CN103454528A - Method for detecting and identifying electric system element fault based on form singular entropy - Google Patents

Abstract

The invention discloses a method for detecting and identifying an electric system element fault based on form singular entropy. The method comprises the following steps that a three-phase voltage signal and a direct current offset zero sequence voltage signal are respectively used as input of a multi-scale morphological filter and four corresponding feature matrixes are constructed by output of the morphological filter; each feature matrix is decomposed in sequence according to the singular value decomposition technology, each set of singular values are obtained, and a large singular value is screened from each set of singular values; the singular value obtained by screening is calculated so that the corresponding morphological singular entropy can be obtained; a fault classification indicator corresponding to each signal is calculated, the fault classification indicator corresponding to each signal is compared with a preset threshold value in sequence, and whether the fault exists or not is detected; if the fault exists, a fault phase is identified and whether the fault is a ground fault or not is judged. According to the method, the mathematical morphology theory, the singular value decomposition theory and the information entropy theory are combined, fault detection and identification are conducted in a short-time window, response is fast, and the calculated amount is small.

Description

Power system component fault detect and recognition methods based on the form singular entropy

Technical field

The present invention relates to a kind of power system component fault detect and recognition methods, especially a kind of power system component fault detect and recognition methods based on the form singular entropy, belong to the power system automation technology field.

Background technology

When power system component breaks down, should excise as early as possible fault element, to guarantee the operation of power system security.Fast, correctly detect and recognition system in the fault that exists be the key of dealing with problems.Tradition fault detect and recognition methods are based on stable state power frequency amount, and its principle is clear, have passed through the long-term test of system operation.Yet in recent years, the development of electric system is very rapid, it is very complicated that the operating condition of system becomes, and brought new problem and challenge to traditional fault detect and recognition methods; On the other hand, along with the raising of system interconnect and electric pressure, the response speed of fault detect and recognition methods is also had higher requirement.Therefore, the performance of existing fault detect and recognition methods still remains to be improved.

The electric power system fault transient signal has contained abundant information, takes full advantage of these information, to realizing the application such as Power System Faults Detection, Fault Identification, relay protection, has important theory and realistic meaning.In the analysis and application of fault transient signal, the key issue existed is how effectively to extract reliably the fault characteristic information lain in the fault transient signal, on the one hand, because the fault transient signal is mingled in steady-state quantity usually, and, because energy is low, amplitude is little, often easily by steady-state quantity and system noise, flooded; On the other hand, the failure message comprised due to the fault transient signal often magnanimity, irregular data information, be difficult to for directly reaching the purpose of Fault Identification research.The reason of this two aspect, caused difficulty just to effective extraction of fault-signal characteristic information.In current published patent and document, many scholars also conduct extensive research for performance how to improve detection and recognition methods, many methods are proposed, also obtain certain effect, but also existed slow such as calculation of complex, response speed in reliability not high defect in these methods.

Summary of the invention

Purpose of the present invention, in order to solve the defect of above-mentioned prior art, a kind of employing short time window is provided, the theory of mathematical morphology, svd and information entropy is combined, and be swift in response, power system component fault detect and recognition methods based on the form singular entropy that calculated amount is little.

Purpose of the present invention can reach by taking following technical scheme:

Power system component fault detect and recognition methods based on the form singular entropy, it is characterized in that: described method realizes by the combination of mathematical morphology, svd and information entropy, comprise the following steps: gather three-phase voltage signal, calculate direct current biasing residual voltage signal, and using three-phase voltage signal and direct current biasing residual voltage signal respectively as the input of Multi-Scale Morphological Filtering device, go out four character pair matrixes by the output construction of morphological filter; Adopt based on singularity value decomposition successively each eigenmatrix to be decomposed, obtain respectively one group of singular value, and filter out larger singular value in every group of singular value; Calculated by the singular value that screening is obtained, obtained the probability corresponding with each singular value, and then obtained corresponding form singular entropy; Calculate three-phase voltage signal and failure modes index corresponding to direct current biasing residual voltage signal, and corresponding failure modes index compares with predetermined threshold value successively by each signal, detect and whether have fault; If the fault of existence, identify fault phase and determine whether earth fault.

Concrete, described direct current biasing residual voltage signal is calculated by following formula:

$V_0^{\′}=\frac{1}{3}(V_A+V_B+V_C)+{0.1V}_{\mathrm{\φ}}$

Wherein, V _a, V _band V _cbe respectively the voltage signal of A phase, B phase and C phase, V _φfor the phase voltage ratings.

Concrete, the construction process of described eigenmatrix is as follows:

Get the sliding data windows that length is n, the sample sequence of the three-phase voltage signal in sliding data windows and direct current biasing residual voltage signal processed with following morphological filter successively:

Wherein,

with being respectively gray scale expands and the gray scale erosion operation;

To the signal of each input, structure obtains matrix as follows:

Wherein, { ψ _i(1), ψ _i(2) ..., ψ _i(n) } be the output sequence of morphological filter when the length of structural element is i, the maximal value that m is structural element length.

Concrete, adopt based on singularity value decomposition to be decomposed the eigenmatrix of structure, can obtain one group of corresponding singular value to each eigenmatrix, as follows to the screening process of the singular value that obtains:

Set a percentage threshold ε, give up the less singular value of discontented foot formula:

σ _i/σ ₁＞ε

Wherein, σ ₁for first singular value, be also maximum singular value simultaneously.

Concrete, to the larger singular value screened, the computation process of corresponding form singular entropy is as follows with it:

If total k of the singular value screened is calculated probability corresponding to each singular value according to the following formula:

$p_i={\mathrm{\σ}}_i/\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\mathrm{\σ}}_i$

Calculate respectively three-phase voltage signal and form singular entropy corresponding to DC offset voltage signal by following formula:

$\mathrm{MSE}=-\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{p}_i\·{\ln p}_i.$

Concrete, the failure modes index that described three-phase voltage signal and direct current biasing residual voltage signal are corresponding is calculated by following formula:

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000012.png"\; state="\{\{state\}\}">E</figure-callout>}}_0={2\&CenterDot;\mathrm{MSE}}_0\\ {\mathrm{<figure-callout\; id="1"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">E</figure-callout>}}_1=\frac{{2\&CenterDot;\mathrm{MSE}}_1^2}{{\mathrm{MSE}}_2+{\mathrm{MSE}}_3}\\ {\mathrm{<figure-callout\; id="2"\; label="E"\; filenames="HDA0000374671750000022.png,HDA0000374671750000023.png"\; state="\{\{state\}\}">E</figure-callout>}}_2=\frac{{2\&CenterDot;\mathrm{MSE}}_2^2}{{\mathrm{MSE}}_1+{\mathrm{MSE}}_3}\\ {\mathrm{<figure-callout\; id="3"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">E</figure-callout>}}_3=\frac{{2\&CenterDot;\mathrm{MSE}}_3^2}{{\mathrm{MSE}}_1+{\mathrm{MSE}}_2}\end{array}\right.$

Wherein, MSE ₀, MSE ₁, MSE ₂and MSE ₃for form singular entropy corresponding to direct current biasing residual voltage signal, corresponding form singular entropy and form singular entropy corresponding to C phase voltage signal of form singular entropy, B phase voltage signal that the A phase voltage signal is corresponding; E ₀, E ₁, E ₂and E ₃be respectively corresponding failure modes index and failure modes index corresponding to C phase voltage signal of failure modes index, B phase voltage signal corresponding to failure modes index, A phase voltage signal that direct current biasing residual voltage signal is corresponding.

Concrete, whether the failure modes index that described each signal is corresponding compares with predetermined threshold value respectively, detect and have fault phase to be undertaken by following formula:

$L_i=\left\{\begin{array}{cc}1& E_i>\mathrm{\&gamma;}\\ 0& E_i<\mathrm{\&gamma;}\end{array}\right.$

Wherein, 0≤i≤3; γ is predetermined threshold value;

If failure modes index E _ibe greater than predetermined threshold value γ, output logic L _i=1, if failure modes index E _ibe less than predetermined threshold value γ, output logic L _i=0; Wherein, L ₁, L ₂, L ₃be respectively used to judge whether A, B, C three-phase are fault phase, L ₀for taking a decision as to whether earth fault.

Concrete, described three-phase voltage signal is sampled by the voltage transformer (VT) and the A/D conversion equipment that are arranged on protected element.

The present invention has following beneficial effect with respect to prior art:

1, the inventive method has proposed the concept of form singular entropy, carrys out structural matrix by selecting multiple dimensioned morphological filter, and the singular value calculated is reasonably screened, thereby calculate the entropy of the frequency content that characterizes analyzed signal.

2, the inventive method is based on fault detect and the recognition methods of the fault transient component in voltage signal, only need the data window of 1/8th cycles, with the method for wavelet singular entropy based on the half-wave data window, compare, response speed is faster, and calculated amount is also less simultaneously.

3, the inventive method utilizes the output of morphological filter to carry out the structural attitude matrix, when can realize suppressing noise and the effect of retention fault transient state component, make the inventive method there is higher sensitivity, and morphological filter only need to simply add/subtract computing and ask and be worth computing most, and without carrying out complicated multiplication and division arithmetic.

The accompanying drawing explanation

The power system component fault detect that Fig. 1 is embodiment 1 and the schematic flow sheet of recognition methods.

The electric system simulation illustraton of model that Fig. 2 is embodiment 2.

Fig. 3 a is the oscillogram that A phase earth fault front and back voltage signal occurs embodiment 2; Fig. 3 b is the oscillogram that A phase earth fault front and back form singular entropy occurs embodiment 2; Fig. 3 c is the oscillogram that A phase earth fault front and back failure modes index occurs embodiment 2.

Fig. 4 a is the oscillogram that A, B double earthfault front and back voltage signal occur embodiment 2; Fig. 4 b is the oscillogram that AB double earthfault front and back form singular entropy occurs embodiment 2; Fig. 4 c is that AB double earthfault front and back failure modes index occurs embodiment 2.

Fig. 5 a is the oscillogram that A, B, C three phase short circuit fault front and back voltage signal occur embodiment 2; Fig. 5 b is the oscillogram that A, B, C three phase short circuit fault front and back form singular entropy occur embodiment 2; Fig. 5 c is the oscillogram that A, B, C three phase short circuit fault front and back failure modes index occur embodiment 2.

Embodiment

Embodiment 1:

The power system component fault detect of the present embodiment and recognition methods are by the combination realization of mathematical morphology, svd and information entropy, and the characteristics of mathematical morphology, svd and information entropy are specific as follows:

1, mathematical morphology

Mathematical morphology is proposed by G.Matheron and J.Serra, and a kind of nonlinear properties that grow up from set theory and integral geometry are processed and analysis tool.Basic thought is to utilize " probe " that is called structural element (structuring element) to gather information.When probe constantly moves, just can investigate the mutual relationship between the information various piece, as the structural element of probe, can directly carry the design feature that knowledge (form, size etc.) is surveyed institute's research information.

Dilation and erosion is two Mathematical Morphology operators the most basic; Be located in Euclidean space two S set, G are arranged, mean respectively signal to be analyzed and structural element, so, the computing of two-value dilation and erosion can be expressed as respectively:

$S\&CirclePlus;G=\underset{g\&Element;G}{\&cup;}(S+g)---\left(1\right)$

Because most of signals are not two-values, in order in signal is processed, to use mathematical morphology, the two-value morphological operations first must be expanded to the gray scale morphological operations; If f (x), g (s) represent respectively grey scale signal and structural element, and the length of g (s) is less than the length of f (x), and gray scale expands and the gray scale erosion operation can be expressed as respectively:

$(f\&CirclePlus;g)\left(x\right)=\underset{s}{\max }\{f(x+s)+g\left(s\right)\}---\left(3\right)$

Wherein, (x+s), (x-s) ∈ D _f, s ∈ D _g, D _fand D _git is respectively the field of definition of f (x) and g (s).

Other Mathematical Morphology computings, as opening operation, closed operation, hit, other several common operation such as refinement and alligatoring, all being based on these two kinds of computings of dilation and erosion develops, and the form mean filter is exactly a kind of morphological filter be used widely, it is the average by the dilation and erosion computing, can be expressed as:

2, svd

Svd is a kind of important matrix decomposition in linear algebra, in fields such as signal processing, statistics, important application is arranged.Svd is similar with symmetric matrix or the diagonalization of Hermite matrix based on proper vector in some aspects.Although yet these two kinds of matrix decomposition have its correlativity, but obvious difference is arranged.The basis of symmetrical matrix eigendecomposition is analysis of spectrum, and svd is the popularization of spectrum analysis theory on Arbitrary Matrix.Svd can be by a matrix

be decomposed into two orthogonal matrix U ∈ R ^{m * m}, V ∈ R ^{n * n}with a diagonal matrix V ∈ R ^{n * n}product, that is:

A＝U∑V ^T (6)

Wherein,

$\mathrm{\&Sigma;}=\left[\begin{array}{cc}\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="HDA0000374671750000011.png,HDA0000374671750000012.png"\; state="\{\{state\}\}">S</figure-callout>}& 0\\ 0& 0\end{array}\right]---\left(7\right)$

All diagonal element in the matrix ∑, comprise nonzero element and neutral element, is called as singular value.Apparently, the number of non-zero singular value is relevant with the frequency content of analyzed signal, and the composition of signal is more complicated, and the number of non-zero singular value is also just more; So singular value can be used as a kind of objective reaction of signal content.

3, information entropy

Information entropy is a kind of tolerance to the chaos of a system and disordered state.In general, entropy is less, and unstable, the chaos of system and uncertain degree are just less.Supposing the system has n state, and the probability of each state is p _i(i=1,2 ..., n), the entropy of system can be expressed as:

$H=-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_i{\ln p}_i---\left(8\right)$

Wherein, 0≤p _i≤ 1,

The realization flow of the present embodiment as shown in Figure 1, comprises the following steps:

1) by the voltage transformer (VT) and the A/D conversion equipment that are installed in protected element, three-phase voltage signal is sampled, obtained A phase voltage signal V _a, B phase voltage signal V _bwith C phase voltage signal V _c;

2) calculate the direct current biasing residual voltage, the result of calculation at residual voltage adds a constant of 0.1 times that equals rated voltage, is shown below:

${\mathrm{<figure-callout\; id="0"\; label="V"\; filenames="HDA0000374671750000011.png,HDA0000374671750000012.png"\; state="\{\{state\}\}">V</figure-callout>}}_0^{\&prime;}=\frac{1}{3}(V_A+V_B+V_C)+{0.1V}_{\mathrm{\&phi;}}---\left(9\right)$

The characteristics of direct current biasing residual voltage: in non-fault or while occurring not produce the fault of zero-sequence current, the principal ingredient that DC component is signal, noise is very little to the contribution of form singular entropy; And when asymmetric earth fault occurs, the principal ingredient that residual voltage is signal, DC component and noise are all very little to the contribution of form singular entropy;

3) get the data window that length is n, to the three-phase voltage signal in the data window and direct current biasing residual voltage signal sample sequence, use respectively the morphological filter of formula (5) expression to be processed, obtain matrix as follows:

Wherein, { ψ _i(1), ψ _i(2) ..., ψ _i(n) } be the output sequence of morphological filter when the length of structural element is i, the maximal value of the length that m is structural element;

4) adopting based on singularity value decomposition to step 3) matrix A of structure processed, and to each eigenmatrix, can obtain one group of corresponding singular value; Set a percentage threshold ε, according to the following formula singular value screened, give up the less singular value of discontented foot formula:

σi/σ ₁＞ε (11)

Wherein, σ ₁for first singular value, be also maximum singular value simultaneously.

5) establish total k of the singular value that meets formula (11), calculate according to the following formula probability corresponding to each singular value:

$p_i={\mathrm{\&sigma;}}_i/\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_i---\left(12\right)$

6) calculate respectively three-phase voltage signal and form singular entropy corresponding to DC offset voltage signal by following formula:

$\mathrm{MSE}=-\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}{p}_i\&CenterDot;{\ln p}_i---\left(13\right)$

7) calculate three-phase voltage signal and failure modes index corresponding to direct current biasing residual voltage signal by following formula:

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000012.png"\; state="\{\{state\}\}">E</figure-callout>}}_0={2\&CenterDot;\mathrm{MSE}}_0\\ {\mathrm{<figure-callout\; id="1"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">E</figure-callout>}}_1=\frac{{2\&CenterDot;\mathrm{<figure-callout\; id="1"\; label="MSE"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_1^2}{{\mathrm{<figure-callout\; id="2"\; label="MSE"\; filenames="HDA0000374671750000022.png,HDA0000374671750000023.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_2+{\mathrm{MSE}}_3}\\ {\mathrm{<figure-callout\; id="2"\; label="E"\; filenames="HDA0000374671750000022.png,HDA0000374671750000023.png"\; state="\{\{state\}\}">E</figure-callout>}}_2=\frac{{2\&CenterDot;\mathrm{<figure-callout\; id="2"\; label="MSE"\; filenames="HDA0000374671750000022.png,HDA0000374671750000023.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_2^2}{{\mathrm{<figure-callout\; id="1"\; label="MSE"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_1+{\mathrm{MSE}}_3}\\ {\mathrm{<figure-callout\; id="3"\; label="E"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">E</figure-callout>}}_3=\frac{{2\&CenterDot;\mathrm{<figure-callout\; id="3"\; label="MSE"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_3^2}{{\mathrm{<figure-callout\; id="1"\; label="MSE"\; filenames="HDA0000374671750000011.png,HDA0000374671750000022.png"\; state="\{\{state\}\}">MSE</figure-callout>}}_1+{\mathrm{MSE}}_2}\end{array}\right.---\left(14\right)$

Wherein, E ₀, E ₁, E ₂and E ₃be respectively corresponding failure modes index and failure modes index corresponding to C phase voltage signal of failure modes index, B phase voltage signal corresponding to failure modes index, A phase voltage signal that direct current biasing residual voltage signal is corresponding;

8) by step 7) failure modes index corresponding to each signal that calculate compare with predetermined threshold value γ respectively, and comparative result is shown below:

$L_i=\left\{\begin{array}{cc}1& E_i>\mathrm{\&gamma;}\\ 0& E_i<\mathrm{\&gamma;}\end{array}\right.---\left(15\right)$

Wherein, 0≤i≤3;

If failure modes index E _ibe greater than predetermined threshold value γ, output logic L _i=1, if failure modes index E _ibe less than predetermined threshold value γ, output logic L _i=0; Wherein, L ₁, L ₂, L ₃be respectively used to judge whether A, B, C three-phase are fault phase, L ₀for taking a decision as to whether earth fault.

Power system component fault detect and the recognition methods cardinal principle of the present embodiment are: under the electric system normal operating condition, bus voltage signal only comprises sinusoidal fundamental wave component and the random noise of power frequency, by mathematical morphology filter with to the screening of singular value, can suppress noise, the form singular entropy now calculated is very little, close to zero; And when electric system is broken down, in voltage signal, gone out to contain outside fundamental frequency component and noise, also contain a large amount of high frequency transient components, these components can be all by the morphological filter filtering, thereby make the difference of singular value diminish, therefore the unusual Entropy Changes of form is large, obviously is greater than zero, so the variation of the frequency content of form singular entropy in can the response voltage signal.

Owing to having coupling and other disturbances, the form singular entropy of healthy phases also may have small size rising.Therefore, in order to obtain more reliable Failure detection and identification result, after calculating the form singular entropy that direct current biasing residual voltage signal and three-phase voltage signal are corresponding, also introduced suc as formula the failure modes index shown in (14), this failure modes index makes the difference of fault phase and healthy phases more obvious.

Embodiment 2:

As shown in Figure 2, wherein MN is protected circuit to the electric system simulation model of the present embodiment, the electric system simulation model parameter: frequency is 50Hz, and line length is 100km, and the two ends supply voltage is respectively E _m=230 0 ° of ∠ and E _n=230 20 ° of ∠, source impedance is Z _m=Z _n=9.186+j40.192 Ω, circuit unit length zero sequence resistance is R ₀=0.3000 Ω/km, circuit unit length positive sequence resistance is R ₁=0.0346 Ω/km, circuit unit length zero sequence inductance is L ₀=3.6340mH/km, circuit unit length positive sequence inductance is L ₁=1.3482mH/km, circuit unit length zero sequence electric capacity is C ₀=0.0062 μ F/km, circuit unit length positive sequence electric capacity is C ₁=0.0086 μ F/km; Voltage transformer (VT), the A/D conversion equipment of the installing of circuit M side are sampled to three-phase voltage signal with the sample frequency of 20kHz; Relative parameters setting is: sliding data windows length n=50, the maximal value m=5 of the length of structural element, percentage threshold ε=1%, predetermined threshold value γ=0.5.

If at t=205ms constantly, A phase earth fault occurs in transmission line of electricity MN mid point f place, obtains three-phase voltage signal in the M side by sampling, and calculates direct current biasing residual voltage V according to formula (9) ₀', as shown in Figure 3 a; Form singular entropy by calculating, failure modes index are respectively as shown in Fig. 3 b and Fig. 3 c.From Fig. 3 a-3c, before fault occurs, although contain noise in voltage signal, because morphological filter and svd all have noise inhibiting ability, so the value of the form singular entropy that calculates before occurring of fault is all the time close to zero.And after fault occurs, contain a large amount of fault transient high-frequency signals in A phase voltage signal and direct current biasing residual voltage signal, the value of corresponding form singular entropy significantly rises rapidly, and in B phase voltage signal and C phase voltage signal, also have a small amount of high-frequency signal, corresponding form singular entropy also after fault, to produce small size rising; Through type (14) is introduced the failure modes index, as shown in Figure 3 c, makes the difference between fault phase (A phase) and healthy phases (B phase, C phase) become larger.All be greater than predetermined threshold value γ (0.5) after the 0.3ms of failure modes index after fault corresponding to A phase voltage signal and direct current biasing residual voltage signal, and B phase voltage signal and failure modes corresponding to C phase voltage signal refer to that target value approaches zero all the time, so fault type judges is A phase earth fault.The inventive method energy fast detecting and identification fault phase are described, the differential protection correct operation, and there is very high sensitivity.

If at t=205ms constantly, transmission line of electricity MN mid point f place generation A, B double earthfault, by the three-phase voltage signal sampled, and calculate direct current biasing residual voltage V according to formula (9) in the M side ₀', as shown in Fig. 4 a; Form singular entropy by calculating, failure modes index respectively as shown in Fig. 4 b and Fig. 4 c.From Fig. 4 a-4c, after fault occurs, contain a large amount of fault transient high-frequency signals in A phase voltage signal, B phase voltage signal and direct current biasing residual voltage signal, the value of corresponding form singular entropy significantly rises rapidly, and the high-frequency signal in the C phase voltage signal seldom, the variation of corresponding form singular entropy is very little; Through type (14) is introduced the failure modes index, as shown in Fig. 4 c, makes the difference between fault phase (A phase, B phase) and healthy phases (C phase) become larger.By failure modes index and predetermined threshold value γ (0.5) are made comparisons, just can judge that fault type is as A, B phase earth fault.

If at t=205ms constantly, the generation of transmission line of electricity MN mid point f place A, B, C three phase short circuit fault, by the three-phase voltage signal sampled, and calculate direct current biasing residual voltage V according to formula (9) in the M side ₀', as shown in Figure 5 a; Form singular entropy by calculating, failure modes index respectively as shown in Fig. 5 b and Fig. 5 c.From Fig. 5 a-5c, after fault occurs, contain a large amount of fault transient high-frequency signals in A phase voltage signal, B phase voltage signal and C phase voltage signal, the value of corresponding form singular entropy significantly rises rapidly, and owing to being symmetric fault, do not produce zero-sequence current in the direct current biasing residual voltage, DC component is still the principal ingredient of signal, so its corresponding form singular entropy changes hardly before and after fault.Through type (14) is introduced the failure modes index, and itself and predetermined threshold value γ (0.5) are made comparisons, and just can judge that fault type is as A, B, C phase short trouble.

The above; it is only patent preferred embodiment of the present invention; but the protection domain of patent of the present invention is not limited to this; patent of the present invention is equally applicable to other systems of using three-phase alternating current and fault detect and the identification of equipment; anyly be familiar with those skilled in the art in the disclosed scope of patent of the present invention; according to the present invention, the technical scheme of patent and patent of invention design thereof are equal to replacement or are changed, and all belong to the protection domain of patent of the present invention.

Claims (8)

1. power system component fault detect and the recognition methods based on the form singular entropy, it is characterized in that: described method realizes by the combination of mathematical morphology, svd and information entropy, comprise the following steps: gather three-phase voltage signal, calculate direct current biasing residual voltage signal, and using three-phase voltage signal and direct current biasing residual voltage signal respectively as the input of Multi-Scale Morphological Filtering device, go out four character pair matrixes by the output construction of morphological filter; Adopt based on singularity value decomposition successively each eigenmatrix to be decomposed, obtain respectively one group of singular value, and filter out larger singular value in every group of singular value; Calculated by the singular value that screening is obtained, obtained the probability corresponding with each singular value, and then obtained corresponding form singular entropy; Calculate three-phase voltage signal and failure modes index corresponding to direct current biasing residual voltage signal, and corresponding failure modes index compares with predetermined threshold value successively by each signal, detect and whether have fault; If the fault of existence, identify fault phase and determine whether earth fault.

2. power system component fault detect and the recognition methods based on the form singular entropy according to claim 1 is characterized in that: described direct current biasing residual voltage signal is calculated by following formula:

$V_0^{\&prime;}=\frac{1}{3}(V_A+V_B+V_C)+{0.1V}_{\mathrm{\&phi;}}$

Wherein, V _a, V _band V _cbe respectively the voltage signal of A phase, B phase and C phase, V _φfor the phase voltage ratings.

3. power system component fault detect and the recognition methods based on the form singular entropy according to claim 1, it is characterized in that: the construction process of described eigenmatrix is as follows:

Get the sliding data windows that length is n, with following morphological filter, the sample sequence of the three-phase voltage signal in sliding data windows and direct current biasing residual voltage signal processed respectively:

Wherein,

with

be respectively gray scale and expand and the gray scale erosion operation, f and g mean respectively processed signal and structural element;

Signal to each input adopts multiple dimensioned flat-structure element to be processed, and by the wave filter output construction, obtains matrix as follows:

Wherein, { ψ _i(1), ψ _i(2) ..., ψ _i(n) } be the output sequence of morphological filter when the length of structural element is i, the maximal value that m is structural element length.

4. power system component fault detect and the recognition methods based on the form singular entropy according to claim 1, it is characterized in that: adopt based on singularity value decomposition to be decomposed constructed eigenmatrix, can obtain one group of corresponding singular value to each eigenmatrix, as follows to the screening process of the singular value that obtains:

Set a percentage threshold ε, give up discontented foot formula less singular value "

σ _i/σ ₁＞ε

σ wherein ₁for first singular value, be also maximum singular value simultaneously.

5. power system component fault detect and the recognition methods based on the form singular entropy according to claim 1 is characterized in that: to the larger singular value screened, the computation process of corresponding form singular entropy is as follows with it:

If total k of the singular value screened is calculated probability corresponding to each singular value according to the following formula:

$p_i={\mathrm{\&sigma;}}_i/\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_i$

Calculate respectively according to the following formula the form singular entropy that three-phase voltage signal and DC offset voltage signal are corresponding:

$\mathrm{MSE}=-\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}{p}_i\&CenterDot;{\ln p}_i.$

6. power system component fault detect and the recognition methods based on the form singular entropy according to claim 1, it is characterized in that: the failure modes index that described three-phase voltage signal and direct current biasing residual voltage signal are corresponding is calculated by following formula:

$\left\{\begin{array}{c}{E}_0={2\&CenterDot;\mathrm{MSE}}_0\\ E_1=\frac{{2\&CenterDot;\mathrm{MSE}}_1^2}{{\mathrm{MSE}}_2+{\mathrm{MSE}}_3}\\ E_2=\frac{{2\&CenterDot;\mathrm{MSE}}_2^2}{{\mathrm{MSE}}_1+{\mathrm{MSE}}_3}\\ E_3=\frac{{2\&CenterDot;\mathrm{MSE}}_3^2}{{\mathrm{MSE}}_1+{\mathrm{MSE}}_2}\end{array}\right.$

Wherein, MSE ₀, MSE ₁, MSE ₂and MSE ₃for form singular entropy corresponding to direct current biasing residual voltage signal, corresponding form singular entropy and form singular entropy corresponding to C phase voltage signal of form singular entropy, B phase voltage signal that the A phase voltage signal is corresponding; E ₀, E ₁, E ₂and E ₃be respectively corresponding failure modes index and failure modes index corresponding to C phase voltage signal of failure modes index, B phase voltage signal corresponding to failure modes index, A phase voltage signal that direct current biasing residual voltage signal is corresponding.

7. power system component fault detect and the recognition methods based on the form singular entropy according to claim 6, whether it is characterized in that: the failure modes index that described each signal is corresponding compares with predetermined threshold value respectively, detect and exist fault to be undertaken by following formula:

$L_i=\left\{\begin{array}{cc}1& E_i>\mathrm{\&gamma;}\\ 0& E_i<\mathrm{\&gamma;}\end{array}\right.$

Wherein, 0≤i≤3; γ is predetermined threshold value: if failure modes index E _ibe greater than predetermined threshold value γ, output logic L _i=1, if failure modes index E _ibe less than predetermined threshold value γ, output logic L _i=0; Wherein, L ₁, L ₂, L ₃be respectively used to judge whether A, B, C three-phase are fault phase, L ₀for taking a decision as to whether earth fault.

8. according to described power system component fault detect and the recognition methods based on the form singular entropy of claim 1-7 any one, it is characterized in that: described three-phase voltage signal is sampled by the voltage transformer (VT) and the A/D conversion equipment that are arranged on protected element.

Images