Power equipment fault diagnosis method and system based on GIS partial discharge atlas

Abstract

The invention discloses a power equipment fault diagnosis method and system based on a GIS partial discharge map, and relates to the technical field of fault detection.The method comprises the steps that data of partial discharge maps in different areas are extracted to construct feature vectors, global statistics are calculated, and the feature vectors of the different areas are standardized through the global statistics; the divergence value is judged to trigger sensor calibration, and power equipment fault primary judgment is carried out through calibrated feature data; analyzing the fault through a physical model, and judging the fault type of the power equipment; calculating a real-time insulation degradation index of the power equipment, and uploading the real-time insulation degradation index to the cloud; aggregating eigenvector statistics in all areas, and updating global statistics to form a closed loop; and the cloud end formulates a dynamic threshold adjustment rule according to the received real-time insulation degradation index and the insulation degradation index, and updates an insulation degradation index threshold by using the dynamic threshold adjustment rule.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and specifically to a method and system for diagnosing faults of power equipment based on GIS partial discharge patterns. Background Art

[0002] Power equipment faces the dual challenges of safe operation and optimization transformation. Gas insulated switchgear (GIS), as the main substation equipment of the power grid, its safe operation is an important foundation for ensuring the reliable operation of the power system. GIS has the advantages of small occupied space, high operation reliability, long maintenance cycle, convenient transportation and installation, etc. However, once a fault occurs in GIS, it may lead to a large-area power outage of the system, causing significant economic losses. With the development of sensor technology, various signals generated by partial discharge can be measured more precisely, such as ultra-high frequency (UHF) signals, etc. By processing and analyzing these signals, partial discharge patterns are drawn, such as phase-resolved partial discharge patterns (PRPD), etc. Researchers have begun to attempt to identify different types of partial discharge faults by observing and analyzing the characteristics of these patterns, such as distinguishing tip discharge, particle discharge, floating discharge and air gap discharge, etc.

[0003] And in the current detection and diagnosis of partial discharge of power equipment, due to the large working area of power equipment, data in multiple regions cannot be shared, resulting in the diagnosis model in each place can only be based on local data and cannot utilize information in other regions, thus affecting the accuracy and generalization ability of the model. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for diagnosing faults of power equipment based on GIS partial discharge patterns to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: A method for diagnosing faults of power equipment based on GIS partial discharge patterns, the method comprising the following steps: S100. Count all regions where the power equipment works, collect partial discharge patterns in different regions, extract data of partial discharge patterns in different regions to construct feature vectors and calculate global statistics, and standardize the feature vectors in different regions using the global statistics; Further, the specific steps for standardizing the feature data in different regions using the global mean are as follows: S101. Count all regions where the power equipment works, collect partial discharge patterns in different regions, and select data related to discharge physical quantities in the partial discharge patterns in different regions, including discharge pulse voltage V p , pulse density D per unit time p , pulse rise time T rand the average discharge charge Q avg ; Using four types of data to construct a feature vector F = [V p , D p , T r , Q avg within different regions, calculating local feature vector statistics within each region, where the statistics include the mean and standard deviation; aggregating the statistics of different regions to obtain global feature vector statistics; S102. Standardize the feature data of different regions extracted using the global feature vector statistics. The formula is: ;

[0006] In the formula, F i k ’ represents the i-th feature vector in the k-th region after standardization, F i k represents the i-th feature vector in the k-th region, u k represents the mean of the feature vectors in the k-th region, σ k represents the standard deviation of the feature vectors in the k-th region, σ global represents the global standard deviation, u global represents the global mean.

[0007] By statistically analyzing all regions where the power equipment operates and collecting partial discharge patterns in different regions, the operating state information of the equipment can be comprehensively obtained. Extracting pattern data to construct feature vectors and calculating global statistics helps convert complex discharge pattern information into quantifiable and analyzable feature data, providing a basis for subsequent analysis and judgment. Standardizing the feature vectors using global statistics can eliminate the influence of the data dimension and scale of different regions, making the data of different regions comparable and improving the accuracy of subsequent analysis and calculation.

[0008] S200. Calculate the divergence value of the phase-resolved partial discharge (PRPD) distribution of different regions and the global after standardization, judge the divergence value to trigger sensor calibration, and perform a preliminary judgment on the power equipment failure through the calibrated feature data; Furthermore, the specific steps for performing a preliminary judgment on the power equipment failure through the calibrated feature data are as follows: S201. Judge regional faults by collecting data in real time according to the time window. In different regions, statistically calculate the regional probability distribution of the local PRPD distribution according to the phase interval. The formula is: regional probability distribution = number of pulses in the phase interval of the region / total number of pulses in the region. Weight the probability distribution of each region in the cloud and take the average to generate the global probability distribution, where the weight value is set manually; calculate the divergence value of the regional and global PRPD distributions using the regional probability distribution and the global probability distribution. The formula is: ;

[0009] In the formula, D JS k represents the divergence value between region k and the global PRPD distribution, and P k represents the probability distribution of region k, and P global represents the global probability distribution; M represents the average probability distribution between region k and the global, and D KL represents extracting the DL divergence within the parentheses; the calculation formula for M is: ; S202. When D JS k < 0.1, it is determined that the PRPD distributions of region k and the global are aligned, and the sensor does not need to be calibrated. When D JS k ≥0.1, it is determined that the PRPD distributions of region k and the global are not aligned, and the sensor calibration is started; Calculating the divergence value of the PRPD distributions of different regions and the global after standardization can measure the difference degree between the partial discharge patterns of different regions and the global pattern. By judging the divergence value to trigger sensor calibration, the accuracy and reliability of the collected data can be ensured, providing a more accurate basis for the preliminary judgment of power equipment faults.

[0010] S203. Extract the mean and standard deviation of the discharge pulse voltage under the current global normal working state from the standardized feature vector statistics for inspection, and calculate the regional outlier value. The formula is: ; In the formula, Z represents the regional outlier value, n represents the number of pulse sampling points within the current time window, represents the mean of the standardized discharge pulse voltage within the current time window, σVnormal represents the standard deviation of the discharge pulse voltage under the global normal working state, and u V normal represents the mean of the discharge pulse voltage under the global normal working state; S203. Set the regional outlier threshold Zy using the normal distribution. When |Z| > Zy, it is determined that the region has a fault and is marked. The same judgment is made for all regions, and the fault marks are output.

[0011] The output result directly drives the subsequent fault type recognition (S300) and global optimization (S400 - S500), forming a closed-loop diagnosis process.

[0012] Based on the calibrated feature data for preliminary fault judgment, potential fault hazards can be detected in a timely manner before obvious faults occur in power equipment, realizing early warning, which helps to take maintenance measures in advance and avoid equipment damage or power outage accidents caused by the further development of faults.

[0013] S300. After initially determining a power equipment fault, analyze the fault through a physical model to determine the type of power equipment fault; calculate the real-time insulation deterioration index of the power equipment and upload the real-time insulation deterioration index to the cloud. Further, the specific steps for calculating the real-time insulation deterioration index of the power equipment are as follows: S301. Construct the physical model of the power equipment as: ; In the formula, V p represents the discharge pulse voltage, E represents the electric field strength at the fault location, h represents the material constant, τ represents the ionization time constant, t r represents the pulse rise time, and d represents the size of the fault location; S302. Extract the discharge pulse voltage, material constant, ionization time constant, pulse rise time, and the size of the abnormal location within the determined fault area and input them into the physical model to calculate the electric field strength at the fault location. According to the characteristics of different partial discharge types in the field of power equipment, compare the calculated real-time fault electric field strength with the characteristics of different partial discharge types to obtain the fault type of the real-time fault area. After initially determining a power equipment fault, analyzing the fault through a physical model can deeply explore the causes and mechanisms of the fault, accurately determine the fault type of the power equipment, and provide strong support for formulating a targeted maintenance plan.

[0014] S303. Construct the insulation deterioration index model as: ; In the model, S(t) represents the insulation deterioration index, V p j represents the jth discharge pulse voltage, Q j represents the charge of the jth discharge, E a represents the activation energy of the insulating material, k B represents the Boltzmann constant, T(t) represents the real-time temperature of the equipment, and N represents the number of discharges; set the fault levels to S th1 and S th2 ; use the fault level to judge the real-time fault degree. When S(t) < S th1 , judge it as a minor fault. When S th1 ≤S(t) < S th2 , judge it as a moderate fault. When S(t) ≥ S th2 , judge it as a severe fault; combine the fault type and the fault degree to obtain the comprehensive fault type and upload the real-time insulation deterioration index to the cloud.

[0015] Calculating the real-time insulation deterioration index of power equipment and uploading it to the cloud can reflect the change of the equipment's insulation performance in real time. Maintenance personnel can timely understand the insulation status of the equipment based on the insulation deterioration index, providing an important reference for the maintenance and overhaul of the equipment and ensuring the safe and reliable operation of the equipment.

[0016] S400. Aggregate the eigenvector statistics in all regions and update the global statistics to form a closed loop; Furthermore, the specific steps for updating the global statistics to form a closed loop are as follows: S401. Extract the eigenvector statistics in different regions for aggregation and update the global statistics. The formula for the global eigenvector mean is: ; In the formula, u global represents the global eigenvector mean, A k represents the number of samples in region k, u k represents the eigenvector mean of region k, and K represents the total number of working regions of the power equipment; S402. Calculate the within-group variance in a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global eigenvector variance. The formula is: ; ; ; In the formula, (σ in ) 2 represents the within-group variance in a single region, (σ be ) 2 represents the between-group variance between different regions, (σ global ) 2 represents the global eigenvector variance; u k represents the eigenvector average in the kth region; calculate the global eigenvector statistics according to the time series to assign values and update the global eigenvector statistics in S102 to form a closed loop; S403. Extract the normal discharge records in all regions, extract all the data related to the physical model in the records, calculate the material constant in each record, and obtain the mean value of the material constants in all records. Take the mean value as the new material constant h new Update the material constants of the physical models in all regions.

[0017] Aggregating the eigenvector statistics in all regions and updating the global statistics to form a closed loop can enable the system to continuously self-adjust and optimize according to newly collected data. This can timely reflect the change of the operation state of power equipment, improve the system's monitoring and diagnostic capabilities for equipment faults, and form a dynamic and adaptive monitoring system.

[0018] After the S500 and the cloud receive the real-time insulation deterioration index, they formulate a dynamic threshold adjustment rule based on the insulation deterioration index, and use the dynamic threshold adjustment rule to update the insulation deterioration index threshold.

[0019] Furthermore, the specific steps for updating the insulation deterioration index threshold using the dynamic threshold adjustment rule are as follows: S501. After the cloud receives the insulation deterioration indexes in different regions, it calculates the global insulation deterioration index G using the cumulative distribution function global , and calculates the fault level using the global insulation deterioration index. The formula is: ; in the formula, S th1 global and S th2 global respectively represent the global fault levels; S502. Localize the calculated global fault level for each region. The formula is: ; in the formula, S th1 k and S th2 k represent the fault level after localization in the kth region; α represents the adjustment coefficient, which is set manually; use the fault level after localization to assign values and update the fault levels in different regions in the S303 to form a closed loop.

[0020] The cloud formulates a dynamic threshold adjustment rule based on the received real-time insulation deterioration index and updates the insulation deterioration index threshold, which can flexibly adjust the threshold for fault judgment according to the actual operating conditions of the equipment and the change of insulation performance. It avoids the problems of misjudgment or missed judgment that may be caused by a fixed threshold, and improves the accuracy and reliability of fault diagnosis.

[0021] A power equipment fault diagnosis system based on a GIS partial discharge pattern. The power equipment fault diagnosis system includes a data acquisition module, a data preprocessing module, a fault preliminary judgment module, a fault type judgment module, a global data update module, and a dynamic threshold update module; The data acquisition module is used to acquire partial discharge patterns in different regions, select data related to discharge physical quantities in the partial discharge patterns in different regions, and construct a feature vector; The data preprocessing module is used to calculate the feature vector statistics of different regions for the feature vector and perform standardization; The fault preliminary judgment module is used to calculate the divergence value of the standardized PRPD distribution of different regions and the global phase-resolved partial discharge map, judge the divergence value to trigger sensor calibration, and perform a preliminary judgment on the power equipment fault through the calibrated feature data. The fault type judgment module is used to analyze the fault through a physical model after initially judging the power equipment fault, judge the fault type of the power equipment; calculate the real-time insulation deterioration index of the power equipment, and upload the real-time insulation deterioration index to the cloud; The global data update module is used to aggregate the feature vector statistics in all regions and update the global statistics to form a closed loop; The dynamic threshold update module is used for the cloud to formulate a dynamic threshold adjustment rule according to the received real-time insulation deterioration index after receiving it, and use the dynamic threshold adjustment rule to update the insulation deterioration index threshold.

[0022] The fault initial judgment module includes a PRPD distribution alignment unit and a fault marking unit; The PRPD distribution alignment unit is used to calculate the divergence value of the regional and global PRPD distributions using the regional probability distribution and the global probability distribution, and judge whether the sensor needs to be calibrated; The fault marking unit is used to extract the mean and standard deviation of the discharge pulse voltage in the current global normal working state from the standardized feature vector statistics for inspection, calculate the regional outlier, and judge whether there is a fault in the region and mark it.

[0023] The fault type judgment module includes a fault type unit and a fault degree unit; The fault type unit is used to calculate the electric field strength at the fault location using a physical model, obtain the characteristics of different partial discharge types according to the professional index in the field of power equipment, and judge the real-time fault type according to the characteristics of different partial discharge types; The fault degree unit is used to calculate the insulation deterioration index at the fault location and use the fault level to judge the fault degree of the real-time fault location.

[0024] The global data update module includes a global mean unit and a global variance unit; The global mean unit is used to extract the feature vector statistics in different regions for aggregation and update the global statistics; The global variance unit is used to calculate the within-group variance in a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global feature vector variance.

[0025] Compared with the prior art, the beneficial effects of the present invention are: 1. Through statistical distribution alignment and global parameter aggregation, each region diagnoses based on a unified benchmark, solving the data island problem. And the global parameter aggregation does not require the exchange of private data such as device data between different regions, protecting data security while solving the data island problem.

[0026] 2. In the present invention, the dynamic threshold is adjusted according to the global aging index distribution, and combined with the regional mutual verification mechanism to filter false alarms, greatly reducing the probability of false alarms of power equipment failures.

[0027] 3. When adjusting the local fault level in the present invention, by calculating the global fault level and making local adjustments to the local fault levels in different regions based on the global fault level, not only the comprehensiveness of the fault level in each region is increased to ensure that there are no unique anomalies in the local area of the region, but also the adjustment is made in line with the characteristics of different regions, making the judgment more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a module distribution diagram of the power equipment fault diagnosis system based on the GIS partial discharge pattern in the present invention; Figure 2 It is a schematic diagram of the steps of the power equipment fault diagnosis method based on the GIS partial discharge pattern in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0030] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution, A power equipment fault diagnosis method based on a GIS partial discharge pattern, the method comprising the following steps: S100. Statistically analyze all regions where the power equipment operates, collect partial discharge patterns in different regions, extract data from the partial discharge patterns in different regions to construct a feature vector and calculate global statistics, and standardize the feature vectors in different regions using the global statistics; The specific steps for standardizing the characteristic data in different regions using the global mean are: S101. Statistically analyze all regions where the power equipment operates, collect partial discharge patterns in different regions, select data related to the discharge physical quantity in the partial discharge patterns in different regions, including the discharge pulse voltage V p , the pulse density D per unit time p , the pulse rise time T r and the average discharge charge amount Q avg ; use the four types of data to construct a feature vector F = [V p , D p , T r , Qavg , local feature vector statistics are calculated within each region, and the statistics include the mean value and the standard deviation; the global feature vector statistics are obtained by aggregating the statistics of different regions; S102. Standardize the feature data of different regions extracted using the global feature vector statistics. The formula is: ; In the formula, F i k ’ represents the i-th feature vector in the k-th region after standardization, and F i k represents the i-th feature vector in the k-th region, u k represents the mean value of the feature vectors in the k-th region, σ k represents the standard deviation of the feature vectors in the k-th region, σ global represents the global standard deviation, and u global represents the global mean value.

[0031] By statistically analyzing all regions where the power equipment operates and collecting partial discharge patterns in different regions, the operating state information of the equipment can be comprehensively obtained. Extracting pattern data to construct feature vectors and calculating global statistics helps to convert complex discharge pattern information into quantifiable and analyzable feature data, providing a basis for subsequent analysis and judgment. Standardizing the feature vectors using global statistics can eliminate the influence of the data dimension and scale of different regions, making the data of different regions comparable and improving the accuracy of subsequent analysis and calculation.

[0032] S200. Calculate the divergence value of the phase-resolved partial discharge map (PRPD) distribution of different regions and the global one after standardization, judge the divergence value to trigger sensor calibration, and perform a preliminary fault judgment on the power equipment using the calibrated feature data; The specific steps for performing a preliminary fault judgment on the power equipment using the calibrated feature data are as follows: S201. Judge regional faults by collecting data in real time according to the time window. In different regions, statistically calculate the regional probability distribution of the local PRPD distribution according to the phase interval. The formula is: regional probability distribution = number of pulses in the region within the phase interval / total number of pulses in the region. Weight the probability distribution of each region in the cloud and take the average to generate the global probability distribution. The weight value is set manually; calculate the divergence value of the regional and global PRPD distributions using the regional probability distribution and the global probability distribution. The formula is: ; In the formula, D JS k represents the divergence value of the PRPD distributions of region k and the global one, P k represents the probability distribution of region k, and P globalrepresents the global probability distribution; M represents the average probability distribution between region k and the global, D KL represents extracting the DL divergence within the brackets; the calculation formula for M is: ; S202. When D JS k < 0.1, it is determined that the PRPD distributions of region k and the global are aligned, and the sensor does not need to be calibrated. When D JS k ≥ 0.1, it is determined that the PRPD distributions of region k and the global are not aligned, and the sensor calibration is started; Calculating the divergence value of the PRPD distributions of different regions and the global after standardization can measure the difference degree between the partial discharge patterns of different regions and the global pattern. By judging the divergence value to trigger sensor calibration, the accuracy and reliability of the collected data can be ensured, providing a more accurate basis for the initial judgment of power equipment faults.

[0033] S203. Extract the mean and standard deviation of the discharge pulse voltage in the current global normal working state from the standardized eigenvector statistics for inspection, and calculate the regional outlier. The formula is: ; In the formula, Z represents the regional outlier, n represents the number of pulse sampling points in the current time window, represents the mean of the standardized discharge pulse voltage in the current time window, σ V normal represents the standard deviation of the discharge pulse voltage in the global normal working state, u V normal represents the mean of the discharge pulse voltage in the global normal working state; S203. Set the regional outlier threshold Zy using the normal distribution. When |Z| > Zy, it is determined that there is a fault in the region and it is marked. The same judgment is made for all regions, and the fault marks are output.

[0034] The output result directly drives the subsequent fault type identification (S300) and global optimization (S400 - S500), forming a closed-loop diagnosis process.

[0035] Based on the calibrated characteristic data for initial fault judgment, potential fault hazards can be detected in a timely manner before obvious faults occur in power equipment, realizing early warning, which helps to take maintenance measures in advance and avoid equipment damage or power outage accidents caused by the further development of faults.

[0036] S300. After initially judging the power equipment fault, analyze the fault through a physical model to determine the power equipment fault type; calculate the real-time insulation deterioration index of the power equipment and upload the real-time insulation deterioration index to the cloud; The specific steps for calculating the real-time insulation deterioration index of electrical equipment are as follows: S301. Construct the physical model of the electrical equipment as: ; In the formula, V p represents the discharge pulse voltage, E represents the electric field strength at the fault location, h represents the material constant, τ represents the ionization time constant, t r represents the pulse rise time, and d represents the size of the fault location; S302. Extract the discharge pulse voltage, material constant, ionization time constant, pulse rise time, and the size of the abnormal location within the judged fault area and input them into the physical model to calculate the electric field strength at the fault location. According to the characteristics of different partial discharge types in the field of electrical equipment, compare the calculated real-time fault electric field strength with the characteristics of different partial discharge types to obtain the fault type of the real-time fault area; After initially judging the fault of the electrical equipment, analyzing the fault through the physical model can deeply explore the causes and mechanisms of the fault, accurately judge the fault type of the electrical equipment, and provide strong support for formulating a targeted maintenance plan.

[0037] S303. Construct the insulation deterioration index model as: ; In the model, S(t) represents the insulation deterioration index, V p j represents the j-th discharge pulse voltage, Q j represents the charge of the j-th discharge, E a represents the activation energy of the insulating material, k B represents the Boltzmann constant, T(t) represents the real-time temperature of the equipment, and N represents the number of discharges; set the fault levels as S th1 and S th2 ; use the fault level to judge the real-time fault degree. When S(t) < S th1 , judge it as a minor fault. When S th1 ≤S(t) < S th2 , judge it as a moderate fault. When S(t) ≥ S th2 , judge it as a serious fault; combine the fault type and the fault degree to obtain the comprehensive fault type, and upload the real-time insulation deterioration index to the cloud.

[0038] Calculating the real-time insulation deterioration index of electrical equipment and uploading it to the cloud can reflect the change of the equipment's insulation performance in real time. The operation and maintenance personnel can understand the insulation status of the equipment in time according to the insulation deterioration index, provide an important reference for the maintenance and repair of the equipment, and ensure the safe and reliable operation of the equipment.

[0039] S400. Aggregate the eigenvector statistics in all regions, update the global statistics to form a closed loop; The specific steps for forming a closed loop by updating the global statistics are as follows: S401. Extract the eigenvector statistics in different regions for aggregation to update the global statistics. The formula for the global eigenvector mean is: ; In the formula, u global represents the global eigenvector mean, A k represents the number of samples in region k, u k represents the eigenvector mean of region k, and K represents the total number of working regions of the power equipment; S402. Calculate the within-group variance within a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global eigenvector variance. The formula is: ; ; ; In the formula, (σ in ) 2 represents the within-group variance within a single region, (σ be ) 2 represents the between-group variance between different regions, and (σ global ) 2 represents the global eigenvector variance; u k represents the average value of the eigenvectors in the kth region; calculate the global eigenvector statistics according to the time series to assign values to and update the global eigenvector statistics in S102 to form a closed loop; S403. Extract the normal discharge records in all regions, extract all the data related to the physical model in the records, calculate the material constant for each record, find the average value of the material constants of all records, and use the average value as the new material constant h new to update the material constants of the physical models in all regions.

[0040] Aggregating the eigenvector statistics in all regions and updating the global statistics to form a closed loop can enable the system to continuously self-adjust and optimize according to newly acquired data. This can timely reflect the changes in the operating state of the power equipment, improve the system's monitoring and diagnostic capabilities for equipment failures, and form a dynamic and adaptive monitoring system.

[0041] S500. After the cloud receives the real-time insulation degradation index, formulate a dynamic threshold adjustment rule based on the insulation degradation index, and use the dynamic threshold adjustment rule to update the insulation degradation index threshold.

[0042] The specific steps for updating the insulation degradation index threshold using the dynamic threshold adjustment rule are as follows: S501. After the cloud receives the insulation degradation indices in different regions, it calculates the global insulation degradation index G using the cumulative distribution function, and calculates the fault level using the global insulation degradation index. The formula is: global , where: ; In the formula, S th1 global and S th2 global respectively represent the global fault levels; S502. Localize the calculated global fault level for each region. The formula is: ; In the formula, S th1 k and S th2 k represent the fault levels after localization in the k-th region; α represents the adjustment coefficient, which is set manually. Assign and update the fault levels in different regions in S303 using the localized fault levels to form a closed loop. The cloud formulates a dynamic threshold adjustment rule based on the received real-time insulation degradation index and updates the insulation degradation index threshold, which can flexibly adjust the fault judgment threshold according to the actual operating conditions and insulation performance changes of the equipment. This avoids misjudgment or missed judgment problems that may be caused by fixed thresholds and improves the accuracy and reliability of fault diagnosis.

[0043] A power equipment fault diagnosis system based on GIS partial discharge patterns. The power equipment fault diagnosis system includes a data acquisition module, a data preprocessing module, a fault preliminary judgment module, a fault type judgment module, a global data update module, and a dynamic threshold update module; The data acquisition module is used to acquire partial discharge patterns in different regions, select data related to discharge physical quantities in the partial discharge patterns in different regions, and construct feature vectors; The data preprocessing module is used to calculate the feature vector statistics of different regions for the feature vectors and perform standardization; The fault preliminary judgment module is used to calculate the divergence values of the normalized PRPD distributions of different regions and the global phase-resolved partial discharge patterns, judge the divergence values to trigger sensor calibration, and perform a preliminary judgment of power equipment faults using the calibrated feature data; The fault type judgment module is used to analyze the fault through a physical model after initially judging a power equipment fault, judge the power equipment fault type; calculate the real-time insulation degradation index of the power equipment and upload the real-time insulation degradation index to the cloud; The global data update module is used to aggregate the feature vector statistics in all regions, update the global statistics to form a closed loop; The dynamic threshold update module is used for the cloud to formulate a dynamic threshold adjustment rule according to the received real-time insulation degradation index, and use the dynamic threshold adjustment rule to update the insulation degradation index threshold.

[0044] The initial fault judgment module includes a PRPD distribution alignment unit and a fault marking unit; The PRPD distribution alignment unit is used to calculate the divergence value between the regional and global PRPD distributions using the regional probability distribution and the global probability distribution, and judge whether the sensor needs to be calibrated; The fault marking unit is used to extract the mean and standard deviation of the discharge pulse voltage under the current global normal working state from the standardized feature vector statistics for inspection, calculate the regional outliers, and judge whether there is a fault in the region and mark it.

[0045] The fault type judgment module includes a fault type unit and a fault degree unit; The fault type unit is used to calculate the electric field strength at the fault location using a physical model, obtain the characteristics of different partial discharge types according to the professional indexes in the field of power equipment, and judge the real-time fault type according to the characteristics of different partial discharge types; The fault degree unit is used to calculate the insulation degradation index at the fault location and use the fault level to judge the fault degree of the real-time fault location.

[0046] The global data update module includes a global mean unit and a global variance unit; The global mean unit is used to extract the feature vector statistics in different regions for aggregation to update the global statistics; The global variance unit is used to calculate the within-group variance within a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global feature vector variance.

[0047] Example 1: When initially judging the fault of partial discharge of power equipment, the characteristics of different partial discharge types are:

[0048] Let the data input into the physical model and the insulation degradation model be: V p = 70mv, t r = 1.5 μs, T(t) = 313K, h = 1.2, τ = 0.2 μs, d = 0.05mm, Q = 50; In the physical model, E = 12kv / mm is calculated, and the calculated insulation degradation index S(t) = 1500; The initial judgment of the fault type is surface discharge; Let the fault levels be S th1 = 1000, S th2= 2500; Determine that the degree of the fault is a medium fault; Finally, the fault type is obtained as "surface discharge - medium fault".

[0049] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A power equipment fault diagnosis method based on GIS partial discharge patterns, characterized in that: The method includes the following steps: S100. Statistically analyze all regions where the power equipment operates, collect partial discharge patterns in different regions, extract data from the partial discharge patterns in different regions to construct feature vectors and calculate global statistics, and standardize the feature vectors in different regions using the global statistics; S200. Calculate the divergence values of the PRPD (Phase Resolved Partial Discharge) distributions of different regions and the global phase after standardization, judge and trigger sensor calibration based on the divergence values, and preliminarily judge the faults of the power equipment through the calibrated feature data; S300. After preliminarily judging the faults of the power equipment, analyze the faults through a physical model to determine the types of power equipment faults; calculate the real-time insulation deterioration index of the power equipment and upload the real-time insulation deterioration index to the cloud; S400. Aggregate the feature vector statistics in all regions and update the global statistics to form a closed loop; S500. After the cloud receives the real-time insulation deterioration index, formulate a dynamic threshold adjustment rule based on the insulation deterioration index, and use the dynamic threshold adjustment rule to update the threshold of the insulation deterioration index.

2. The power equipment fault diagnosis method based on the GIS partial discharge pattern according to claim 1, wherein: The specific steps of standardizing the feature data in different regions using the global mean in S100 are as follows: S101. Statistically analyze all regions where the power equipment operates, collect partial discharge patterns in different regions, and select data related to discharge physical quantities from the partial discharge patterns in different regions, including discharge pulse voltage V p , pulse density D per unit time p , pulse rise time T r and average discharge charge Q avg ; Use these four types of data to construct a feature vector F = [V p , D p , T r , Q avg in different regions, calculate the local feature vector statistics in each region, and the statistics include mean value and standard deviation; Aggregate the statistics of different regions to obtain the global feature vector statistics; S102. Standardize the feature data in different regions extracted using the global feature vector statistics. The formula is: ； In the formula, F i k ' represents the i-th eigenvector in the k-th region after standardization, and F i k represents the i-th eigenvector in the k-th region, u k represents the average value of the eigenvectors in the k-th region, and σ k represents the standard deviation of the eigenvectors in the k-th region, and σ global represents the global standard deviation, and u global represents the global average value.

3. The power equipment fault diagnosis method based on the GIS partial discharge pattern according to claim 2, characterized in that: The specific steps of preliminarily judging the faults of the power equipment through the calibrated feature data in S200 are as follows: S201. Judge regional faults by collecting data in real time according to the time window. In different regions, statistically analyze the regional probability distribution of the local PRPD distribution according to the phase interval. The formula is: regional probability distribution = number of pulses in the phase interval of the region / total number of pulses in the region. Weight the probability distribution of each region in the cloud and take the average to generate the global probability distribution. The weight value is set manually; calculate the divergence value of the regional and global PRPD distributions using the regional probability distribution and the global probability distribution. The formula is: ； In the formula, D JS k represents the divergence value between region k and the global PRPD distribution, and P k represents the probability distribution of region k, and P global represents the global probability distribution; M represents the average probability distribution between region k and the global, and D KL represents extracting the DL divergence within the parentheses; the calculation formula for M is: ; S202. When D JS k < 0.1, it is determined that the PRPD distributions of region k and the global are aligned, and the sensor does not need to be calibrated. When D JS k ≥ 0.1, it is determined that the PRPD distributions of region k and the global are not aligned, and sensor calibration is initiated; S203. Extract the mean and standard deviation of the discharge pulse voltage in the current globally normal operating state from the standardized feature vector statistics for inspection, and calculate the regional outlier value. The formula is: ； In the formula, Z represents the regional outlier, n represents the number of pulse sampling points within the current time window, and `V p represents the mean value of the normalized discharge pulse voltage within the current time window, and σ V normal represents the standard deviation of the discharge pulse voltage under the global normal working state, and u V normal represents the mean value of the discharge pulse voltage under the global normal working state; S203. Set the regional anomaly threshold Zy using the normal distribution. When |Z| > Zy, judge that the region has a fault and mark it. Make the same judgment for all regions and output the fault mark.

4. The power equipment fault diagnosis method based on the GIS partial discharge pattern according to claim 3, characterized in that: The specific steps of calculating the real-time insulation deterioration index of the power equipment in S300 are as follows: S301. Construct the physical model of the power equipment as: ； In the formula, V p represents the discharge pulse voltage, E represents the electric field strength at the fault, h represents the material constant, τ represents the ionization time constant, and t r represents the pulse rise time, and d represents the size at the fault; S302. Extract the discharge pulse voltage, material constant, ionization time constant, pulse rise time, and size at the abnormal location in the judged fault region and input them into the physical model to calculate the electric field strength at the fault location. According to the characteristics of different partial discharge types in the power equipment field, compare the calculated real-time fault electric field strength with the characteristics of different partial discharge types to obtain the fault type of the real-time fault region; S303. Construct the insulation deterioration index model as: ； In the model, S(t) represents the insulation degradation index, V p j represents the voltage of the j-th discharge pulse, Q j represents the charge of the j-th discharge, E a represents the activation energy of the insulating material, k B represents the Boltzmann constant, T(t) represents the real-time temperature of the device, and N represents the number of discharges; the fault levels are set to be S th1 and S th2 ; Judge the real-time fault degree by using the fault level. When S(t) < S th1 , judge it as a minor fault. When S th1 ≤ S(t) < S th2 , judge it as a moderate fault. When S(t) ≥ S th2 , judge it as a serious fault; Obtain the comprehensive fault type by combining the fault type and the fault degree, and upload the real-time insulation deterioration index to the cloud.

5. The power equipment fault diagnosis method based on the GIS partial discharge pattern according to claim 4, characterized in that: The specific steps of updating the global statistics to form a closed loop in S400 are as follows: S401. Extract the eigenvector statistics in different regions for aggregation to update the global statistics. The formula for the global eigenvector mean is as follows: ； In the formula, u global represents the mean of the global feature vectors, A k represents the number of samples in region k, u k represents the mean of the feature vectors in region k, and K represents the total number of working regions of the power equipment; S402. Calculate the within-group variance within a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global eigenvector variance. The formula is as follows: ； ； ； In the formula, (σ in ) 2 represents the within-group variance within a single region, (σ be ) 2 represents the between-group variance between different regions, (σ global ) 2 represents the global feature vector variance; u k represents the average value of the feature vectors in the k-th region; calculate the global feature vector statistic according to the time series to assign and update the global feature vector statistic in S102 to form a closed loop; S403. Extract the normal discharge records in all regions, extract all the data related to the physical model in the records, calculate the material constants for each record, obtain the mean value of the material constants of all records, and use the mean value as the new material constant h. new Update the material constants of the physical models for all regions.

6. The power equipment fault diagnosis method based on the GIS partial discharge pattern according to claim 5, characterized in that: The specific steps for updating the insulation degradation index threshold using the dynamic threshold adjustment rule in S500 are as follows: S501. After the cloud receives the insulation degradation indices in different regions, the global insulation degradation index G is calculated using the cumulative distribution function. global , and the fault level is calculated using the global insulation degradation index. The formula is as follows: ； In the formula, S th1 global and S th2 global respectively represent the global failure level; S502. Localize the calculated global fault level for each region, with the formula: ; In the formula, S th1 k and S th2 k represent the fault levels after localization of the k-th region; α represents an adjustment coefficient, which is set manually; the fault levels in different regions in the S303 are assigned and updated using the fault levels after localization to form a closed loop.

7. A power equipment fault diagnosis system based on GIS partial discharge patterns, characterized in that: The power equipment fault diagnosis system includes a data acquisition module, a data preprocessing module, a fault preliminary judgment module, a fault type judgment module, a global data update module, and a dynamic threshold update module; The data acquisition module is used to collect the partial discharge patterns in different regions, select the data related to the discharge physical quantity in the partial discharge patterns in different regions, and construct eigenvectors; The data preprocessing module is used to calculate the eigenvector statistics in different regions for the eigenvectors and perform standardization; The fault preliminary judgment module is used to calculate the divergence value of the PRPD distribution of different regions and the global phase-resolved partial discharge map (PRPD) after standardization, judge to trigger sensor calibration, and perform a preliminary judgment on the power equipment fault through the calibrated characteristic data; The fault type judgment module is used to analyze the fault through a physical model after initially judging the power equipment fault, and judge the power equipment fault type; calculate the real-time insulation degradation index of the power equipment and upload the real-time insulation degradation index to the cloud; The global data update module is used to aggregate the eigenvector statistics in all regions and update the global statistics to form a closed loop; The dynamic threshold update module is used for the cloud to formulate a dynamic threshold adjustment rule according to the received real-time insulation degradation index, and update the insulation degradation index threshold using the dynamic threshold adjustment rule.

8. The power equipment fault diagnosis system based on the GIS partial discharge pattern according to claim 7, characterized in that: The fault preliminary judgment module includes a PRPD distribution alignment unit and a fault marking unit; The PRPD distribution alignment unit is used to calculate the divergence value of the regional and global PRPD distributions using the regional probability distribution and the global probability distribution, and judge whether the sensor needs to be calibrated; The fault marking unit is used to extract the mean and standard deviation of the discharge pulse voltage under the current global normal working state from the standardized eigenvector statistics for inspection, calculate the regional outliers, and judge whether there is a fault in the region and mark it.

9. The power equipment fault diagnosis system based on the GIS partial discharge pattern according to claim 7, characterized in that: The fault type judgment module includes a fault type unit and a fault degree unit; The fault type unit is used to calculate the electric field strength at the fault using a physical model, obtain the characteristics of different partial discharge types according to the professional indexes in the field of power equipment, and judge the real-time fault type according to the characteristics of different partial discharge types; The fault degree unit is used to calculate the insulation degradation index at the fault and use the fault level to judge the fault degree at the real-time fault location.

10. The power equipment fault diagnosis system based on the GIS partial discharge pattern according to claim 7, characterized in that: The global data update module includes a global mean unit and a global variance unit; The global mean unit is used to extract the eigenvector statistics in different regions for aggregation to update the global statistics; The global variance unit is used to calculate the within-group variance within a single region and the between-group variance between different regions respectively, and combine the two variances to obtain the global eigenvector variance.

Images