CN103995221A - Ultrasonic receiving array orientation method for transformer partial discharge positioning - Google Patents

Abstract

The invention discloses an orientation method of an ultrasonic receiving array for localizing partial discharge of a transformer. The method is aimed at an L-shaped or cross-shaped ultrasonic receiving array arranged along the edge or central line of the planar array. The present invention combines broadband focusing and array expansion, firstly performs focusing operation on broadband ultrasonic array receiving signals, divides broadband ultrasonic signals generated by partial discharge into time segments, uses noise-free data to calculate the focusing matrix, and focuses the signals to the same frequency segment; then use the fourth-order cumulant algorithm to virtually expand the array, and the data covariance matrix of the array after virtual expansion is replaced by the fourth-order cumulant matrix formed in order by the original unexpanded array; finally, after virtual expansion The array uses a multiple signal classification algorithm to estimate the direction of arrival of partial discharges.

Description

The orientation method of the ultrasound wave receiving array of locating for partial discharge of transformer

[technical field]

The invention belongs to high voltage installation and experimental technique field, specifically a kind of orientation method of the ultrasound wave receiving array for partial discharge of transformer location.

[background technology]

Power transformer is the important component part of electric system.Statistics shows, insulation defect is the first cause that causes transformer fault, and shelf depreciation is the important sign of reflection insulation degradation.The detection of shelf depreciation is the important content that power transformer and other power equipments are carried out to all kinds of tests, and to the accurate positioning relation of shelf depreciation to reflecting more accurately the insulation status of equipment and formulating corresponding maintenance policy, therefore shelf depreciation is positioned and is significant.

At present, for partial discharge of transformer, set up but partial discharge positioning method that always can successful implementation has Electric Method, supercritical ultrasonics technology and the combination of the two.Along with the development of detection technique, people also bring into use ultrahigh-frequency signal to detect and locate shelf depreciation, and the more important thing is begins one's study is applied to shelf depreciation location by array technique.Adopting the advantage of array technique is that positioning precision is high, and accuracy is good, in the situation that sensor does not rotate, realizes the detection to many Partial Discharge Sources simultaneously.Therefore, research shelf depreciation location array technique and related algorithm thereof, have practical meaning in engineering.

At present, adopt array to estimate that shelf depreciation arrival bearing mainly contains two kinds of algorithms, the disposal route of one class based on coherent signal, once there is document to set up 4 * 4 supersonic array, with signal subspace conversion SST algorithm pair array signal, focus on, then carry out direction finding in conjunction with Fast Subspace direction finding FDOA algorithm.Yet the very few array of array number can cause orientation accuracy low.Equations of The Second Kind is the array extension method based on Higher Order Cumulants, once developed the miniature compound sensor that comprises a L shaped supersonic array of 13 array element and a uhf sensor, and the compound sensor that comprises 13 array element cruciform ultrasound arrays and 2 * 2 array element uhf sensors composition, and utilize fourth order cumulant algorithm to carry out after virtual extended supersonic array wherein, Virtual array number has reached respectively 61 and 97, use again multiple signal classification MUSIC algorithm to carry out the orientation of Partial Discharge Sources, and obtain ultrasound-transmissive time delay in conjunction with ultrahigh frequency, realized Partial Discharge Sources has been located.Yet the equivalent covariance matrix that virtual array expansion obtains, only utilizes arrowband spatial spectral estimation algorithm to process, and the broad band ultrasonic signal directional effect of shelf depreciation radiation is remained to deficiency.Therefore realize supersonic array technology practical target in partial discharge of transformer location, on algorithm, must realize the adaptation of broad band ultrasonic signal is combined with array extension.

[summary of the invention]

The object of the invention is to for the deficiencies in the prior art, a kind of orientation method of the ultrasound wave receiving array for partial discharge of transformer location is provided, and the method is for along planar array array edge or center line is that arrange, L-shaped or criss-cross ultrasound wave receiving array.The method receives signal to the supersonic array in broadband and carries out focusing operation, the narrow band signal that formation comprises all information, form again Higher Order Cumulants L shaped or cruciform array are carried out to virtual extended, the covariance matrix that obtains focusing on the array extending of rear narrow band signal, carries out Estimation of Spatial Spectrum to incoming wave and reaches the accurate pointing to shelf depreciation.

For achieving the above object, the present invention adopts following technical scheme:

The orientation method of the ultrasound wave receiving array of locating for partial discharge of transformer, comprises the following steps:

1) the ultrasonic array signal that utilizes ultrasonic array sensor receiving transformer shelf depreciation to produce, is divided into J time period X by the ultrasonic array signal receiving ₀, X ₁..., X _j-1, calculate the covariance matrix of each time period noise free data, its computing formula is as follows:

$P_j=R_j-{\mathrm{\σ}}_j^{2}I---\left(1\right)$

In formula:

j＝0,1,...J-1；

R _jfor X _jcovariance matrix;

for R _jlittle eigenwert and;

I is unit matrix;

2) focussing matrix of selected segmented array signal, its computing formula is as follows:

wherein, j=1 ... J-1 (2)

In formula:

with respectively P ₀and P _jeigenvector;

3) by step 2) the focussing matrix T that obtains _jpremultiplication X _jevery section of narrow-band array signal Y after can focusing on _j(f ₀), its computing formula is as follows:

Y _j(f ₀)＝T _jX _j j＝1,...J-1 (3)

In formula:

F ₀reference frequency for array signal;

4) calculate and focus on rear every section of narrow-band array signal Y _j(f ₀) fourth order cumulant matrix R _{4, j}, and calculate its mean value Rcum, its computing formula is as follows:

$\begin{array}{c}{R}_{4,j}=E\{{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})}^H\\ -E\left\{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})\right\}\·E\left\{{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})}^H\right\}\\ -E\left\{\left(Y_j\left(f_0\right)Y_j{\left(f_0\right)}^H\right)\right\}\⊗E\left\{{\left(Y_j\left(f_0\right)Y_j{\left(f_0\right)}^H\right)}^{*}\right\}\end{array}---\left(4\right)$

In formula:

E is for asking matrix simple crosscorrelation;

for asking the Kronecker product of matrix;

* for asking matrix transpose;

H is for asking Matrix Conjugate transposition;

$\mathrm{Rcum}=\frac{1}{J}\underset{j=0}{\overset{J-1}{\mathrm{\Σ}}}{R}_{4,j}---\left(5\right)$

5) Rcum is carried out to feature decomposition, obtain noise subspace En, and carry out spectrum peak search by the spectrum estimation formulas of formula (6), obtain the corresponding arrival bearing of maximal point be local discharge signal arrival bearing, its computing formula is as follows:

In formula:

represent that noise signal exists the inverse of the spatial spectrum energy of direction;

θ and represent respectively arrival bearing's the angle of pitch and position angle;

popular vector for supersonic array.

Compared with prior art, the present invention has following technique effect:

The present invention uses supersonic array technology to position adopted algorithm to Partial Discharge Sources or is the Wideband Signal Processing algorithm, or the arrowband algorithm based on Higher Order Cumulants.The Wideband Signal Processing algorithm is owing to failing to realize array extension, so array aperture is little, and the very few array of array number can cause orientation accuracy low.Although the arrowband algorithm based on Higher Order Cumulants has been realized array extension, but existing algorithm can only utilize a part of narrow band information in broadband signal to carry out arrival bearing's estimation, lost bulk information, therefore the broad band ultrasonic signal directional effect Shortcomings of existing algorithm to shelf depreciation radiation.

The invention solves the problem of supersonic array technology in partial discharge of transformer position application, Wideband Focusing of the present invention need not estimated in signal arrival bearing situation, there is the ability of processing well broadband signal, fourth order cumulant algorithm can be realized the virtual extended of array, increase array number, there is the ability that suppresses additive white Gaussian noise and Gauss's coloured noise simultaneously.The present invention realizes supersonic array technology to lay a good foundation in the practicality of partial discharge of transformer location.

[accompanying drawing explanation]

The three array element expansion schematic diagram that Fig. 1 (a) is ultrasound array sensor, the L shaped array extension schematic diagram that Fig. 1 (b) is ultrasound array sensor, the cruciform array extension schematic diagram that Fig. 1 (c) is ultrasound array sensor;

Space spectrogram based on fourth order cumulant arrowband algorithm when Fig. 2 (a) is single Partial Discharge Sources, space spectrogram based on TCT algorithm when Fig. 2 (b) is single Partial Discharge Sources, the space spectrogram based on fourth order cumulant and burst focusing algorithm when Fig. 2 (c) is single Partial Discharge Sources;

Space spectrogram based on fourth order cumulant arrowband algorithm when Fig. 3 (a) is two Partial Discharge Sources, space spectrogram based on TCT algorithm when Fig. 3 (b) is two Partial Discharge Sources, the space spectrogram based on fourth order cumulant and burst focusing algorithm when Fig. 3 (c) is two Partial Discharge Sources.

[embodiment]

Below in conjunction with the drawings and specific embodiments, the present invention will be further described in detail.

The present invention is for the orientation method of the ultrasound wave receiving array of partial discharge of transformer location, the method is for the supersonic array for transformer shelf depreciation arrival bearing orientation, and array is L-shaped or cruciform etc. along face a burst of row edge or center line arrange, see Fig. 1.

(1) choose the focusing that focussing matrix is realized broadband signal

It is to utilize the relation between the noise free data on not overlapping Frequency point in frequency band to choose focussing matrix at structure focussing matrix herein that broadband signal is assembled crucial, and the signal space on each frequency is focused on to reference frequency point.

First the array signal receiving being divided into J time period is X ₀, X ₁..., X _j-1, calculate each time period noise free data covariance matrix,

$P_j=R_j-{\mathrm{\σ}}_j^{2}I,j=\mathrm{0,1},...J-1---\left(1\right)$

In formula: R _jbe the covariance matrix of j segment data, be the little eigenwert of covariance matrix and.

Focussing matrix

$T_j={\stackrel{\‾}{Q}}_0{\stackrel{\‾}{Q}}_j,j=1,...,J-1---\left(2\right)$

with respectively P ₀and P _jeigenvector.The focussing matrix T trying to achieve by formula (2) _jpremultiplication X _jarray signal Y after can focusing on _j(f ₀).

Y _j(f ₀)＝T _jX _j j＝1,...J-1 (3)

So, utilize focusing algorithm by the data focusing transform of broadband signal to reference frequency f ₀the narrow band data at place.

(2) array extension based on fourth order cumulant

Utilize Higher Order Cumulants treatment technology to carry out virtual extended to former cruciform or L shaped array, set up the mathematical model of supersonic array, array element is positioned in X, Y-axis and equidistantly arranges, and P point is partial discharge position, θ with represent respectively arrival bearing's the angle of pitch and position angle.Each array element coordinate subtracts the coordinate that just can obtain Virtual array mutually, and as shown in Figure 1, filled circles is actual element position to the position before and after expansion, and open circles is Virtual array position.As seen from Figure 1, after expansion, array element is face battle array.

Carrying out the array extension step that also realization positions shelf depreciation is.

Calculate and focus on rear every section of narrow-band array signal Y _j(f ₀) the mean value Rcum of fourth order cumulant matrix;

Rcum is carried out to feature decomposition, obtain noise subspace En, and carry out spectrum peak search by the spectrum estimation formulas of formula (4), obtain maximal point corresponding be Partial discharge signal arrival bearing.

Take 25 array elements cruciform array and be example.Adopt respectively the arrowband algorithm of TCT segmentation focusing algorithm, fourth order cumulant and the integration algorithm of this paper to broadband array signal is carried out to Estimation of Spatial Spectrum, the space spectrogram obtaining as shown in Figures 2 and 3.Comparison diagram 2 (a) and Fig. 2 (b) and (c), can find, the space spectrogram acuteness of Fig. 2 (a) is the poorest, the broadband signal bad adaptability of fourth order cumulant algorithm to shelf depreciation radiation.Comparison diagram 2 (b) can obtain with Fig. 2 (a) with (c), and the pseudo-peak in the former figure is more, illustrates that TCT Wideband Focusing algorithm is poor to the rejection of noise.The situation of locating from two Partial Discharge Sources of Fig. 3, above conclusion is more obvious.In a word, from the result to single Partial Discharge Sources or two Partial Discharge Sources, integration algorithm with only implement fourth order cumulant arrowband algorithm or TCT segmentation focusing algorithm is compared, in space, spectrogram acuteness is directional effect, there is to better effect the aspects such as inhibition of Noise and Interference.

At present, the algorithm that is applied to shelf depreciation supersonic array location has, array extension and Wideband Focusing both direction.Fourth order cumulant array extension algorithm is arrowband spatial spectral estimation algorithm, can not adapt to the broadband signal of Partial Discharge Sources radiation; And TCT Wideband Focusing algorithm can not be implemented effectively expansion by pair array, and poor to the inhibition of noise.Integration algorithm has advantages of fourth order cumulant algorithm and TCT algorithm, has realized array extension and Wideband Focusing to combination, has increased array element quantity, good to the inhibition of noise, can well adapt to the broadband signal of shelf depreciation simultaneously.

Claims (1)

1. the orientation method of the ultrasound wave receiving array of locating for partial discharge of transformer, is characterized in that, comprises the following steps:

1) the ultrasonic array signal that utilizes ultrasonic array sensor receiving transformer shelf depreciation to produce, is divided into J time period X by the ultrasonic array signal receiving ₀, X ₁..., X _j-1, calculate the covariance matrix of each time period noise free data, its computing formula is as follows:

$P_j=R_j-{\mathrm{\σ}}_j^{2}I---\left(1\right)$

In formula:

j＝0,1,...J-1；

R _jfor X _jcovariance matrix;

for R _jlittle eigenwert and;

I is unit matrix;

2) focussing matrix of selected segmented array signal, its computing formula is as follows:

wherein, j=1 ... J-1 (2)

In formula:

with respectively P ₀and P _jeigenvector;

3) by step 2) the focussing matrix T that obtains _jpremultiplication X _jevery section of narrow-band array signal Y after can focusing on _j(f ₀), its computing formula is as follows:

Y _j(f ₀)＝T _jX _j j＝1,...J-1 (3)

In formula:

F ₀reference frequency for array signal;

4) calculate and focus on rear every section of narrow-band array signal Y _j(f ₀) fourth order cumulant matrix R _{4, j}, and calculate its mean value Rcum, its computing formula is as follows:

$\begin{array}{c}{R}_{4,j}=E\{{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})}^H\\ -E\left\{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})\right\}\·E\left\{{(Y_j\left(f_0\right)\⊗Y_j{\left(f_0\right)}^{*})}^H\right\}\\ -E\left\{\left(Y_j\left(f_0\right)Y_j{\left(f_0\right)}^H\right)\right\}\⊗E\left\{{\left(Y_j\left(f_0\right)Y_j{\left(f_0\right)}^H\right)}^{*}\right\}\end{array}---\left(4\right)$

In formula:

E is for asking matrix simple crosscorrelation;

for asking the Kronecker product of matrix;

* for asking matrix transpose;

H is for asking Matrix Conjugate transposition;

$\mathrm{Rcum}=\frac{1}{J}\underset{j=0}{\overset{J-1}{\mathrm{\Σ}}}{R}_{4,j}---\left(5\right)$

5) Rcum is carried out to feature decomposition, obtain noise subspace En, and carry out spectrum peak search by the spectrum estimation formulas of formula (6), obtain the corresponding arrival bearing of maximal point be local discharge signal arrival bearing, its computing formula is as follows:

In formula:

represent that noise signal exists the inverse of the spatial spectrum energy of direction;

θ and represent respectively arrival bearing's the angle of pitch and position angle;

popular vector for supersonic array.