CN103323747A - Method for single-phase earth fault line selection of small current grounding system - Google Patents

Abstract

The invention relates to the field of power systems and the automation of the power systems, in particular to a method for single-phase earth fault line selection of a small current grounding system. According to the method, the zero sequence circuit of a single-phase earth fault is analyzed, the frequency characteristic of the transient state characteristic quantity of the fault is obtained, an optimal FIR filter is designed according to the frequency characteristic to extract the components of transient state capacitive zero sequence currents with obvious fault characteristics, and the interference of noise and the interference of the unbalanced operation condition of a power grid are eliminated. A hierarchical clustering algorithm is used, the polarity and the amplitude characteristics of the transient state capacitive zero sequence currents of the initial ends of all feed lines are integrated to carry out classification on the feed lines, and a single feed line with the change trend of the transient state capacitive zero sequence current easier than the transient state capacitive zero sequence currents of other feed lines is selected as a fault line. The method improves the reliability of the line selection, lowers signal sampling frequency needs, avoids multi-layer signal decomposition, saves computing resources, and is easy to achieve.

Description

A kind of fault line selection method for single-phase-to-ground fault of small current neutral grounding system

Technical field

The present invention relates to the Power System and its Automation field, be specifically related to a kind of fault line selection method for single-phase-to-ground fault of small current neutral grounding system.

Background technology

Single-phase ground fault line selecting method of small-electric current grounding system, although passed through for a long time widely research, but because not obvious, the interference of noise of fault signature, the problems such as impact of power distribution network operation conditions (such as asymmetry), existing fault-line selecting method Shortcomings still aspect reliability and universality.

Based on the difference of the characteristic quantity of institute's foundation, at present, fault line selection method for single-phase-to-ground fault can be divided three classes: signal injection method, steady-state quantity route selection method and transient route selection method.Wherein, signal injection method and steady-state quantity route selection method all exist characteristic signal faint, the insecure problem of route selection result.And large several times even tens times of the transient state aspect ratio steady-state value of transient route selection method institute foundation, and be not subjected to the impact of arc suppression coil, therefore have higher reliability and using value.

Yet, except fault-signal, also mixed the polymorphic type signal in the transient, directly utilize, easily cause erroneous judgement.Therefore, effective extraction of fault transient feature with take full advantage of, just become problem demanding prompt solution.Wavelet transformation, powerful tool as Time-Frequency Localization, multiresolution analysis, in emulation experiment, be widely used in the extraction of singlephase earth fault one of transient characteristic quantity, Effective Raise the failure line selection reliability, but still Shortcomings: at first, feature band is to determine by the principle of energy maximum, under serious noise or unbalanced power supply running status, the feature band that generates might not be for effective fault transient measure feature frequency band, causes erroneous judgement; In addition, feature band is based on the collection signal Analysis deterrmination, can not know in advance, according to Shannon sampling law, can only improve as far as possible sample frequency, and is required to satisfy unknown feature band, improved the requirement to hardware device; Simultaneously, need to successively extract in a large number and separate sampled data, the multiplying amount be huge, is unfavorable for realizing in the embedded system; In addition, according to the qualitative differentiation fault of signal polarity and non-fault line, only rely on artificial judgement, lack quantitative scientific description, can cause erroneous judgement equally.Therefore, at present, not high based on the fault line selection method for single-phase-to-ground fault degree of being practical of wavelet transformation.

Summary of the invention

For the deficiencies in the prior art, the fault line selection method for single-phase-to-ground fault that the purpose of this invention is to provide a kind of small current neutral grounding system, the method is by the analysis to singlephase earth fault zero sequence circuit, obtain the frequency characteristic of fault transient characteristic quantity, devise optimum FIR wave filter accordingly, with the obvious transient state capacitive of extraction fault signature zero-sequence current component, and the interference of removing noise and unbalanced power supply operation conditions; Utilize hierarchical clustering algorithm, comprehensively polarity and the amplitude Characteristics of each feeder line top transient state capacitive zero-sequence current are classified to feeder line, and the single feeder line of finally selecting transient state capacitive zero-sequence current variation tendency to be easy to other feeder line is faulty line.The method has improved the route selection reliability, has reduced the signal sampling frequency needs, has avoided multi-level signal decomposition, saves computational resource, is easy to realize.

The objective of the invention is to adopt following technical proposals to realize:

The invention provides a kind of fault line selection method for single-phase-to-ground fault of small current neutral grounding system, its improvements are, described method comprises the steps: based on optimum FIR wave filter and hierarchical clustering

(1) preliminary work of the fault line selection method for single-phase-to-ground fault of small current neutral grounding system;

(2) operation information of Real-time Collection electrical network, and judge whether the bus residual voltage surpasses threshold value;

(3) the feeder line top zero-sequence current that collects is input to the FIR filter filtering, obtains transient state capacitive zero-sequence current component;

(4) to the classification of n bar feeder line, make the class number k=n on the cluster level;

(5) determine the distance between any two classes in k the class, merge two classes corresponding to minor increment, form new cluster level;

(6) the pseudo-F on the cluster level in the determining step (5);

(7) judge that whether all classes gather is a class;

(8) based on cluster result, determine fault feeder.

Wherein, in the described step (1), preliminary work comprises:

＜1〉design FIR wave filter;

＜2〉electricity grid substation device configuration.

Wherein, described step＜1〉in, the optimum FIR bandpass filter of design 300～3000Hz comprises the steps:

A, the minimum exponent number of FIR wave filter are estimated:

FIR wave filter low frequency transitional zone is designed to 280～300Hz, according to the Kaiser equation, determines that the minimum exponent number of FIR wave filter is that 500, Kaiser equation is as follows:

$N\≈\frac{-20\lg \left(\sqrt{(1-{10}^{{-\mathrm{\α}}_P/20})\left({10}^{{-\mathrm{\α}}_S/20}\right)}\right)-13}{14.6({\mathrm{\ω}}_s-{\mathrm{\ω}}_p)/2\mathrm{\π}}$ ④；

Wherein, N is the minimum exponent number of FIR wave filter, ω _pWith ω _sBe respectively normalization passband and stopband boundary angle frequency, α _pBe peak value passband ripple, α _SBe minimum stop-band attenuation;

B, determine optimum FIR filter parameter, determine that namely exponent number is 500 FIR filter coefficient:

By Constraint least square algorithm, the peak value of the absolute difference of following expression formula is minimized, realize that the actual frequency response H of FIR wave filter (j ω) approaches desirable frequency response D (j ω) to greatest extent, that is:

$\min \underset{\mathrm{\ω}\∈R}{\max }|H\left(\mathrm{j\ω}\right)-D\left(\mathrm{j\ω}\right)|$ ⑤；

Frequency signal has 0.04s, namely after filtering

Fixed delay; F _TThe frequency of expression FIR wave filter.

Wherein, described step＜2〉in, at the transformer station place intellectual analysis main website, optimum FIR wave filter, bus residual voltage monitoring device, each feeder line top zero-sequence current harvester and storer and communication device are set.

Wherein, in the described step (2), the intellectual analysis main website is by bus residual voltage monitoring device and each feeder line top zero-sequence current harvester Real-Time Monitoring operation of power networks information, and judgement bus residual voltage U ₀Whether surpass threshold value, if bus residual voltage U ₀Surpass threshold value, then record this event and occur constantly, and enter step (3), otherwise, this step repeated; Described threshold value is 0.35 times bus rated voltage.

Wherein, in the described step (3), each feeder line top zero-sequence current I that fault collects in the 0.06s time range after occuring constantly gets in the intellectual analysis main website _0iAnalyze, through optimum FIR filter filtering, obtain transient state capacitive zero-sequence current component I _C0i, i=1 wherein, 2 ..., n, n are the feeder line number.

Wherein, in the described step (4), based on I _C0i, according to the hierarchical clustering method, feeder line is classified: n is the feeder line number, and a corresponding n primitive class makes k=n, and wherein k is the class number on this cluster level.

Wherein, in the described step (5), determine the distance between any two classes in k the class according to following expression formula, obtain

Individual distance value merges two classes corresponding to minor increment, forms new cluster level, a corresponding k-1 class;

$D_{\mathrm{AB}}={\left|\right|\stackrel{\‾}{X_A}-\stackrel{\‾}{X_B}\left|\right|}^2+W_A/N_A+W_B/N_B$ ⑥；

Wherein, X _iBe i object;

With

Be respectively category-A and category-B object mean vector;

Be the poor quadratic sum of category-A inter-object distance; Be the poor quadratic sum of category-B inter-object distance; N _AAnd N _BBe respectively the object number in A and the category-B.

Wherein, in the described step (6), according to the pseudo-F value on the cluster level in the following expression formula determining step (5):

PSF＝((W _j-P _G)/(G-1))/(P _G/(N-G)) ⑦；

Wherein, G is the class number on this cluster level;

N is the object number in the j class;

$P_G=\underset{j=1}{\overset{G}{\mathrm{\Σ}}}{W}_j;$

Described pseudo-F is used for estimating the Clustering Effect when being classified as G class, and the larger indicated object of its value is classified as G class, and corresponding cluster numbers was best cluster numbers when peak value appearred in pseudo-F.

Wherein, in the described step (7), make k=k-1, whether check k is 1, and namely whether all objects are poly-is a class, if then enter step (8); If not, then return step (5); Be the corresponding cluster level of pseudo-F maximum value according to pedigree Cluster Assessment index, determine best cluster number m.

Wherein, in the described step (8), if m=2, and wherein a class only comprises a feeder line, and then this feeder line is fault feeder; If above-mentioned condition does not satisfy and the bus residual voltage continues to surpass threshold value (threshold value is 0.35 times bus rated voltage), then is bus-bar fault.

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

For single-phase grounded malfunction in grounded system of low current, owing to not obvious, the noise of fault signature, being subject to the problem such as the off-center operation situation of electrical network impact own, the accuracy of existing fault line selection device is generally not high.The present invention adopts two technology head it offs: 1. by optimum FIR wave filter image data is carried out filtering, extract effective fault characteristic signals, and remove the interference of noise and unbalanced power supply operation conditions; 2. utilize the hierarchical clustering method, based on waveform after the filtering, to the feeder line classification, obtain faulty line.The interference of each factor in the electrical network actual motion that the method has been passed through the wave filter establishment.In addition, utilize the FIR wave filter to carry out signal and process, the data sampling frequency only need reach that 6kHz is above to be got final product, and is lower than the sample frequency demand of fault-line selecting method based on wavelet analysis commonly used, conserve system resources.Simultaneously, based on clustering algorithm, determine fault feeder according to pedigree Cluster Assessment index, guaranteed route selection result's reliability.

Description of drawings

Fig. 1 is the zero sequence circuit diagram behind the typical small current neutral grounding system generation singlephase earth fault provided by the invention;

Fig. 2 is the zero sequence equivalent circuit diagram behind the typical small current neutral grounding system generation singlephase earth fault provided by the invention;

Fig. 3 is optimum FIR bandpass filter amplitude-frequency oscillogram provided by the invention;

Fig. 4 is optimum FIR bandpass filter phase frequency oscillogram provided by the invention;

Fig. 5 is the single-phase ground fault line selecting method of small-electric current grounding system process flow diagram based on optimum FIR wave filter and hierarchical clustering provided by the invention.

Embodiment

Below in conjunction with accompanying drawing the specific embodiment of the present invention is described in further detail.

The present invention is according to the frequency characteristic of singlephase earth fault one of transient characteristic quantity in the small current neutral grounding system, and devise optimum FIR wave filter is with the obvious transient state capacitive of extraction fault signature zero-sequence current component, and the interference of removing noise and unbalanced power supply operation conditions; Utilize hierarchical clustering algorithm, based on the variation tendency of each feeder line top transient state capacitive zero-sequence current component feeder line is classified, determine fault feeder.The method of the invention has improved the reliability of failure line selection, and be applicable to all kinds of singlephase earth fault situations (and fault state that stake resistance larger less such as the fault switching angle, seriously measure fault state under the serious off-center operation state of fault state, intermittent arc-earth fault, power distribution network under the noise etc.), conserve system resources is easy to realize.

The present invention carries out on the basis of following technology:

(1) optimum FIR design of filter

Optimum FIR design of filter is the prerequisite that realizes fault line selection method for single-phase-to-ground fault of the present invention.Based on the frequency distribution characteristic of each component in the fault transient zero-sequence current as can be known, be to extract the obvious transient state capacitive of fault signature zero-sequence current component, need the bandpass filter of a 300～3000Hz of design, desirable ideal frequency response is as follows:

$\left\{\begin{array}{cc}\left|D\left(\mathrm{j\&omega;}\right)\right|=1& 300\&le;\frac{\mathrm{\&omega;}}{2\mathrm{\&pi;}}\&le;3000\\ \left|D\left(\mathrm{j\&omega;}\right)\right|=0& \frac{\mathrm{\&omega;}}{2\mathrm{\&pi;}}<300,\frac{\mathrm{\&omega;}}{2\mathrm{\&pi;}}>3000\end{array}\right.$ ①；

The transport function of FIR wave filter is:

$H\left(z\right)=\frac{Y\left(z\right)}{X\left(z\right)}=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}h\left(n\right)z^{-n}$ ②；

Wherein, N is filter order.The impulse response h (n) of this wave filter is only at n=0, and 1 ..., on limited the point of N-1 value is arranged.For input x (n), it is output as:

$y\left(n\right)=h\left(0\right)x\left(n\right)+h\left(1\right)x(n-1)+\&CenterDot;\&CenterDot;\&CenterDot;+h(N-1)x(n-N+1)$

③；

$=\underset{m=0}{\overset{N-1}{\mathrm{\&Sigma;}}}h\left(m\right)x(n-m)=h\left(n\right)\&CircleTimes;x\left(n\right)$

For realizing that maximum error minimizes between actual frequency response H (j ω) and the desirable frequency response D (j ω), i.e. actual frequency response can be adopted optimum FIR filter design method substantially near desirable frequency response, and what obtain is optimum FIR wave filter.

(2) singlephase earth fault zero sequence circuit analysis:

Typical zero-sequence network behind the small current neutral grounding system generation singlephase earth fault as shown in Figure 1.Among Fig. 1, suppose that singlephase earth fault occurs the F point place on feeder line i, namely be equivalent to produce a zero sequence power supply at the F point.R _0k, L _0k, C _0kBe respectively feeder line k unit length zero sequence resistance, zero sequence inductance and zero sequence distributed capacitance, L is the arc suppression coil inductance, and n is the feeder line number.K switch disconnects, and this network is isolated neutral system, and K is closed, is the neutral by arc extinction coil grounding system.

Obviously, according to each feeder line top zero-sequence current direction and amplitude, can carry out failure line selection.But in the small current neutral grounding system, the zero sequence path resistance value that singlephase earth fault forms is larger, and the stable status zero-sequence current amplitude is less, and is lower based on its failure line selection reliability of carrying out.For deep excavation zero sequence flow characteristic, Fig. 1 is converted into equivalent electrical circuit shown in Figure 2 analyzes.

Consider complex situations, namely K is closed, and has the situation of measuring noise and power distribution network off-center operation, can get:

i _0i＝i _0L+i _0C+i _ub+n ⑧；

Wherein, i _0iThe zero-sequence current that collects for fault feeder top; i _0LBe the nasty zero sequence component; i _0CBe the capacitive zero-sequence component; i _UbZero-sequence component for power distribution network self off-center operation initiation; N is for measuring noise.In conjunction with Fig. 1 as can be known, the zero-sequence current that collects of non-fault feeder top is except not comprising i _0LOutward, the same i of all the other components _0i

Set up the differential equation by Fig. 2, find the solution in conjunction with starting condition:

Wherein, i' _0LBe transient state nasty zero sequence current component; I'' _0LBe stable state nasty zero sequence current component; ω is the power frequency angular frequency,

Be fault close angle; τ _CTime constant for the nasty zero sequence loop.

Wherein, i' _0CBe transient state capacitive zero-sequence current component; I'' _0CBe stable state capacitive zero-sequence current component;

Be i' _0CAngular frequency; τ _CTime constant for capacitive zero sequence loop.

According to actual power distribution network R ₀, L ₀, C ₀And the span of L, convolution 9., 10. calculate as can be known: i' _0CThe natural frequency of vibration (namely

) scope is 300～3000Hz, with i' _0LCan not mutually compensate, and amplitude is larger, is steady-state value i'' _0CSeveral times in addition tens times (namely ), but decay is very fast, and generally continuing 0.5～1 power frequency period (is τ _C).And i _Ub, i'' _0LWith i'' _0CBe power frequency component, and the measurement noise concentrates on below the 250Hz mainly also.

To sum up, behind the small current neutral grounding system generation singlephase earth fault, all comprise shorter but transient state capacitive zero-sequence current component i' that amplitude is larger of duration in the zero-sequence current that each feeder line top produces _0C, fault signature is obvious, according to its direction and amplitude, can effectively distinguish fault and non-fault line.In addition, i' _0CHave the band characteristic that is different from other component in the zero-sequence current, namely with other component frequency domain on without occuring simultaneously, thereby can realize accordingly the extraction of signal.

(3) hierarchical clustering method:

Utilize optimum FIR wave filter can effectively extract transient state capacitive zero-sequence current component.By the circuit analysis of singlephase earth fault zero sequence as can be known, no matter faulty line top transient state capacitive zero-sequence current all is different from non-fault line from amplitude or phase place aspect.Yet in some cases, transient state capacitive zero-sequence current amplitude integral body is less than normal and vibration is frequent, only qualitatively judges artificially easily initiation erroneous judgement.For this reason, can adopt the hierarchical clustering method, comprehensive transient state capacitive zero-sequence current amplitude and phase information are classified to feeder line, draw the single feeder line that variation tendency is different from other feeder line, are fault feeder.If do not exist singlephase earth fault (be that the bus residual voltage surpasses certain threshold value, be generally 0.35 times bus rated voltage) not occuring in this type of feeder line and the definite electrical network, then is bus-bar fault.

Hierarchical clustering be with each object as a class, merge two classes corresponding to minimum between class distance on this level, afterwards, on new cluster level, repeating above-mentioned class merging process, is a class until all objects gather, and forms the pedigree chart of a close and distant relation of expression.At last, utilize the Cluster Assessment index (such as pseudo-F value) on each level, pedigree chart is sheared, determine the class number, finish classification.

In the situation that singlephase earth fault occurs in feeder line, estimate the feeder line cluster result, it should be the most rational being divided into two classes, the single feeder line that is classified as separately a class is fault feeder.

Single-phase ground fault line selecting method of small-electric current grounding system process flow diagram based on optimum FIR wave filter and hierarchical clustering provided by the invention comprises the steps: as shown in Figure 5

(1) preliminary work of the fault line selection method for single-phase-to-ground fault of small current neutral grounding system:

＜1〉design FIR wave filter:

The minimum exponent number of FIR wave filter is decided by the narrowest transition band width.Basis signal extracts demand, and the filtration of low frequency signal is had relatively high expectations, and needs steady-state signal and the following noise signal of 250Hz of filtering 50Hz.Consider the passband frequency range design requirement of 300～3000Hz and keep adequate allowance, the low frequency transitional zone can be designed to 280～300Hz, accordingly, and in conjunction with power distribution network common signal frequency acquisition F _T=6.4kHz, according to the Kaiser equation, calculating the minimum exponent number of FIR wave filter is 500, namely with the 500 design exponent numbers as optimum FIR wave filter.

The optimum FIR bandpass filter of design 300～3000Hz comprises the steps:

A, the minimum exponent number of FIR wave filter are estimated:

The minimum exponent number of FIR wave filter is decided by the narrowest transition band width.Basis signal extracts demand, and the filtration of low frequency signal is had relatively high expectations, and needs steady-state signal and the following noise signal of 250Hz of filtering 50Hz.Consider the passband frequency range design requirement of 300～3000Hz and keep adequate allowance, the low frequency transitional zone can be designed to 280～300Hz, accordingly, and in conjunction with power distribution network common signal frequency acquisition F _T=6.4kHz, according to the Kaiser equation, calculating the minimum exponent number of FIR wave filter is that 500, Kaiser equation is as follows:

$N\&ap;\frac{-201g\left(\sqrt{(1-{10}^{{-\mathrm{\&alpha;}}_P/20})\left({10}^{{-\mathrm{\&alpha;}}_S/20}\right)}\right)-13}{14.6({\mathrm{\&omega;}}_s-{\mathrm{\&omega;}}_p)/2\mathrm{\&pi;}}$ ④；

Wherein, N is the minimum exponent number of FIR wave filter, ω _pWith ω _sBe respectively normalization passband and stopband boundary angle frequency, α _pBe peak value passband ripple, α _SBe minimum stop-band attenuation.

Because in the situation that satisfy filtering performance, filter order is more little more easily to be realized, arithmetic speed is also faster, therefore take minimum exponent number---the 500 design exponent numbers as optimum FIR wave filter of FIR wave filter.

B, determine optimum FIR filter parameter:

This problem can be converted into optimization problem and find the solution.Namely find the solution exponent number and be 500 FIR filter coefficient, so that the peak value of the absolute difference of formula shown in 5. minimizes, realize that the actual frequency response H of FIR wave filter (j ω) approaches desirable frequency response D (j ω) to greatest extent, that is:

$\min \underset{\mathrm{\&omega;}\&Element;R}{\max }|H\left(\mathrm{j\&omega;}\right)-D\left(\mathrm{j\&omega;}\right)|$ ⑤；

This optimization problem can adopt Constraint least square algorithm to find the solution.Finally, obtain having the as shown in Figure 3 and Figure 4 optimum FIR wave filter of frequency characteristic.Because this filter order is 500, therefore, signal has 0.04s, namely after filtering

Fixed delay, F _TThe frequency of expression FIR wave filter.

＜2〉electricity grid substation device configuration:

The present invention need to arrange intellectual analysis main website, the optimum FIR wave filter that designs, bus residual voltage monitoring device, each feeder line top zero-sequence current harvester and storer and communication device at the transformer station place.

(2) the intellectual analysis main website is by bus residual voltage monitoring device and each feeder line top zero-sequence current harvester Real-Time Monitoring operation of power networks information, if bus residual voltage U ₀Surpass certain threshold value (0.35 times bus rated voltage), then record this event and occur constantly, and enter step (3), otherwise, this step repeated.

(3) each feeder line top zero-sequence current I that fault collects in the 0.06s time range after occuring constantly gets in the intellectual analysis main website _0i(i=1,2 ..., n, n are the feeder line number) analyze, through optimum FIR filter filtering, obtain transient state capacitive zero-sequence current component I _C0i(i=1,2 ..., n).

(4) to the classification of n bar feeder line, make the class number k=n on the cluster level:

Based on I _C0i(i=1,2 ..., n), according to the hierarchical clustering method, feeder line is classified.N bar circuit, a corresponding n primitive class (be k=n, wherein k is the class number on this cluster level).

(5) determine the distance between any two classes in k the class according to following expression formula, obtain

Individual distance value merges two classes corresponding to minor increment, forms new cluster level, a corresponding k-1 class;

$D_{\mathrm{AB}}={\left|\right|\stackrel{\&OverBar;}{X_A}-\stackrel{\&OverBar;}{X_B}\left|\right|}^2+W_A/N_A+W_B/N_B$ ⑥；

Wherein, X _iBe i object;

With

Be respectively category-A and category-B object mean vector;

Be the poor quadratic sum of category-A inter-object distance;

Be the poor quadratic sum of category-B inter-object distance; N _AAnd N _BBe respectively the object number in A and the category-B.

(6) the pseudo-F on the cluster level in the determining step (5):

According to the pseudo-F value on the cluster level in the following expression formula determining step (5):

PSF＝((W _j-P _G)/(G-1))/(P _G/(N-G)) ⑦；

Wherein, G is the class number on this cluster level; N is the object number in the j class;

$P_G=\underset{j=1}{\overset{G}{\mathrm{\&Sigma;}}}{W}_j;$

Described pseudo-F is used for estimating the Clustering Effect when being classified as G class, and the larger indicated object of its value is classified as G class, and corresponding cluster numbers was best cluster numbers when peak value appearred in pseudo-F.

(7) judge that whether all classes gather is a class:

Make k=k-1, whether check k is 1, and namely whether all objects are poly-is a class, if then enter step (8); If not, then return step (5); Be the corresponding cluster level of pseudo-F maximum value according to pedigree Cluster Assessment index, determine best cluster number m.

(8) based on cluster result, determine fault feeder:

If m=2, and wherein a class only comprises a feeder line, and then this feeder line is fault feeder; If above-mentioned condition does not satisfy and the bus residual voltage continues to surpass threshold value (threshold value is 0.35 times bus rated voltage), then is bus-bar fault.

Should be noted that at last: above embodiment is only in order to illustrate that technical scheme of the present invention is not intended to limit, although with reference to above-described embodiment the present invention is had been described in detail, those of ordinary skill in the field are to be understood that: still can make amendment or be equal to replacement the specific embodiment of the present invention, and do not break away from any modification of spirit and scope of the invention or be equal to replacement, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (11)

1. the fault line selection method for single-phase-to-ground fault of a small current neutral grounding system is characterized in that, described method comprises the steps: based on optimum FIR wave filter and hierarchical clustering

(1) preliminary work of the fault line selection method for single-phase-to-ground fault of small current neutral grounding system;

(2) operation information of Real-time Collection electrical network, and judge whether the bus residual voltage surpasses threshold value;

(3) the feeder line top zero-sequence current that collects is input to the FIR filter filtering, obtains transient state capacitive zero-sequence current component;

(4) to the classification of n bar feeder line, make the class number k=n on the cluster level;

(5) determine the distance between any two classes in k the class, merge two classes corresponding to minor increment, form new cluster level;

(6) the pseudo-F on the cluster level in the determining step (5);

(7) judge that whether all classes gather is a class;

(8) based on cluster result, determine fault feeder.

2. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (1), preliminary work comprises:

＜1〉design FIR wave filter;

＜2〉electricity grid substation device configuration.

3. fault line selection method for single-phase-to-ground fault as claimed in claim 2 is characterized in that, described step＜1〉in, the optimum FIR bandpass filter of design 300～3000Hz comprises the steps:

A, the minimum exponent number of FIR wave filter are estimated:

FIR wave filter low frequency transitional zone is designed to 280～300Hz, according to the Kaiser equation, determines that the minimum exponent number of FIR wave filter is that 500, Kaiser equation is as follows:

$N\&ap;\frac{-20\lg \left(\sqrt{(1-{10}^{{-\mathrm{\&alpha;}}_P/20})\left({10}^{{-\mathrm{\&alpha;}}_S/20}\right)}\right)-13}{14.6({\mathrm{\&omega;}}_s-{\mathrm{\&omega;}}_p)/2\mathrm{\&pi;}}$ ④；

Wherein, N is the minimum exponent number of FIR wave filter, ω _pWith ω _sBe respectively normalization passband and stopband boundary angle frequency, α _pBe peak value passband ripple, α _SBe minimum stop-band attenuation;

B, determine optimum FIR filter parameter, determine that namely exponent number is 500 FIR filter coefficient:

By Constraint least square algorithm, the peak value of the absolute difference of following expression formula is minimized, realize that the actual frequency response H of FIR wave filter (j ω) approaches desirable frequency response D (j ω) to greatest extent, that is:

$\min \underset{\mathrm{\&omega;}\&Element;R}{\max }|H\left(\mathrm{j\&omega;}\right)-D\left(\mathrm{j\&omega;}\right)|$ ⑤；

Frequency signal has 0.04s, namely after filtering

Fixed delay; F _TThe frequency of expression FIR wave filter.

4. fault line selection method for single-phase-to-ground fault as claimed in claim 2, it is characterized in that, described step＜2〉in, at the transformer station place intellectual analysis main website, optimum FIR wave filter, bus residual voltage monitoring device, each feeder line top zero-sequence current harvester and storer and communication device are set.

5. fault line selection method for single-phase-to-ground fault as claimed in claim 1, it is characterized in that, in the described step (2), the intellectual analysis main website is by bus residual voltage monitoring device and each feeder line top zero-sequence current harvester Real-Time Monitoring operation of power networks information, and judgement bus residual voltage U ₀Whether surpass threshold value, if bus residual voltage U ₀Surpass threshold value, then record this event and occur constantly, and enter step (3), otherwise, this step repeated; Described threshold value is 0.35 times bus rated voltage.

6. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (3), each feeder line top zero-sequence current I that fault collects in the 0.06s time range after occuring constantly gets in the intellectual analysis main website _0iAnalyze, through optimum FIR filter filtering, obtain transient state capacitive zero-sequence current component I _C0i, i=1 wherein, 2 ..., n, n are the feeder line number.

7. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (4), based on I _C0i, according to the hierarchical clustering method, feeder line is classified: n is the feeder line number, and a corresponding n primitive class makes k=n, and wherein k is the class number on this cluster level.

8. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (5), determines the distance between any two classes in k the class according to following expression formula, obtains Individual distance value merges two classes corresponding to minor increment, forms new cluster level, a corresponding k-1 class;

$D_{\mathrm{AB}}={\left|\right|\stackrel{\&OverBar;}{X_A}-\stackrel{\&OverBar;}{X_B}\left|\right|}^2+W_A/N_A+W_B/N_B$ ⑥；

Wherein, X _iBe i object;

With

Be respectively category-A and category-B object mean vector;

Be the poor quadratic sum of category-A inter-object distance; Be the poor quadratic sum of category-B inter-object distance; N _AAnd N _BBe respectively the object number in A and the category-B.

9. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (6), according to the pseudo-F value on the cluster level in the following expression formula determining step (5):

PSF＝((W _j-P _G)/(G-1))/(P _G/(N-G)) ⑦；

Wherein, G is the class number on this cluster level;

N is the object number in the j class;

$P_G=\underset{j=1}{\overset{G}{\mathrm{\&Sigma;}}}{W}_j;$

Described pseudo-F is used for estimating the Clustering Effect when being classified as G class, and the larger indicated object of its value is classified as G class, and corresponding cluster numbers was best cluster numbers when peak value appearred in pseudo-F.

10. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (7), makes k=k-1, and whether check k is 1, and namely whether all objects are poly-is a class, if then enter step (8); If not, then return step (5); Be the corresponding cluster level of pseudo-F maximum value according to pedigree Cluster Assessment index, determine best cluster number m.

11. fault line selection method for single-phase-to-ground fault as claimed in claim 1 is characterized in that, in the described step (8), if m=2, and wherein a class only comprises a feeder line, and then this feeder line is fault feeder; If above-mentioned condition does not satisfy and the bus residual voltage continues to surpass threshold value, then is bus-bar fault.

Images