A method and system for locating multiple partial discharge sources in a large-scale substation

Abstract

The present invention relates to the field of partial discharge source localization, and specifically to a method and system for localizing multiple partial discharge sources in a large substation. The method is as follows: A sensor array is composed of a movable ultrasonic sensor and a UHF sensor to collect ultrasonic signals and UHF signals generated by partial discharge sources respectively; the collected discharge signals are denoised, and the distances between the partial discharge sources and the sensor array are calculated; the denoised ultrasonic array signals are divided into multiple narrowband signals by using the discrete Fourier transform, the weighted covariance matrix after focusing transformation of all narrowband signals is averaged to obtain the weighted covariance matrix at the focused frequency points, the weighted covariance matrix at the focused frequency points is subjected to unitary transformation, and the number of partial discharge sources is estimated according to the signal source number estimation equation; spectral peak search is performed at a preset step size within the specified spatial parameter range to obtain the spatial parameters corresponding to each partial discharge source, and the coordinates of the partial discharge sources in three-dimensional space are obtained.

Description

Technical Field

[0001] The present invention relates to the field of partial discharge source location, and specifically to a method and system for locating multiple partial discharge sources in a large substation. Background Art

[0002] Partial discharge refers to an electrical discharge phenomenon that occurs in insulating materials due to local defects or non-uniformities. It usually occurs in high-voltage equipment and is an early indication of the aging and failure of the insulation system. Its monitoring and analysis are crucial for ensuring the safety and reliability of equipment.

[0003] Currently, the commonly used partial discharge location methods mainly include the electrical location method and the time difference of arrival method. The electrical location method measures the pulse current generated by partial discharge and analyzes it to obtain the magnitude and location of the discharge. This method has a simple principle and is easy to implement. However, due to the large electromagnetic interference often present at the discharge site, the measured current information is easily affected, reducing the accuracy of the location result. The time difference of arrival method locates by measuring the time difference between the arrival of the discharge signal at different sensors. By substituting the obtained time difference information into the three-dimensional space location model of partial discharge, the location of the partial discharge source can be inversely solved. This method has high accuracy for locating a single discharge source.

[0004] However, in a complex large substation, the number of partial discharge sources is usually not one. As a result, the pulse current measured by the electrical location method cannot be accurately distinguished, and the traditional time difference of arrival method cannot effectively identify the signals of each discharge source. Especially in the case where the signal propagation paths are similar, the measurement error of the time difference will significantly affect the location result. Therefore, there is an urgent need for an effective method to locate multiple discharge sources. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for locating multiple partial discharge sources in a large substation, which can efficiently judge the signals of each partial discharge source. Especially in the case where the signal propagation paths are similar, it can also accurately identify the number of partial discharge sources. Combining the azimuth and elevation angle information obtained by spatial spectrum scanning, the location result can be three-dimensionally coordinated, making the location result more intuitive and clear.

[0006] The method and system for locating multiple partial discharge sources in a large substation provided by this application are implemented as follows:

[0007] A method for locating multiple partial discharge sources in a large substation specifically includes:

[0008] S1. A sensor array is composed of a movable ultrasonic sensor and a UHF sensor to respectively collect the ultrasonic signal and the UHF signal generated by the partial discharge source, and the ultrasonic signal and the UHF signal are collectively referred to as the discharge signal;

[0009] The distance between the sensor array and the electrical equipment in the substation is less than 10 meters. An 8 ultrasonic sensors and 1 UHF sensor are fixed by brackets to form a sensor array. The sensor array is used to collect discharge signals from the partial discharge source. Among them, the UHF sensor is located at the center of the sensor array. Two ultrasonic sensors are grouped together and distributed around the UHF sensor, forming a cross-shaped sensor array with the UHF sensor. The array composed of 8 ultrasonic sensors alone is called the ultrasonic array.

[0010] The UHF sensor antenna in the sensor array uses a half-wave dipole antenna with a length of 150 mm, and the detection frequency band is between 300 MHz and 1500 MHz. The ultrasonic sensor antenna uses a spherical antenna with a diameter of 10 mm, and the detection frequency band is ultrasonic signals of 20 KHz - 200 KHz. The UHF sensor and the ultrasonic sensor are electrically connected to an oscilloscope through a feeder line.

[0011] The spacing between each array element increases linearly as the distance between the sensor array and the high-voltage electrical equipment increases. The specific calculation formula is:

[0012] ;

[0013] Among them, represents the spacing between the array elements of the sensor array, D represents the distance between the sensor array and the high-voltage electrical equipment, and the distance D satisfies , with the unit of meter.

[0014] S2. Perform wavelet decomposition on the collected discharge signals, use a dynamic threshold function to quantize the wavelet coefficients obtained by wavelet decomposition, and reconstruct the quantized wavelet coefficients through inverse wavelet transform to obtain the denoised discharge signals;

[0015] According to the time difference between the arrival of the denoised ultrasonic signal at the reference ultrasonic sensor and the arrival of the denoised UHF signal at the UHF sensor, calculate the distance between the partial discharge source and the sensor array.

[0016] The reference ultrasonic sensor in S2 is specifically;

[0017] During the process of collecting discharge signals, the ultrasonic sensor that first receives the discharge signal in the sensor array is the reference ultrasonic sensor.

[0018] S3. Use the discrete Fourier transform method to divide the denoised ultrasonic array signal into multiple narrowband signals, wherein the narrowband signal at the center frequency of the denoised ultrasonic array signal is the central narrow frequency band, calculate the focusing matrix, signal model and covariance matrix of each narrowband signal, perform focusing transformation, assign different inertia weights to the covariance matrix obtained after the focusing transformation of each narrowband signal, and obtain the weighted covariance matrix of each narrowband signal after the focusing transformation, and perform mean calculation on the weighted covariance matrices after the focusing transformation of all narrowband signals to obtain the weighted covariance matrix of the focusing frequency point.

[0019] In S3, different inertia weights are assigned to the covariance matrix obtained after focusing transformation of each narrowband signal. The specific operation is as follows:

[0020] Assume that the center frequency of the denoised ultrasonic array signal is F0, define the narrowband signal where F0 is located as the central narrow frequency band, set the position of the central narrow frequency band to m, that is, the mth narrowband signal, and assign different inertia weights to the covariance matrix obtained after focusing transformation of each narrowband signal. The specific expression is:

[0021] ;

[0022] in, is the inertia weight, j is where the narrowband signal is located, represents the lower limit of weight, J This is the location of the last narrowband signal.

[0023] In S3, the focus transformation is used to calculate the The weighted covariance matrix corresponding to the narrowband signal after focusing transformation is The specific method is:

[0024] ,

[0025] in, is the inertia weight, Indicates the frequency The covariance matrix at , represents the focusing matrix and T represents the conjugate transpose.

[0026] S4. Perform a unitary transformation on the weighted covariance matrix of the focused frequency points to obtain a unitary matrix. Based on the Gale circle criterion, perform a similarity transformation on the unitary matrix using the center and radius of the Gale circle in the unitary matrix, and estimate the number of local discharge sources according to the signal source number estimation equation.

[0027] In S4, the weighted covariance matrix of the focused frequency point is transformed into a unitary transformation, and the transformation matrix for:

[0028] ;

[0029] Among them, is a unitary matrix composed of the eigenvectors of is a sub-matrix obtained by removing the last row and the last column from the weighted covariance matrix at the focused frequency points. N is the dimension of the weighted covariance matrix at the obtained focused frequency points.

[0030] S5. Perform eigenvalue decomposition on the weighted covariance matrix at the focused frequency points, construct a spatial spectrum function using the multiple signal classification algorithm, and perform spectrum peak search at a preset step within the specified spatial parameter range to obtain the spatial parameters corresponding to each local discharge source.

[0031] The specified spatial parameters in S5 include azimuth angle and elevation angle.

[0032] The specified spatial parameter range in S5 is elevation angle 0 - 90°, azimuth angle 0 - 180°.

[0033] S6. Perform coordinate transformation based on the distance between the local discharge source and the sensor array and the spatial parameters of the local discharge source to obtain the coordinates of the local discharge source in three-dimensional space.

[0034] The present invention also provides a multi-local discharge source positioning system for a large substation, including:

[0035] A sensor array; a sensor array is composed of a movable ultrasonic sensor and a UHF sensor, and is used to collect ultrasonic signals and UHF signals generated by local discharge sources respectively, and the ultrasonic signals and UHF signals are collectively referred to as discharge signals;

[0036] A discharge signal denoising module; perform wavelet decomposition on the collected discharge signals, perform quantization processing on the wavelet coefficients obtained by wavelet decomposition using a dynamic threshold function, and reconstruct the quantized wavelet coefficients through inverse wavelet transform to obtain the denoised discharge signals;

[0037] A local discharge source distance obtaining module; calculate the distance between the local discharge source and the sensor array according to the time difference between the denoised ultrasonic signal reaching the reference ultrasonic sensor and the denoised UHF signal reaching the UHF sensor;

[0038] Wideband signal focusing module; using the discrete Fourier transform method to divide the denoised ultrasonic array signal into multiple narrowband signals. Among them, the narrowband signal where the center frequency of the denoised ultrasonic array signal is located is the central narrow frequency band. Calculate the focusing matrix, signal model, and covariance matrix of each narrowband signal, perform focusing transformation, assign different inertia weights to the covariance matrix obtained after the focusing transformation of each narrowband signal, obtain the weighted covariance matrix of each narrowband signal after the focusing transformation, and calculate the weighted covariance matrix of the focusing frequency point by taking the mean of the weighted covariance matrices of all narrowband signals after the focusing transformation;

[0039] Unitary transformation module; perform unitary transformation on the weighted covariance matrix of the focusing frequency point to obtain a unitary matrix. Based on the Gerschgorin circle criterion, use the center and radius of the Gerschgorin circles in the unitary matrix to perform similarity transformation on the unitary matrix, and estimate the number of partial discharge sources according to the signal source number estimation equation;

[0040] Spatial parameter acquisition module; perform eigenvalue decomposition on the weighted covariance matrix of the focusing frequency point, construct a spatial spectrum function using the multiple signal classification algorithm, and perform spectrum peak search at a preset step size within the specified spatial parameter range to obtain the spatial parameters corresponding to each partial discharge source;

[0041] Coordinate acquisition module; perform coordinate transformation based on the distance between the partial discharge source and the sensor array and the spatial parameters of the partial discharge source to obtain the coordinates of the partial discharge source in three-dimensional space.

[0042] A method for locating multiple partial discharge sources in a large substation provided by the present invention has the beneficial effects as follows:

[0043] In a complex environment, by using multiple ultrasonic sensors, signals can be collected from more angles and directions, increasing the comprehensiveness of signal capture; the combination of wavelet decomposition and dynamic threshold function can improve the signal-to-noise ratio of the signal in the case of high noise, ensuring that the positioning signal is clearer; through focusing transformation, the array manifold matrix at different frequencies is transformed to the same specific frequency, enabling narrowband signal processing methods to also process broadband signals accordingly; by performing weight assignment through inertia weights, the covariance matrix obtained after focusing the narrowband signals near the center frequency point can obtain a higher weight, ensuring that these covariance matrices play a dominant role in the subsequent analysis results and better reflecting the true characteristics of the signal; using the Gerschgorin circle criterion to reasonably estimate the number of signal sources, thereby realizing the judgment of the number of signal sources; by jointly using ultra-high frequency sensors and ultrasonic sensors, the distance between the partial discharge source and the sensor array can be calculated; combined with the azimuth and elevation angle information obtained by spatial spectrum scanning, the positioning result can be three-dimensionally coordinated, making the positioning result more intuitive and clear. Description of the Drawings

[0044] By reading the following detailed description of the preferred embodiments, the solutions and advantages of the present application will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0045] In the drawings:

[0046] Figure 1 It is a schematic diagram of the position of the sensor array in the embodiment;

[0047] Figure 2 It is a time difference estimation diagram of the UHF signal and the ultrasonic signal;

[0048] Figure 3 It is an ultrasonic signal image before denoising;

[0049] Figure 4 It is an ultrasonic signal image after denoising;

[0050] Figure 5 It is a spatial spectrum scanning result diagram. Specific Embodiments

[0051] Example 1

[0052] The purpose of the present application is to provide a method and system for locating multiple partial discharge sources in a large substation, which can efficiently judge the signals of each partial discharge source. Especially in the case where the signal propagation paths are similar, it can also accurately identify the number of partial discharge sources. Combining the azimuth and elevation angle information obtained by spatial spectrum scanning, the positioning result can be three-dimensionally coordinated, making the positioning result more intuitive and clear.

[0053] The method and system for locating multiple partial discharge sources in a large substation provided by the present application are implemented as follows:

[0054] A method for locating multiple partial discharge sources in a large substation specifically includes:

[0055] S1. A sensor array is composed of a movable ultrasonic sensor and a UHF sensor to collect the ultrasonic signal and the UHF signal generated by the partial discharge source respectively. The ultrasonic signal and the UHF signal are collectively referred to as the discharge signal.

[0056] The distance between the sensor array and the electrical equipment in the substation is less than 10 meters. 8 ultrasonic sensors and 1 UHF sensor are fixed by brackets to form a sensor array. The discharge signal of the partial discharge source is collected through the sensor array. Among them, the UHF sensor is located at the center of the sensor array, and two ultrasonic sensors are in a group, distributed around the UHF sensor, forming a cross-shaped sensor array with the UHF sensor as shown in Figure 1 . The array composed of 8 ultrasonic sensors alone is called the ultrasonic array.

[0057] In the sensor array, the UHF sensor antenna selects a half-wave dipole antenna with a length of 150 mm, and the detection frequency band is between 300 MHz and 1500 MHz. The ultrasonic sensor antenna selects a spherical antenna with a diameter of 10 mm, and the detection frequency band is ultrasonic signals of 20 KHz - 200 KHz; the UHF sensor and the ultrasonic sensor are electrically connected to an oscilloscope through a feeder.

[0058] The spacing between each array element increases linearly as the distance between the sensor array and the high-voltage electrical equipment increases. The specific calculation formula is:

[0059] ;

[0060] where, represents the spacing between the array elements of the sensor array, D represents the distance between the sensor array and the high-voltage electrical equipment, and the distance D satisfies , with the unit of meter.

[0061] Due to the range limitation of the sensor bandwidth, the signals collected here are all broadband signals.

[0062] S2. Perform wavelet decomposition on the collected discharge signals, use a dynamic threshold function to quantize the wavelet coefficients obtained from the wavelet decomposition, and reconstruct the quantized wavelet coefficients through inverse wavelet transform to obtain the denoised discharge signals. The ultrasonic signal before denoising is shown in Figure 3 , and the ultrasonic signal after denoising is shown in Figure 4 ;

[0063] According to the time difference between the arrival of the denoised ultrasonic signal at the reference ultrasonic sensor and the arrival of the denoised UHF signal at the UHF sensor, refer to Figure 2 , and calculate the distance between the partial discharge source and the sensor array.

[0064] It should be noted that when calculating the distance in S2, there is a situation where there are partial discharge sources with the same distance from the sensor array but different positions and discharging simultaneously. When this situation occurs, the specific number of partial discharge sources cannot be accurately known.

[0065] The reference ultrasonic sensor in S2 is specifically;

[0066] During the process of collecting discharge signals, the ultrasonic sensor in the sensor array that first receives the discharge signal is the reference ultrasonic sensor.

[0067] Since the propagation speed of UHF signals is extremely fast, usually propagating at the speed of light, it is considered that the moment when the UHF sensor receives the signal is the moment when partial discharge occurs. The propagation speed of ultrasonic signals is relatively slow. Taking the first ultrasonic sensor that receives the discharge signal as a reference, by detecting the time difference between the signal reaching the UHF sensor and the reference ultrasonic sensor, the distance between the discharge source and the sensor array can be calculated.

[0068] The collected discharge signals are decomposed using wavelet decomposition technology, and the decomposition formula is:

[0069] ,

[0070] where, represents the u th v wavelet coefficient on the scale obtained by decomposition, is the discharge signal, represents complex conjugate, t is the signal time variable, representing the position of the signal in the time domain, and <·> represents the inner product operation.

[0071] In this embodiment, the "dB8" wavelet basis is selected as the wavelet basis function, and the decomposition level is selected as three layers.

[0072] The approximate coefficients obtained by wavelet decomposition of each layer represent the low-frequency components of the discharge signal, and these low-frequency components usually contain the main information of the discharge signal.

[0073] The detail coefficients obtained by wavelet decomposition of each layer represent the high-frequency components of the discharge signal, and these high-frequency components usually contain the noise part of the discharge signal.

[0074] The wavelet coefficients obtained by decomposition are processed using a dynamic threshold function.

[0075] The expression of the dynamic threshold function is:

[0076] ,

[0077] In the formula, is the adjustment factor, is the fixed threshold, , is the variance of the noise, Q is the length of the discharge signal, is the processed wavelet coefficient.

[0078] Wavelet inverse transform is performed on the wavelet coefficients processed by the dynamic threshold function, and the inverse transform formula is:

[0079] ;

[0080] Among them, Z represents an integer, and the denoised discharge signal is reconstructed through inverse wavelet transform .

[0081] Let the time difference between the denoised ultrasonic signal reaching the reference ultrasonic sensor and the denoised UHF signal reaching the UHF sensor be , and calculate the distance between the discharge source and the sensor array d , and the specific calculation formula is as follows:

[0082] ;

[0083] Among them, c is the propagation speed of ultrasonic waves in the air, and the value is .

[0084] S3. Use the discrete Fourier transform method to divide the denoised ultrasonic array signal into multiple narrowband signals. Among them, the narrowband signal where the center frequency of the denoised ultrasonic array signal is located is the central narrow frequency band. Calculate the focusing matrix, signal model, and covariance matrix of each narrowband signal, perform focusing transformation, assign different inertia weights to the covariance matrix obtained after focusing transformation of each narrowband signal, obtain the weighted covariance matrix after focusing transformation of each narrowband signal, and calculate the weighted covariance matrix of the focusing frequency point by taking the mean of the weighted covariance matrices after focusing transformation of all narrowband signals.

[0085] Perform discrete Fourier transform on the denoised ultrasonic array signal at points, divide the broadband ultrasonic array signal into narrowband signals, and the model of each narrowband signal can be written as:

[0086] ,

[0087] Among them, , , respectively represent the discrete Fourier transforms of the narrowband signal, signal source, and noise at f j frequency, represents f j the array manifold matrix at frequency, f j is the frequency of the j th narrowband signal, .

[0088] The expressions of the focusing matrices of each narrowband signal are as follows:

[0089] ,

[0090] Among them, represents the focusing matrix, and are respectively the matrix composed of the left singular value vector and the right singular value vector of represents the array manifold matrix of the focusing frequency points, is the focusing frequency point, and T represents the conjugate transpose operation.

[0091] The specific operation of assigning different inertia weights to the covariance matrix after the focusing transformation of each narrowband signal in S3 is as follows;

[0092] Let the center frequency of the denoised ultrasonic array signal be F 0 , and F 0 the narrowband signal where it is located is defined as the central narrow frequency band, and the position where the central narrow frequency band is located is set as m , that is, the m th narrowband signal. Different inertia weights are assigned to the covariance matrix after the focusing transformation of each narrowband signal. The specific expression is:

[0093] ,

[0094] Among them, is the inertia weight, j is the position where the narrowband signal is located, represents the weight lower limit, J is the position where the last narrowband signal is located.

[0095] By performing weight assignment through the inertia weight, the covariance matrix obtained after focusing the narrowband signals near the center frequency point can obtain a higher weight, ensuring that these covariance matrices play a major role in the subsequent analysis results and better reflecting the true characteristics of the signals.

[0096] Through the focusing transformation, the weighted covariance matrix corresponding to the th narrowband signal after the focusing transformation is

[0097] ,

[0098] Among them, is the inertia weight, represents the covariance matrix at the frequency , represents the focusing matrix, and T represents the conjugate transpose.

[0099] The weighted covariance matrix of the final focusing frequency point Yes The mean of the weighted covariance matrix of narrowband signals:

[0100] .

[0101] S4. Perform a unitary transformation on the weighted covariance matrix at the focused frequency point to obtain a unitary matrix. Based on the Gerschgorin circle criterion, use the center and radius of the Gerschgorin circles in the unitary matrix to perform a similarity transformation on the unitary matrix, and estimate the number of partial discharge sources according to the signal source number estimation equation.

[0102] In S4, perform a unitary transformation on the weighted covariance matrix at the focused frequency point, and the transformation matrix is:

[0103] ;

[0104] where is a unitary matrix composed of the eigenvectors of is the sub-matrix obtained by removing the last row and the last column from the weighted covariance matrix at the focused frequency point, and N is the dimension of the obtained weighted covariance matrix at the focused frequency point.

[0105] After the unitary transformation, we can get:

[0106] ;

[0107] is the unitary transformation matrix. According to the Gerschgorin circle criterion, , respectively represent the center and radius of the Gerschgorin circle, and are constants, i represents the i th Gerschgorin circle.

[0108] Construct a diagonal matrix P :

[0109] ,

[0110] where , obviously , that is .

[0111] Perform a similarity transformation on , and after the transformation, we can get:

[0112] ,

[0113] where is The matrix after similarity transformation. Through similarity transformation, on the premise that the centers of the Gerschgorin circles remain unchanged, the radii of the Gerschgorin circles are compressed according to their respective centers. Generally, the smaller the radius, the greater the compression ratio. Thus, the radii of the noise Gerschgorin circles can be made far away from those of the signal Gerschgorin circles. Combining with the minimum description length criterion, the estimated equation for the number of signal sources is as follows:

[0114] ,

[0115] where, is the new Gerschgorin circle radius obtained, N is the dimension of Ry, k is an integer taking values from small to large, and when k makes the above formula the minimum value, k is the number of partial discharge sources.

[0116] The number of partial discharge sources k includes the cases of partial discharge sources that are at equal distances from the sensor array but have different positions and discharge simultaneously.

[0117] S5. Perform eigen - decomposition on the weighted covariance matrix at the focused frequency points, construct the spatial spectrum function using the multiple signal classification algorithm, and search for spectral peaks with a preset step size within the specified spatial parameter range to obtain the spatial parameters corresponding to each partial discharge source.

[0118] Perform eigen - decomposition on the weighted covariance matrix at the focused frequency points, then we have:

[0119] ,

[0120] After eigen - decomposition, the signal subspace is denoted as , and the noise subspace is denoted as , represents the diagonal matrix composed of the eigenvalues corresponding to the eigen - vectors constituting the signal subspace, represents the diagonal matrix composed of the eigenvalues corresponding to the eigen - vectors constituting the noise subspace.

[0121] The expression of the constructed spatial spectrum function is:

[0122] ;

[0123] G represents the spatial spectrum function, is the column vector of the array manifold matrix (i.e., the steering vector). By spatial scanning, the maximum value is found, that is, the spectral peak position. The position where the maximum value appears reflects the azimuth angle and the elevation angle information of the partial discharge source.

[0124] Based on the number of local discharge sources estimated in S4 and the maximum points obtained from the spatial spectrum scanning, search for spectral peaks to achieve the positioning of multiple local discharge sources. Specifically, taking the maximum points as the reference, search for spectral peaks downward, and obtain k spectral peaks.

[0125] The specified spatial parameters in S5 include azimuth angle and elevation angle.

[0126] The range of the specified spatial parameters in S5 is elevation angle 0 - 90°, azimuth angle 0 - 180°. The spectral peak search results are as Figure 5 shown.

[0127] S6. Perform coordinate transformation based on the distance between the local discharge source and the sensor array and the spatial parameters of the local discharge source to obtain the coordinates of the local discharge source in three-dimensional space.

[0128] When performing three-dimensional coordinate transformation by combining the distance between the local discharge source and the sensor array obtained in S2 with the specified spatial parameters obtained in S5, the specific transformation formula is:

[0129] ,

[0130] ,

[0131] ,

[0132] where x represents the abscissa of the local discharge source, y represents the ordinate of the local discharge source, and z represents the height of the local discharge source.

[0133] By converting the spatial parameters into three-dimensional coordinates, the positioning result becomes more intuitive and clear.

[0134] The present invention also provides a multi-local discharge source positioning system for a large substation, including:

[0135] A sensor array; The sensor array is composed of a movable ultrasonic sensor and a UHF sensor, and respectively collects the ultrasonic signals and UHF signals generated by the local discharge source, and collectively refers to the ultrasonic signals and UHF signals as discharge signals;

[0136] A discharge signal denoising module; Perform wavelet decomposition on the collected discharge signals, use a dynamic threshold function to quantize the wavelet coefficients obtained from the wavelet decomposition, and reconstruct the quantized wavelet coefficients through inverse wavelet transform to obtain the denoised discharge signals;

[0137] A local discharge source distance obtaining module; Calculate the distance between the local discharge source and the sensor array according to the time difference between the arrival of the denoised ultrasonic signal at the reference ultrasonic sensor and the arrival of the denoised UHF signal at the UHF sensor;

[0138] Wideband signal focusing module; using the discrete Fourier transform method to divide the denoised ultrasonic array signal into multiple narrowband signals, where the narrowband signal with the center frequency of the denoised ultrasonic array signal is the central narrow frequency band, calculating the focusing matrix, signal model and covariance matrix of each narrowband signal, performing focusing transformation, assigning different inertial weights to the covariance matrix obtained after the focusing transformation of each narrowband signal to obtain the weighted covariance matrix of each narrowband signal after the focusing transformation, and calculating the weighted covariance matrix of the focusing frequency point by taking the mean of the weighted covariance matrices of all narrowband signals after the focusing transformation;

[0139] Unitary transformation module; performing unitary transformation on the weighted covariance matrix of the focusing frequency point to obtain a unitary matrix, based on the Gerschgorin circle criterion, using the center and radius of the Gerschgorin circles in the unitary matrix to perform similarity transformation on the unitary matrix, and estimating the number of partial discharge sources according to the signal source number estimation equation;

[0140] Spatial parameter acquisition module; performing eigenvalue decomposition on the weighted covariance matrix of the focusing frequency point, constructing a spatial spectrum function using the multiple signal classification algorithm, and performing spectrum peak search within a specified spatial parameter range with a preset step size to obtain the spatial parameters corresponding to each partial discharge source;

[0141] Coordinate acquisition module; performing coordinate transformation based on the distance between the partial discharge source and the sensor array and the spatial parameters of the partial discharge source to obtain the coordinates of the partial discharge source in three-dimensional space.

Claims

1. A method for locating multiple partial discharge sources in a large-scale substation, characterized in that, The specific steps are as follows: S1. A sensor array is composed of a movable ultrasonic sensor and a UHF sensor to collect ultrasonic signals and UHF signals generated by a partial discharge source respectively. The ultrasonic signals and UHF signals are collectively referred to as discharge signals; S2. The collected discharge signals are subjected to wavelet decomposition. The wavelet coefficients obtained by wavelet decomposition are quantized using a dynamic threshold function, and the quantized wavelet coefficients are reconstructed through inverse wavelet transform to obtain the denoised discharge signals; Based on the time difference between the denoised ultrasonic signal reaching the reference ultrasonic sensor and the denoised UHF signal reaching the UHF sensor, the distance between the partial discharge source and the sensor array is calculated; S3. The denoised ultrasonic array signals are divided into multiple narrowband signals using the discrete Fourier transform method, and the weighted covariance matrix of the focused frequency points is calculated based on the narrowband signals; S4. The weighted covariance matrix of the focused frequency points is subjected to unitary transformation to obtain a unitary matrix. Based on the Gerschgorin circle criterion, the unitary matrix is subjected to similarity transformation using the center and radius of the Gerschgorin circles in the unitary matrix, and the number of partial discharge sources is estimated according to the signal source number estimation equation; S5. The weighted covariance matrix of the focused frequency points is subjected to eigenvalue decomposition, and the multiple signal classification algorithm is used to construct a spatial spectrum function, and spectral peak search is performed at a preset step size within the specified spatial parameter range to obtain the spatial parameters corresponding to each partial discharge source; S6. Coordinate transformation is performed based on the distance between the partial discharge source and the sensor array and the spatial parameters of the partial discharge source to obtain the coordinates of the partial discharge source in three-dimensional space.

2. A method for locating multiple partial discharge sources in a large-scale substation according to claim 1, characterized in that, The specific operation of assigning different inertia weights to the covariance matrices obtained by focusing and transforming each narrowband signal in S3 is as follows; Let the center frequency of the denoised ultrasonic array signal be F 0 , and define the narrowband signal where F 0 is located as the central narrow frequency band, and set the position where the central narrow frequency band is located as m , that is, the m th narrowband signal. Assign different inertia weights to the covariance matrix after focusing transformation for each narrowband signal. The specific expression is as follows: ； Among them, is the inertia weight, j is the position where the narrowband signal is located, represents the lower limit of the weight, J is the position where the last narrowband signal is located.

3. A method for locating multiple partial discharge sources in a large-scale substation according to claim 1, characterized in that, In S4, a unitary transformation is performed on the weighted covariance matrix of the focused frequency points, and the transformation matrix is as follows: ； Among them, is a unitary matrix composed of eigenvectors, is a submatrix obtained by removing the last row and the last column from the weighted covariance matrix of the focusing frequency points. N is the dimension of the obtained weighted covariance matrix of the focusing frequency points.

4. A method for locating multiple partial discharge sources in a large substation according to claim 1, characterized in that, In step S3, through focusing transformation, calculate the weighted covariance matrix corresponding to the th narrowband signal after focusing transformation The specific method is as follows: ， Among them, is the inertia weight, represents the covariance matrix at the frequency and represents the focusing matrix, and T represents the conjugate transpose.

5. A method for locating multiple partial discharge sources in a large substation according to claim 1, characterized in that, The distance between the sensor array and the electrical equipment in the substation is less than 10 meters. An 8 ultrasonic sensors and 1 UHF sensor are fixed by brackets to form a sensor array. The sensor array is used to collect discharge signals from the partial discharge source. Among them, the UHF sensor is located at the center of the sensor array. Two ultrasonic sensors are grouped together and distributed around the UHF sensor, forming a cross-shaped sensor array with the UHF sensor. The array composed of 8 ultrasonic sensors alone is called the ultrasonic array.

6. A method for locating multiple partial discharge sources in a large substation according to claim 5, characterized in that, The antenna of the UHF sensor in the sensor array uses a half-wave dipole antenna with a length of 150mm, and the detection frequency band is between 300MHz and 1500MHz. The antenna of the ultrasonic sensor uses a spherical antenna with a diameter of 10mm, and the detected ultrasonic signal has a frequency band of 20KHz - 200KHz; The UHF sensor and the ultrasonic sensor are electrically connected to an oscilloscope through a feeder line; ​ ； Among them, represents the spacing between the elements of the sensor array, and D represents the distance between the sensor array and the high-voltage electrical equipment. The distance D satisfies , with the unit of meter.

7. A method for locating multiple partial discharge sources in a large-scale substation according to claim 1, characterized in that, ​ ​ 8. A method for locating multiple partial discharge sources in a large-scale substation according to claim 1, characterized in that, The denoised ultrasonic array signals are divided into multiple narrowband signals by using the discrete Fourier transform method. The specific operation of calculating the weighted covariance matrix of the focusing frequency points based on the narrowband signals is as follows: The narrowband signal where the center frequency of the denoised ultrasonic array signals is located is the central narrow frequency band. Calculate the focusing matrix, signal model, and covariance matrix of each narrowband signal, perform the focusing transformation, assign different inertia weights to the covariance matrix obtained after the focusing transformation of each narrowband signal to obtain the weighted covariance matrix of each narrowband signal after the focusing transformation, and calculate the mean value of the weighted covariance matrices of all narrowband signals after the focusing transformation to obtain the weighted covariance matrix of the focusing frequency points.

9. A method for locating multiple partial discharge sources in a large-scale substation according to claim 8, characterized in that, The specified spatial parameter range described in S5 is the elevation angle of 0 - 90° and the azimuth angle of 0 - 180°.

10. A multi-local partial discharge source positioning system for a large substation, characterized in that, Including: Sensor array; The sensor array is composed of a movable ultrasonic sensor and a very high frequency sensor to collect the ultrasonic signals and very high frequency signals generated by the partial discharge source respectively, and the ultrasonic signals and very high frequency signals are collectively referred to as discharge signals; Discharge signal denoising module; The collected discharge signals are subjected to wavelet decomposition, the wavelet coefficients obtained by the wavelet decomposition are quantized by using the dynamic threshold function, and the wavelet coefficients after the quantization process are reconstructed by the inverse wavelet transform to obtain the denoised discharge signals; Local discharge source distance obtaining module; According to the time difference between the denoised ultrasonic signal reaching the reference ultrasonic sensor and the denoised very high frequency signal reaching the very high frequency sensor, calculate the distance between the local discharge source and the sensor array; Wideband signal focusing module; The denoised ultrasonic array signals are divided into multiple narrowband signals by using the discrete Fourier transform method. Among them, the narrowband signal where the center frequency of the denoised ultrasonic array signals is located is the central narrow frequency band. Calculate the focusing matrix, signal model, and covariance matrix of each narrowband signal, perform the focusing transformation, assign different inertia weights to the covariance matrix obtained after the focusing transformation of each narrowband signal to obtain the weighted covariance matrix of each narrowband signal after the focusing transformation, and calculate the mean value of the weighted covariance matrices of all narrowband signals after the focusing transformation to obtain the weighted covariance matrix of the focusing frequency points; Unitary transformation module; Perform unitary transformation on the weighted covariance matrix of the focusing frequency points to obtain a unitary matrix. Based on the Gerschgorin circle criterion, use the center and radius of the Gerschgorin circles in the unitary matrix to perform similarity transformation on the unitary matrix, and estimate the number of local discharge sources according to the signal source number estimation equation; Spatial parameter obtaining module; Perform eigenvalue decomposition on the weighted covariance matrix of the focusing frequency points, construct a spatial spectrum function by using the multiple signal classification algorithm, and perform spectrum peak search at a preset step size within the specified spatial parameter range to obtain the spatial parameters corresponding to each local discharge source; Coordinate obtaining module; Based on the distance between the local discharge source and the sensor array and the spatial parameters of the local discharge source, perform coordinate transformation to obtain the coordinates of the local discharge source in the three-dimensional space.

Images