Discharge degree detection method based on sound source frequency statistical characteristics

Abstract

The invention relates to the technical field of partial discharge detection, in particular to a discharge degree detection method based on sound source frequency statistical characteristics. The method comprises the following steps: performing framing processing on continuously collected discharge sound signals, and performing unilateral Fourier transform on each frame; each frame of frequency spectrum is converted into a two-dimensional single-channel image with the same size, the width of the image corresponds to a frequency scale, the height of the image corresponds to a frequency amplitude, in the image generation process, pixels below a frequency spectrum curve are set to be 1, other pixels are set to be 0, binarized spectrograms are formed, pixel-level accumulation is conducted on all the spectrograms, and the frequency spectrum is obtained. According to the method, a single-channel frequency characteristic statistical graph is generated, numerical values corresponding to frequencies in the image reflect the discharge degree, quantitative characterization of the discharge intensity is achieved through statistics of the frequency domain characteristics, misjudgment caused by normal corona discharge is remarkably reduced, and the accuracy and robustness of discharge sound source positioning are improved.

Description

Technical Field

[0001] This invention relates to the field of partial discharge detection technology, specifically a method for detecting the degree of discharge based on the statistical characteristics of sound source frequencies. Background Technology

[0002] In power facilities such as substations and cable tunnels, partial discharge is an important indicator reflecting the insulation status of electrical equipment. Traditional partial discharge detection methods mainly rely on electrical measurements or infrared and ultraviolet imaging, while acoustic signal detection methods have advantages such as being non-contact, interference-resistant, and highly applicable, and are gradually becoming a research hotspot.

[0003] However, current acoustic detection methods mostly employ time-domain or single-frame spectral analysis, which is weakly resistant to interference from environmental noise and normal corona discharge, easily leading to false alarms or missed alarms. Especially when there are multiple repetitive discharges or weak signal continuity, single-frame spectral analysis can easily mask characteristics. Therefore, it is necessary to design a method that can statistically comprehensively express the frequency distribution of discharge behavior to reflect the degree of discharge corresponding to different frequency components. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention proposes a discharge degree detection method based on the statistical characteristics of sound source frequencies. By accumulating and statistically enhancing the features of multiple consecutive frames of sound signal spectrum images, a frequency-intensity distribution map is established, enabling joint determination of discharge frequency components and discharge degree. This effectively improves the system's ability to identify partial discharge in complex environments, as detailed below.

[0005] The technical solution adopted by this invention to solve its technical problem is: a method for detecting the degree of discharge based on the statistical characteristics of sound source frequency, comprising the following steps: S1: Use an acoustic sensor to collect data on a continuous length of... The audio signal is divided into frames, with the number of frames being... The length of each frame is ,and ; S2: Utilize a hardware processor to process each frame of the acquired audio signal. (No. frame, The amplitude spectrum is obtained after performing a one-sided Fourier transform. Then convert the amplitude spectrum into a binary image. ; S3: For each frame of binary image To perform superposition, the superposition expression is: ; S4: Will After subtracting the mean of the corresponding row from each row, normalize the result. The expression is: ,in for The OK, for The first in pixel value, and These represent images. row and column indexes, This represents the frequency image after subtracting the mean, followed by normalization. ,in This is a frequency map after feature extraction; S5: According to To determine whether the discharge behavior is normal.

[0006] Preferably, the acoustic sensor is applied at the site of the electrical equipment being tested to collect the sound in real time.

[0007] Preferably, in step S1, The duration of the acquired acoustic signal. It is a positive integer.

[0008] Preferably, in step S2, the hardware processor model is determined based on the size of each frame of audio signal data in S1.

[0009] Preferably, in step S2, the width of the image in the binary image Set as .

[0010] Preferred, The sampling rate of the system is high. And it was collected in S1 The theoretical maximum value of the spectrum amplitude of the frame sound signal.

[0011] Preferred, The results in the table represent the single-sided spectrum of each frame. Represents a single-sided spectrum. Indicates the first The single-sided spectrum corresponding to the frame sound signal Indicates frequency, and exist between.

[0012] Preferably, in step S3, This is a single-channel frequency image after overlay. For the first Frame sound signal, This represents the total number of audio signal frames.

[0013] Preferably, in step S4, This represents the image after normalizing the mean of each row of P in S3. Indicates taking The minimum value in, Indicates taking The maximum value in.

[0014] Preferably, the determination criterion for step S5 is: The corresponding region contains abnormal discharges at certain frequencies, and the degree of discharge is within that region. The corresponding maximum value.

[0015] The beneficial effects of this invention are as follows: This invention establishes a frequency-intensity distribution map by accumulating and statistically analyzing the spectrum images of multiple consecutive frames of acoustic signals and enhancing their features. This enables the joint determination of discharge frequency components and discharge degree, effectively improving the system's ability to identify partial discharge in complex environments. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 This is the overall flowchart of the present invention; Figure 2 It is in S2 of the present invention A schematic diagram showing the conversion of a frame audio signal from the time domain to a binary representation with frequency meaning; Figure 3 This is a binary diagram of the frequency-image used to verify the feasibility of the technical solution in an embodiment of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0019] like Figure 1 As shown, this invention provides a method for detecting the degree of discharge based on the statistical characteristics of sound source frequencies, comprising the following steps: S1: The acoustic sensor acquires a continuous acoustic signal of length T and divides it into frames. The number of frames is... The length of each frame is ,in The duration of the acquired acoustic signal can also represent the number of sampling points. It is a positive integer, and , It should be noted that the sound sensor is used at the site of the electrical equipment being tested and needs to collect the sound at the site in real time, that is, to start collecting the sound from the beginning, regardless of the volume of the sound. Therefore, it can acquire the sound at the site of the electrical equipment and provide standard-sized audio data for subsequent analysis by the processor.

[0020] S2: Utilize a hardware processor to process each frame of the acquired audio signal. (No. frame, The amplitude spectrum is obtained after performing a one-sided Fourier transform. Then convert the amplitude spectrum into a binary image. In the binary image, the width W of the image is set to... ,in The system's sampling rate, the image's high... , and is the theoretical maximum value calculated from the spectral amplitude of the K-frame acoustic signal collected in S1; It should be noted that the hardware processor model in this invention needs to be determined based on the size of each frame of audio signal data in S1. For example, if the data volume is too large, a more advanced model may be required. Generally speaking, an ARM architecture chip or an FPGA chip can be used. Furthermore, the one-sided Fourier transform is a mature mathematical operation that can be directly performed by the processor. This invention will not elaborate on this further. in, The results in the table represent the single-sided spectrum of each frame. Represents a single-sided spectrum. Indicates the first The single-sided spectrum corresponding to the frame sound signal Indicates frequency, and exist Between, thus realizing the transfer of S2 A schematic diagram illustrating the transformation of a frame audio signal from the time domain to a frequency-meaning binary representation, as shown below. Figure 2 As shown.

[0021] S3: For each frame of binary image To perform superposition, the superposition expression is: , This is a single-channel frequency image after overlay.

[0022] in, This is a single-channel frequency image after overlay. For the first Frame sound signal, This represents the total number of audio signal frames, thus enabling the processing of data in S2. Summing the frequency images yields a total image. This results in image-level overlay, such as... The value of the first pixel in the image is... The result is obtained by adding up the first pixel values ​​of all the images.

[0023] S4: Will After subtracting the mean of the corresponding row from each row, normalize the result. The expression is: ,in for The hth line, for The first in pixel value, and These represent images. row and column indexes, This represents the frequency image after subtracting the mean, followed by normalization. ,in This is a frequency map after feature extraction.

[0024] Among them, in the above expression This represents the image after normalizing the mean of each row of P in S3, which can also be understood as "removing the DC component". Indicates taking The minimum value in, Indicates taking The maximum value in the value is used to normalize P. However, unlike the previous "0-1" normalization, a "DC component removal" operation is performed before the normalization. It is the standard "0-1", a mature normalization method, often used in deep learning, which will not be elaborated on in this invention.

[0025] S5: According to To make a judgment, it is believed that The corresponding region contains abnormal discharges at certain frequencies, and the degree of discharge is within that region. The corresponding maximum value.

[0026] Furthermore, in order to verify the feasibility of the technical solution described in this invention, the present invention utilizes the following examples to detect partial discharge signals in industrial electrical systems, as detailed below.

[0027] The parameter settings are as follows: Acoustic sensor sampling duration: Second; System sampling rate: ; Length of each frame: Second; Number of sampling points per frame: ; Frame rate: ; Sampling precision: 16-bit ADC, sampling value range [-32768, 32767].

[0028] More specifically: First, the signal is acquired in frames, dividing the 2-second raw signal into... Frames, each frame is [length missing] Seconds, to obtain the frame signal sequence , .

[0029] Secondly, amplitude spectrum calculation and image conversion are performed, including for each frame of signal. Perform a Fast Fourier Transform (FFT) and compute its one-sided amplitude spectrum, converting the amplitude spectrum into an image. The image width W = 960 / 2 = 480, and the horizontal resolution is... Since the theoretical maximum value of FFT is 32767 = 960 * 32767 ≈ 31,849,920, the vertical resolution is 31,849,920 / 480 = 66,354. Setting the numerical value corresponding to each frequency to 1 yields the following result: Figure 3 The image in the image shows that the shaded area is 1 and the rest is 0.

[0030] Then, image overlay is performed, combining each frame of binary images. By performing point-by-point accumulation, a single-channel frequency image is obtained. .

[0031] Next, the frequency image is normalized, that is, the frequency image is normalized. Remove the mean from each row: .

[0032] After normalization, the frequency characteristic map is obtained. .

[0033] Finally, anomaly detection is performed on the normalized image. In the analysis, the pixel values ​​in rows 40-50 and columns 0-101 were found to be significantly higher than 0.5, indicating a significant energy mutation in this frequency range. Combined with the frequency range analysis, the frequency in this region is concentrated in 2-2.5kHz, which matches the frequency band of a typical arc discharge signal. It is determined to be a local arc discharge event, and the partial discharge amount is the maximum value corresponding to columns 0-101.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the degree of discharge based on the statistical characteristics of sound source frequency, characterized in that: Includes the following steps: S1: Acquire a continuous sound signal of length T using a sound sensor, and divide it into frames, where the number of frames is K, and the length of each frame is L. ; S2: Utilize a hardware processor to process each frame of the acquired audio signal. (frame k, The amplitude spectrum is obtained after performing a one-sided Fourier transform. Then convert the amplitude spectrum into a binary image. ; S3: For each frame of binary image To perform superposition, the superposition expression is: ; S4: Normalize P by subtracting the mean of the corresponding row from each row. The expression is: ,in For the h-th row of P, For the first in P The pixel value, where h and w represent the row and column indices of image P, respectively. This represents the frequency image after subtracting the mean, followed by normalization. ,in This is a frequency map after feature extraction; S5: According to To determine whether the discharge behavior is normal.

2. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: Acoustic sensors are used at the site of the electrical equipment being tested to collect the sound in real time.

3. The discharge degree detection method based on the statistical characteristics of sound source frequency as described in claim 1, characterized in that: In step S1, T is the duration of the acquired acoustic signal, and K is a positive integer.

4. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: In step S2, the hardware processor model is determined based on the size of each frame of audio signal data in S1.

5. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: In step S2, the width of the image in the binary image Set as .

6. The discharge degree detection method based on the statistical characteristics of sound source frequency as described in claim 5, characterized in that: The sampling rate of the system is high. , collected in S1 The theoretical maximum value of the spectrum amplitude of the frame sound signal.

7. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: The results in the table represent the single-sided spectrum of each frame. Represents a single-sided spectrum. Indicates the first The single-sided spectrum corresponding to the frame sound signal Indicates frequency, and exist between.

8. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: In step S3, P is the superimposed single-channel frequency image. For the first Frame sound signal, This represents the total number of audio signal frames.

9. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: In step S4, This represents the image after normalizing the mean of each row of P in S3. Indicates taking The minimum value in, Indicates taking The maximum value in.

10. The method for detecting the degree of discharge based on the statistical characteristics of the sound source frequency as described in claim 1, characterized in that: The judgment criteria for step S5 are as follows: The corresponding region contains abnormal discharges at certain frequencies, and the degree of discharge is within that region. The corresponding maximum value.

Images