Partial discharge identification method and system suitable for direct current GIS

Abstract

The invention relates to the technical field of direct-current GIS fault detection, in particular to a partial discharge identification method and system suitable for a direct-current GIS, and the method comprises the steps: collecting a partial discharge signal of an unknown defect on the direct-current GIS, and obtaining a pulse sequence representation diagram according to the relation between the discharge time and the discharge capacity in the partial discharge signal of the unknown defect; inputting the pulse sequence representation diagram into the trained GIS partial discharge mode recognition model, and performing defect recognition to obtain a recognition result; and comparing the known defect with an unknown defect to make a manual judgment according to the manually observed partial discharge spectrum characteristics of the known defect, and combining a manual judgment result with an identification result to obtain a judgment result of the unknown defect. According to the method, the partial discharge phenomenon occurring in the operation process of the direct-current GIS can be quickly classified, the specific defect type can be identified, and stable operation of the direct-current GIS is facilitated.

Description

Technical Field

[0001] The invention relates to the technical field of direct current GIS fault detection, and in particular to a partial discharge identification method and system suitable for direct current GIS. Background Art

[0002] As the demand for power transmission continues to grow, high-voltage direct current (HVDC) technology is becoming increasingly important in the field of high-voltage systems. As an important part of the DC transmission system, DC GIS equipment also has these advantages. DC GIS equipment can significantly reduce the space occupied by the DC field of offshore platforms and onshore converter stations by up to 90%. This space optimization is particularly important for arranging power equipment in limited space.

[0003] Partial discharge (PD) of DC GIS equipment is a localized discharge phenomenon that only partially crosses the insulator, causing the insulation to gradually degrade, which may eventually cause the failure of the power equipment. Therefore, the detection and characterization of PD on-site and online is a key link to ensure the stable operation of the equipment. Through PD measurement, insulation defects can be identified before the final breakdown occurs, which can significantly improve the reliability and availability of equipment, reduce the unplanned downtime of the power system, extend the operation cycle, and reduce maintenance costs. At the same time, due to the lower risk of explosion and combustion accidents, this also ensures the safety of asset managers and service personnel. In addition, environmental risks such as leakage of harmful gases and greenhouse gases can be reduced.

[0004] Partial discharge measurement under AC voltage is one of the important means of diagnosis and monitoring of high-voltage components. The measurement systems, devices, procedures and interpretation methods have been fully developed. Compared with the standardized methods for AC partial discharge measurement (IEC60270) and related evaluation methods (PRPD mode, etc.), there are no useful test procedures and meaningful acceptance criteria defined to evaluate HVDC components, especially GIS and GIL. In addition, due to the lack of phase angle of PD pulse, there is no available evaluation method for DC voltage. Therefore, it is necessary to establish a partial discharge signal characterization method suitable for DC, which does not rely on phase information and can accurately describe the partial discharge signals under different typical defects, accurately identify the typical defects of GIS under DC voltage, and thus achieve safe and stable operation of DC GIS. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a partial discharge identification method and system suitable for DC GIS in view of the deficiencies in the above-mentioned prior art, so as to solve the technical problem that the existing partial discharge signals of different types of GIS defects cannot be accurately identified.

[0006] The objective of the present invention is achieved by the following technical solutions: In a first aspect, the present invention provides a partial discharge identification method applicable to a DC GIS, comprising: Collect the partial discharge signal of the unknown defect on the DC GIS, and obtain the pulse sequence representation diagram according to the relationship between the discharge time and discharge amount in the partial discharge signal of the unknown defect; The pulse sequence characterization diagram is input into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain the recognition result; According to the partial discharge spectrum characteristics of known defects observed manually, manual judgment is made by comparing them with unknown defects, and the manual judgment results are combined with the recognition results to obtain the judgment results for unknown defects; The GIS partial discharge pattern recognition model is obtained by training according to a spectrum feature data set, and the spectrum feature data set is obtained based on partial discharge signals of DC GIS with different types of defects.

[0007] As a further improvement of the present invention, the collecting of the partial discharge signal of the unknown defect on the DC GIS and obtaining the pulse sequence characterization diagram according to the relationship between the discharge time and the discharge amount in the partial discharge signal of the unknown defect specifically include: Use partial discharge test and detection system to obtain partial discharge signals of unknown defects in high-voltage DC GIS; According to the partial discharge signal of the unknown defect, the discharge pulse sequence under different DC voltages is obtained; Using the relationship between the discharge time and the discharge amount in the discharge pulse sequence, constructing discharge key parameters, wherein the discharge key parameters include discharge time, discharge amount, discharge time interval and discharge amount difference; Using the key discharge parameters, three types of pulse sequence characterization spectra are obtained; the pulse sequence characterization spectra include a pulse sequence characterization spectra based on discharge amount difference, a pulse sequence characterization spectra of discharge amount difference-discharge time, and a pulse sequence characterization spectra based on discharge time.

[0008] As a further improvement of the present invention, the pulse sequence characterization spectrum is used to represent a thermal map through a color coding mode, and is used to characterize the spectrum characteristics of the defect based on the dispersion of data points in the spectrum.

[0009] As a further improvement of the present invention, the GIS local discharge pattern recognition model adopts a YOLOv5 convolutional neural network model.

[0010] As a further improvement of the present invention, the training process of the GIS partial discharge pattern recognition model is as follows: The partial discharge signal data of typical defects of high-voltage DC GIS are obtained through partial discharge test and detection system; the typical defects of DC GIS include needle tip defects, surface contamination defects, metal particle defects and air gap defects; Determine the discharge starting voltage of different typical defects, and select DC GIS typical defects in turn to carry out single typical defect discharge test; Based on the measured discharge starting voltage, the discharge pulse sequences of different typical defects under different DC voltages are collected; According to the discharge pulse sequence of different typical defects, the corresponding discharge key parameters are obtained by using the relationship between discharge time and discharge amount, and the pulse sequence characterization spectrum is obtained by using the discharge key parameters; Based on the pulse sequence characterization spectra corresponding to each typical defect, a standard library of typical defect fault characteristics based on partial discharge pulse sequences is established. The spectra in the standard library of typical defect fault characteristics are annotated and used as the spectrum feature data set to train the YOLOv5 convolutional neural network to obtain the GIS partial discharge pattern recognition model.

[0011] As a further improvement of the present invention, the discharge starting voltage of different typical defects is measured, and typical defects of DC GIS are selected in turn to carry out a single typical defect discharge test, specifically including: Starting from the initial voltage, step-by-step voltage is applied. After each voltage increase to the first set voltage value, a set time interval is maintained. Within the set time interval, when partial discharge is observed and a set number of partial discharge pulses are detected, the voltage value at this moment is taken as the discharge starting voltage. Starting from the voltage level of the discharge starting voltage, the voltage is increased step by step according to the set voltage interval, and the pulse sequence of partial discharge at each voltage value after the voltage is increased is collected respectively until the DC GIS breaks down.

[0012] As a further improvement of the present invention, typical defects of DC GIS are selected in sequence to carry out a single typical defect discharge test, specifically including: the single typical defect discharge test process is used to be carried out in a high-voltage shielding room.

[0013] As a further improvement of the present invention, the partial discharge signal data of typical defects of high-voltage DC GIS are obtained by the partial discharge test and detection system, specifically including: the partial discharge test and detection system uses a pulse current method to obtain the partial discharge signal of the typical defect.

[0014] As a further improvement of the present invention, the partial discharge test and detection system includes a withstand voltage test control box, a transformer, a rectifier silicon stack, a current limiting resistor, a coupling capacitor and a partial discharge instrument; A withstand voltage test control box, a transformer, a rectifier silicon stack, a current limiting resistor, and a DC GIS sample with typical defects are connected in series in sequence to form a first circuit. In the first circuit, the withstand voltage test control box is used to adjust and control the size of the test voltage and the voltage boost rate, and the transformer is used to perform a voltage application operation according to the output result of the withstand voltage test control box. The AC voltage is converted into a DC voltage through the rectifier silicon stack and applied to the DC GIS sample after passing through the current limiting resistor to stimulate partial discharge and generate a partial discharge signal. A coupling capacitor is arranged on the line between the current limiting resistor and the DC GIS sample, and the coupling capacitor is also connected in series with a partial discharge meter. The coupling capacitor is used to couple the partial discharge signal from the high-voltage side of the DC GIS sample to the partial discharge meter on the low-voltage side. The partial discharge meter is used to collect the partial discharge signal and form a pulse sequence of partial discharge.

[0015] In a second aspect, the present invention provides a partial discharge identification system applicable to a DC GIS, which is used to implement the above-mentioned partial discharge identification method applicable to a DC GIS, comprising: The pulse sequence characterization acquisition module collects the partial discharge signal of the unknown defect on the DC GIS, and obtains the pulse sequence characterization diagram according to the relationship between the discharge time and the discharge amount in the partial discharge signal of the unknown defect; The first defect recognition module inputs the pulse sequence characterization diagram into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain recognition results; The defect determination module makes a manual decision based on the partial discharge spectrum characteristics of known defects observed manually and compared with unknown defects. The manual decision result is combined with the recognition result to obtain the determination result for the unknown defect. The GIS partial discharge pattern recognition model is obtained by training according to a spectrum feature data set, and the spectrum feature data set is obtained based on partial discharge signals of DC GIS with different types of defects.

[0016] The beneficial effects of the present invention are as follows: the partial discharge identification method applicable to DC GIS provided by the present invention uses the pulse sequence of DC partial discharge signal to generate a variety of signal distribution maps representing defects, does not require the phase information of DC missing, and can be directly used for defect pattern recognition of DC signal; uses the trained GIS partial discharge pattern recognition model to perform defect recognition on the pulse sequence characterization map, and quickly obtains a preliminary recognition result. This model-based automatic processing method can process a large number of unknown defect partial discharge signals in a short time, greatly improving the efficiency of defect determination. The pattern recognition model can capture subtle features that are difficult for the human eye to detect, improve the accuracy of defect identification, and the introduction of manual judgment further improves the reliability of identification and avoids the limitations of a single method. The manual judgment result is combined with the recognition result of the pattern recognition model to obtain the final unknown defect determination result. This combination further improves the accuracy and reliability of identification. The spectrum representation differences of partial discharge signals under different defects and pattern recognition and classification can quickly and accurately classify the partial discharge phenomenon occurring during the operation of DC GIS and identify the specific defect type, which is conducive to the stable operation of DC GIS.

[0017] Furthermore, by constructing key parameters such as discharge time, discharge amount, discharge time interval, and discharge amount difference, the timing and amplitude characteristics of partial discharge can be fully reflected. The pulse sequence characterization spectrum based on the key discharge parameters can more accurately reflect the detailed characteristics of partial discharge and improve the accuracy of defect identification. Different types of pulse sequence characterization spectrum can show the discharge characteristics from different angles and provide a more comprehensive analysis basis.

[0018] Furthermore, the heat map provides an intuitive and visual analysis tool that allows technicians to quickly identify and understand the timing and amplitude characteristics of partial discharge signals. The color coding mode allows for a more intuitive observation of the density and amplitude of data points, improving the intuitiveness and readability of the analysis. The heat map can fully characterize the spectral characteristics of different types of defects, including key parameters such as discharge time, discharge amount, discharge time interval, and discharge amount difference. The color coding mode can more comprehensively display the detailed characteristics of partial discharge signals, improving the comprehensiveness of the analysis.

[0019] Furthermore, the present invention can be applied to GIS equipment of different types and specifications, and has strong scalability. By continuously updating the atlas feature data set, the model can continuously learn and adapt to new defect types, improving the comprehensiveness of recognition. By training the YOLOv5 convolutional neural network, the automatic recognition of GIS partial discharge patterns can be achieved, improving the efficiency and accuracy of recognition.

[0020] Furthermore, the defect discharge test process is carried out in a high-voltage shielding room, which can effectively shield noise interference such as lights, mobile phone signals, and external electromagnetic signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A flow chart of a method for constructing a partial discharge pattern recognition system for a DC GIS according to the present invention; Figure 2 This is a circuit connection block diagram required for the acquisition of DC partial discharge signals of typical defects in GIS according to the present invention; Figure 3a The partial discharge pulse sequence is collected by using the partial discharge experimental platform for constructing DC GIS described in the present invention.

[0023] Figure 3b A pulse sequence analysis diagram under a pinpoint defect constructed by using the partial discharge pattern recognition system for constructing a DC GIS according to the present invention; Figure 3c A pulse sequence analysis diagram under an air gap defect constructed by using the partial discharge pattern recognition system for constructing a DC GIS according to the present invention; Figure 3d A pulse sequence analysis diagram under insulator contamination defect constructed by using the partial discharge pattern recognition system for DC GIS according to the present invention; Figure 3e A pulse sequence analysis diagram under metal particle defects constructed by using the partial discharge pattern recognition system for constructing a DC GIS according to the present invention; Figure 4 The invention discloses the recognition result of random graphs by using the neural network pattern recognition to construct the partial discharge pattern recognition system of DC GIS.

[0024] In the figure: 1. Withstand voltage test control box; 2. Transformer; 3. Rectifier silicon stack; 4. Current limiting resistor; 5. Protection resistor; 6. Coupling capacitor; 7. Filter capacitor; 8. Voltage divider; 9. DC GIS sample; 10. Oscilloscope; 11. Display; 12-14. Partial discharge meter. DETAILED DESCRIPTION

[0025] In order to make the purpose and technical solution of the present invention clearer and easier to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments, wherein the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0027] Example 1 like Figure 1-Figure 4 As shown, this embodiment provides a partial discharge identification method applicable to DC GIS. Based on the DC partial discharge experimental platform of GIS typical defects, experiments are carried out using equivalent models of four typical GIS defects, DC GIS partial discharge signal data is obtained and its partial discharge pulse sequence is collected, and a method for characterizing the partial discharge pulse sequence is given. The following is a specific implementation method.

[0028] The partial discharge signal of the unknown defect on the DC GIS is collected, and the pulse sequence characterization diagram is obtained according to the relationship between the discharge time and discharge amount in the partial discharge signal of the unknown defect.

[0029] Specifically, the partial discharge test and detection system is used to obtain the partial discharge signal of unknown defects in the high-voltage DC GIS; According to the partial discharge signal of the unknown defect, the discharge pulse sequence under different DC voltages is obtained; The relationship between the discharge time and the discharge amount in the discharge pulse sequence is used to construct the key discharge parameters, which mainly include the discharge time t i , discharge capacity Q i , discharge time interval △t i And the discharge difference △Q i Among them, as another implementation method, it is also necessary to construct the next discharge time t i+1 , the next discharge amount Q i+1 , next discharge time interval △ t i+1 , the next discharge difference △Q i+1 Key parameters. In this embodiment, the discharge time interval is the difference between the next discharge time and the current discharge time ( △T i =T i+1 -T i ), the discharge difference is the difference between the next discharge and the current discharge ( △Q i = Q i+1-Q i ).

[0030] Using the key discharge parameters, three types of pulse sequence characterization spectra are obtained, including: △Q i+1 =f(△Q i )、△Q i =f (△t i )、△t i+1 =f(△t i ) ; The pulse sequence characterization spectrum includes a pulse sequence characterization spectrum based on discharge amount difference, a pulse sequence characterization spectrum of discharge amount difference-discharge time, and a pulse sequence characterization spectrum based on discharge time.

[0031] In addition, the pulse sequence characterization spectrum is used to represent the heat map through a color coding mode, which is used to characterize the spectral characteristics of the defects according to the dispersion of the data points in the spectrum. Specifically, key parameters such as discharge time, discharge amount, discharge time interval and discharge amount difference are extracted from the pulse sequence characterization spectrum as data points. The data points are mapped to the heat map using the color coding mode. The color depth indicates the density and amplitude of the data points. The spectral characteristics of different types of defects are characterized by observing the dispersion, color distribution and pattern changes of the data points in the heat map. Select a suitable color mapping scheme (such as blue to red, white to black, etc.) to map the density and amplitude of the data points to the color depth. By observing the dispersion of the data points in the heat map, the discharge patterns of different defect types can be identified.

[0032] The pulse sequence characterization diagram is input into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain the recognition result; The GIS partial discharge pattern recognition model adopts a convolutional neural network model, and specifically, the YOLOv5 convolutional neural network model is adopted in this embodiment.

[0033] The GIS partial discharge pattern recognition model is trained based on the spectrum feature data set, which is obtained based on the partial discharge signals of DC GIS with different types of defects. This embodiment mainly involves typical defects of DC GIS, which include needle tip defects, surface contamination defects, metal particle defects and air gap defects. The specific training process of the GIS partial discharge pattern recognition model is as follows: Obtain the partial discharge signal data of typical defects of high-voltage DC GIS through partial discharge test and detection system; Determine the discharge inception voltage (PDIV) of different typical defects, and select typical defects of DC GIS to conduct single typical defect discharge test; Based on the measured discharge starting voltage, the discharge pulse sequences of different typical defects under different DC voltages are collected; According to the discharge pulse sequence of different typical defects, the corresponding discharge key parameters are obtained by using the relationship between discharge time and discharge amount, and the pulse sequence characterization spectrum is obtained by using the discharge key parameters; Based on the pulse sequence characterization spectra corresponding to each typical defect, a standard library of typical defect fault characteristics based on partial discharge pulse sequences is established. The spectra in the standard library of typical defect fault characteristics are annotated and used as the spectrum feature data set to train the YOLOv5 convolutional neural network to obtain the GIS partial discharge pattern recognition model.

[0034] The partial discharge test and detection system includes a withstand voltage test control box 1, a transformer 2, a rectifier silicon stack 3, a current limiting resistor 4, a coupling capacitor 6 and partial discharge instruments 12-14; The withstand voltage test control box 1, transformer 2, rectifier silicon stack 3, current limiting resistor 4, and DC GIS sample 9 with typical defects are connected in series in sequence to form the first circuit. In the first circuit, the withstand voltage test control box 1 is used to adjust and control the size of the test voltage and the voltage boost rate, and the transformer 2 is used to perform a voltage application operation according to the output result of the withstand voltage test control box. The AC voltage is converted into a DC voltage through the rectifier silicon stack 3 and the current limiting resistor 4, and then applied to the DC GIS sample to stimulate partial discharge and generate a partial discharge signal. The coupling capacitor 6 is set on the line between the current limiting resistor and the DC GIS sample, and the coupling capacitor 6 is also connected in series with a partial discharge meter 14; the coupling capacitor 6 is used to couple the partial discharge signal from the high voltage side of the DC GIS sample 9 to the partial discharge meter on the low voltage side; the partial discharge meter is used to collect the partial discharge signal and form a pulse sequence of partial discharge.

[0035] In addition, the system also includes a protective resistor 5, which is arranged in series between the current limiting resistor 4 and the DC GIS sample 9. It has the same function as the current limiting resistor 4 and is used to limit the current in the test circuit, thereby protecting other equipment. A filter capacitor 7 and a voltage divider 8 are also arranged in parallel on the line between the current limiting resistor 4 and the protective resistor 5. The filter capacitor 7 is used to filter out the ripple component in the test voltage. The voltage divider 8 is used to measure the test voltage. A display 11 is also provided on the DC GIS sample 9, which is connected to the oscilloscope 10 through the display 11. During the acquisition of the DC partial discharge signal of the typical defect of GIS, the oscilloscope 10 can receive the partial discharge signal transmitted from the coupling capacitor and the test voltage signal measured by the voltage divider, and then display these signals in the form of a visualized waveform. The display 11 is used to visualize the signal waveform displayed by the oscilloscope and other related information (such as test parameters, measurement results, etc.).

[0036] The partial discharge instrument used in this embodiment is the MPD800 partial discharge instrument. In the process of collecting DC partial discharge signals of typical defects in GIS, the MPD800 partial discharge instrument receives partial discharge signals transmitted from coupling capacitors.

[0037] The power supply of the experimental platform includes: 5kVA / 50kV AC withstand voltage test control box, 5Kva / 50kV oil-immersed high-voltage test transformer, AC100kV / 380kΩ high-voltage current limiting resistor, 50kV / 0.5μF filter capacitor, 60k AC / DC high-voltage divider with a voltage divider ratio of 1000:1, 50kV / 5kΩ protection resistor, 50kV / 800pF coupling capacitor. The partial discharge test and detection system adopts the pulse current method, using Omicron MPD 800 and measuring impedance cpl1 for PD pulse detection. The bandwidth of the measurement system is set to 300khz and the center frequency is 250kHz. The detection impedance, coupling capacitor and other equipment are in compliance with international standards and instrument specifications such as IEC60270, and the detection sensitivity can reach 10fC. The oscilloscope used is Teledyne high-speed oscilloscope wafesurfer510, with a bandwidth of 1GHz and a maximum sampling rate of 10GS / s.

[0038] In addition, the typical defect discharge test in this embodiment is carried out in a high-voltage shielded room, which can effectively shield interference such as light, mobile phone signals, and external electromagnetic signals. All equipment and the shielded room are grounded separately, and the background noise of the partial discharge instrument is less than 0.5pC. In order to eliminate the influence of environmental noise and line noise and obtain a partial discharge signal that can better characterize typical defects, the threshold of the discharge amount acquisition is adjusted to 10pC.

[0039] Specifically, the parameter initialization process includes setting various parameters such as the center frequency fs, bandwidth ∆f, partial discharge measurement threshold, and the accuracy of each spectrum generation of the partial discharge measurement system. The MPD800 partial discharge instrument is calibrated for charge and voltage respectively to ensure measurement accuracy, so that the MPD800 can directly display the applied voltage and discharge charge of the partial discharge.

[0040] Determine the initial discharge voltage (PDIV) of the plate-plate electrode, needle tip defect, air gap defect, insulator surface contamination defect, and metal particle defect. The plate-plate electrode is to ensure that the local discharge signals of different defects collected later are not local discharges of the line itself. The specific method is to step up the voltage from 1kV, increase the voltage by 1kV each time, and maintain the voltage for 30s after the boost is completed to observe whether there is local discharge until three PD pulses are detected within 30 seconds. The voltage value at this time is the PDIV of the defect model; if not, continue to increase the voltage. A single discharge during the boost process is not considered to be the initial local discharge. Go to step 3 to collect the local discharge signal of the typical GIS defect needle tip defect under DC.

[0041] Install a needle tip defect in the cavity. Next, measure the voltage level of PDIV and increase it by 10% each step. If there is no breakdown, collect the pulse sequence of partial discharge under the current voltage to obtain 100,000 data points, output the csv file, continue to increase the voltage, collect data points, until the line breaks down, replace the air gap defect, insulator surface contamination defect, and metal particle defect in turn and repeat, and obtain the partial discharge pulse sequence of four typical GIS defects under DC.

[0042] According to the pulse sequence csv file collected in step 3, the discharge time is obtained respectively t i , discharge capacity Q i , discharge time interval △t i , discharge capacity difference △Q i , Subsequent discharge time t i+1 , subsequent discharge Q i+1 , six key parameters, among which △Q i =Q i+1 - Q i 、△T i =T i+1 -T iThree different pulse sequence characterization spectra were obtained using four key parameters, including △Q i+1 =f(△ Q i )、△Q i =f(△t i )、△t i+1 =f(△t i ) , the distribution of data points is represented in the form of a heat map.

[0043] According to the drawn pulse sequence analysis heat map, data annotation is performed, and the three pulse sequence characterization diagrams are obtained to establish a standard library of typical defect fault features based on local discharge pulse sequences. The spectral feature data set is randomly allocated to build a training set: test set database with a ratio of 8:2. The training set is input into the convolutional neural network for training, and the test set is used as the input into the trained convolutional neural network to obtain the local discharge pattern classification accuracy, and a GIS local discharge pattern neural network recognition model is constructed.

[0044] Finally, based on the manual observation of the partial discharge spectrum characteristics of known defects and comparison with unknown defects, manual judgment is made, and the manual judgment results are combined with the recognition results to obtain the judgment results for the unknown defects.

[0045] The construction method of the partial discharge pattern recognition system applied to the actual DC GIS in this embodiment is described. Figure 2 As shown, the operation process of partial discharge pattern recognition of actual DC GIS by the method of the present invention is explained.

[0046] By applying a high-voltage DC voltage, local discharge will occur in the typical defect model 9, generating a pulse current. The pulse current passes through the coupling capacitor 6 and the measuring impedance in the MPD800 measurement system connected in series with it, and is collected by the local discharge measuring instrument to form a pulse sequence of local discharge, which is output through the display.

[0047] In actual detection, partial discharge development in HVDC GIS follows the same physical processes as those observed under AC voltage stress. However, the behavior of DC discharge events, such as amplitude and repetition rate, is very different from that of AC discharge events due to the constant electric field, the continuous directional motion of charge carriers, and the generation of space and surface charges. Therefore, the methods and findings related to AC PD classification cannot be directly applied to HVDC GIS. For example, the mature AC GIS partial discharge detection methods rely on measurements in the ultra-high frequency (UHF) range and the evaluation of phase-resolved partial discharge (PRPD) maps by human experts, which are not suitable for distinguishing DC partial discharge source signals due to the lack of necessary phase information. Therefore, the requirements for the construction of partial discharge pattern recognition systems for DC GIS are more stringent in order to utilize DC partial discharge signals so that they can accurately express the discharge characteristics under different typical defects.

[0048] The corresponding DC GIS partial discharge pattern recognition system must abandon the phase information in the AC, and use the amplitude and time relationship of continuous partial discharge pulses to construct a pulse sequence analysis spectrum. The overall shape of the three characteristic spectra and the distribution of data points can characterize the partial discharge information of different defects, and ultimately realize the identification of DC GIS partial discharge defect types, helping maintenance personnel to quickly discover and find defects, and ensure the safe and stable operation of the power system GIS.

[0049] Combination Figure 3a ~ Figure 3e , the practical application of the method for constructing the partial discharge pattern recognition system of DC GIS of the present invention is explained.

[0050] Obtain partial discharge signals according to the center frequency and bandwidth requirements of the IEC standard, such as Figure 3a As shown, the bandwidth of the measurement system is set to 300khz and the center frequency is 250kHz. This experiment is a plate-plate electrode experiment. When the voltage is increased from 11kV to 12kV, three pulse signals appear within 30s. Therefore, the PDIV is determined to be 11.2kV, and the voltage level signal is collected. The voltage is then increased to 1.1 times the PDIV, that is, 12.3kV. It is found that there is no breakdown, and the partial discharge signal is collected until the line breaks down and the collection ends. The above steps are performed on the needle tip defect, air gap defect, insulator surface contamination defect, and metal particle defect. The PDIV under positive DC voltage is 5.2kV, 6kV, 3kV, and 5.2kV, respectively, and the PDIV under negative DC voltage is -6.3kV, -4.5kV, -2kV, and -3.5kV, respectively.

[0051] Figure 3b ~ Figure 3e The pulse sequence analysis diagrams under different typical defects. The first row is under positive voltage. △ti+1 =f (△t i )、△Q i+1 =f(△Q i )、△Q i =f(△t i )， The second line is under negative voltage △t i+1 =f(△t i )、△Q i+1 =f(△ Q i )、△Q i =f(△t i )。

[0052] Figure 3b The pulse sequence analysis diagram under the pinpoint defect constructed by the partial discharge pattern recognition system for DC GIS according to the present invention is shown in the figure. △t i+1 =f(△t i ) The spectrum appears to radiate outward from the zero point. △Q i+1 =f (△Q i ) The graph is a triangle with a small amount of triangles extending out from three blocks of data points, and the center point of the data set is relatively round. △Q i =f(△t i ) The spectrum shows that a large amount of particles are concentrated near the zero point, and a small amount of particles are concentrated in parallel above and below.

[0053] Figure 3c The pulse sequence analysis diagram under the air gap defect constructed by the partial discharge pattern recognition system for DC GIS according to the present invention is shown in the figure. △t i+1 =f(△t i ) The spectrum appears to radiate outward from the zero point. △Q i+1 = f(△Q i )The spectrum shows a triangular distribution, but the points in the data set are distributed in three radial lines. △Q i =f(△t i ) The spectrum shows that a large amount of data is concentrated near the zero point.

[0054] Figure 3d The pulse sequence analysis diagram under the insulator contamination defect constructed by the partial discharge pattern recognition system for DC GIS according to the present invention is shown in the figure. △t i+1 =f(△t i ) The spectrum appears to radiate outward from the zero point. △ Q i+1 =f(△Q i ) The map is represented by a triangle plus more triangles extending from three areas where points from multiple data sets converge. △Q i =f(△t i ) The spectrum shows that a large amount of particles are concentrated near the zero point, and a small amount of particles are concentrated in parallel above and below.

[0055] Figure 3e The pulse sequence analysis diagram under the condition of metal particle defects constructed by the partial discharge pattern recognition system for DC GIS according to the present invention is shown in the figure. △t i+1 =f(△t i ) The spectrum appears to radiate outward from the zero point. △ Q i+1 =f(△Q i ) The spectrum is shown as multiple clusters, with the overall shape of petals. △Q i =f(△t i ) The spectrum shows that there is a gap at the zero point, and the rest is roughly divided into three convergence points, diverging from the y-axis to the positive direction of the x-axis.

[0056] Figure 4The recognition results of random graphs by using the neural network pattern recognition of the partial discharge pattern recognition system for DC GIS constructed by the present invention. The two graphs on the left are graphs drawn by collecting partial discharge signals from metal particle defects, and the results of neural network recognition are also particle defects; the upper right is a graph drawn by collecting partial discharge signals from insulator surface contamination defects, and the results of neural network recognition are also contamination defects; the lower right is a graph drawn by collecting partial discharge signals from air gap defects, and the results of neural network recognition are also air gap defects; the accuracy and effectiveness of neural network recognition of the graphs drawn by the present invention are verified.

[0057] like Figure 1 As shown, the method for constructing a partial discharge pattern recognition system for a DC GIS provided by the present invention combines a partial discharge signal pulse sequence and utilizes the pulse sequence of the DC partial discharge signal to generate a variety of signal distribution spectra characterizing defects. It does not require the phase information of the DC missing and can be directly used for defect pattern recognition of the DC signal, and can accurately detect partial discharge phenomena inside the power equipment.

[0058] As shown in Figures 3 and 4, the method of constructing the partial discharge signal characterization spectrum proposed by the present invention has obvious differences under different typical defects. Most of them use human experts to learn the collected defects and directly determine the types of unknown defects. In addition to manual means, defects can also be intelligently identified through neural network algorithms. Machine learning models can quickly process and analyze large amounts of data and give preliminary diagnostic results in a short time. This greatly reduces the time and energy required for manual diagnosis. At the same time, manual diagnosis can be further verified and confirmed on this basis to ensure the accuracy and completeness of the diagnosis. By combining manual diagnosis and machine learning models, continuous monitoring and predictive maintenance of equipment status can be achieved. This helps to discover potential faults in advance and take appropriate measures to repair them, thereby avoiding production losses and maintenance costs caused by GIS downtime due to faults.

[0059] Example 2 This embodiment provides a partial discharge identification system applicable to DC GIS, which is mainly used to implement the partial discharge identification method applicable to DC GIS shown in Example 1. The system mainly includes: The pulse sequence characterization acquisition module collects the partial discharge signal of the unknown defect on the DC GIS, and obtains the pulse sequence characterization diagram according to the relationship between the discharge time and the discharge amount in the partial discharge signal of the unknown defect; The first defect recognition module inputs the pulse sequence characterization diagram into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain recognition results; The defect determination module makes a manual decision based on the partial discharge spectrum characteristics of known defects observed manually and compared with unknown defects. The manual decision result is combined with the recognition result to obtain the determination result for the unknown defect. The GIS partial discharge pattern recognition model is trained based on a spectrum feature data set, and the spectrum feature data set is obtained based on partial discharge signals of DC GIS with different types of defects.

[0060] The working process and principle of each module in this system have been described in Example 1 and will not be repeated here.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A partial discharge identification method suitable for DC GIS, characterized in that: include: Collect the partial discharge signal of the unknown defect on the DC GIS, and obtain the pulse sequence representation diagram according to the relationship between the discharge time and discharge amount in the partial discharge signal of the unknown defect; The pulse sequence characterization diagram is input into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain the recognition result; According to the partial discharge spectrum characteristics of known defects observed manually, manual judgment is made by comparing them with unknown defects, and the manual judgment results are combined with the recognition results to obtain the judgment results for unknown defects; The GIS partial discharge pattern recognition model is obtained by training according to a spectrum feature data set, and the spectrum feature data set is obtained based on partial discharge signals of DC GIS with different types of defects.

2. The partial discharge identification method applicable to DC GIS according to claim 1 is characterized in that: The collecting of the partial discharge signal of the unknown defect on the DC GIS and obtaining the pulse sequence characterization diagram according to the relationship between the discharge time and the discharge amount in the partial discharge signal of the unknown defect specifically include: Use partial discharge test and detection system to obtain partial discharge signals of unknown defects in high-voltage DC GIS; According to the partial discharge signal of the unknown defect, the discharge pulse sequence under different DC voltages is obtained; Using the relationship between the discharge time and the discharge amount in the discharge pulse sequence, constructing discharge key parameters, wherein the discharge key parameters include discharge time, discharge amount, discharge time interval and discharge amount difference; Using the key discharge parameters, three types of pulse sequence characterization spectra are obtained; the pulse sequence characterization spectra include a pulse sequence characterization spectra based on discharge amount difference, a pulse sequence characterization spectra of discharge amount difference-discharge time, and a pulse sequence characterization spectra based on discharge time.

3. The partial discharge identification method applicable to DC GIS according to claim 2 is characterized in that: The pulse sequence characterization map is used to represent the heat map through a color-coded pattern to characterize the spectral characteristics of the defect based on the dispersion of the data points in the map.

4. The partial discharge identification method applicable to DC GIS according to claim 1 is characterized in that: The GIS local discharge pattern recognition model adopts the YOLOv5 convolutional neural network model.

5. The partial discharge identification method applicable to DC GIS according to claim 4 is characterized in that: The training process of GIS partial discharge pattern recognition model is as follows: The partial discharge signal data of typical defects of high-voltage DC GIS are obtained through partial discharge test and detection system; the typical defects of DC GIS include needle tip defects, surface contamination defects, metal particle defects and air gap defects; Determine the discharge starting voltage of different typical defects, and select DC GIS typical defects in turn to carry out single typical defect discharge test; Based on the measured discharge starting voltage, the discharge pulse sequences of different typical defects under different DC voltages are collected; According to the discharge pulse sequence of different typical defects, the corresponding discharge key parameters are obtained by using the relationship between discharge time and discharge amount, and the pulse sequence characterization spectrum is obtained by using the discharge key parameters; Based on the pulse sequence characterization spectra corresponding to each typical defect, a standard library of typical defect fault characteristics based on partial discharge pulse sequences is established. The spectra in the standard library of typical defect fault characteristics are annotated and used as the spectrum feature data set to train the YOLOv5 convolutional neural network to obtain the GIS partial discharge pattern recognition model.

6. The partial discharge identification method applicable to DC GIS according to claim 5 is characterized in that: Determine the discharge starting voltage of different typical defects, select typical defects of DC GIS in turn to carry out single typical defect discharge test, including: Starting from the initial voltage, step-by-step voltage is applied. After each voltage increase to the first set voltage value, a set time interval is maintained. Within the set time interval, when partial discharge is observed and a set number of partial discharge pulses are detected, the voltage value at this moment is taken as the discharge starting voltage. Starting from the voltage level of the discharge starting voltage, the voltage is increased step by step according to the set voltage interval, and the pulse sequence of partial discharge at each voltage value after the voltage is increased is collected respectively until the DC GIS breaks down.

7. The partial discharge identification method applicable to DC GIS according to claim 6 is characterized in that: Typical defects of DC GIS are selected in turn to carry out single typical defect discharge test, specifically including: the single typical defect discharge test process is used to be carried out in a high-voltage shielding room.

8. The partial discharge identification method applicable to DC GIS according to claim 5 is characterized in that: The method of obtaining local discharge signal data of typical defects of high-voltage DC GIS by using a local discharge test and detection system specifically includes: the local discharge test and detection system uses a pulse current method to obtain local discharge signals of typical defects.

9. The partial discharge identification method applicable to DC GIS according to claim 8, characterized in that: The partial discharge test and detection system includes a withstand voltage test control box, a transformer, a rectifier silicon stack, a current limiting resistor, a coupling capacitor and a partial discharge instrument; A withstand voltage test control box, a transformer, a rectifier silicon stack, a current limiting resistor, and a DC GIS sample with typical defects are connected in series in sequence to form a first circuit. In the first circuit, the withstand voltage test control box is used to adjust and control the size of the test voltage and the voltage boost rate, and the transformer is used to perform a voltage application operation according to the output result of the withstand voltage test control box. The AC voltage is converted into a DC voltage through the rectifier silicon stack and applied to the DC GIS sample after passing through the current limiting resistor to stimulate partial discharge and generate a partial discharge signal. A coupling capacitor is arranged on the line between the current limiting resistor and the DC GIS sample, and the coupling capacitor is also connected in series with a partial discharge meter. The coupling capacitor is used to couple the partial discharge signal from the high-voltage side of the DC GIS sample to the partial discharge meter on the low-voltage side. The partial discharge meter is used to collect the partial discharge signal and form a pulse sequence of partial discharge.

10. A partial discharge identification system applicable to a DC GIS, used to implement the partial discharge identification method applicable to a DC GIS as claimed in any one of claims 1 to 9, characterized in that: include: The pulse sequence characterization acquisition module collects the partial discharge signal of the unknown defect on the DC GIS, and obtains the pulse sequence characterization diagram according to the relationship between the discharge time and the discharge amount in the partial discharge signal of the unknown defect; The first defect recognition module inputs the pulse sequence characterization diagram into the trained GIS partial discharge pattern recognition model to perform defect recognition and obtain recognition results; The defect determination module makes a manual decision based on the partial discharge spectrum characteristics of known defects observed manually and compared with unknown defects. The manual decision result is combined with the recognition result to obtain the determination result for the unknown defect. The GIS partial discharge pattern recognition model is obtained by training according to a spectrum feature data set, and the spectrum feature data set is obtained based on partial discharge signals of DC GIS with different types of defects.

Images