A Transformer Fault Hierarchical Diagnosis Method Based on Extended Three-Ratio Method and Association Rules

Abstract

The present invention discloses a hierarchical diagnosis method for transformer faults based on the extended three-ratio method and association rules. First, based on on-line or off-line dissolved gas data in oil, the extended three-ratio method is used for primary diagnosis of transformer faults, for judging different degrees of overheating, discharge or other fault natures and their early warning; then, combined with the transformer fault tree, classification of fault natures and corresponding fault subsets is realized; on-line monitoring test characteristic quantities are extracted to exclude certain fault types in the fault subsets; finally, association rules are used to analyze the association relationship between fault types and fault characteristic quantities, so as to construct a refined diagnosis model for transformer faults and realize dynamic evaluation of fault natures, fault types and fault credibility. The present invention can overcome the problems that traditional transformer fault diagnosis methods rely too much on on-site test data, the accuracy of fault diagnosis results is not high and it is difficult to guide on-site operation and maintenance, etc., and can improve the scientificity and practicability of transformer fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer fault diagnosis, and particularly relates to a hierarchical diagnosis method for transformer faults based on an extended three-ratio method and association rules. Background Art

[0002] A transformer is one of the key devices in a power system, and its operating state directly affects the safe and stable operation of the power system. Most power operation and maintenance units collect various operation data of transformers through means such as on-line monitoring, off-line testing, and daily operation and maintenance to further understand and master their operating states. However, due to the large number and relative independence of the existing characteristic quantities used to characterize the operating state of transformers, the state evaluation and fault diagnosis of transformers often rely on the experience of operation and maintenance personnel, which is prone to misjudgment or missed judgment, resulting in the fault outage of transformers. Therefore, if accurate, rapid, and intelligent diagnosis of transformer faults can be achieved, it is of great significance for improving the safe and stable operation ability of transformers and reducing the operation and maintenance pressure of power units.

[0003] For the problem of transformer fault diagnosis, according to the different basic data used, transformer fault diagnosis methods can be roughly divided into the following two categories: (1) relying on dissolved gas analysis (DGA) in transformer oil, using artificial intelligence algorithms such as the three-ratio method, Duval triangle method, or neural network to identify internal overheating or discharge defects in transformers; (2) based on transformer operation and test data, using methods such as association rules and case-based reasoning to directly judge the transformer fault type. There are still many problems in the field application of the above methods, mainly including: (1) The transformer fault diagnosis method based on DGA is difficult to accurately locate the transformer fault location, resulting in overly conservative operation and maintenance decisions; (2) The transformer fault diagnosis method based on operation and test data is overly dependent on on-site data and is difficult to play a role when the on-site data is not sufficient; (3) The above methods still have certain limitations in special requirements or application scenarios such as the integrity of transformer fault types, real-time fault diagnosis, fault traceability, and multiple concurrent faults in the same device, which restricts their further popularization and application in the power system. Summary of the Invention

[0004] Aiming at the problems existing in the prior art, the purpose of the present invention is to provide a hierarchical diagnosis method for transformer faults based on an extended three-ratio method and association rules. This method can quickly diagnose the transformer fault type on the basis of a small amount of operation and maintenance data, realize the evaluation of transformer fault types and probabilities, and improve the scientificity and accuracy of transformer fault diagnosis.

[0005] To achieve the purpose of this invention, the present invention adopts the following technical solutions. A hierarchical diagnosis method for transformer faults based on an extended three-ratio method and association rules includes the following steps:

[0006] S1: Collect the on-line monitoring, off-line test, operation, maintenance and repair data of the diagnosed transformer; among them, at least one of the on-line monitoring or off-line test data should include the component type, content or gas production rate of the dissolved gases in the transformer oil.

[0007] S2: Make a prior judgment on the existence of faults based on the content or gas production rate of the dissolved gases in the oil, and preliminarily classify the internal faults of the transformer into one of the following three types of faults: overheating fault, discharge fault or other faults.

[0008] S3: For the internal discharge or overheating faults determined in the transformer, combined with the on-line or off-line dissolved gas data in the transformer oil, use the extended three-ratio method to complete the primary fault diagnosis of the transformer, and the diagnosis result is one of the following six fault natures: low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge or high-energy discharge.

[0009] S4: Based on the transformer fault tree, sort out the main fault types of the transformer, establish the corresponding relationship between the fault nature and the fault type of the transformer, obtain the fault subset corresponding to the above fault nature, extract the on-line monitoring or off-line test characteristic quantities, and further screen the fault types in the fault subset;

[0010] S5: Use association rule analysis to obtain the fault characteristic quantities and their corresponding weights corresponding to each fault type. Combined with the on-line monitoring, off-line test, operation, maintenance and repair data of the transformer, use the traversal method to conduct secondary fault diagnosis on the screened fault subset, and calculate the transformer fault type, fault location and corresponding fault credibility; if all fault credibilities are zero, output the diagnosis result as normal and end the process.

[0011] Furthermore, the dissolved gas sample data in the oil in S1 should include the contents of H2, CO, CO2 and the contents of four hydrocarbon gases, namely CH4, C2H6, C2H4 and C2H2.

[0012] Furthermore, the principle for judging the existence of internal faults of the transformer in S2 mainly includes taking the set of sample data closest to the current time as the standard. If the contents of each component gas simultaneously meet the following three conditions: H2 is less than 150 μL / L, C2H2 is less than 1 μL / L, and the total content of the four hydrocarbon gases is less than 150 μL / L, the diagnosis result is other faults. If the above conditions are not met, it is determined that there are discharge or overheating fault natures.

[0013] Furthermore, the extended three-ratio method in S3 specifically includes the following steps:

[0014] S3.1: If it is preliminarily determined that there is an overheating or discharge fault inside the transformer, calculate the three ratios corresponding to the sample data of dissolved gases in oil: CH4 / H2, C2H2 / C2H4, and C2H4 / C2H6, and encode the sample data using the following rules.

[0015]

[0016]

[0017] S3.2: According to the obtained coding combination, query the following fault type judgment criterion table to determine the nature of the transformer fault as low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge, or high-energy discharge.

[0018]

[0019] S3.3: Combining the correspondence between multiple dissolved gas samples in oil and the nature of transformer faults, summarize the determination attribution table for the missing code 011 as follows.

[0020]

[0021] Where: TG represents the total of four hydrocarbon gases, and M(*) represents the gas content in the parentheses.

[0022] Furthermore, the correspondence between the nature of the fault and the fault type described in S4 is

[0023]

[0024]

[0025] Furthermore, the extraction of characteristic quantities from on-line or off-line tests for fault type exclusion described in S4 specifically includes the following steps:

[0026] S4.1: When extracting on-line or off-line test characteristic quantities, it should include the CO2 gas production rate, the gas production rate ratio CO2 / CO, the C2H2 content, the total hydrocarbon content, the H2 content, the CH4 content, and the core grounding current.

[0027] S4.2: Combining the correspondence between the fault types and sample data of multiple faulty transformers, summarize the following fault exclusion criterion table.

[0028]

[0029] Furthermore, the use of association rule analysis to obtain the fault characteristic quantities and their weight coefficients corresponding to each fault type described in S5 specifically includes the following steps:

[0030] S5.1: Collect the on-line monitoring, off-line test, operation, maintenance and repair data of the diagnosed transformer, extract the state parameters that can reflect the transformer fault types, and construct the transformer fault feature set.

[0031] S5.2: Calculate the support and confidence of the fault feature quantities corresponding to each fault type. Use the fault cases of the transformer to construct the transaction database N. Denote the number of all non-empty fault subsets in N as |N|, the number of occurrences of fault type A in a certain subset as |X|, and the number of abnormal occurrences of feature quantity B as |Y|. Then the support of A→B is

[0032]

[0033] Set the support to be greater than 70%, and feature quantity B is the fault feature quantity of fault type A. The confidence of A→B is

[0034]

[0035] S5.3: After normalizing the confidence of each fault feature quantity in the fault type, the corresponding weight coefficient can be calculated.

[0036] Advantages of the present invention:

[0037] A transformer fault hierarchical diagnosis method based on the extended triple ratio and association rules provided by the present invention can combine the on-line monitoring, off-line test, operation, maintenance and repair data of the transformer, further refine the transformer fault diagnosis from the fault nature to the fault type, evaluate the corresponding fault location and fault credibility, and can overcome the problems that the traditional transformer fault diagnosis method relies too much on on-site test data, the accuracy of the fault diagnosis result is not high and it is difficult to guide on-site operation and maintenance, etc., and can improve the scientificity and practicability of the transformer fault diagnosis. Description of the drawings

[0038] Attached Figure 1 is the flow chart of the transformer fault hierarchical diagnosis method based on the extended triple ratio and association rules in the present invention.

[0039] Attached Figure 2 is the fault tree diagram of the transformer winding and lead in the present invention.

[0040] Attached Figure 3 is the fault result diagram of the verification case in the present invention; (a) medium voltage coil; (b) inter-turn short circuit point. Detailed implementation manners

[0041] The following further describes the present invention in conjunction with embodiments. It is necessary to point out here that the following embodiments are only used to further illustrate the present invention and should not be construed as limiting the protection scope of the present invention. Some non-essential improvements and adjustments made by those skilled in the art based on the above invention content still fall within the protection scope of the present invention.

[0042] As Figure 1 shown, the specific implementation manner of the present invention provides a hierarchical diagnosis method for transformer faults based on extended triple ratios and association rules, including the following steps:

[0043] S1: Collect on-line monitoring, off-line test and operation and maintenance data of the transformer to be diagnosed; among them, at least one of the on-line monitoring or off-line test data should include the contents of H2, CO, CO2 and the contents of four hydrocarbon gases, namely CH4, C2H6, C2H4, and C2H2, in the dissolved gas in the transformer oil.

[0044] S2: Based on the content of dissolved gas in the oil or the gas production rate, make a preliminary judgment on the existence of faults, and classify the internal faults of the transformer into one of the following three faults: overheating fault, discharge fault or other faults.

[0045] In this embodiment, the preliminary judgment on the existence of internal faults of the transformer mainly includes taking the set of sample data closest to the current time as the standard. If the contents of each component gas simultaneously meet the following three conditions: H2 is less than 150 μL / L, C2H2 is less than 1 μL / L, and the total content of the four hydrocarbon gases is less than 150 μL / L, the diagnosis result is other faults. If the above conditions are not met, it is determined that there is a discharge or overheating fault property;

[0046] S3: For the internal faults of the transformer determined to have a discharge or overheating fault, combine the on-line or off-line dissolved gas data in the transformer oil of the transformer, and use the extended triple ratio to complete the primary fault diagnosis of the transformer. The diagnosis result is one of the following six fault properties: low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge or high-energy discharge.

[0047] In this embodiment, the primary fault diagnosis of the transformer based on the extended triple ratio specifically includes the following steps:

[0048] S3.1: If it is preliminarily determined that there is an overheating or discharge fault inside the transformer, then calculate the three ratios corresponding to the sample data of the dissolved gas in the oil: CH4 / H2, C2H2 / C2H4, and C2H4 / C2H6, and use the following rules to encode the sample data.

[0049]

[0050] S3.2: According to the calculated coding combinations, query the following transformer fault type judgment criterion table to determine that the nature of the transformer fault is low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge or high-energy discharge.

[0051]

[0052]

[0053] S3.3: Combining the corresponding relationships between multiple dissolved gas samples in oil and the nature of transformer faults, summarize the determination attribution table for the missing code 011 as follows. Where TG represents total hydrocarbons, and M(*) represents the content of the gas in the parentheses.

[0054]

[0055] S4: Based on the transformer fault tree, sort out the main transformer fault types, establish the corresponding relationship between the nature of the transformer fault and the fault type, obtain the fault subset corresponding to the above-mentioned fault nature, extract the on-line monitoring test characteristic quantities, and further screen the fault types in the fault subset;

[0056] In this embodiment, according to the transformer structure and the functions of each group of components, it is divided into 6 groups of components: winding and lead wire, iron core, insulating oil, oil tank, tap changer, and bushing. As Figure 2 is the fault tree of the winding and lead wire. Based on the fault tree, a total of 21 transformer fault types that can be directly reflected by the state quantity are sorted out.

[0057]

[0058] Combined with on-site operation experience, there is no strict boundary between different fault natures, and the fault nature may also change at different development stages of the transformer fault. In this regard, this embodiment classifies some fault types into different fault sets at the same time to improve the effect of the transformer fault diagnosis model. Based on the above transformer fault tree, the following classification is made for the fault nature and fault type of the transformer.

[0059]

[0060] For the in-service transformer to be diagnosed, considering that it is relatively difficult to obtain some off-line test data, this embodiment gives priority to extracting the characteristic quantities from the on-line monitoring test to achieve the screening of the fault type, which specifically includes the following steps:

[0061] S4.1: Extract the on-line or off-line test characteristic quantities, which should include the CO2 gas production rate, the gas production rate ratio CO2 / CO, the C2H2 content, the total hydrocarbon content, the H2 content, the CH4 content, and the iron core grounding current.

[0062] S4.2: Combining the correspondence between the fault types of multiple faulty transformers and the sample data, the following troubleshooting criterion table is summarized.

[0063]

[0064] S5: Using association rule analysis to obtain the fault characteristic quantities and their corresponding weights for each fault type, combining the on-line monitoring, off-line test and operation and maintenance data of the transformer, and using the traversal method to perform secondary fault diagnosis on the selected fault subset, calculating the transformer fault type, fault location and corresponding fault credibility; if all fault credibilities are zero, the diagnosis result is output as normal and the process ends.

[0065] In this embodiment, using association rule analysis to obtain the fault characteristic quantities and their credibilities for each fault type specifically includes the following steps:

[0066] S5.1: Collect the on-line monitoring, off-line test and operation and maintenance data of the transformer to be diagnosed, extract the state parameters that can reflect the transformer fault type, and construct a transformer fault characteristic set.

[0067] S5.2: Calculate the support and confidence of the fault characteristic quantities corresponding to each fault type. Using the fault cases of the transformer to construct a transaction database N, recording the number of all non-empty fault subsets in N as |N|, the number of occurrences of fault type A in a certain subset as |X|, and the number of abnormal occurrences of characteristic quantity B as |Y|, then the support of A→B is

[0068]

[0069] It is set that the support is greater than 70%, and the characteristic quantity B is the fault characteristic quantity of the fault type A. The confidence of A→B is

[0070]

[0071] S5.3: After normalizing the confidence of each fault characteristic quantity in the fault type, the corresponding weight coefficient can be calculated.

[0072] Taking the abnormal partial discharge of the winding as an example to calculate the weight coefficient of the corresponding fault characteristic quantity. In the following table, N represents the number of occurrences of the i-th type of fault, M represents the number of abnormal occurrences of the j-th characteristic quantity in the i-th fault type, K represents the total number of abnormal occurrences of the j-th characteristic quantity, X 1.1 is the C2H2 content, X 1.2 is the winding insulation resistance, X 1.3 is the H2 content, X 1.4 is the winding DC resistance, X 1.5 is the CO2 gas production rate.

[0073]

[0074] Record the thing database D = {winding partial discharge anomaly}, |D| = 137, feature quantity X 1.1 , X 1.2 , X 1.3 , X 1.4 , X 1.5 In the winding partial discharge anomaly fault, the number of anomalies is 120, 115, 112, 111, 99 respectively, and in the total sample, the number of anomalies is 312, 291, 421, 312, 258 respectively. X can be calculated 1.1 , X 1.2 , X 1.3 , X 1.4 , X 1.5 The support degrees in the winding partial discharge anomaly fault are: S1.1 = 87.6%, S1.2 = 83.9%, S1.3 = 81.8%, S1.4 = 81.0%, S1.5 = 72.3%, and the results are all greater than 70%, indicating that there is a certain correlation between the 5 feature quantities in the above table and the winding partial discharge anomaly fault. The confidence degrees of the feature quantities are: C1.1 = 38.3%, C1.2 = 39.6%, C1.3 = 26.8, C1.4 = 35.8%, C1.5 = 38.4. After normalizing the confidence degrees, the weights of the fault feature quantities can be obtained as w 1.1 = 0.215, w 1.2 = 0.220, w 1.3 = 0.149, w 1.4 = 0.199, w 1.5 = 0.215.

[0075] In the transformer fault types, there are feature quantity support degrees exceeding 98%. For example, the excessive furfural content can reflect the winding insulation aging fault, and the abnormal core grounding current can reflect the multi-point grounding fault of the core. For such strongly correlated feature quantities, in this embodiment, their weights are set to 1 separately and do not participate in the weight calculation of other feature quantities, so as to avoid the increase of weight calculation errors caused by the too small confidence degrees of other feature quantities. The calculation results of the weights of the fault feature quantities corresponding to some fault types are shown in the following table.

[0076]

[0077] Based on the state of the transformer fault feature quantity, fault diagnosis can be realized. In this embodiment, the feature quantity has two states: normal and abnormal. Record y ij = 0 indicates that the feature quantity is normal, y ij = 1 indicates that the feature quantity is abnormal, where y ij is the state of the jth fault feature quantity in the ith fault type. The credibility Q formula of the ith fault type is

[0078]

[0079]

[0080] Verification example:

[0081] For the on-line oil chromatogram monitoring device of phase C of the 500 kV main transformer in a certain regional substation, an alarm was reported for excessive total hydrocarbon content. Subsequently, the chromatogram tracking frequency was increased until June, the main transformer was taken out of service, and relevant off-line tests and disassembly inspections were carried out. The on-line monitoring data on April 16 and the off-line test data on June 6 were collected, as shown in Table 1.

[0082] Table 1 Operating and test data of 500 kV main transformer

[0083] Characteristic quantity April 16th June 6th Threshold value <![CDATA[H2 content / (μL / L)]]> 35 54.00 150 <![CDATA[CH4 content / (μL / L)]]> 61.40 155.50 100 <![CDATA[C2H4 content / (μL / L)]]> 65 191.10 50 <![CDATA[C2H6 content / (μL / L)]]> 18.40 55.30 65 <![CDATA[C2H2 content / (μL / L)]]> 0.20 0.51 1 Total hydrocarbon content / (μL / L) 180 456.40 150 CO gas production rate / mL / d 104.70 123.70 100 <![CDATA[CO2 gas production rate / mL / d]]> 264.80 351.20 200 Core grounding current / A 0.02 0.02 0.1 Core insulation resistance / (MΩ) — 1800 100 Winding dielectric loss / % — 0.54 0.50 Initial value difference of winding capacitance / % — 1.20 0.30 Mutual difference of winding DC resistance / % — 2.30 2

[0084] Hierarchical diagnosis of transformer faults was carried out on the two tests in the above table. In the sample on April 16, the abnormal characteristic quantities were: CH4 content, C2H4 content, total hydrocarbon content, CO gas production rate, and CO2 gas production rate. The primary diagnosis result of transformer faults based on the extended three-ratio method was high-temperature overheating. Since the iron core grounding current was less than 1 A, the CO2 gas production rate was abnormal, and the total hydrocarbon content exceeded the standard. Combining with the fault set corresponding to the nature of the high-temperature overheating fault, the fault types that could be screened out were overheating of the current circuit and inter-turn short circuit. Combining with the association rules, the credibility of the inter-turn short circuit was calculated to be 36.6%, and the credibility of the overheating of the current circuit was 60.2%. Similarly, for the test sample on June 6, the primary diagnosis result was high-temperature overheating, and the secondary diagnosis results were inter-turn short circuit and overheating of the current circuit, with credibilities of 79.3% and 80.2% respectively.

[0085] Table 2 Diagnosis results of fault cases

[0086]

[0087] Calculated according to the insulating oil density of 0.88 t / m 3 , the C2H2 gas production rate of a single main transformer with a total oil volume of 50.2 t was 34 mL / d, exceeding the threshold of 0.2 mL / d. The C2H2 content was showing an increasing trend, and the credibility of the inter-turn short circuit fault was still increasing continuously.

[0088] Combining the hierarchical diagnosis results of the two transformer faults, it was speculated that an inter-turn short circuit might occur in the transformer winding part, resulting in overheating of the main circuit and possibly accompanied by a certain degree of insulation carbonization. The results of the transformer disassembly inspection are as Figure 3 shown. It can be found that there was an inter-turn short circuit in the medium-voltage coil of this main transformer, which caused a circulating current resulting in local high-temperature overheating, and at the same time, obvious carbonization appeared in the surrounding insulating paper, which was consistent with the hierarchical diagnosis results.

[0089] To further verify the applicability and effectiveness of the transformer hierarchical diagnosis model proposed in the embodiments, 100 groups of transformer test samples were used to evaluate the accuracy rate of the hierarchical diagnosis model, and the results are shown in Table 3.

[0090] Table 3 Evaluation of the Hierarchical Diagnosis Model

[0091]

[0092] It can be seen from this that the accuracy rate of the transformer fault hierarchical diagnosis model of the present invention is 87%, and its diagnostic effect has been improved to a certain extent compared with the traditional three-ratio method. On normal samples, the diagnostic accuracy of the transformer fault hierarchical diagnosis model is 88.1%, which reduces the misjudgment rate of the transformer to a certain extent. In dealing with single faults and concurrent multiple faults, due to the limitations of the traditional three-ratio method, it is impossible to diagnose the specific fault type and fault location of the transformer, while the fault hierarchical diagnosis model can well solve this problem and enhance the practicality of on-site applications.

[0093] The above are only embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are included in the scope of the claims of the present invention pending approval.

Claims

1. A hierarchical fault diagnosis method for transformers based on extended triple ratios and association rules, characterized in that It includes the following steps: S1: Collect the on-line monitoring, off-line test and operation and maintenance data of the diagnosed transformer; among them, at least one of the on-line monitoring or off-line test data should include the component type, content or gas production rate of the dissolved gases in the transformer oil; S2: Make a prior judgment on the existence of faults based on the content or gas production rate of the dissolved gases in the oil, and preliminarily classify the internal faults of the transformer into one of the following three types of faults: overheating fault, discharge fault or other faults; S3: For the internal discharge or overheating faults determined in the transformer, combined with the on-line or off-line dissolved gas data in the oil of the transformer, use the extended three-ratio method to complete the primary fault diagnosis of the transformer, and the diagnosis result is one of the following six fault natures: low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge or high-energy discharge; S4: Based on the transformer fault tree, sort out the main fault types of the transformer, establish the corresponding relationship between the fault nature and the fault type of the transformer, obtain the fault subset corresponding to the above fault nature, extract the on-line monitoring or off-line test characteristic quantities, and further screen the fault types in the fault subset; The corresponding relationship between the fault nature and the fault type in S4 is: The on-line or off-line test characteristic quantities extracted in S4 include the CO2 gas production rate, the gas production rate ratio CO2 / CO, the C2H2 content, the total hydrocarbon content, the H2 content, the CH4 content and the core grounding current; S5: Use association rule analysis to obtain the fault characteristic quantities and their corresponding weights corresponding to each fault type, combined with the on-line monitoring, off-line test and operation and maintenance data of the transformer, use the traversal method to perform secondary fault diagnosis on the screened fault subset, and calculate the transformer fault type, fault location and corresponding fault credibility; if all fault credibilities are zero, output the diagnosis result as normal and end the process.

2. The transformer fault hierarchical diagnosis method based on the extended three-ratio method and association rules according to claim 1, wherein: The dissolved gas sample data in the oil in S1 should include the contents of H2, CO, CO2 and the four hydrocarbon gases CH4, C2H6, C2H4, C2H2.

3. The hierarchical fault diagnosis method for transformers based on the extended three-ratio method and association rules according to claim 2, wherein: The principle for judging the existence of internal faults of the transformer in S2 mainly includes taking the set of sample data closest to the current time as the standard. If the contents of each component gas simultaneously meet the following three conditions: H2 is less than 150 μL / L, C2H2 is less than 1 μL / L, and the total content of the four hydrocarbon gases is less than 150 μL / L, the diagnosis result is other faults. If the above conditions are not met, it is determined that there are discharge or overheating fault natures.

4. The hierarchical fault diagnosis method for transformers based on the augmented three-ratio method and association rules according to claim 3, wherein: The extended three-ratio method in S3 specifically includes the following steps: S3.1: If it is preliminarily determined that there are overheating or discharge faults inside the transformer, calculate the three ratios corresponding to the dissolved gas sample data in the oil: CH4 / H2, C2H2 / C2H4 and C2H4 / C2H6, and use the following rules to encode the sample data; S3.2: According to the calculated coding combination, query the following transformer fault type judgment criterion table to determine the transformer fault nature as low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge or high-energy discharge; S3.3: Combine the corresponding relationships between multiple dissolved gas in oil samples and transformer fault natures, and summarize the determination attribution table for the missing code 011 as follows; Where: TG represents the total of four hydrocarbon gases, and M(*) represents the content of the gas in the parentheses.

5. A transformer fault hierarchical diagnosis method based on an extended three-ratio method and association rules according to claim 1, characterized in that: In S4, extract the characteristic quantities of on-line or off-line tests for fault type exclusion. Combine the corresponding relationships between the fault types and sample data of multiple faulty transformers, and summarize to obtain the following fault exclusion criterion table; Further screen the fault types in the fault subset based on the fault exclusion criterion table.

6. The hierarchical diagnosis method for transformer faults based on the extended three-ratio method and association rules according to claim 1, characterized in that: In S5, the fault characteristic quantities and their weight coefficients corresponding to each fault type are obtained by using association rule analysis, which specifically includes the following steps: S5.1: Collect the on-line monitoring, off-line test and operation and maintenance data of the transformer to be diagnosed, extract the state parameters that can reflect the transformer fault type, and construct a transformer fault feature set; S5.2: Calculate the support and confidence of the fault characteristic quantities corresponding to each fault type; construct a transaction database N with the fault cases of the transformer. Denote the number of all non-empty fault subsets in N as |N|, the number of occurrences of fault type A in a certain subset as |X|, and the number of abnormal times of characteristic quantity B as |Y|. Then the support of A→B is Set the support to be greater than 70%, and the characteristic quantity B is the fault characteristic quantity of fault type A; the confidence of A→B is S5.3: After normalizing the confidence of each fault characteristic quantity in the fault type, the corresponding weight coefficient can be calculated.

Images