Background
Partial Discharge (PD) is an electrical phenomenon caused by local electric field concentration inside or on the surface of an insulating material, and PD is an important symptom of insulation degradation of a dry reactor, and long-term PD may cause insulation degradation, threaten the safe operation of the dry reactor, cause a fault in operation, and affect the safe and stable operation of a power system. Through the online monitoring and positioning of the PD, the latent defect of the internal insulation of the dry-type reactor can be found in time, the degree of the internal insulation degradation of the dry-type reactor is judged, the position of the PD source is visualized, great convenience is provided for the overhaul of the dry-type reactor, and the occurrence of sudden insulation faults of the dry-type reactor is avoided.
Physical phenomena and chemical reactions, such as sound, light, electricity, heat, etc., occur with PD. Accordingly, there are various methods for detecting partial discharge, such as ultrasonic detection, pulse current method, optical method, ultrahigh frequency method, and chemical detection method. The dry-type reactor mostly operates under a large-current working condition, acoustic signals caused by electromagnetic vibration of the dry-type reactor have obvious phase amplitude characteristics, and the characteristics of the generated acoustic signals can change along with insulation faults in the dry-type reactor, so that the ultrasonic detection method is commonly used for fault monitoring of the dry-type reactor, has high sensitivity and simple and convenient equipment use, can be used for defect positioning through point-by-point measurement of the equipment, and is suitable for being applied to live-line inspection and short-term online monitoring of the dry-type reactor.
However, in the practical application process, it is found that when the dry reactor PD is positioned by an ultrasonic detection method, due to the influence of various interference noises, environmental states, measurement errors and other factors, in the process of solving a PD source time difference positioning equation by using a grid search Newton iterative algorithm recommended by the Power Transformer PD ultrahigh frequency detection method (DL/T-2016) in the Power industry standard, the problem that one PD source of the dry reactor can be positioned by a plurality of coordinates or cannot be positioned exists, and the positioning precision cannot meet the requirement; in addition, in order to make the positioning device simple in structure, save space and reduce cost, the sound pressure sensor array is generally arranged on a horizontal plane, according to the principle of the grid search newton iterative algorithm, when the sound pressure sensor is arranged on the same horizontal plane, when the X, Y, Z coordinates are subjected to equal-step optimization, a large error exists in the positioning result of the Z, and the positioning result of the PD source of the dry reactor is not accurate enough.
The reasons stated above in two aspects can reduce the positioning accuracy of the dry reactor PD source by the traditional grid search Newton iteration algorithm. Therefore, it is necessary to improve the existing positioning algorithm to provide a PD source positioning algorithm with higher accuracy to solve the PD source of the dry reactor.
Disclosure of Invention
Aiming at the problems in the background art, the invention provides a PD source positioning algorithm based on a gradient approximation type dry reactor.
In order to solve the technical problems, the invention adopts the following technical scheme: a PD source positioning algorithm based on a gradient approximation type dry reactor is characterized in that a traditional grid search Newton iteration method is improved, and the improved algorithm is in a step shape and comprises steps 1 and 2; step 1, roughly positioning a PD source by utilizing a grid search Newton iteration method; step 2, carrying out distance and value minimum value approximation type optimization on the initial positioning coordinate X, Y, and substituting X, Y after optimization into a time difference positioning equation to obtain a Z coordinate; the method comprises the following steps:
step 1, step 1 comprises the following specific steps:
step 1.1, receiving an acoustic signal generated by a PD source of the dry type reactor by a sound pressure sensor array;
step 1.2, according to a time difference column PD source positioning time difference equation of the acoustic signal acquired by the acoustic pressure sensor compared with a reference acoustic pressure sensor:
wherein V is the propagation speed of the acoustic signal,
step 1.3, substituting the collected PD source sound signal data of the dry reactor into a PD source positioning time difference equation to form a coefficient equation matrix related to variables x, y and z;
step 1.4, carrying out preliminary solution on the positioning time difference equation matrix by utilizing a grid search Newton iteration method to obtain a rough N groups of coordinate point sets P of the initial PD sourcek(Xk,Yk),k=1,2,3,…;
Step 2, step 2:
step 2.1, establishing N groups of coordinate point sets Pk(Xk,Yk) With different sound pressure sensor coordinates Sl(Xl,Yl) L x N linear equation sets among the acoustic pressure sensors, wherein l is the number of the acoustic pressure sensors, and N is the number of initial values of the samples;
step 2.2, solving the position of the true PD source to be solved as P '(X', Y '), and solving the sum of vertical distances from P' to an l X N linear equation according to the plane geometric knowledge;
step 2.3, assuming a plurality of coordinates P ', and carrying out global search and minimum approximation type optimization on the vertical distances and values under different P' to obtain the minimum value d of the sum of the vertical distancesmin;
Step 2.4, minimize dminCorresponding coordinate P0(X0,Y0) Obtaining Z by the equation of time difference between back substitution and positioning0Positioning the fault position;
step 2.5, finishing the PD source coordinate positioning of the dry type reactor with the coordinate P0(X0,Y0,Z0)。
Compared with the prior art, the invention provides a PD source positioning algorithm based on a gradient approximation type dry reactor, which improves the traditional grid search Newton iteration method, the improved algorithm is in a step shape, and the steps are as follows, step 1: roughly positioning the PD source by utilizing a grid search Newton iteration method; step 2: and (3) performing approximate optimization of the distance and value minimum value on the initial positioning coordinate X, Y, and substituting X, Y after optimization into the time difference positioning equation to obtain the Z coordinate. The improved algorithm reduces the positioning error of the Z coordinate to a certain extent and improves the PD source positioning precision of the dry type reactor.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The present invention is further illustrated by the following examples, which are not to be construed as limiting the invention.
When a grid search Newton iterative algorithm is used for solving the PD source of the acoustic signal dry reactor, the problem that a plurality of PD source coordinates are positioned or the solution is not effective exists, and when the sound pressure sensor is arranged on the same horizontal plane, when the algorithm is used for carrying out equal-step optimization on X, Y, Z coordinates, a large error exists in the positioning result of Z, and the real PD source positioning result of the dry reactor is not accurate enough. The PD source positioning algorithm based on the gradient approximation type dry reactor is an improved algorithm based on a grid search Newton iteration algorithm, a plurality of PD source positioning coordinates are combined through a vertical distance minimum value approximation type optimization method, the optimization result is substituted back to a positioning time difference equation to obtain a Z coordinate, the problem that the grid search Newton iteration algorithm has a large Z positioning error can be solved, and the PD source positioning accuracy of the dry reactor is effectively improved.
The embodiment is realized by the following technical scheme, and the PD source positioning algorithm based on the gradient approximation type dry reactor specifically comprises the following steps:
step 1:
s1.1: the sound pressure sensor array always receives sound signals generated at the PD source of the dry-type reactor, and sound signals at the PD source collected by different sound pressure sensors have a certain time difference;
s1.2: according to the time difference of the acoustic signal acquired by the acoustic pressure sensor compared with the reference acoustic pressure sensor, a column PD source positioning time difference equation is shown as a formula 1:
where V is the speed of propagation of the acoustic signal,
s1.3: substituting the acquired acoustic signal data into a positioning time difference equation to form a coefficient equation matrix;
s1.4: the grid search Newton iteration method is utilized to carry out preliminary solution on the positioning time difference equation matrix to obtain N groups of coordinate positions P of the rough initial PD sourcek(Xk,Yk)。
Step 2:
s2.1: establishing N groups of PkWith different sound pressure sensor coordinates Sl(Xl,Yl) L x N linear equation sets in between;
s2.2: assuming that the position of a true PD source to be solved is P '(X', Y '), and solving the sum of vertical distances from P' to an l X N linear equation according to the plane geometric knowledge;
s2.3: assuming a plurality of coordinates P 'and carrying out global search and minimum approximation type optimization on vertical distances and values under different P' to obtain a minimum value d of the sum of the vertical distancesmin;
S2.4: will dminTo what is providedCorresponding coordinate P0(X0,Y0) back to the positioning time difference equation to obtain Z0;
S2.5: completing the PD source coordinate positioning of the dry type reactor with the coordinate of P0(X0,Y0,Z0)。
In the embodiment, the PD source positioning algorithm flow of the dry type reactor is improved based on the traditional grid search Newton iteration method, the improved algorithm is in a step shape, and in a step 2, the initial rough positioning coordinate P in the step 1 is subjected tok(Xk,Yk) Carrying out approximate optimization on the value by the vertical distance and the minimum value, and carrying out P after optimizationkAnd then the Z coordinate is calculated in the equation of time difference positioning. The improved dry reactor PD source positioning algorithm has the basic idea that the PD source coordinate P with higher precision is obtained through the approximation optimization of the minimum value of the vertical distance and the value0(X0,Y0) And solving the influence of the variables x and y on the variable Z in the process of solving the hyperboloid time difference positioning method by solving a univariate equation F (Z) about Z to eliminate the influence of the variables x and y on the variable Z, wherein the step 1 is rough positioning, and the step 2 is improved fine positioning based on the step 1. The improved algorithm reduces the positioning error of the Z coordinate to a certain extent, and improves the positioning precision of the PD source of the dry reactor.
In specific implementation, a specific flow and principle of a PD source location algorithm based on a gradient approximation dry reactor is shown in fig. 1.
Step 1:
step 1: the sound pressure sensor array always receives sound signals generated by the PD source of the dry-type reactor, and sound signals at the PD source acquired by different sound pressure sensors have certain time difference due to different space positions of the sound pressure sensors;
step 2: as shown in fig. 2, according to the sound pressure sensor S1、S2、S3、S4Time difference of collected acoustic signal, and sound pressure sensor1The time of the collected acoustic signal is a PD source positioning time difference equation of a reference time point column, and the equation is as the formula 1:
where V is the speed of sound of propagation of the acoustic signal,
and step 3: substituting the collected PD source acoustic signal data of the dry reactor into a positioning time difference equation to form a coefficient equation matrix related to the variables x, y and z;
and 4, step 4: the method comprises the steps of utilizing a grid search Newton iteration method to carry out preliminary solution on a positioning time difference equation matrix to obtain N groups of coordinate point sets P of a rough initial PD sourcek(Xk,Yk) (k-1, 2,3 …), P in fig. 21、…、Pn。
Step 2:
step 1: as shown in FIG. 2, P may be assumed1To PnRandomly distributed around the center of the real PD source, which is P ' (X ', Y ') point in FIG. 21To PnCoordinates S of sound pressure sensorl(Xl,Yl) L x N linear equations (the number of sound pressure sensors l is 4 in this embodiment) between the two, and is represented by P1To PnThe straight line determined with the sound pressure sensor may be regarded as an intersection line with the circle P';
step 2: the intersection line has a chord center distance from P', and the 4 sound pressure sensors have 4 x N chord center distances, which are summed as:
wherein k is 1,2, …, and N is the initial number of samples; the number of sensors is 1,2,3 and 4.
Using knowledge of the space geometry, the chord-center distance and the function d can be rewritten as follows:
wherein S islAs a sound pressure sensor coordinate Sl(Xl,Yl)。
And step 3: assuming a plurality of coordinates P', there will be a plurality of sum functions d, and the formula can be expressed as:
in the formula, r is the number of samples in the optimization sum value d, and the larger r is, the higher the progress is. In the sample set there is one drThe minimum value of (d) is obtained by adopting global searching and minimum value approximate optimization;
and 4, step 4: coordinate P corresponding to the optimized minimum d value0(X0,Y0) Substituting into the positioning time difference equation to obtain Z0. For the dry reactor arranged in the shape of a Chinese character pin as shown in FIG. 3, the coordinate P is obtained0That is, the fault dry-type reactor can be positioned, and Z can be further obtained0The location of the fault on the dry reactor can be determined; for the dry type reactor mounted in a stacked manner as shown in fig. 4, Z is required0The fault reactor can be found out and the fault position can be accurately positioned;
and 5: completing the PD source coordinate positioning of the dry type reactor with the coordinate of P0(X0,Y0,Z0)。
In the embodiment, a dry reactor PD source positioning algorithm flow is improved based on a traditional grid search Newton iteration method, and in a step 2, an initial rough dry reactor PD source positioning coordinate P in a step 1 is subjected tok(Xk,Yk) And performing approximate optimization of the minimum value of the vertical distance and the value, and replacing X, Y after optimization into a time difference positioning equation to obtain a Z coordinate. The improved PD source positioning algorithm based on the gradient approximation type dry reactor is based on the basic idea that the PD source coordinate P with higher precision is obtained by optimizing the minimum value of the vertical distance and the value through the particle swarm optimization0(X0,Y0) Then, solving the multivariate optimization function by solving the univariate equation F (Z) elimination time difference positioning method about ZThe influence of the variables x, y on the variable z during the number F (x, y, z), step 1 is a coarse positioning and step 2 is a refined fine positioning based on step 1. The improved algorithm reduces the positioning error of the Z coordinate to a certain extent, and improves the positioning precision of the PD source of the dry reactor.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.