMD decomposition is performed on the original signal to obtain IMF components of each order;

(3) Add different white noises to the original signal, repeating steps (1) and (2);

(4) Utilize the zero-mean principle of Gaussian white noise spectrum to eliminate the influence of Gaussian white noise as a time-domain distribution reference structure, and decompose the original signal;

(5) Perform Hilbert transform on each IMF obtained from the decomposition, and determine the marginal spectrum according to the obtained Hilbert transform spectrum to represent the cumulative amplitude distribution of each frequency point.

In described step (2), the concrete process of EMD decomposition comprises:

(2-1) Find the average value of its upper envelope and lower envelope according to the maximum point and minimum point of the original signal function;

(2-2) Calculate the difference between the original signal function and the average value obtained;

(2-3) regard the difference as a new original signal function, repeat steps (2-2)-(2-3), and iterate until the calculated difference satisfies the IMF condition;

(2-4) The difference when meeting the IMF condition is regarded as a new IMF, and the difference between the original signal function and the new IMF is obtained as a new original signal function, and steps (2-1)-(2) are repeated -4), calculate in turn the decomposed order IMF and the remaining components.

In the described step (3), the added Gaussian white noise number of times obeys: the error value between the original signal and the IMF of each order is equal to the difference between the amplitude of the Gaussian white noise and the square root of the overall number .

In the step (4), the original signal is decomposed into the sum of each order IMF and the remaining components.

In the step (5), Hilbert transform is performed on each IMF obtained through EEMD to obtain an analysis signal, and the instantaneous frequency is calculated according to the phase of the analysis signal.

In the step (5), each IMF is subjected to Hilbert transformation, and the transformed structure is recorded as a Hilbert spectrum, and the marginal spectrum is defined as an integral for the Hilbert spectrum.

Compared with prior art, the beneficial effect of the present invention is:

(1) The present invention is based on the time-domain analysis of the average empirical mode decomposition (EEMD), obtains the eigenmode function group, can obtain each component amplitude and phase situation of the vibration signal, and avoids false components and modal aliasing phenomenon appears, making the result more accurate;

(2) The present invention is based on the improved HHT analysis, can obtain the analysis of vibration signal amplitude and frequency of three working conditions, and obtain the characteristic criterion of vibration signal under different faults;

(3) Time-frequency analysis of GIS mechanical vibration signals by HT transformation after preprocessing by EEMD algorithm can effectively process GIS vibration signals, thereby establishing a GIS mechanical fault diagnosis database, in order to realize on-site live detection of GIS mechanical faults Provide a theoretical basis.

Description of drawings

The accompanying drawings constituting a part of the present application are used to provide further understanding of the present application, and the schematic embodiments and descriptions of the present application are used to explain the present application, and do not constitute improper limitations to the present application.

Figure 1(a), (b), and (c) are the time-domain diagrams of vibration signals collected by the oscilloscope under three working conditions;

Figure 2(a), (b), and (c) are the time-domain diagrams of the vibration signals of the three working conditions obtained by EEMD decomposition;

Figure 3(a), (b), and (c) are the HHT marginal spectra of the vibration signals of the three working conditions respectively;

Fig. 4 is a schematic flow chart of the present invention.

Detailed ways:

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

It should be pointed out that the following detailed description is exemplary and intended to provide further explanation to the present application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It should be noted that the terminology used here is only for describing specific implementations, and is not intended to limit the exemplary implementations according to the present application. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that when the terms "comprising" and/or "comprising" are used in this specification, they mean There are features, steps, operations, means, components and/or combinations thereof.

As introduced in the background technology, the fault research of GIS in the current stage still focuses on the direction of partial discharge, and the vibration signals collected are mostly electromagnetic wave signals with high frequency, which is difficult for a wide range of mechanical equipment with low frequency. Fault signal research is less problem, a GIS mechanical vibration signal time-frequency analysis method based on improved HHT algorithm is proposed.

In a typical implementation of the present application, as shown in Figure 4, the present invention measures the vibration signal on the surface of the GIS equipment, and constructs two common GIS mechanical failures, namely, screw loosening and transformer oscillation based on winding deformation. The GIS vibration signals under three working conditions (normal signal, screw loosening, and transformer oscillation) are detected at the same time, and the improved HHT method is introduced to process the GIS mechanical vibration signals, and the time-frequency analysis is performed on the three vibration signals, so that the type can be targeted at The processing method of nonlinear and non-stationary signal can be widely and deeply applied in the field of vibration signal processing.

The empirical mode decomposition method is to decompose complex signals into several Intrinsic Mode Functions (IMF), so that the instantaneous frequency obtained by HT (Hilbert Transform) can be applied to practice. The IMF must meet the following two conditions, namely: ①In the entire data segment, the number of extreme points and zero-crossing points should be equal or have a difference of at most 1; ②At any point, the upper envelope formed by the local maximum point and the local The mean value of the lower envelope formed by the minimum point is zero, that is, the signal is locally symmetrical about the time axis. The specific processing method is:

First, according to the maximum point and minimum point of the original signal function s(t), the average value of its upper envelope v ₁ (t) and lower envelope v ₂ (t) is calculated

Then find the difference h between s(t) and m, that is

s(t)-m=h (2)

Then regard h as a new s(t) to repeat the above operations and iterate until h satisfies the IMF condition. At this time, it is recorded as

c ₁ =h (3)

Treat c ₁ as a new IMF, as

s(t)-c ₁ =r (4)

Treat r as a new s(t), repeat the above process, and obtain c ₂ , c ₃ , c ₄ ... until r(t) basically shows a monotonic trend or |r(t)| is very small, which can be regarded as measurement error to stop. at this time,

That is, the original signal has been decomposed into n IMFs: c ₁ , c ₂ , c ₃ , c ₄ . . . , c _n , and a residual component r.

There are still many problems in empirical mode decomposition, among which the more serious ones are prone to false components and mode mixing. Therefore, on the basis of the EMD algorithm, the EEMD algorithm is proposed.

Aiming at the problem of EMD, it is found that except for the first-order component, the power spectrum of each order IMF presents the same band-pass characteristics, and the average frequency of the previous order IMF is approximately twice that of the latter order; at the same time, it is also found that the signal will Frequency aliasing occurs. Based on the above analysis, it is proposed to add white noise to the decomposed signal to supplement some missing scales, and thus the idea of overall average empirical mode decomposition is proposed.

Therefore, the main algorithm structure of EEMD is basically the same as that of the EMD algorithm: the original signal is sieved into an intrinsic mode function, and a Gaussian white noise signal w is added to the signal during the sieving process due to the modal aliasing phenomenon (t), providing a consistent reference structure for the temporal distribution of the remaining components after each EMD decomposition.

1.3 Specific steps of overall average empirical mode decomposition

1) Obtain an overall signal by superimposing a set of Gaussian white noise signals w(t) on the original signal x(t):

X(t)=x(t)+w(t) (6)

2) Perform EMD decomposition on X(t) to obtain the IMF components of each order:

3) Add different white noise w _i (t) to the original signal, and repeat steps 1) and 2).

4) Using the zero-mean principle of Gaussian white noise spectrum to eliminate the influence of Gaussian white noise as a time-domain distribution reference structure, at this time the IMF component c _n (t) can be expressed as

Gaussian white noise times added by EEMD obey formula (10)

In the formula: ε is the amplitude of Gaussian white noise; N is the total number; ε _n represents the error between the original signal and the IMF of each order after addition. In order to ensure the rapidity of the harmonic detection algorithm, ε is generally selected as 0.01, and N=200.

5) Therefore, the final original signal x(t) can be decomposed into

Hilbert transform

Hilbert transform is performed on each IMF obtained by EEMD,

x _i (t) = c _i (t) (12)

get the parsed signal,

z(t)=x _i (t)+iy _i (t)=a(t)e ^iθ(t) (14)

Where a(t)——instantaneous amplitude, θ(t)——phase,

The instantaneous frequency is calculated according to the formula

Do HT for each IMF to get

At this time, the residual term is ignored, and the above formula is called the Hilbert spectrum, denoted as

Further defining the marginal spectrum

The marginal spectrum can statistically represent the cumulative amplitude distribution of each frequency point of the entire set of data.

The test of the present invention utilizes a complete set of 110kV single-phase sub-box GIS equipment of a certain switch factory. Set the abnormal vibration source at the connecting screw and transformer of GIS circuit breaker operating mechanism and disconnector, including loosening of screws and vibration of transformer based on winding deformation. Fault 1 is loose screws, and fault 2 is deformation of transformer windings. Use the external piezoelectric acceleration sensor to detect three types of vibration signals (normal, loose screws, and transformer vibration) on the lower side of the GIS peephole.

Figure 1(a), (b), and (c) are the time-domain diagrams of the vibration signals collected by the oscilloscope when the GIS is normal, fault 1, and fault 2, respectively. From the time-domain diagrams, there are two types of vibration signals: screw loosening and transformer oscillation. The vibration signal amplitude under all conditions is greater than the normal vibration signal, and the vibration signal amplitude increases significantly under the screw loose fault. Intuitively, the vibration signal intensity under the transformer oscillation is the highest, and it is speculated that there is a superposition of frequency multiplied signals.

After simple denoising processing of the three groups of vibration signals, the eigenmode function groups of the GIS vibration signals under the three working conditions (Fig. From top to bottom, the figure shows each intrinsic mode function (imf) and residual component (res.). The eigenmode component correspondingly contains components of different frequency bands from high to low, and changes with the original signal.

After the Hilbert transform, the Hilbert yellow spectrum and marginal spectrum can be obtained. The variation of the amplitude with the frequency distribution can be seen more intuitively through the marginal spectrum. Compared with the HHT marginal spectrum of the vibration signal under normal operation (Fig. 3a), the fault characteristics can be seen very clearly in the HHT marginal spectrum of the GIS vibration signal under fault 1 (Fig. 3b) and fault 2 (Fig. 3c). Loose faults have oscillations at the fundamental frequency of 100Hz, and the amplitude is significantly higher than that during normal operation. The energy of the transformer oscillation signal is mainly concentrated in the range of 140-160Hz and 440-460Hz, and its amplitude is much higher than that during normal operation. Lower the amplitude of the base frequency around 100Hz. Combining the Hilbert yellow spectrum analysis, the characteristic criterion of two kinds of faults can be obtained.

The invention uses the improved HHT analysis method to analyze and process the vibration signal of the GIS equipment. Using the HHT method, in the laboratory, by setting two common mechanical faults (ie, screw looseness and transformer oscillation) for GIS equipment, this method is used to test the GIS equipment under three working conditions (normal, fault 1, fault 2) Analyze vibration signals. Concluded as follow:

(1) Based on the time-domain analysis of the average empirical mode decomposition (EEMD), the intrinsic mode function group is obtained, and the amplitude and phase of each component of the vibration signal can be obtained, avoiding the appearance of false components and modal aliasing, make the result more accurate.

(2) Based on the improved HHT analysis, the analysis of the amplitude and frequency of vibration signals under three working conditions can be obtained, and the characteristic criterion of vibration signals under different faults can be obtained.

(3) The frequency spectrum of the normal vibration signal is distributed around 100Hz, with a narrow frequency band; the frequency spectrum of the screw loose fault vibration signal is also distributed around 100Hz, with a wide frequency band, and the amplitude is significantly higher than the normal vibration signal; the transformer vibration fault vibration signal The frequency spectrum is distributed in the range of 140-160Hz and 440-460Hz.

The above shows that the method based on the improved HHT algorithm is effective for the time-frequency analysis of the mechanical vibration signal of the GIS equipment. By simulating different types of mechanical faults, a GIS mechanical fault diagnosis database can finally be established.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, there may be various modifications and changes in the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included within the protection scope of this application.

Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it does not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.

Claims (5)

1. A GIS mechanical vibration signal time-frequency analysis method based on the improved HHT algorithm, is characterized in that: measure the GIS normal operation vibration signal and the GIS vibration signal under the fault, utilize the overall set empirical mode decomposition to extract the characteristic quantity of the vibration signal, and then Perform time-frequency analysis on the GIS mechanical vibration signal through the Hilbert transform of the signal obtained through EEMD preprocessing; The specific process of using the overall set empirical mode decomposition method includes: (1) Obtain an overall signal by superimposing a group of Gaussian white noise signals on the original signal; (2) EMD decomposition is performed on the original signal to obtain IMF components of each order; (3) Add different white noises to the original signal, repeat steps (1) and (2), the number of Gaussian white noise added obeys: the error value between the original signal and the IMF of each order is equal to Gaussian white The difference between the magnitude of the noise and the square root of the population; (4) Using the zero-mean principle of Gaussian white noise spectrum to eliminate the influence of Gaussian white noise as a time-domain distribution reference structure, decompose the original signal into the sum of IMF and residual components of each order; (5) carry out Hilbert transform to each IMF obtained by decomposing, determine the marginal spectrum according to the obtained Hilbert transform spectrum, to characterize the cumulative amplitude distribution of each frequency point; This method is based on the time-domain analysis of the average empirical mode decomposition EEMD, obtains the intrinsic mode function group, obtains the amplitude and phase of each component of the vibration signal, and avoids the appearance of false components and modal aliasing; Based on the analysis of the improved HHT algorithm, the analysis of the amplitude and frequency of the vibration signal under three working conditions is obtained, and the characteristic criterion of the vibration signal under different faults is obtained; Set abnormal vibration sources at the connection screws and transformers of GIS circuit breaker operating mechanism and disconnector, including loose screws and transformer vibration based on winding deformation; The sensor is used to detect the vibration signals under the three conditions of normal, loose screw and transformer oscillation on the lower side of the GIS peephole. 2. A kind of GIS mechanical vibration signal time-frequency analysis method based on improved HHT algorithm as claimed in claim 1, is characterized in that: original signal is sieved into intrinsic mode function, adds a Gaussian white to signal in sieving process A noise signal that provides a consistent reference structure for the temporal distribution of the remaining components after each EMD decomposition. 3. a kind of GIS mechanical vibration signal time-frequency analysis method based on improved HHT algorithm as claimed in claim 1, is characterized in that: in described step (2), the concrete process of EMD decomposition comprises: (2-1) Find the average value of its upper envelope and lower envelope according to the maximum point and minimum point of the original signal function; (2-2) Calculate the difference between the original signal function and the average value obtained; (2-3) regard the difference as a new original signal function, repeat steps (2-2)-(2-3), and iterate until the calculated difference satisfies the IMF condition; (2-4) The difference when meeting the IMF condition is regarded as a new IMF, and the difference between the original signal function and the new IMF is obtained as a new original signal function, and steps (2-1)-(2) are repeated -4), calculate in turn the decomposed order IMF and the remaining components. 4. a kind of GIS mechanical vibration signal time-frequency analysis method based on improved HHT algorithm as claimed in claim 1, is characterized in that: in described step (5), each IMF that obtains through EEMD is done Hilbert transformation, obtains Analyze the signal, and calculate the instantaneous frequency according to the phase of the analyzed signal. 5. a kind of GIS mechanical vibration signal time-frequency analysis method based on improved HHT algorithm as claimed in claim 1, is characterized in that: in described step (5), carry out Hilbert transform to each IMF, transform The final structure is recorded as the Hilbert spectrum, and the marginal spectrum is defined as the integral of the Hilbert spectrum.