CN113640566A - FOCT drift fault feature extraction method - Google Patents

Abstract

The invention discloses a FOCT drift fault feature extraction method, which comprises the steps of obtaining an output signal x (t) of an optical fiber current transformer containing a drift error signal; decomposing the output signal x (t) into a plurality of intrinsic scale components containing error information and a residual signal containing error information using a modified local feature scale decomposition algorithm; respectively calculating sample entropies of the intrinsic scale components and the residual signals to form an error signal component data set; superposing error signal components in the error signal component data set to obtain an error signal; and selecting a peak structure with monotonically rising amplitude within a certain time in the error signal as the characteristic of the drift fault according to the time domain image of the error signal. The method is suitable for extracting the fault characteristics of the optical fiber current transformer, and improves the accuracy and the real-time performance of characteristic extraction, thereby improving the fault research and judgment speed.

Description

FOCT drift fault feature extraction method

Technical Field

The invention relates to a FOCT drift fault feature extraction method, and belongs to the technical field of fault detection of optical fiber current transformers.

Background

The optical fiber current transformer (FOCT) based on the Faraday effect has the advantages of good insulating property, high reliability, broadband frequency domain and the like, and is widely applied to high-voltage direct-current transmission engineering. However, the optical fiber current transformer is affected by the complex environment in the substation in the long-term operation process, and the performance of the optical fiber current transformer is degraded and even causes operation accidents.

The faults of the optical fiber current transformer are mainly reflected as complete failure faults, fixed deviation faults, drifting deviation faults, precision reduction faults and the like. The drift deviation fault is the most common FOCT fault, and the fault reason can be analyzed by collecting fault signals and extracting fault characteristics, so that a basis is provided for quick diagnosis and positioning of the fault.

The signal feature extraction method for the optical fiber current transformer is generally divided into three categories: time domain signal analysis, frequency domain signal analysis and time frequency signal analysis. The existing mature time-frequency signal analysis methods mainly comprise Hilbert transform, wavelet transform, empirical mode decomposition and the like. However, the methods have the phenomena of over-enveloping and under-enveloping, end point response, modal aliasing and the like, so that errors exist in the decomposition result, the operation process is too complex, the real-time performance of the extraction result is poor, and the subsequent fault discrimination speed is influenced.

Disclosure of Invention

The purpose is as follows: the invention provides a FOCT drift fault feature extraction method, aiming at solving the problems of less envelope of envelope, end point response, complex operation and the like in the prior art.

The technical scheme is as follows: in order to solve the technical problems, the technical scheme adopted by the invention is as follows:

the invention discloses a FOCT drift fault feature extraction method, which comprises the following steps:

step 1: and acquiring an output signal x (t) of the optical fiber current transformer containing the drift error signal.

Step 2: the output signal x (t) is decomposed into a plurality of intrinsic scale components containing error information and a residual signal containing error information using a modified local feature scale decomposition algorithm.

And step 3: respectively calculating the sample entropies of the intrinsic scale components and the residual signals to obtain the sample entropies of the intrinsic scale components and the residual signal sample entropies, taking the sample entropies of the intrinsic scale components larger than a threshold value as error signal components, taking the sample entropies of the residual signals larger than the threshold value as error signal components, and combining the error signal components to form an error signal component data set.

And 4, step 4: and overlapping the error signal components in the error signal component data set to obtain an error signal.

And 5: and selecting a peak structure with monotonically rising amplitude within a certain time in the error signal as the characteristic of the drift fault according to the time domain image of the error signal.

Preferably, the output signal x (t) of the optical fiber current transformer containing the drift error signal is calculated according to the following formula:

x(t)＝0.5K[1+cos(Δθ+ωt)]+K₁t'

wherein: k denotes the parameter of the opto-electronic circuit, K ═ K_pLI₀，K_pIs the photoelectric conversion coefficient of the photodetector, L is the optical path loss, I₀For the light intensity transmitted by the light source, Δ θ is the faraday effect, Δ θ is 4VNI, V is the Verdet constant of the optical fiber, N is the number of sensing loops of the optical fiber, I is the current to be measured, ω represents the signal frequency, and t is the time K₁The drift deviation coefficient is shown, and t' shows the time when the deviation occurs.

Preferably, the improved local feature scale decomposition algorithm includes the following steps:

(2-1) calculating all extreme points X of the output signal X (t)_kAnd the corresponding time tau_kM, where k is 1,2, M is the number of extreme points.

(2-2) calculating any two adjacent maximum or minimum extreme points X_kAnd X_k+2Formed straight line L_kM-2, find all adjacent maximum or minimum extremum points X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1And calculating τ_k+1The function value of (A)_k+1And corresponding L_k+1The value of (c):

L_k+1＝αA_k+1+(1-α)X_k+1,k＝1,2,...M-2

wherein:

A_k+1represents the extreme point X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1The function value of (c); l is_k+1Is represented by A_k+1And X_k+1Mean, maximum or minimum extreme point X of_k+2Corresponding time τ_k+2And α represents a scale factor, and typically, α is 0.5.

(2-3) A finally obtained_kAnd L_kThe subscript of (1) is 2₁And M₁Performing mirror extension estimation to obtain the left and right endsExtreme point (τ)₀,X₀)，(τ_M+1,X_M+1) Separately find A₁,A_MAnd L₁,L_M。

(2-4) interpolation L using piecewise cubic Hermite₂,L₃,...L_M-1Fitting all the resulting baseline curves BL₁(t) separately fitting L using piecewise linear transformation₁-L₂,L_M-1-L_M，BL₁(t) denotes the baseline signal

(2-5) separating the baseline signal BL from the output signal x (t)₁(t) obtaining a new signal h₁(t) is

h₁(t)＝x(t)-BL₁(t)

(2-6) judgment of h₁(t) whether or not the condition for discriminating the ISC component is satisfied:

strict monotonicity exists between any two adjacent maximum values and minimum values in the whole data section of the output signal; the ratio of the function value corresponding to the extreme value point between any two maximum (small) value points in the whole data segment to the extreme value corresponding to the function value is kept unchanged.

If so, let ISC1 be h₁(t) output as a first ISC component; repeating the steps (2-1) - (2-6) k times until h_k(t) satisfies ISC component condition, i.e., ISCK ═ h_k(t)。

(2-7) separating ISC1 from the output signal to obtain a new signal r₁(t)：

r₁(t)＝x(t)-ISC1

(2-8) removing r₁(t) repeating the steps (2-1) to (2-7) k times as a residual component until r_k(t) is constant or monotonic to obtain k residual components, and the k residual components are summed to obtain a residual component r (t).

(2-9) the final decomposition of the output signal x (t) into k ISCs_j(t) a component and a remaining component r (t), i.e.

As a preferred scheme, the method for calculating sample entropy by using the intrinsic scale component and the residual signal comprises the following steps:

(3-1) decomposing the intrinsic scale component and the residual signal into a vector sequence X (i) with a one-dimensional array of m

X(i)＝[y₁,y₂,...,y_i...,y_m],i＝1,2,...,m

y_iIs the ith one-dimensional array of the decomposition.

(3-2) defining vector y_iAnd y_jA distance d (i, j) therebetween, wherein j is 1,2,.., m (i ≠ j)

d(i,j)＝|y_i-y_j|

(3-3) setting a threshold value r, counting the number b of distances d (i, j) smaller than r, and then making a ratio of the number b to the total number m of the distances to be recorded as

Namely, it is

In the formula: 1, 2.

(3-4)B^m(r) is

Average value of (2)

(3-5) increasing the number of dimensions to m +1, repeating the steps (3-1) to (3-4) to obtain B^m+1(r)

(3-6) sample entropy formulas of the intrinsic scale component and the residual signal are as follows:

has the advantages that: according to the FOCT drift fault feature extraction method, the problems of envelope under-enveloping, end point response, complex operation and the like existing in the existing fault feature extraction algorithm are solved, and the fitting precision is improved by using three times of Hermite (Hermite) interpolation; and the left end point and the right end point respectively use piecewise linear change, so that the end point effect is reduced. And decomposing the output signal of the optical fiber current transformer into infinite harmonic components by using a Bessel decomposition algorithm, and solving primary current by using the ratio of fundamental waves to second harmonics. And extracting fault components in the current by using a local feature scale decomposition algorithm, calculating sample entropies of the components, and judging error features according to values of the sample entropies to obtain fault feature vectors. The method is suitable for extracting the fault characteristics of the optical fiber current transformer, improves the accuracy and the real-time performance of characteristic extraction, and further improves the fault research and judgment speed.

Drawings

FIG. 1 is a flow chart of the disclosed method.

FIG. 2 is a flow chart of an improved local feature scale decomposition algorithm in the method disclosed by the invention.

FIG. 3 is a diagram illustrating the error signal recombination result under drift deviation.

Detailed Description

The present invention will be further described with reference to the following examples.

As shown in FIG. 1, the invention discloses a FOCT drift fault feature extraction method, which comprises the following steps:

step 1: and acquiring an output signal x (t) of the optical fiber current transformer containing the drift error signal.

Step 2: the output signal x (t) is decomposed into a plurality of intrinsic scale components containing error information and a residual signal containing error information using a modified local feature scale decomposition algorithm (LCD).

And step 3: respectively calculating the sample entropies of the intrinsic scale components and the residual signals to obtain the sample entropies of the intrinsic scale components and the residual signal sample entropies, taking the sample entropies of the intrinsic scale components larger than a threshold value as error signal components, taking the sample entropies of the residual signals larger than the threshold value as error signal components, and combining the error signal components to form an error signal component data set.

And 4, step 4: and overlapping the error signal components in the error signal component data set to obtain an error signal.

And 5: and selecting a peak structure with monotonically rising amplitude within a certain time in the error signal as the characteristic of the drift fault according to the time domain image of the error signal.

As shown in fig. 2, the output signal x (t) of the fiber current transformer containing the drift error signal is calculated according to the following formula:

x(t)＝0.5K[1+cos(Δθ+ωt)]+K₁t'

wherein: k denotes the parameter of the opto-electronic circuit, K ═ K_pLI₀，K_pIs the photoelectric conversion coefficient of the photodetector, L is the optical path loss, I₀For the light intensity transmitted by the light source, Δ θ is the faraday effect, Δ θ is 4VNI, V is the Verdet constant of the optical fiber, N is the number of sensing loops of the optical fiber, I is the current to be measured, ω represents the signal frequency, and t is the time K₁The drift deviation coefficient is shown, and t' shows the time when the deviation occurs.

The improved local feature scale decomposition algorithm comprises the following steps:

(2-1) calculating all extreme points X of the output signal X (t)_kAnd the corresponding time tau_kM, where k is 1,2, M is the number of extreme points.

(2-2) calculating any two adjacent maximum or minimum extreme points X_kAnd X_k+2Formed straight line L_kM-2, find all adjacent maximum or minimum extremum points X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1And calculating τ_k+1The function value of (A)_k+1And corresponding L_k+1The value of (c):

L_k+1＝αA_k+1+(1-α)X_k+1,k＝1,2,...M-2

wherein:

A_k+1represents the extreme point X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1The function value of (c); l is_k+1Is represented by A_k+1And X_k+1Mean, maximum or minimum extreme point X of_k+2Corresponding time τ_k+2And α represents a scale factor, and typically, α is 0.5.

(2-3) A finally obtained_kAnd L_kThe subscript of (1) is 2₁And M₁Carrying out mirror image continuation estimation to obtain extreme points (tau) at the left end and the right end₀,X₀)，(τ_M+1,X_M+1) Separately find A₁,A_MAnd L₁,L_M。

(2-4) interpolation L using piecewise cubic Hermite₂,L₃,...L_M-1Fitting all the resulting baseline curves BL₁(t) separately fitting L using piecewise linear transformation₁-L₂,L_M-1-L_M，BL₁(t) denotes the baseline signal

(2-5) separating the baseline signal BL from the output signal x (t)₁(t) obtaining a new signal h₁(t) is

h₁(t)＝x(t)-BL₁(t)

(2-6) judgment of h₁(t) whether or not the condition for discriminating the ISC component is satisfied:

strict monotonicity exists between any two adjacent maximum values and minimum values in the whole data section of the output signal; the ratio of the function value corresponding to the extreme value point between any two maximum (small) value points in the whole data segment to the extreme value corresponding to the function value is kept unchanged.

If so, let ISC1 be h₁(t) output as a first ISC component; repeating the steps (2-1) - (2-6) k times until h_k(t) satisfies ISC component condition, i.e., ISCK ═ h_k(t)。

(2-7) separating ISC1 from the output signal to obtain a new signal r₁(t)：

r₁(t)＝x(t)-ISC1

(2-8) removing r₁(t) repeating the steps (2-1) to (2) as a residual component7) k times until r_k(t) is constant or monotonic to obtain k residual components, which are summed to obtain a residual signal r (t).

(2-9) the final decomposition of the output signal x (t) into k ISCs_j(t) a component and a residual signal r (t), i.e.

The method for calculating the sample entropy by the intrinsic scale component and the residual signal comprises the following steps:

(3-1) decomposing the intrinsic scale component and the residual signal into a vector sequence X (i) with a one-dimensional array of m

X(i)＝[y₁,y₂,...,y_i...,y_m],i＝1,2,...,m

y_iIs the ith one-dimensional array of the decomposition.

(3-2) defining vector y_iAnd y_jA distance d (i, j) therebetween, wherein j is 1,2,.., m (i ≠ j)

d(i,j)＝|y_i-y_j|

(3-3) setting a threshold value r, counting the number b of distances d (i, j) smaller than r, and then making a ratio of the number b to the total number m of the distances to be recorded as

Namely, it is

In the formula: 1, 2.

(3-4)B^m(r) is

Average value of (2)

(3-5) increasing the number of dimensions to m +1, repeating the steps (3-1) to (3-4) to obtain B^m+1(r)

(3-6) sample entropy formulas of the intrinsic scale component and the residual signal are as follows:

and taking the sample entropy which is greater than a threshold value 1 in the intrinsic scale component and the sample entropy of the residual signal as an error signal component.

As shown in FIG. 3, in the time domain image of the error signal, at the time when the time is less than 0.3s, the amplitude of the error signal starts to monotonically increase, the rising amplitude is greater than the average amplitude, the time lasts for about 0.4s, the peak is reached, and then the peak starts to fall, so that a sharp corner structure appears, which indicates that the FOCT has a drift fault.

The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (8)

1. A FOCT drift fault feature extraction method is characterized by comprising the following steps: the method comprises the following steps:

step 1: acquiring an output signal x (t) of the optical fiber current transformer containing the drift error signal;

step 2: decomposing the output signal x (t) into a plurality of intrinsic scale components containing error information and a residual signal containing error information using a modified local feature scale decomposition algorithm;

and step 3: respectively calculating the sample entropies of the intrinsic scale components and the residual signals to obtain the sample entropies of the intrinsic scale components and the sample entropies of the residual signals, taking the sample entropies of the intrinsic scale components larger than a threshold value as error signal components, taking the sample entropies of the residual signals larger than the threshold value as error signal components, and combining the error signal components to form an error signal component data set;

and 4, step 4: superposing error signal components in the error signal component data set to obtain an error signal;

and 5: and selecting a peak structure with monotonically rising amplitude within a certain time in the error signal as the characteristic of the drift fault according to the time domain image of the error signal.

2. The FOCT drift fault feature extraction method according to claim 1, characterized in that: the output signal x (t) of the optical fiber current transformer containing the drift error signal is calculated according to the following formula:

x(t)＝0.5K[1+cos(Δθ+ωt)]+K₁t'

wherein: k denotes the parameter of the photoelectric circuit, Delta theta denotes the Faraday effect, omega denotes the signal frequency, t denotes the time, K₁The drift deviation coefficient is shown, and t' shows the time when the deviation occurs.

3. The FOCT drift fault feature extraction method according to claim 2, characterized in that: k is_pLI₀，K_pIs the photoelectric conversion coefficient of the photodetector, L is the optical path loss, I₀Transmitting light intensity to the light source.

4. The FOCT drift fault feature extraction method according to claim 2, characterized in that: and the delta theta is 4VNI, V is a Verdet constant of the optical fiber, N is the sensing loop number of the optical fiber, and I is the current to be measured.

5. The FOCT drift fault feature extraction method according to claim 1, characterized in that: the improved local feature scale decomposition algorithm comprises the following steps:

(2-1) calculating all extreme points X of the output signal X (t)_kAnd the corresponding time tau_kK is 1,2,. M, M is the number of extreme points;

(2-2) calculating any two adjacent maximum or minimum extreme points X_kAnd X_k+2Formed straight line L_k，k＝1,2,..M-2, finding all adjacent maximum or minimum extremum points X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1And calculating τ_k+1The function value of (A)_k+1And corresponding L_k+1The value of (c):

L_k+1＝αA_k+1+(1-α)X_k+1,k＝1,2,...M-2

wherein:

A_k+1represents the extreme point X_kAnd X_k+2Extreme point X in between_k+1Corresponding time τ_k+1The function value of (c); l is_k+1Is represented by A_k+1And X_k+1Mean, maximum or minimum extreme point X of_k+2Corresponding time τ_k+2And α represents a scale factor;

(2-3) A finally obtained_kAnd L_kThe subscript of (1) is 2₁And M₁Carrying out mirror image continuation estimation to obtain extreme points (tau) at the left end and the right end₀,X₀)，(τ_M+1,X_M+1) Separately find A₁,A_MAnd L₁,L_M；

(2-4) interpolation L using piecewise cubic Hermite₂,L₃,...L_M-1Fitting all the resulting baseline curves BL₁(t) separately fitting L using piecewise linear transformation₁-L₂,L_M-1-L_M；

(2-5) separating the baseline signal BL from the output signal x (t)₁(t) obtaining a new signal h₁(t) is

h₁(t)＝x(t)-BL₁(t)

(2-6) judgment of h₁(t) whether or not the condition for discriminating the ISC component is satisfied:

strict monotonicity exists between any two adjacent maximum values and minimum values in the whole data section of the output signal; the ratio of the function value corresponding to the extreme value point between any two maximum (small) value points in the whole data segment to the extreme value corresponding to the function value is kept unchanged;

if the determination condition is satisfied, let ISC1 be h₁(t) output as a first ISC component; repeating the steps (2-1) - (2-6) k times until h_k(t) satisfies ISC component condition, i.e., ISCK ═ h_k(t)；

(2-7) separating ISC1 from the output signal to obtain a new signal r₁(t)：

r₁(t)＝x(t)-ISC1

(2-8) removing r₁(t) repeating the steps (2-1) to (2-7) k times as a residual component until r_k(t) obtaining k residual components until a constant or monotonic function, and summing the k residual components to obtain a residual signal r (t);

(2-9) the final decomposition of the output signal x (t) into k ISCs_j(t) a component and a residual signal r (t), i.e.

6. The FOCT drift fault feature extraction method according to claim 5, characterized in that: the alpha is 0.5.

7. The FOCT drift fault feature extraction method according to claim 1, characterized in that: the method for calculating the sample entropy by the intrinsic scale component and the residual signal comprises the following steps:

(3-1) decomposing the intrinsic scale component and the residual signal into a vector sequence X (i) with a one-dimensional array of m

X(i)＝[y₁,y₂,...,y_i...,y_m],i＝1,2,...,m

y_iIs the ith one-dimensional array of the decomposition;

(3-2) defining vector y_iAnd y_jA distance d (i, j) therebetween, wherein j is 1,2,.., m (i ≠ j)

d(i,j)＝|y_i-y_j|

(3-3) setting a threshold value r, counting the number b of distances d (i, j) smaller than r, and then making a ratio of the number b to the total number m of the distances to be recorded as

Namely, it is

In the formula: 1,2,. m;

(3-4)B^m(r) is

Average value of (2)

(3-5) increasing the number of dimensions to m +1, repeating the steps (3-1) to (3-4) to obtain B^m+1(r)

(3-6) sample entropy formulas of the intrinsic scale component and the residual signal are as follows:

8. the FOCT drift fault feature extraction method according to claim 1, characterized in that: the threshold is set to 1.

Images