Transformer internal voiceprint detection abnormity positioning method

Abstract

The invention relates to a transformer internal voiceprint detection abnormity positioning method. The method comprises the following steps: S1, acquiring voiceprint signals of different position points in a transformer through a distributed optical fiber acoustic sensing technology; s2, converting the voiceprint signal of each position point into a voiceprint frequency domain signal to obtain a sound pressure representation vector of the corresponding position point; synthesizing the sound pressure characterization vectors of all the position points to construct a sound pressure distribution feature vector in the transformer; and S3, on the basis of the internal sound pressure distribution feature vector of the transformer, constructing a target function for reflecting sound pressure variation features, and then finding an individual corresponding to the maximum sound pressure variation by adopting a genetic algorithm, the position point corresponding to the individual being an abnormal position point, and a certain spatial range taking the abnormal position point as the center being an abnormal region. According to the method, the abnormal position in the transformer can be accurately positioned, so that the operation reliability and safety of the transformer are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment detection, and in particular to a method for locating abnormalities in transformer internal voiceprint detection. Background Art

[0002] In the prior art, transformer voiceprint detection generally uses external detection. When the sensor performs voiceprint detection externally, it will be interfered by the complex and changeable surrounding environmental sound, and certain anti-interference measures need to be taken. The sound waves also need to pass through complex media such as transformer oil, transformer oil tank wall, and air. The changes in the sound signals generated by the windings and cores in the transformer may distort the sound signals, resulting in inaccurate detection of the transformer's operating status.

[0003] However, most of the existing transformer built-in sensor devices can only detect a single point or a limited number of points. Not only can the accuracy and reliability of the detection not be guaranteed, but also the location where the abnormality or fault may occur in the transformer cannot be located, which brings certain difficulties to maintenance and troubleshooting. Summary of the invention

[0004] The purpose of the present invention is to provide a method for locating abnormalities in transformer internal voiceprint detection, which is conducive to accurately locating abnormal positions in the transformer, thereby improving the reliability and safety of transformer operation.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a method for locating abnormalities in transformer internal voiceprint detection, comprising the following steps:

[0006] S1. Obtain the voiceprint signals at different positions in the transformer through distributed optical fiber acoustic sensing technology;

[0007] S2. Convert the voiceprint signal of each position point into a voiceprint frequency domain signal to obtain a sound pressure characterization vector of the corresponding position point; and construct a transformer internal sound pressure distribution feature vector by combining the sound pressure characterization vectors of all position points;

[0008] S3. Based on the characteristic vector of the sound pressure distribution inside the transformer, an objective function is constructed to reflect the characteristics of the sound pressure variation. Then, a genetic algorithm is used to find the individual corresponding to the maximum sound pressure variation. The position point corresponding to the individual is the abnormal position point, and a certain spatial range centered on the abnormal position point is the abnormal area.

[0009] Furthermore, the implementation method of step S1 is:

[0010] Multiple optical fibers are laid on the inner wall of the transformer, outside the important components of the transformer, and at possible fault locations in the transformer. They are then led out of the transformer and connected to the DAS demodulation device to form a distributed optical fiber acoustic sensing system.

[0011] The distributed optical fiber acoustic sensing system uses an optical fiber as a sensor, and transmits laser pulses to the optical fiber at intervals through a demodulation device to sense the sound vibrations at intervals along the extension direction of the optical fiber, and then periodically collects the phase information of the Rayleigh scattered light of each measuring point along the extension direction of the optical fiber, and analyzes and obtains the soundprint signals of all measuring points on the optical fiber; the soundprint signals of all measuring points on all optical fibers are combined to obtain the soundprint signals of different positions in the transformer.

[0012] Further, after the optical fiber is laid out, a coordinate system is established to locate the coordinates of each measurement position point;

[0013] After locating the coordinates of each measuring point, determine the adjacent relationship of each measuring point: for a measuring point, find all adjacent measuring points within a certain spatial range centered on the measuring point through spatial distance calculation.

[0014] Furthermore, the time length of each voiceprint signal collection is 100ms, and the interval between each collection cycle is 60s.

[0015] Furthermore, the implementation method of step S2 is:

[0016] Perform fast Fourier transform and homogenization on the voiceprint signal at the spatial position (x, y, z) to obtain the sound pressure representation vector X at the spatial position (x, y, z) x,y,z (f), which is calculated as:

[0017]

[0018] Where, X x,y,z (f) is the sound pressure characterization vector at the spatial position (x, y, z) after the homogenization process. A(f) is the amplitude spectrum at the frequency f obtained by the fast Fourier transform of the voiceprint signal. f is the signal frequency, which ranges from 50 Hz to 20 kHz with an interval of 50 Hz, that is, only the amplitude of the integer multiple frequency point of 50 Hz is taken for calculation. f j is the signal frequency, which ranges from 50 Hz to 20 kHz, with a value interval of 50 Hz;

[0019] The sound pressure distribution characteristic vector P inside the transformer is obtained by integrating the sound pressure characterization vectors of all positions. t =[X x1,y1,z1 (f),X x2,y2,z2 (f),…,X xi,yi,zi (f),…,X xn,yn,zn (f)], where X xi,yi,zi (f) is the sound pressure characterization vector of the i-th position point, and there are n position points in total; t is the acquisition time point; by setting the cyclic acquisition period, the sound pressure distribution characterization vector that changes with time can be obtained.

[0020] Furthermore, in step S3, the objective function F of the genetic algorithm is constructed as follows:

[0021] F=a*|R(x,y,z)|+b*R'(x,y,z)+c*R”(x,y,z)

[0022] Wherein, R(x,y,z) is the sound pressure characterization vector of the position point (x,y,z), |R(x,y,z)| represents the modulus value of the calculation vector R(x,y,z), R'(x,y,z) is the first-order difference of the sound pressure characterization vector of the position point (x,y,z) and its adjacent position points, and R"(x,y,z) is the second-order difference of the sound pressure characterization vector of the position point (x,y,z) and its adjacent position points; a, b, c are the set coefficients;

[0023] When the objective function F reaches its maximum value, the position point corresponding to the corresponding individual is the abnormal position point.

[0024] Furthermore, the calculation method of the first-order difference R'(x, y, z) of the sound pressure characterization vector of the position point (x, y, z) and its adjacent position points is:

[0025]

[0026] Among them, k is the number of adjacent points of the position point (x, y, z), R(x i ,y i ,z i ) is the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z);

[0027] The calculation method of the second-order difference R" (x, y, z) of the sound pressure characterization vector of the position point (x, y, z) and its adjacent position points is:

[0028]

[0029] Among them, R'(x i ,y i ,z i ) is the first-order difference of the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z).

[0030] Furthermore, in step S3, the implementation method of the genetic algorithm is:

[0031] First, initialize: initialize the evolutionary algebra counter t = 0, set the maximum evolutionary algebra T, take all measurement positions in the transformer as individuals, and randomly generate M individuals as the initial group;

[0032] Then the non-dominated sorting and crowding degree calculation are performed;

[0033] After non-dominated sorting and crowding calculation, individuals are selected, crossed, and mutated, and then the population is merged. The results of non-dominated sorting and crowding calculation are judged. If they do not meet the requirements, the iteration is continued until the maximum evolutionary generation is reached. The individual corresponding to the largest sound pressure variation is output, and the position point corresponding to this individual is the abnormal position point.

[0034] Furthermore, the implementation method of the non-dominated sorting is:

[0035] Initialization: The non-dominated level of all individuals is set to 0;

[0036] Non-dominated hierarchy partitioning: traverse the individuals in the population, for each individual i: check whether other individuals j dominate individual i; if individual j dominates individual i, the non-dominated hierarchy of individual i increases by 1; record all individuals that are not dominated by other individuals, that is, individuals with a non-dominated hierarchy of 0;

[0037] Construct a non-dominated hierarchy: remove all individuals with a non-dominated hierarchy of 0 to form a new population, and repeat the above process for the remaining individuals until all individuals are assigned a non-dominated hierarchy;

[0038] Sorting is completed: Finally, all individuals will be assigned to different non-dominated levels. The smaller the level value, the higher the position of the individual in the non-dominated sorting;

[0039] Then the congestion degree is calculated, and the implementation method is as follows:

[0040] Initialization: First, determine the solution set for which the congestion degree needs to be calculated, which is usually a solution set of one or more non-dominated levels obtained after non-dominated sorting;

[0041] Boundary value processing: For each objective function, the crowding degree of the solutions on the boundary of the solution set is set to infinity;

[0042] Normalization: In order to eliminate the impact of dimensional differences between different objectives, the value of each objective function is normalized;

[0043] Calculate the crowding degree: For each value of the objective function, sort the solution set by the value of the objective function, and then calculate the crowding degree of each solution;

[0044] Selection operation: During the selection process, solutions with lower non-dominated levels and higher congestion are preferentially selected to enter the next generation population.

[0045] Furthermore, after finding the individual corresponding to the largest sound pressure variation through the genetic algorithm, it is determined whether the objective function value corresponding to the individual exceeds the set threshold and the magnitude of the excess threshold, and the probability of the transformer failing at the corresponding position is determined, thereby determining the intervention measures to be taken.

[0046] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for locating abnormalities in transformer internal soundprint detection, which uses distributed fiber optic acoustic sensing technology to obtain soundprint signals at different positions in the transformer space, and then establishes the sound pressure distribution feature vector in the three-dimensional space inside the transformer through time-frequency transformation. On this basis, an objective function that can characterize the characteristics of sound pressure variation is constructed, and the location point with the largest sound pressure variation is found through a genetic algorithm to accurately locate the abnormal position in the transformer. According to the calculation results of this method, the operation and maintenance personnel can take corresponding intervention measures in a timely manner to avoid the occurrence or further development of abnormalities or faults, thereby greatly improving the reliability and safety of transformer operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a flow chart of a method implementation of an embodiment of the present invention;

[0048] Figure 2 is a schematic structural diagram of a distributed optical fiber acoustic sensing system in an embodiment of the present invention;

[0049] Figure 3 1 is a time domain waveform and a frequency domain waveform diagram of a voiceprint signal in an embodiment of the present invention;

[0050] Figure 4 It is a flow chart of the implementation of the genetic algorithm in the embodiment of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0054] like Figure 1 As shown, this embodiment provides a method for locating abnormalities in transformer internal voiceprint detection, including the following steps:

[0055] S1. Obtain the soundprint signals at different locations in the transformer through distributed fiber optic acoustic sensing technology.

[0056] First, multiple optical fibers are laid on the inner wall of the transformer, outside the important components of the transformer, and at possible fault locations in the transformer. Then, they are led out of the transformer and connected to the DAS demodulation device to form a distributed optical fiber acoustic sensing system. Figure 2 The optical fiber laying plan is determined according to the actual situation of the transformer, including the internal space of the transformer, important components that need to be monitored, and components or locations that are more prone to failure.

[0057] The distributed optical fiber acoustic sensing system uses optical fiber as a sensor, making full use of the characteristics of continuous spatial distribution of optical fiber to achieve the integration of "sensing" and "sensing". The demodulation device emits laser pulses to the optical fiber at intervals, which can sense the sound vibration at intervals along the extension direction of the optical fiber, and then periodically collects the phase information of Rayleigh scattered light at each measuring point along the extension direction of the optical fiber, and analyzes and obtains the soundprint signals of all measuring points on the optical fiber. In this embodiment, considering that the transformer runs for 20ms per cycle, the time length of each soundprint signal collection is set to 100ms, and the interval between each collection cycle is 60s.

[0058] By combining the soundprint signals of all measuring points on all optical fibers, we can get the soundprint signals of different positions in the transformer. According to the accuracy required for detection and positioning, the density of optical fibers and measuring points can be increased to achieve high-precision positioning.

[0059] In order to subsequently establish the characteristic vector of the sound pressure distribution inside the transformer, after the optical fiber is laid, a rectangular coordinate system needs to be established to locate the coordinates of each measurement position point.

[0060] After locating the coordinates of each measuring point, determine the adjacent relationship of each measuring point: for a measuring point, find all adjacent measuring points within a certain spatial range centered on the measuring point through spatial distance calculation. For an area where measuring points are densely arranged on a surface, normally, except for the edge measuring points, the number of adjacent measuring points of the remaining measuring points is 8. For an area where measuring points are densely arranged in a three-dimensional space, normally, except for the edge measuring points, the number of adjacent measuring points of the remaining measuring points is 26. The number of adjacent measuring points of measuring points at different positions is related to the actual spatial situation and is determined according to the actual situation. For a measuring point that is not at the end of an optical fiber, it has at least 2 adjacent measuring points.

[0061] S2. The voiceprint signal obtained in step S1 is a time domain signal. The voiceprint signal of each position point is converted into a frequency domain signal to obtain a sound pressure characterization vector of the corresponding position point; the sound pressure characterization vectors of all position points are combined to construct a transformer internal sound pressure distribution feature vector.

[0062] Perform fast Fourier transform and homogenization on the voiceprint signal at the spatial position (x, y, z) to obtain the sound pressure representation vector X at the spatial position (x, y, z) x,y,z (f), which is calculated as:

[0063]

[0064] Where, X x,y,z (f) is the sound pressure characterization vector at the spatial position (x, y, z) after the homogenization process. A(f) is the amplitude spectrum at the frequency f obtained by the fast Fourier transform of the voiceprint signal. f is the signal frequency, which ranges from 50 Hz to 20 kHz with an interval of 50 Hz, that is, only the amplitude of the integer multiple frequency point of 50 Hz is taken for calculation. f j is the signal frequency, which ranges from 50Hz to 20kHz, with a value interval of 50Hz. The time domain waveform of the acquired voiceprint signal and the converted frequency domain waveform are shown in Figure 3 shown.

[0065] The sound pressure distribution characteristic vector P inside the transformer is obtained by integrating the sound pressure characterization vectors of all positions. t =[X x1,y1,z1 (f),X x2,y2,z2 (f),…,X xi,yi,zi (f),…,X xn,yn,zn (f)], where X xi,yi,zi (f) is the sound pressure characterization vector of the i-th position point, and there are n positions in total; t is the acquisition time point. The sound pressure distribution feature vector reflects the sound pressure distribution characteristics of each frequency component at different positions in the transformer within a certain sampling period.

[0066] By setting the cyclic acquisition period, the sound pressure distribution characterization vector that changes with time can be obtained. When a component closest to the spatial position (x, y, z) inside the transformer is abnormal, the sound pressure characterization vector at the spatial position (x, y, z) will change. By constructing an objective function that can reflect the characteristics of sound pressure variation and using a certain algorithm to find the maximum point of sound pressure variation, the location of the variation can be located.

[0067] S3. Based on the characteristic vector of the sound pressure distribution inside the transformer, an objective function is constructed to reflect the characteristics of the sound pressure variation. Then, a genetic algorithm is used to find the individual corresponding to the maximum sound pressure variation. The position point corresponding to the individual is the abnormal position point, and a certain spatial range centered on the abnormal position point is the abnormal area.

[0068] Considering that the local changes in the sound pressure distribution variables reflect the information of local sound pressure anomalies to a certain extent, the objective function F constructed by this method is as follows:

[0069] F=a*|R(x,y,z)|+b*R'(x,y,z)+c*R”(x,y,z)

[0070] Among them, R(x,y,z) is the sound pressure characterization vector of the position point (x,y,z), that is, R(x,y,z)=X x,y,z (f); |R(x,y,z)| represents the modulus of the calculated vector R(x,y,z); R'(x,y,z) is the first-order difference of the sound pressure characterization vector between the position point (x,y,z) and its adjacent position points; R"(x,y,z) is the second-order difference of the sound pressure characterization vector between the position point (x,y,z) and its adjacent position points; a, b, c are the set coefficients. When the objective function F reaches the maximum value, the position point corresponding to the corresponding individual is the abnormal position point.

[0071] The calculation method of the first-order difference R'(x, y, z) of the sound pressure characterization vector between the position point (x, y, z) and its adjacent position points is:

[0072]

[0073] Among them, k is the number of adjacent points of the position point (x, y, z), R(x i ,y i ,z i ) is the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z). In this calculation, the characterization vectors are first subtracted to obtain the difference vector, and then the first-order difference value is obtained by calculating the square sum and other calculations.

[0074] The calculation method of the second-order difference R" (x, y, z) of the sound pressure characterization vector of the position point (x, y, z) and its adjacent position points is:

[0075]

[0076] Among them, R'(x i ,y i ,z i ) is the first-order difference of the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z).

[0077] like Figure 4 As shown, in this embodiment, the implementation method of the genetic algorithm is:

[0078] First, initialization is performed: the evolutionary algebra counter t=0 is initialized, the maximum evolutionary algebra T is set, all measurement positions in the transformer are taken as individuals, and M individuals are randomly generated as the initial group.

[0079] Then non-dominated sorting and congestion calculation are performed.

[0080] The implementation method of the non-dominated sorting is:

[0081] Initialization: The non-dominated levels of all individuals are set to 0.

[0082] Non-dominated hierarchy partitioning: traverse the individuals in the population, for each individual i: check whether other individuals j dominate individual i (that is, individual j is not worse than individual i in all objectives, and is better in at least one objective); if individual j dominates individual i, the non-dominated hierarchy of individual i increases by 1; record all individuals that are not dominated by other individuals, that is, individuals with a non-dominated hierarchy of 0.

[0083] Construct a non-dominated hierarchy: remove all individuals with a non-dominated hierarchy of 0 to form a new population, and repeat the above process for the remaining individuals until all individuals are assigned a non-dominated hierarchy.

[0084] Sorting is completed: In the end, all individuals will be assigned to different non-dominated levels. The smaller the level value, the higher the position of the individual in the non-dominated sorting.

[0085] Then the congestion degree is calculated. The greater the congestion degree, the more unique and important the solution is in the solution set. On the contrary, such a solution may not be particularly important in the solution set. The specific implementation method is:

[0086] Initialization: First determine the solution set for which the congestion needs to be calculated, usually a solution set of one or more non-dominated levels obtained after non-dominated sorting.

[0087] Boundary value processing: For each objective function, the congestion of the solutions on the boundary of the solution set (i.e., the solutions corresponding to the maximum and minimum values ​​of the objective) is set to infinity. This is because the solutions on the boundary are already optimal for the objective and there is no need to further consider the impact of other solutions.

[0088] Normalization: In order to eliminate the impact of dimensional differences between different objectives, it is usually necessary to normalize the value of each objective function. The normalized value range is usually 0 to 1.

[0089] Calculate the crowding: For each value of the objective function, sort the solution set by the value of the objective function and then calculate the crowding of each solution.

[0090] Selection operation: In the selection process, solutions with lower non-dominated levels and higher congestion are preferred to enter the next generation population. This can ensure the quality of the solution set and maintain the diversity of the solution set.

[0091] After non-dominated sorting and crowding calculation, individuals are selected, crossed, and mutated, and then the population is merged. The results of non-dominated sorting and crowding calculation are judged. If they do not meet the requirements, the iteration is continued until the maximum evolutionary generation is reached. The individual corresponding to the largest sound pressure variation is output, and the position point corresponding to this individual is the abnormal position point.

[0092] After finding the individual corresponding to the largest sound pressure variation through the genetic algorithm, it is determined whether the objective function value corresponding to the individual exceeds the set threshold and the magnitude of the excess threshold, and the probability of the transformer failing at the corresponding position is determined, and then the intervention measures to be taken are determined.

[0093] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0094] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0097] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for locating abnormalities in transformer internal voiceprint detection, characterized in that: The following steps are involved: S1. Obtain the voiceprint signals at different positions in the transformer through distributed optical fiber acoustic sensing technology; S2. Convert the voiceprint signal of each position point into a voiceprint frequency domain signal to obtain a sound pressure characterization vector of the corresponding position point; and construct a transformer internal sound pressure distribution feature vector by combining the sound pressure characterization vectors of all position points; S3. Based on the characteristic vector of the sound pressure distribution inside the transformer, an objective function is constructed to reflect the characteristics of the sound pressure variation. Then, a genetic algorithm is used to find the individual corresponding to the maximum sound pressure variation. The position point corresponding to the individual is the abnormal position point, and a certain spatial range centered on the abnormal position point is the abnormal area.

2. A transformer internal voiceprint detection abnormality positioning method according to claim 1, characterized in that: The implementation method of step S1 is: Multiple optical fibers are laid on the inner wall of the transformer, outside the important components of the transformer, and at possible fault locations in the transformer. They are then led out of the transformer and connected to the DAS demodulation device to form a distributed optical fiber acoustic sensing system. The distributed optical fiber acoustic sensing system uses an optical fiber as a sensor, and transmits laser pulses to the optical fiber at intervals through a demodulation device to sense the sound vibrations at intervals along the extension direction of the optical fiber, and then periodically collects the phase information of the Rayleigh scattered light of each measuring point along the extension direction of the optical fiber, and analyzes and obtains the soundprint signals of all measuring points on the optical fiber; the soundprint signals of all measuring points on all optical fibers are combined to obtain the soundprint signals of different positions in the transformer.

3. A transformer internal voiceprint detection abnormality positioning method according to claim 2, characterized in that: After the optical fiber is laid, a coordinate system is established to locate the coordinates of each measurement location point; After locating the coordinates of each measuring point, determine the adjacent relationship of each measuring point: for a measuring point, find all adjacent measuring points within a certain spatial range centered on the measuring point through spatial distance calculation.

4. The method for locating abnormalities in transformer internal voiceprint detection according to claim 2, characterized in that: The time length for each voiceprint signal collection is 100ms, and the interval between each collection cycle is 60s.

5. The method for locating abnormalities in transformer internal voiceprint detection according to claim 1, characterized in that: The implementation method of step S2 is: Perform fast Fourier transform and homogenization on the voiceprint signal at the spatial position (x, y, z) to obtain the sound pressure representation vector X at the spatial position (x, y, z) x,y,z (f), which is calculated as: Where, X x,y,z (f) is the sound pressure characterization vector at the spatial position (x, y, z) after the homogenization process. A(f) is the amplitude spectrum at the frequency f obtained by the fast Fourier transform of the voiceprint signal. f is the signal frequency, which ranges from 50 Hz to 20 kHz with an interval of 50 Hz, that is, only the amplitude of the integer multiple frequency point of 50 Hz is taken for calculation. f j is the signal frequency, which ranges from 50 Hz to 20 kHz, with a value interval of 50 Hz; The sound pressure distribution characteristic vector P inside the transformer is obtained by integrating the sound pressure characterization vectors of all positions. t =[X x1,y1,z1 (f),X x2,y2,z2 (f),…,X xi,yi,zi (f),…,X xn,yn,zn (f)], where X xi,yi,zi (f) is the sound pressure characterization vector of the i-th position point, and there are n position points in total; t is the acquisition time point; by setting the cyclic acquisition period, the sound pressure distribution characterization vector that changes with time can be obtained.

6. The method for locating abnormalities in transformer internal voiceprint detection according to claim 1, characterized in that: In step S3, the objective function F of the genetic algorithm is constructed as follows: F=a*|R(x,y,z)|+b*R'(x,y,z)+c*R”(x,y,z) Wherein, R(x,y,z) is the sound pressure characterization vector of the position point (x,y,z), |R(x,y,z)| represents the modulus value of the calculation vector R(x,y,z), R'(x,y,z) is the first-order difference of the sound pressure characterization vector of the position point (x,y,z) and its adjacent position points, and R"(x,y,z) is the second-order difference of the sound pressure characterization vector of the position point (x,y,z) and its adjacent position points; a, b, c are the set coefficients; When the objective function F reaches its maximum value, the position point corresponding to the corresponding individual is the abnormal position point.

7. A transformer internal voiceprint detection abnormality positioning method according to claim 6, characterized in that: The calculation method of the first-order difference R'(x, y, z) of the sound pressure characterization vector between the position point (x, y, z) and its adjacent position points is: Among them, k is the number of adjacent points of the position point (x, y, z), R(x i ,y i ,z i ) is the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z); The calculation method of the second-order difference R" (x, y, z) of the sound pressure characterization vector of the position point (x, y, z) and its adjacent position points is: Among them, R'(x i ,y i ,z i ) is the first-order difference of the sound pressure characterization vector of the ith adjacent position point of the position point (x, y, z).

8. The method for locating abnormalities in transformer internal voiceprint detection according to claim 1, characterized in that: In step S3, the implementation method of the genetic algorithm is: First, initialize: initialize the evolutionary algebra counter t = 0, set the maximum evolutionary algebra T, take all measurement positions in the transformer as individuals, and randomly generate M individuals as the initial group; Then the non-dominated sorting and crowding degree calculation are performed; After non-dominated sorting and crowding calculation, individuals are selected, crossed, and mutated, and then the population is merged. The results of non-dominated sorting and crowding calculation are judged. If they do not meet the requirements, the iteration is continued until the maximum evolutionary generation is reached. The individual corresponding to the largest sound pressure variation is output, and the position point corresponding to this individual is the abnormal position point.

9. A transformer internal voiceprint detection abnormality positioning method according to claim 8, characterized in that: The implementation method of the non-dominated sorting is: Initialization: The non-dominated level of all individuals is set to 0; Non-dominated hierarchy partitioning: traverse the individuals in the population, for each individual i: check whether other individuals j dominate individual i; If individual j dominates individual i, the non-dominance level of individual i increases by 1; record all individuals that are not dominated by other individuals, that is, individuals with a non-dominance level of 0; Construct a non-dominated hierarchy: remove all individuals with a non-dominated hierarchy of 0 to form a new population, and repeat the above process for the remaining individuals until all individuals are assigned a non-dominated hierarchy; Sorting is completed: Finally, all individuals will be assigned to different non-dominated levels. The smaller the level value, the higher the position of the individual in the non-dominated sorting; Then the congestion degree is calculated, and the implementation method is as follows: Initialization: First, determine the solution set for which the congestion degree needs to be calculated, which is usually a solution set of one or more non-dominated levels obtained after non-dominated sorting; Boundary value processing: For each objective function, the crowding degree of the solutions on the boundary of the solution set is set to infinity; Normalization: In order to eliminate the impact of dimensional differences between different objectives, the value of each objective function is normalized; Calculate the crowding degree: For each value of the objective function, sort the solution set by the value of the objective function, and then calculate the crowding degree of each solution; Selection operation: During the selection process, solutions with lower non-dominated levels and higher congestion are preferentially selected to enter the next generation population.

10. The method for locating abnormalities in transformer internal voiceprint detection according to claim 1, characterized in that: After finding the individual corresponding to the largest sound pressure variation through the genetic algorithm, it is determined whether the objective function value corresponding to the individual exceeds the set threshold and the magnitude of the excess threshold, and the probability of the transformer failing at the corresponding position is determined, and then the intervention measures to be taken are determined.

Images