Method for classifying partial discharge types of GIS equipment based on partial discharge time domain signals

Abstract

The application provides a GIS device partial discharge type classification method based on a partial discharge time domain signal, and belongs to the partial discharge detection classification field; solves the problem that a partial discharge type judgment method based on a partial discharge mode spectrum is invalid due to the failure to obtain an externally applied voltage signal in a field detection process; the method comprises the following steps: collecting and preprocessing time domain data sets of partial discharge light of a sample in the occurrence processes of three partial discharge phenomena of corona discharge, surface discharge and floating discharge through an experimental platform; performing double-threshold peak searching on the data in the three preprocessed time domain data sets of partial discharge light respectively by using a double-threshold peak searching method, obtaining corresponding partial discharge peak value data sets, and processing the partial discharge peak value data sets into corresponding partial discharge light signal amplitude difference spectrum diagrams respectively; and realizing partial discharge type classification through the partial discharge light signal amplitude difference spectrum diagrams; and the application applies the GIS device partial discharge detection.

Description

Technical Field

[0001] This application relates to the field of partial discharge detection and classification technology, and in particular to a method for classifying partial discharge types in GIS equipment based on partial discharge time-domain signals. Background Technology

[0002] Gas-insulated switchgear (GIS) possesses excellent anti-interference performance and is less affected by harsh conditions such as high altitude, extreme weather, and pollution, leading to its widespread use throughout China. However, defects within the GIS cavity, such as metal burrs on high-voltage conductors, metal particles adhering to basin insulators, and metal debris deposited at the bottom of the cavity, can cause partial discharge. Prolonged and continuous partial discharge can degrade the insulation level of the GIS, jeopardizing its safe and stable operation. Therefore, timely detection of partial discharge within the GIS, identification of the type of partial discharge, and implementation of corresponding maintenance measures can effectively extend the service life of the GIS.

[0003] Partial discharge optical measurement is a method for detecting partial discharge in GIS. Based on silicon photomultiplier (SiPM) sensors, this method offers advantages over other detection methods (ultrasonic detection, pulsed current method, and UHF detection) such as high sensitivity, flexible deployment, and ease of driving. By embedding the SiPM sensor within the detection port of the GIS, partial discharge detection can be achieved within the sensor's effective range. To identify the type of partial discharge, the partial discharge optical time-domain signal needs to be processed to obtain a partial discharge mode spectrum. Traditional partial discharge mode spectra include partial discharge phase angle analytical spectra, partial discharge pulse sequence spectra, and polar coordinate phase distribution partial discharge analytical spectra. These spectra share a common feature: they convert the time information of the partial discharge into phase information by combining the applied power frequency voltage time-domain signal with the partial discharge time-domain signal. However, during field detection, the inability to obtain the applied voltage signal often renders the method of determining the partial discharge type through the partial discharge mode spectrum ineffective, significantly increasing the probability of false detections and misdetections of partial discharge in GIS. Summary of the Invention

[0004] To address the aforementioned technical issues, this application proposes a method for classifying partial discharge types in GIS equipment based on partial discharge time-domain signals. This method can accurately classify the partial discharge types of GIS equipment without relying on externally applied voltage time-domain signals, using only the partial discharge time-domain signals from the sensor end.

[0005] The technical solution adopted in this application is: a method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals, comprising the following steps:

[0006] S1: Build an experimental platform for partial discharge light detection, and collect time-domain datasets of partial discharge light of the sample during the occurrence of three partial discharge phenomena: corona discharge, surface discharge and suspension discharge. Preprocess the time-domain datasets of the three partial discharge phenomena to obtain three preprocessed time-domain datasets of partial discharge light.

[0007] S2: The dual threshold peak finding method is used to perform dual threshold peak finding on the data in the time domain datasets of the three preprocessed partial discharge lights to obtain the corresponding partial discharge peak datasets, and then process them into the corresponding partial discharge light signal amplitude difference spectrum.

[0008] S3: The corona discharge type, which is inconsistent with the graphic characteristics of surface discharge and suspension discharge, can be directly identified by the graphic characteristics of the amplitude difference spectrum of the three partial discharge optical signals.

[0009] S4: For surface discharge and levitation discharge with similar graphic features, the distance features on the amplitude difference spectrum of the partial discharge optical signal are extracted by a distance feature calculation method based on the spectral symmetry axis and center point to obtain a distance feature histogram. The statistical features of the distance feature histogram are calculated and fitted to the characteristic curve of the partial discharge optical signal. The partial discharge is classified using the sum of squared residuals of the characteristic curve.

[0010] Furthermore, the experimental platform includes a partial discharge model, an oscilloscope, a SiPM sensor, and a coupling capacitor C. k Detection impedance R k A protective resistor R1 and a high-voltage source V without partial discharge are connected, with the positive terminal of the high-voltage source V connected to one end of the protective resistor R1, and the other end of the protective resistor R1 connected to the high-voltage terminal of the partial discharge model and the coupling capacitor C. k One end, coupling capacitor C k The other end is connected in series with the sensing impedance R k At one end, one probe of the oscilloscope is connected to the coupling capacitor C. k and detection impedance R k Between, the detection impedance R k The other end is connected in parallel to the low-voltage end of the partial discharge model and the negative terminal of the high-voltage source V without partial discharge, and then grounded. The signal output end of the SiPM sensor is connected to another probe of the oscilloscope. The SiPM sensor and the partial discharge model are placed in a sealed cavity to isolate the sensor from the interference of external ambient light. The partial discharge model includes corona discharge model, surface discharge model and suspension discharge model.

[0011] Furthermore, since the partial discharge optical signal is a negative polarity peaked wave, preprocessing is required before peak finding in the time domain. The preprocessing method is to preprocess all voltage amplitudes U n Voltage amplitude U at data points <0 nTaking the opposite number yields the time-domain dataset of the preprocessed partial discharge light.

[0012] Furthermore, the specific steps for performing double-threshold peak finding on the time-domain datasets of the three preprocessed partial discharge light datasets to obtain the corresponding partial discharge peak datasets are as follows:

[0013] Set a time threshold ΔT and a voltage amplitude threshold ΔU, and use ΔU to filter out the voltage amplitude U in the time-domain data of the preprocessed partial discharge light. n For data points smaller than ΔU, suspicious partial discharge peak points are found through local peak finding, resulting in multiple local maxima. The time interval between adjacent local maxima is calculated. If the time interval is less than ΔT, the smaller value of the adjacent local maxima is removed; if the time interval is greater than ΔT, the two adjacent local maxima are retained, and finally, the partial discharge peak dataset is obtained.

[0014] Furthermore, the steps for processing the partial discharge peak dataset into a partial discharge optical signal amplitude difference spectrum are as follows:

[0015] Calculate the voltage amplitude difference between adjacent partial discharge peak points to obtain the normalized voltage amplitude difference. Using the normalized voltage amplitude difference of the current partial discharge optical signal pulse as the abscissa and the normalized voltage amplitude difference of the next pulse as the ordinate, obtain the normalized partial discharge optical signal amplitude difference spectrum.

[0016] Furthermore, the process of extracting the distance features from the amplitude difference spectrum of the partial discharge optical signal using a distance feature calculation method based on the spectral symmetry axis and center point to obtain the distance feature histogram is as follows:

[0017] With y=-x as the axis of symmetry L and (0,0) as the center point O, calculate the shortest distance from each point i in the amplitude difference spectrum of the partial discharge optical signal to L and O. , The distance feature histogram of distance-cumulative points is obtained based on the shortest distance from the partial discharge point to the origin and the axis of symmetry in the amplitude difference spectrum of the partial discharge optical signal.

[0018] Furthermore, the steps for calculating the statistical characteristics of the distance feature histogram and fitting it to the characteristic curve of the partial discharge optical signal are as follows:

[0019] Calculate the statistical feature skewness S in the distance feature histogram k and kurtosis K u , with skewness S k The x-axis represents the kurtosis K. u Using the ordinate as the vertical axis, we obtain S. k -K u The figure shows the S-values ​​under multiple applied voltage levels fitted using a quadratic polynomial or power function. k -Ku The data in the figure shows the characteristic curve of the partial discharge optical signal.

[0020] Furthermore, the coefficient of determination R of the characteristic curve is used. 2 The sum of squared residuals (SSR) distinguishes between surface discharge and levitation discharge.

[0021] Furthermore, when R 2 When the value is greater than 0.9, the characteristic curve is considered to be able to well characterize S. k -K u Data shows that if SSR > 10, it indicates that the discharge type is surface discharge; if SSR < 10, it indicates that the discharge type is floating discharge.

[0022] Furthermore, when the amplitude difference spectrum of the local discharge light signal shows a circular center point set, an upper vertical point set, and a right-side stripe point set, corona discharge can be classified based on these graphic characteristics.

[0023] The advantages of this application over the prior art are as follows:

[0024] (1) The partial discharge type classification method proposed in this application is a classification method based on the optical signal of the SiPM sensor. Compared with commonly used partial discharge detection methods such as pulse current method, ultrasonic detection method and ultra-high frequency detection method, it has good anti-interference performance and can obtain ideal detection effect without additional software / hardware processing.

[0025] (2) Existing partial discharge classification methods are based on phase analysis of applied voltage signals. The partial discharge type classification method of this application classifies corona discharge, surface discharge and suspension discharge by analyzing the relationship between partial discharge signals without relying on the phase information of applied voltage. This provides a new feasible method for partial discharge classification in the field detection process. Attached Figure Description

[0026] The following description, in conjunction with the accompanying drawings, further illustrates this application:

[0027] Figure 1 A flowchart illustrating the method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals provided in this application embodiment;

[0028] Figure 2 A flowchart for dual-threshold peak finding provided in an embodiment of this application;

[0029] Figure 3 A flowchart illustrating the classification of surface discharge and levitation discharge in embodiments of this application;

[0030] Figure 4 This is a schematic diagram of an experimental platform for partial discharge optical detection provided in an embodiment of this application;

[0031] Figure 5 A schematic diagram of three partial discharge models provided in the embodiments of this application;

[0032] Figure 6 A schematic diagram of a single pulse of partial discharge optical signal and pulsed current signal provided in an embodiment of this application;

[0033] Figure 7 This is a schematic diagram of dual-threshold filtering provided in an embodiment of this application;

[0034] Figure 8 This is a schematic diagram illustrating the calculation of partial discharge amplitude difference provided in an embodiment of this application;

[0035] Figure 9 This is a schematic diagram of the amplitude difference spectrum of a typical partial discharge optical signal provided in an embodiment of this application;

[0036] Figure 10 This is a schematic diagram illustrating the calculation of partial discharge distance characteristics provided in an embodiment of this application;

[0037] Figure 11 Typical surface discharge distance characteristic histograms provided for embodiments of this application;

[0038] Figure 12 A typical histogram of levitational discharge distance characteristics provided for embodiments of this application;

[0039] Figure 13 A schematic diagram of a typical surface discharge fitting curve provided in the embodiments of this application;

[0040] Figure 14 This is a schematic diagram of a typical levitation discharge fitting curve provided in the embodiments of this application;

[0041] In the figure: 1 is a partial discharge model, 2 is an oscilloscope, 3 is a SiPM sensor, 4 is an epoxy resin board, and 5 is metal particles. Detailed Implementation

[0042] like Figures 1 to 14 As shown, this application provides a method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals, including the following steps:

[0043] S1: Construct an experimental platform for partial discharge light detection, and use this platform to collect time-domain datasets of partial discharge light during the occurrence of three partial discharge phenomena: corona discharge, surface discharge, and suspension discharge. Preprocess the time-domain datasets collected for the three partial discharge phenomena to obtain three preprocessed time-domain datasets of partial discharge light.

[0044] like Figure 4As shown, the experimental platform includes a partial discharge model 1, an oscilloscope 2, a SiPM sensor 3, and a coupling capacitor C. k Detection impedance R k The system consists of a protective resistor R1 and a high-voltage source V without partial discharge. The positive terminal of the high-voltage source V is connected to one end of the protective resistor R1, and the other end of the protective resistor R1 is connected in parallel to the high-voltage terminal of partial discharge model 1 and a coupling capacitor C. k One end, coupling capacitor C k The other end is connected in series with the sensing impedance R k At one end, a probe of oscilloscope 2 is connected to coupling capacitor C. k and detection impedance R k Between, the detection impedance R k The other end is connected in parallel to the low-voltage end of partial discharge model 1, the negative terminal of the high-voltage source V without partial discharge, and then grounded. The signal output terminal of SiPM sensor 3 is connected to another probe of oscilloscope 2, and SiPM sensor 3 is placed in a sealed cavity to isolate the sensor from interference from external ambient light. SiPM sensor 3 is used to collect the partial discharge light signal generated during the occurrence of partial discharge phenomenon; coupling capacitor C k and detection impedance R k The partial discharge pulse current signal is used to acquire the partial discharge pulse current signal; the partial discharge-free high voltage source V is used to generate a 50Hz power frequency high voltage; the protection resistor R1 is used to protect the equipment when the sample breaks down; the oscilloscope 2 is used to acquire the time domain signals of the partial discharge light signal and the pulse current signal; the partial discharge model 1 is used to simulate the partial discharge generated by the defect under different conditions.

[0045] Figure 5 Schematic diagrams of corona discharge model, surface discharge model, and suspension discharge model are shown. Figure 5 (a) is a corona discharge model, consisting of copper metal needles, epoxy resin plate 4 and stainless steel plate; Figure 5 (b) is a surface discharge model in which a 5mm thick epoxy resin plate 4 is sandwiched between two cylindrical electrodes to simulate surface discharge. Figure 5 (c) is a suspended discharge model, using a copper wire with a length of 3 mm as the metal particle 5.

[0046] The process of collecting time-domain datasets of different partial discharge lights through the experimental platform is as follows: The applied voltage at the moment when the partial discharge signal first appears in partial discharge model 1 is recorded as the initial discharge voltage (PDIV), and the voltage value at this time is expressed as U. PDIV Then, voltages were gradually applied to the partial discharge model 1 at typical applied voltage levels. The optical signals of the partial discharge were collected using a SiPM sensor 3, and the optical signals and pulse current signals during partial discharge were collected and recorded using an oscilloscope 2. A comparison was made between single pulses of partial discharge optical signals and single pulses of partial discharge pulse current at typical applied voltage levels. Figure 6 As shown, the partial discharge optical signal is a negative polarity peaked wave. Therefore, before finding the peak of the time-domain signal, the data needs to be preprocessed. The processing method is as follows: all voltage amplitudes U n Voltage amplitude U at data points <0 n Taking the opposite number yields the time-domain dataset of the preprocessed partial discharge light, and the processing result can be represented as follows:

[0047] .

[0048] All three preprocessed partial discharge time-domain datasets have a dimension of t*2, where t is the number of data points and 2 represents time and voltage.

[0049] S2: The dual-threshold peak finding method is used to perform dual-threshold peak finding on the data in the time domain datasets of the three preprocessed partial discharge lights to obtain the corresponding partial discharge peak datasets, and then process them into the corresponding partial discharge light signal amplitude difference spectrum.

[0050] like Figure 7 and 8 As shown, the specific implementation principle and steps of the dual-threshold peak finding method are as follows:

[0051] Set the time threshold ΔT and the voltage amplitude threshold ΔU. Set the time threshold ΔT (6μs) and the voltage amplitude threshold ΔU. Use ΔU to filter out the concentrated voltage amplitude U in the time-domain data of the preprocessed partial discharge light. n For data points smaller than ΔU, suspicious partial discharge peak points are identified through local peak finding. Assuming m suspicious partial discharge peak points are obtained, n (n=m-1) time intervals T can be obtained from these. n , will T n Compared with ΔT, if T n If T > ΔT, then retain two adjacent data points; if T n If the voltage amplitude is less than ΔT, then remove the data points with smaller voltage amplitudes to obtain n. peak A partial discharge peak data set is formed by identifying several partial discharge peak points.

[0052] The selection of ΔU affects the effectiveness of dual-threshold peak finding. If ΔU is too large, partial discharge data points will be filtered out, and if ΔU is too small, background noise will not be completely filtered out. Therefore, it is necessary to adjust the ΔU threshold multiple times in combination with the amplitude of the partial discharge time-domain waveform to find the optimal threshold.

[0053] After obtaining the partial discharge peak data set, the voltage amplitude difference between adjacent partial discharge peak data is calculated as follows: Figure 8 As shown.

[0054] Calculate the voltage amplitude difference ΔU between adjacent partial discharge peak points. i The normalized voltage amplitude difference Δu is obtained.i :

[0055] ;

[0056] ;

[0057] In the formula: ΔU m For ΔU i The maximum absolute value; i is the number of data points in the amplitude difference spectrum of the partial discharge optical signal, where i=n peak -2.

[0058] The normalized voltage amplitude difference Δu of the current partial discharge optical signal pulse i The horizontal axis represents the normalized voltage amplitude difference Δu of the next pulse. i+1 Using the vertical axis as the ordinate, the normalized amplitude difference spectrum of the partial discharge optical signal is obtained.

[0059] S3: Corona discharge types inconsistent with the graphical characteristics of surface discharge and suspended discharge can be directly identified by analyzing the graphical features of the amplitude difference spectrum of partial discharge optical signals. Surface discharge and suspended discharge are difficult to distinguish by spectral graphical features and require further analysis.

[0060] S4: For surface discharge and levitation discharge with similar graphic features, the distance features on the amplitude difference spectrum of the partial discharge optical signal are extracted by a distance feature calculation method based on the spectral symmetry axis and center point to obtain a distance feature histogram. The statistical features of the distance feature histogram are calculated and fitted to the characteristic curve of the partial discharge optical signal. The partial discharge is classified using the sum of squared residuals of the characteristic curve.

[0061] To quantify the differences in optical signal amplitude difference spectra among different partial discharge types and to classify partial discharges, this application proposes a method for calculating distance features based on the spectral symmetry axis and center point. Using y=-x as the symmetry axis L and (0,0) as the center point O, the shortest distance from each point i in the partial discharge optical signal amplitude difference spectra to L and O is calculated. , The calculation methods are as follows:

[0062] ;

[0063] ;

[0064] In the formula: A, B, and C are constants on the axis of symmetry y = -x, which are 1, 1, and 0 respectively; x i For Δu i ;y i For Δu i+1 .

[0065] Based on the shortest distances from the partial discharge points to the origin and the axis of symmetry in the amplitude difference spectrum of the partial discharge optical signal, a distance feature histogram is obtained, which is calculated as the distance-cumulative number of points. The statistical skewness (S) in the distance feature histogram is then calculated. k ) and kurtosis (K) u The calculation method is as follows:

[0066] ;

[0067] ;

[0068] In the formula: p i μ and σ are distances d, respectively. i The probability, mean, and standard deviation of occurrence; n and the number of peak points of the partial discharge optical signal. peak Related, n=n peak -1.

[0069] Plotting skewness on the x-axis and kurtosis on the y-axis, we obtain S. k -K u Figure. Using a quadratic polynomial or power function to fit S under multiple applied voltage levels. k -K u The data in the figure yields the characteristic curve of the partial discharge optical signal. The coefficient of determination R0 is used. 2 To distinguish between surface discharge and floating discharge from the residual sum of squares (SSR), the calculation method is as follows:

[0070] ;

[0071] In the formula: y i For S k -K u K in the diagram u The true value; For S k -K u K in the diagram u The average value; For S k -K u K in the diagram u The predicted value.

[0072] When R 2 When the value is greater than 0.9, the characteristic curve is considered to be able to well characterize S. k -K u Data shows that if SSR > 10, the discharge type is surface discharge; if SSR < 10, the discharge type is floating discharge. Combining graphic features to distinguish partial discharge can classify corona discharge, surface discharge, and floating discharge.

[0073] The present application will be further described below with reference to specific embodiments.

[0074] Example 1

[0075] The preprocessed data is subjected to dual threshold peak finding, and the process is as follows: Figure 2 As shown, after removing all data points with voltage amplitudes less than ΔU, the local maximum value (suspected partial discharge peak point) of the time domain signal is searched. The time interval between adjacent local maximum values ​​is calculated. If the time interval is less than ΔT, the smaller value of the adjacent local maximum value is removed; if the time interval is greater than ΔT, the two adjacent local maximum values ​​are retained, and finally the partial discharge peak dataset is obtained.

[0076] After obtaining the partial discharge peak dataset, the amplitude difference between adjacent partial discharge peaks is calculated. At this point, the partial discharge peak dataset contains n. peak Several partial discharge peak data points, consisting of n peak A total of n partial discharge peak data points can be obtained. peak -1 amplitude difference; in the amplitude difference dataset, 1:n peak – Two data points are used as the x-axis, 2:n peak – Using one data point as the vertical axis, construct a spectrum of amplitude difference of partial discharge optical signal.

[0077] The applied voltage is 2.0U. PDIV Taking corona discharge as an example, such as Figure 9 As shown, by comparing the amplitude difference spectra of partial discharge optical signals of surface discharge and suspended discharge under the same applied voltage level, it can be observed that the central point set of the amplitude difference spectrum of corona discharge optical signal exhibits a circular distribution, with vertical point sets and strip-shaped point sets scattered above and to the right of the spectrum, respectively. This shows a significant difference in graphic characteristics compared to surface discharge and suspended discharge. Therefore, when the amplitude difference spectrum of partial discharge optical signal shows a circular central point set, an upper vertical point set, and a right-side strip-shaped point set, corona discharge can be classified based on these graphic characteristics.

[0078] Example 2

[0079] The flowchart of the classification method for the amplitude difference of partial discharge optical signals in surface discharge and levitation discharge is as follows: Figure 3 As shown. The processing flow of time-domain data is as described in Example 1 to obtain the amplitude difference spectrum of the surface discharge optical signal, as follows. Figure 7 As shown, the spectral characteristics of surface discharge are similar to those of suspension discharge under the same applied voltage level.

[0080] The distance characteristics of the amplitude difference spectrum of partial discharge optical signals are calculated as follows: Figure 10 As shown, combined with the above d L-i and d O-i The calculation formula can be used to obtain the distance distribution histograms of surface discharge and levitation discharge, as shown below. Figure 11 and Figure 12 As shown.

[0081] Obtain an external voltage of 1.1U. PDIV 1.3U PDIV 1.5U PDIV 1.7U PDIV ,2.0U PDIV The distance distribution histogram at time is used to calculate S. k K u Features, fitting surface discharge and levitation discharge characteristic curves at a 95% confidence level, are as follows: Figure 13 and Figure 14 As shown in the figure. The characteristic curve R at this time... 2 The SSR and SSR are shown in Table 1. Analysis of Table 1 shows that the R of the characteristic curve... 2 All values ​​are above 0.9, indicating that the characteristic curve represents S. k -K u It has strong capabilities in data points; Figure 13 The SSRs obtained by different distance characteristic calculation methods for partial discharge are 89.546 and 68.964, respectively. Since the SSRs are both greater than 10, the partial discharge can be classified as surface discharge. Figure 14 The SSRs obtained from different distance characteristic calculation methods for partial discharge were 2.077 and 2.383, respectively. Since both SSRs are less than 10, this partial discharge can be classified as a floating discharge. Therefore, by combining the graphical characteristics of the amplitude difference spectrum of the partial discharge optical signal and the SSR judgment of the partial discharge characteristic curve, the classification of partial discharges into corona discharge, surface discharge, and floating discharge can be achieved.

[0082] Table 1: Ri values ​​of the surface discharge and levitation discharge characteristic curves at a 95% confidence level 2 And SSR.

[0083]

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and not to limit them; although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for classifying partial discharge types in GIS equipment based on partial discharge time-domain signals, characterized in that: Includes the following steps: S1: Build an experimental platform for partial discharge light detection, and collect time-domain datasets of partial discharge light of the sample during the occurrence of three partial discharge phenomena: corona discharge, surface discharge and suspension discharge. Preprocess the time-domain datasets of the three partial discharge phenomena to obtain three preprocessed time-domain datasets of partial discharge light. S2: The dual threshold peak finding method is used to perform dual threshold peak finding on the data in the time domain datasets of the three preprocessed partial discharge lights to obtain the corresponding partial discharge peak datasets, and then process them into the corresponding partial discharge light signal amplitude difference spectrum. The specific steps for performing double-threshold peak finding on the time-domain datasets of three preprocessed partial discharge beams to obtain the corresponding partial discharge peak datasets are as follows: Set a time threshold ΔT and a voltage amplitude threshold ΔU, and use ΔU to filter out the voltage amplitude U in the time-domain data of the preprocessed partial discharge light. n For data points smaller than ΔU, suspicious partial discharge peak points are found through local peak finding, resulting in multiple local maxima. The time interval between adjacent local maxima is calculated. If the time interval is less than ΔT, the smaller value of the adjacent local maxima is removed; if the time interval is greater than ΔT, the two adjacent local maxima are retained, and finally, the partial discharge peak dataset is obtained. The steps to process the partial discharge peak dataset into a partial discharge optical signal amplitude difference spectrum are as follows: Calculate the voltage amplitude difference between adjacent partial discharge peak points to obtain the normalized voltage amplitude difference. With the normalized voltage amplitude difference of the current partial discharge optical signal pulse as the abscissa and the normalized voltage amplitude difference of the next pulse as the ordinate, obtain the normalized partial discharge optical signal amplitude difference spectrum. S3: The corona discharge type, which is inconsistent with the graphic characteristics of surface discharge and suspension discharge, can be directly identified by the graphic characteristics of the amplitude difference spectrum of the three partial discharge optical signals. S4: For surface discharge and levitation discharge with similar graphic features, distance features on the amplitude difference spectrum of the partial discharge optical signal are extracted using a distance feature calculation method based on the spectral symmetry axis and center point to obtain a distance feature histogram. The statistical features of the distance feature histogram are calculated and fitted to the characteristic curve of the partial discharge optical signal. The coefficient of determination R of the characteristic curve is then used. 2 Differentiate surface discharge from suspension discharge using residual sum of squares (SSR); When R 2 When the value is greater than 0.9, the characteristic curve is considered to be able to well characterize S. k -K u Data shows that if SSR > 10, it indicates that the discharge type is surface discharge; if SSR < 10, it indicates that the discharge type is floating discharge.

2. The method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals according to claim 1, characterized in that: The experimental platform includes a partial discharge model, an oscilloscope, a SiPM sensor, and a coupling capacitor C. k Detection impedance R k A protective resistor R1 and a high-voltage source V without partial discharge are connected, with the positive terminal of the high-voltage source V connected to one end of the protective resistor R1, and the other end of the protective resistor R1 connected to the high-voltage terminal of the partial discharge model and the coupling capacitor C. k One end, coupling capacitor C k The other end is connected in series with the sensing impedance R k At one end, one probe of the oscilloscope is connected to the coupling capacitor C. k and detection impedance R k Between, the detection impedance R k The other end is connected in parallel to the low-voltage end of the partial discharge model and the negative terminal of the high-voltage source V without partial discharge, and then grounded. The signal output end of the SiPM sensor is connected to another probe of the oscilloscope. The SiPM sensor and the partial discharge model are placed in a sealed cavity to isolate the sensor from the interference of external ambient light. The partial discharge model includes corona discharge model, surface discharge model and suspension discharge model.

3. The method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals according to claim 2, characterized in that: Partial discharge optical signals are negative polarity spiked waves. Before finding the peaks in the time-domain signal, the data needs to be preprocessed. The processing method is as follows: all voltage amplitudes U n Voltage amplitude U at data points <0 n Taking the opposite number yields the time-domain dataset of the preprocessed partial discharge light.

4. The method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals according to claim 1, characterized in that: The process of extracting distance features from the amplitude difference spectrum of the partial discharge optical signal and obtaining the distance feature histogram using a distance feature calculation method based on the spectral symmetry axis and center point is as follows: With y=-x as the axis of symmetry L and (0,0) as the center point O, calculate the shortest distance d from each point i in the amplitude difference spectrum of the partial discharge optical signal to L and O. L-i d O-i The distance feature histogram of distance-cumulative points is obtained based on the shortest distance from the partial discharge point to the origin and the axis of symmetry in the amplitude difference spectrum of the partial discharge optical signal.

5. The method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals according to claim 4, characterized in that: The steps for calculating the statistical characteristics of the distance feature histogram and fitting it to the characteristic curve of the partial discharge optical signal are as follows: Calculate the statistical feature skewness S in the distance feature histogram k and kurtosis K u , with skewness S k The x-axis represents the kurtosis K. u Using the ordinate as the vertical axis, we obtain S. k -K u The figure shows the S-values ​​under multiple applied voltage levels fitted using a quadratic polynomial or power function. k -K u The data in the figure shows the characteristic curve of the partial discharge optical signal.

6. The method for classifying partial discharge types of GIS equipment based on partial discharge time-domain signals according to claim 1, characterized in that: When the amplitude difference spectrum of the partial discharge optical signal shows a circular center point set, an upper vertical point set, and a right-side strip-shaped point set, corona discharge can be classified based on these graphic characteristics.

Images