CN-116430184-B - Transformer partial discharge positioning method for eliminating abnormal time difference based on multiple sensors

CN116430184BCN 116430184 BCN116430184 BCN 116430184BCN-116430184-B

Abstract

The invention discloses a transformer partial discharge positioning method based on multi-sensor abnormal time difference elimination, which comprises the steps of S1, installing a sensor for receiving partial discharge signals on the surface of a transformer, extracting time for the sensor to receive the partial discharge signals, S2, establishing a spherical positioning equation set under a space rectangular coordinate system and linearizing, S3, solving the linearization equation set, screening to obtain an initial solution set, S4, carrying out cluster analysis on the initial solution set by using a DBSCAN clustering algorithm to obtain a plurality of clusters and outliers, S5, classifying the clusters according to residual errors, counting measuring time difference frequencies corresponding to the clusters and the outliers, identifying good time difference, common time difference and abnormal time difference, S6, removing the abnormal time difference, taking measuring noise of the residual time difference into consideration, and establishing a constraint overall least square model, S7, selecting an iterative initial solution, and calculating a final positioning solution by using a Newton iteration method. The invention can eliminate abnormal time difference and reduce noise influence, thereby still having higher positioning precision under higher noise environment.

Inventors

LI BING
WANG MENGNAN
YAN SI
ZUO LEI
YIN BAIQIANG

Assignees

合肥工业大学

Dates

Publication Date: 20260508
Application Date: 20230504

Claims (7)

1. The transformer partial discharge positioning method based on the abnormal time difference elimination of the multiple sensors is characterized by comprising the following steps: s1, mounting a sensor for receiving partial discharge signals on the surface of a transformer, and extracting the time when the sensor receives the partial discharge signals; s2, establishing a spherical positioning equation set under a space rectangular coordinate system and linearizing; s3, solving a linearized equation set, and obtaining an initial solution set after screening; s4, performing cluster analysis on the initial solution set by using a DBSCAN clustering algorithm to obtain a plurality of clusters and outliers; s5, classifying the clusters according to residual errors, counting the measured time difference frequency corresponding to the clusters and the outliers, and identifying good time difference, common time difference and abnormal time difference; S6, eliminating abnormal time difference, and taking measurement noise of the residual time difference into consideration to establish a constraint overall least square model; s7, selecting an iterative initial solution, and calculating a final positioning solution by using a Newton iterative method; The method for classifying clusters according to residual errors in S5 comprises the following steps: s51, calculating average value of all initial solutions in each cluster ,(h = 1,2,...,c); S52, for the measurement time difference τ i , the absolute residual of its corresponding spherical equation is: ; s53, defining the average absolute residual error of the h cluster as ; S54, solving the average value of the h cluster Substituting the absolute residual error into the above formula, and obtaining the average absolute residual error of each cluster; s55, selecting a cluster with the minimum average absolute residual error as a good cluster, and taking an average value solution X g of the cluster as an iteration initial solution; The method for identifying the good time difference, the common time difference and the abnormal time difference in S5 comprises the following steps: s56, taking the cluster with the minimum average residual as a good cluster, taking other clusters as common clusters, and then counting the frequency f i and f i ' of the initial coordinate solutions corresponding to the measurement time difference tau i in the good cluster and the outlier respectively; s57, if the frequency of the measurement time difference tau i meets the following conditions: And is also provided with Considering the measurement time difference tau i as an abnormal measurement time difference, wherein f o is the maximum frequency of the measurement time difference obtained by statistics in an outlier, and f g is the maximum frequency of the measurement time difference obtained by statistics in a good cluster; S58, if the frequency of the measurement time difference tau i meets the following conditions: And is also provided with Then the measurement time difference τ i is considered to be a good time difference; s59, the rest other measurement time differences are common measurement time differences; the method for constructing the constraint overall least square model in the S6 comprises the following steps: S61, eliminating equations corresponding to the abnormal time difference, and combining n-1 linear equations without the abnormal time difference to obtain a linear equation set AX=b, wherein n is the number of the residual measurement time difference; s62 covariance matrix of time difference noise eta subject to normal distribution By Cholesky decomposition Can obtain whitened noise vector A constrained overall least squares model can thus be built for the linear equation set ax=b: ; where u is the whitened noise vector of the noise eta of the measured time difference, And The noise terms of matrix a and vector b, respectively.
2. The method for positioning partial discharge of a transformer based on abnormal time difference elimination of multiple sensors according to claim 1, wherein the sensors in S1 are ultrasonic sensors or ultrahigh frequency sensors, and the number of the sensors is N >6.
3. The method for positioning partial discharge of a transformer based on abnormal time difference elimination of multiple sensors according to claim 1, wherein in S2, a vertex of the transformer is taken as an origin, a space rectangular coordinate system is established, and a spherical positioning equation set of a partial discharge source can be established based on a TDOA principle: ; Wherein N is the number of sensors, P (x, y , z) is the coordinates of the partial discharge source, (x i , y i, z i ) is the coordinates of the sensor S i , t 1 is the time for the partial discharge signal to reach the reference sensor S 1 from the partial discharge source, For the measured time differences between the respective sensor S i and the reference sensor (i=2, 3, v is the isosceles wave velocity.
4. The method for positioning partial discharge of a transformer based on abnormal time difference elimination of multiple sensors according to claim 1, wherein in S3, one other spherical positioning equation is arbitrarily selected to be subtracted from the spherical positioning equation of the reference sensor, so as to obtain a linear equation: ; Wherein: , , , , , ,(i = 2,3,...,N)。
5. The method for positioning partial discharge of a transformer based on abnormal time difference elimination according to claim 4, wherein one reference sensor and any 5 other sensors can be obtained as a solution Is co-available with N sensors Equation set, solving by Gaussian elimination A linear equation set is obtained The initial solution is adopted, the equivalent wave velocity v is obviously beyond the normal wave velocity range or is discarded as the initial positioning solution of the imaginary number, and an initial coordinate solution set containing m samples is obtained 。
6. The transformer partial discharge positioning method based on the abnormal time difference elimination of the multiple sensors according to claim 1, wherein the step of performing cluster analysis on the initial solution set by the DBSCAN clustering algorithm in S4 is as follows: s41, inputting a neighborhood search radius Eps, a minimum point number MinPts in the neighborhood and an initial coordinate solution set H; s42, randomly selecting a coordinate point P i without a cluster label from the initial coordinate solution set H; S43, finding all points from the point P i which can reach the densities of Eps and MinPts; S44, if P i is a core point, namely, taking a point P i as a center, forming a new cluster at the moment and adding new cluster labels to all points in the cluster, and then processing the next core point in the cluster, continuously collecting core points with reachable density and adding the core points into the cluster until no new core points are added, wherein the cluster becomes a complete cluster at the moment; S45, if P i is a boundary point, namely, a point P i is taken as a center, and the number of points contained in a neighborhood with the radius of Eps is less than MinPts, then P i has no point with reachable density; s46, repeating the steps S42-S45 until all points in the initial coordinate solution set H are processed; And S47, after the clustering is completed, c clusters and l outliers are obtained.
7. The transformer partial discharge positioning method based on the abnormal time difference elimination of the multiple sensors according to claim 1, wherein the constraint overall least square method model is solved by adopting a Newton iteration method, and the CTLS solution is a function The solution of the minimum value is taken, and the iterative formula is as follows: ; Wherein: And A gradient matrix and a black plug matrix of the function F (X), respectively; Calculating correction vectors at each iteration Frobenius norm of (F) Condition of When satisfied, the iteration ends, where ε is the precision.

Description

Transformer partial discharge positioning method for eliminating abnormal time difference based on multiple sensors Technical Field The invention relates to the field of transformer fault positioning, in particular to a transformer partial discharge positioning method based on abnormal time difference elimination of multiple sensors. Background Transformers are one of the most central devices in power systems, bear the heavy duty of voltage class conversion, and the running state of the transformers directly affects the safety and stability of the whole system. However, with long-term operation of the transformer, insulation defects may occur, and most of the catastrophic insulation breakdown is developed by the cumulative effect of partial discharge over time. Therefore, in order to ensure reliable operation of the transformer, a rapid and accurate positioning of the discharge source is necessary when partial discharge of the transformer occurs. Currently, the partial discharge positioning method is mainly based on three technologies of time difference of arrival (TIME DIFFERENCE ofArrival, TDOA) positioning, direction of arrival (direction ofarrival, DOA) positioning and received signal strength indication (RECEIVED SIGNAL STRENGTH indication, RSSI) positioning. Because of the simple principle of TDOA and the mature technology, the TDOA is widely applied to various passive positioning problems including the positioning of partial discharge of a transformer. The key to solving the source coordinates by the TDOA technique is to acquire the time difference and calculate a system of nonlinear equations. However, on one hand, because the positioning equation set established based on the method is often highly nonlinear and cannot be directly solved, the solution is generally carried out through a traditional iterative algorithm and a heuristic search algorithm at present. The newton iteration method has the advantages of square convergence speed, and can obtain an accurate solution with fewer iteration times, but has obvious defects that only one initial value exists and the search path is single, so that an initial value close to a real solution needs to be selected, otherwise, no solution or larger solved error can be caused. Compared with an iterative algorithm, the heuristic algorithm has the advantages that an initial value is not needed, but the influence of measurement noise on a result cannot be considered in the optimizing process. On the other hand, the propagation speed of the partial discharge signal is very fast, so the method has higher requirement on the time difference measurement precision, and when the time difference measurement contains noise or even abnormal value, the error of the positioning result is often larger. It can be seen that the quality of the measurement time difference and the noise level will directly influence the positioning accuracy. The existing solving method has the advantages that the measured values are few, so that the time difference measurement quality is generally not identified and screened, even the influence of measurement noise is ignored in the solving process, and the positioning result error of the algorithm solved in the scene containing abnormal time difference and higher measurement noise is larger. Disclosure of Invention Based on the technical problems in the background art, the invention provides the transformer partial discharge positioning method based on the multi-sensor abnormal time difference elimination, which can eliminate the abnormal time difference and reduce the noise influence, so that the transformer partial discharge positioning method still has higher positioning precision in a higher noise environment. The invention provides a transformer partial discharge positioning method based on multi-sensor abnormal time difference elimination, which comprises the following steps: s1, mounting a sensor for receiving partial discharge signals on the surface of a transformer, and extracting the time when the sensor receives the partial discharge signals; s2, establishing a spherical positioning equation set under a space rectangular coordinate system and linearizing; s3, solving a linearized equation set, and obtaining an initial solution set after screening; s4, performing cluster analysis on the initial solution set by using a DBSCAN clustering algorithm to obtain a plurality of clusters and outliers; s5, classifying the clusters according to residual errors, counting the measured time difference frequency corresponding to the clusters and the outliers, and identifying good time difference, common time difference and abnormal time difference; S6, eliminating abnormal time difference, and taking measurement noise of the residual time difference into consideration to establish a constraint overall least square model; s7, selecting an iterative initial solution, and calculating a final positioning solution by using a Newton iteration method. Prefe