High-voltage converter valve operation state voiceprint infrared feature extraction method

Abstract

The invention relates to the technical field of converter valve fault diagnosis, in particular to a high-voltage converter valve operation state voiceprint infrared feature extraction method. The invention discloses a voiceprint infrared feature extraction method for a high-voltage converter valve operation state. The method comprises the following steps: carrying out spectral analysis on a sound signal in the high-voltage converter valve operation state, and judging an abnormal frequency; based on the abnormal frequency, executing a spatial spectrum estimation algorithm to obtain an incoming direction of an abnormal sound source; converting an incoming angle of the abnormal sound source into a rectangular plane coordinate; and matching the abnormal sound source plane coordinates with the infrared image data to obtain spatial fusion features including spatial coordinates, voiceprint frequency and infrared data. According to the method, the voiceprint infrared fusion features of the converter valve operation state can be extracted. The features comprise abundant information such as the spatial position of the abnormal sound source, the abnormal voiceprint frequency, the infrared temperature and the like, a fault recognition model with higher accuracy can be trained, and the reliability of converter valve fault monitoring is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of converter valve fault diagnosis, and in particular to a method for extracting infrared features of soundprints of a high-voltage converter valve operating state. Background Art

[0002] Ultra-high voltage direct current transmission has the technical advantages of long distance, large capacity and low loss. In the high voltage direct current system, the converter valve is the key equipment to realize the conversion of AC and DC power, and its stable operation plays a vital role in the safety of the entire power system.

[0003] Corona discharge is a common fault defect of converter valves. It refers to the local self-sustaining discharge phenomenon that occurs on the surface of a charged body in a gas medium. It often occurs in areas with concentrated field strength in extremely non-uniform electric fields. For high-voltage electrical equipment, although short-term corona discharge will not cause serious consequences, its long-term cumulative effects cannot be ignored.

[0004] Since the converter valve is the core equipment of the converter station, its structure is complex and compact. Therefore, the power department requires that in order not to affect the safe operation of the converter valve, monitoring equipment is not allowed to be installed on the converter valve. Therefore, it is necessary to use non-contact means to monitor the corona discharge of the high-voltage converter valve.

[0005] The Chinese invention patent with application number: 202311731387.3 and name “A detection method for local discharge pattern recognition based on acoustic, optical and electrical fusion spectrum” adopts a scheme of detecting the acoustic signal, optical signal and electromagnetic wave signal generated in the local discharge process, performing NSCT decomposition and inverse transformation to obtain the acoustic, optical and electrical fusion PRPD spectrum, and finally performing local discharge pattern recognition to improve the accuracy of local discharge detection. Summary of the invention

[0006] In view of this, the present invention provides another non-contact method for detecting partial discharge of a converter valve, namely, a method for extracting infrared features of the soundprint of the operating status of a high-voltage converter valve based on an audio-visual device and infrared images. By establishing the plane rectangular coordinates of the sound source and combining the soundprint frequency and infrared temperature, it is possible to detect whether the converter valve has partial discharge, loose components, or other faults whose sound frequency exceeds the normal range.

[0007] A method for extracting infrared features of soundprints of a high-voltage converter valve operating state comprises the following steps:

[0008] Step S1, performing spectrum analysis on the sound signal of the high-voltage converter valve in operation to determine the abnormal frequency;

[0009] Step S2, executing a spatial spectrum estimation algorithm based on the abnormal frequency to obtain the direction of the abnormal sound source;

[0010] Step S3, converting the direction angle of the abnormal sound source into plane rectangular coordinates;

[0011] Step S4, matching the plane coordinates of the abnormal sound source with the infrared image data to obtain a spatial fusion feature including rectangular coordinates, voiceprint frequency, and infrared data.

[0012] Beneficial effects: The present invention can realize the extraction of the soundprint infrared fusion features of the converter valve operation status. The features contain rich information such as the spatial position of the abnormal sound source, the abnormal soundprint frequency, and the infrared temperature, which is conducive to training a more accurate fault identification model and improving the reliability of converter valve fault monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The flowchart of the method for extracting infrared features of voiceprint of high-voltage converter valve operation status is shown in FIG.

[0014] Figure 2 This is the sound signal spectrum diagram of the normal operation state of the converter valve.

[0015] Figure 3 This is a position relationship diagram of the sound source and the camera in the spatial rectangular coordinate system obtained by the two-dimensional spatial spectrum estimation algorithm in step S2; the camera is located at the origin of the spatial rectangular coordinate system, the circle is the position of the sound source, the elevation angle is θ, and the azimuth angle is φ.

[0016] Figure 4 To convert the spatial coordinates into the screen coordinates in the plane rectangular coordinate system of the shooting screen, the camera is the coordinate origin of the shooting screen, the sound source is the red dot coordinate of the shooting screen, and φ is the angle between the line connecting the sound source and the camera and the x-axis.

[0017] Figure 5 This is the picture taken by the camera of the handheld sound source device in the real scene. The red dot is the position drawn on the camera screen according to the coordinates (x, y) after adjusting the coef parameters through the plane coordinate conversion formula. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Please see Figure 1 , the method for extracting infrared features of voiceprint of high-voltage converter valve operation status includes the following steps:

[0020] Step S1, performing spectrum analysis on the sound signal of the high-voltage converter valve in operation to determine the abnormal frequency;

[0021] Step S2, executing a spatial spectrum estimation algorithm based on the abnormal frequency to obtain the direction of the abnormal sound source;

[0022] Step S3, converting the direction angle of the abnormal sound source into plane rectangular coordinates;

[0023] Step S4, matching the plane coordinates of the abnormal sound source with the infrared image data to obtain a spatial fusion feature including the spatial coordinates, voiceprint frequency, and infrared data.

[0024] The sound signal of the high-voltage converter valve in operation is collected by an audio-visual instrument. Specifically, the audio-visual instrument consists of an acoustic array and a camera. The acoustic array is composed of a plurality of microphones in formation. The plurality of microphones synchronously collect the sound signal of the converter valve. The camera is located at the center of the acoustic array to collect image information. The sound signal of the converter valve collected by the audio-visual instrument is processed by FFT (Fast Fourier Transform) to obtain a spectrum diagram. Then, the spectrum analysis of the sound signal of the converter valve in normal operation can be performed to determine the spectrum situation in the normal state.

[0025] For example, Figure 2 The sound signal spectrum of the converter valve under normal conditions collected at a converter station is Figure 1 It can be seen that under normal operating conditions, the sound frequency of the converter valve is concentrated within the range of 400Hz. Then, the real-time collected converter valve sound signal is processed by FFT and spectrum analysis is performed. If a frequency exceeding the normal range is found in the spectrum graph, it can be determined that the real-time collected converter valve sound is an abnormal sound.

[0026] In a preferred embodiment, step S1, performing spectrum analysis on the real-time collected converter valve sound signal to determine the abnormal frequency, includes the following steps:

[0027] Step S11, assuming that the frequency spectrum upper limit of the sound signal is f_max when the converter valve is in a normal state, and the real-time collected converter valve sound signal is x(n), n=1, 2, ..., N,

[0028] Where N is the number of snapshot points for signal analysis;

[0029] Step S12, perform FFT processing on x(n) to obtain a spectrum sequence S(f_i), i=1,2,...,N,

[0030] Where S(f_i) is the spectrum amplitude, and f_i is the frequency value corresponding to the i-th spectrum sequence value;

[0031] Step S13: preset a spectrum amplitude threshold threshold. If f_i>f_max and S(f_i)>threshold exists, then f_i is determined to be an abnormal frequency point.

[0032] In a preferred embodiment, step S2, based on the abnormal frequency, executes a spatial spectrum estimation algorithm to obtain the direction of the abnormal sound source, including the following steps:

[0033] Step S21, after obtaining the abnormal frequency point f_i, assuming that the sound propagation speed is c, the signal wavelength λ=c / f_i is obtained;

[0034] Step S22, based on the coordinates of each microphone array element of the sound imager and the wavelength λ of the abnormal sound source signal, an existing two-dimensional spatial spectrum estimation algorithm, such as MUSIC, ESPRIT, etc., is used to calculate the direction angle of the abnormal sound source, such as Figure 3 As shown, it includes elevation angle θ and azimuth angle φ, where,

[0035] The elevation angle θ is the angle between the line connecting the sound source and the coordinate origin and the z-axis, ranging from 0 to 90°;

[0036] The azimuth angle φ is the angle between the projection line of the line connecting the sound source and the coordinate origin on the xOy plane and the x-axis, ranging from 0 to 360°.

[0037] In a preferred embodiment, step S3, converting the direction angle of the abnormal sound source into plane rectangular coordinates, specifically as follows:

[0038] like Figure 4 As shown, based on the elevation angle θ and azimuth angle φ of the abnormal sound source, combined with the camera image in the sound imager, it is converted into plane rectangular coordinates. The camera is located at the origin of the sound array coordinate system. Figure 4 In the figure, the center of the camera screen is taken as the coordinate origin, the horizontal coordinate is the x-axis, and the vertical coordinate is the y-axis. The plane coordinate conversion formula of the abnormal sound source position is:

[0039] x=Rcosφ

[0040] y=Rsinφ

[0041] R=θθ·coef

[0042] Among them, coef is a parameter that characterizes the angle pixel distance and is obtained through the image calibration method.

[0043] Please see Figure 5 , the process of image calibration method is:

[0044] A handheld sound device is placed in front of the image captured by the camera of the audio-visual device. The sound frequency of the sound device is collected by the audio-visual device. Assuming the sound propagation speed is 340m / s, the wavelength λ of the sound device is obtained by the formula λ=c / f_i. Based on the coordinates of each microphone array element of the audio-visual device, the existing two-dimensional spatial spectrum estimation algorithm is used to obtain the elevation angle θ and azimuth angle φ of the sound device. First, a coef value is evaluated, and then it is substituted into the plane coordinate conversion formula to obtain the evaluation coordinates (x, y), and the coordinates are marked with a red dot on the camera screen; then the coef value is adjusted. When the red dot coincides with the sound device, it means that the coef value is adjusted in place and the red dot can represent the position of the sound source.

[0045] Step S4, matching the plane coordinates of the abnormal sound source with the infrared image data to obtain the spatial fusion features including the spatial coordinates, voiceprint frequency, and infrared data. Specifically,

[0046] The infrared image sensor is also located at the origin of the acoustic array coordinate system, and collects infrared image data. According to the plane coordinates (x, y) of the abnormal sound source obtained in step S3, it is matched with the infrared image data at the same coordinate position. Assuming that the temperature data value of the infrared data at the coordinate (x, y) is temp, the infrared fusion feature data of the converter valve soundprint is obtained as: [(x, y), f_i, temp, status].

[0047] Among them, (x, y) is the coordinate position of the abnormal sound source in the shooting picture; f_i is the frequency point of the abnormal sound source obtained in step S1; temp is the infrared temperature data of the abnormal sound source position; status is the operating status value of the converter valve. For example, it can be agreed that status = 0 in normal state, status = 1 in partial discharge, status = 2 when the component is loose, and so on.

[0048] The converter valve voiceprint infrared fusion features include the position information of the sound source relative to the camera and the infrared temperature information, which are then clarified by the status value. Then, they are trained according to the high-voltage converter valve operating status voiceprint infrared feature extraction method of the present invention to obtain a more accurate coef value, so as to accurately find the fault point of partial discharge or other possible faults of the converter valve, thereby improving the reliability of converter valve fault detection.

[0049] What is disclosed above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for extracting infrared features of voiceprints of high-voltage converter valve operating status, characterized in that: The following steps are involved: Step S1, performing spectrum analysis on the sound signal of the high-voltage converter valve in operation to determine the abnormal frequency; Step S2, executing a spatial spectrum estimation algorithm based on the abnormal frequency to obtain the direction of the abnormal sound source; Step S3, converting the direction angle of the abnormal sound source into plane rectangular coordinates; Step S4, matching the plane rectangular coordinates of the abnormal sound source with the infrared image data to obtain a spatial fusion feature including the rectangular coordinates, voiceprint frequency, and infrared data.

2. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 1, characterized in that: Performing spectrum analysis on the sound signal in the operating state of the high-voltage converter valve to determine the abnormal frequency includes the following steps: Step S11, assuming that the frequency spectrum upper limit of the sound signal is f_max when the converter valve is in a normal state, and the real-time collected converter valve sound signal is x(n), n=1, 2, ..., N, Where N is the number of snapshot points for signal analysis; Step S12, perform FFT processing on x(n) to obtain a spectrum sequence S(f_i), i=1,2,...,N, Where S(f_i) is the spectrum amplitude, and f_i is the frequency value corresponding to the i-th spectrum sequence value; Step S13: preset a spectrum amplitude threshold threshold. If f_i>f_max and S(f_i)>threshold exists, then f_i is determined to be an abnormal frequency point.

3. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 2, characterized in that: Based on the abnormal frequency, a spatial spectrum estimation algorithm is executed to obtain the direction of the abnormal sound source, including the following steps: Step S21, after obtaining the abnormal frequency point f_i, assuming that the sound propagation speed is c, the signal wavelength λ=c / f_i is obtained; Step S22, based on the coordinates of each microphone array element of the sound imager and the wavelength λ of the abnormal sound source signal, a two-dimensional spatial spectrum estimation algorithm is used to calculate the direction angle of the abnormal sound source, including the elevation angle θ and the azimuth angle φ, where: The elevation angle θ is the angle between the line connecting the sound source and the coordinate origin and the z-axis, ranging from 0 to 90°; The azimuth angle φ is the angle between the projection line of the line connecting the sound source and the coordinate origin on the xOy plane and the x-axis, ranging from 0 to 360°.

4. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 3, characterized in that: The direction angle of the abnormal sound source is converted into plane rectangular coordinates. Specifically, based on the obtained elevation angle θ and azimuth angle φ of the abnormal sound source, combined with the camera shooting picture in the sound imager, it is converted into plane rectangular coordinates; where the camera is located at the origin of the sound array coordinate system, the horizontal coordinate is the x-axis, and the vertical coordinate is the y-axis. The plane coordinate conversion formula of the abnormal sound source position is: x=R cosφ y=Rsinφ R=θ·coef Among them, coef is a parameter that characterizes the angle pixel distance and is obtained through the image calibration method.

5. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 4, characterized in that: The image calibration method includes the following steps: Step S31: a handheld sound device is placed in front of the image captured by the camera of the sound imager, and the sound frequency of the sound device is collected by the sound imager, and the wavelength λ of the sound device is obtained by the formula λ=c / f_i; Step S32: based on the coordinates of each microphone array element of the sound imager, using the existing two-dimensional spatial spectrum estimation algorithm, obtain the elevation angle θ and azimuth angle φ of the sound-emitting device; Step S33: Evaluate a coef value, then substitute it into the plane coordinate conversion formula to obtain the evaluation coordinates (x, y), and mark the coordinates with red dots on the camera screen; Step S34: adjust the coef value until the red dot coincides with the sound-emitting device. The final adjusted coef value is the coef value of the plane coordinate conversion formula.

6. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 4, characterized in that: Match the plane coordinates of the abnormal sound source with the infrared image data to obtain the spatial fusion features including rectangular coordinates, voiceprint frequency, and infrared data, including the following steps: Step S41: setting the position of the infrared image sensor as the origin of the acoustic array coordinate system, and collecting infrared image data; Step S42: According to the rectangular coordinates (x, y) of the abnormal sound source plane obtained in step S3, the infrared image data at the same coordinate position is matched to obtain the spatial fusion feature data including the rectangular coordinates, voiceprint frequency, and infrared data: [(x, y), f_i, temp, status]; Among them, (x, y) is the coordinate position of the abnormal sound source in the shooting picture; f_i is the frequency point of the abnormal sound source obtained in step S1; temp is the infrared temperature data of the abnormal sound source position; status is the operating status value of the converter valve.

7. The method for extracting infrared features of voiceprint of high-voltage converter valve operation status according to claim 1, characterized in that: In step S1, the sound signal of the high-voltage converter valve in the operating state is collected by an audio-visual instrument, which consists of an audio array and a camera. The audio array consists of a plurality of microphones in an array. The plurality of microphones synchronously collect the sound signal of the converter valve. The camera is located at the center of the audio array to collect image information.

8. The method for extracting infrared features of voiceprints of high-voltage converter valve operating status according to claim 7, characterized in that: The sound signal of the converter valve collected by the audio-visual instrument is subjected to fast Fourier transform to obtain a spectrum diagram. When the sound frequency of the converter valve exceeds the normal range, it is determined to be an abnormal frequency.

Images