Artificial intelligence based transformer bushing security monitoring system

Abstract

This invention discloses an artificial intelligence-based transformer bushing safety monitoring system, mainly composed of a fault diagnosis expert system and a transformer oil chromatography online analysis system. The fault diagnosis expert system includes a transformer bushing electrical knowledge base, an SQL Server database, an inference engine, a human-computer interface, and a built-in controller. The transformer oil chromatography online analysis system includes a gas source module, an oil-gas separation module, a gas detection module, and a chromatographic data acquisition and processing module. This invention generates a gas concentration distribution map based on the pairwise correlations between gas occurrences, and forms triangles by connecting pairs of concentration values, statistically analyzing the S-t curves of the areas of each triangle. By combining the gas concentration distribution map and the S-t curves for diagnosis, the visualization of various fault diagnoses is significantly improved, and the speed of fault analysis is greatly increased. Furthermore, statistical analysis of data from a period prior to the fault provides a new approach for real-time fault monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformer safety management, and particularly relates to a transformer bushing safety monitoring system based on artificial intelligence. BACKGROUND

[0002] The transformer is the heart of the substation, and it is required to work reliably during operation. Once a fault occurs, it may cause equipment damage or even a fire, which may endanger the safety of the substation. Therefore, it is necessary to detect the thermal defects of the transformer in a timely and effective manner to prevent safety accidents.

[0003] With the rapid development of artificial intelligence technology, intelligent transformer terminals and fault diagnosis expert systems have gradually replaced traditional manual maintenance systems. For example, the recently developed intelligent transformer online monitoring system integrates data maintenance management and analysis diagnosis into one, establishes a systematic, multi-level and visual fault diagnosis expert system, realizes real-time online monitoring of the operating condition of the transformer, and realizes remote management of the monitoring data through network communication technology, thereby ensuring the safe operation of the power grid and laying a foundation for the establishment of smart grids.

[0004] Among them, the transformer oil dissolved gas online monitoring unit is one of the important components of the intelligent transformer online monitoring system, and the online monitoring and analysis of the dissolved gas content in the insulating oil is an effective method for discovering and diagnosing potential internal faults of the transformer.

[0005] The key of the existing various intelligent transformer online monitoring systems lies in the core algorithm, and the focus of the core algorithm lies in how to visually process the analysis data of the dissolved gas in the transformer oil. Common visual processing includes the David triangle method, the IEC three-ratio method and the fuzzy mathematical reasoning method. However, after information integration, the results of various visual processing are still a combination mode of numerical values and data charts. When feeding back information to the fault diagnosis expert system, experts still need to compare and contrast based on artificial experience and knowledge reserves to obtain specific diagnostic conclusions. Although it has made great progress compared with the manual maintenance system, it still needs multiple data integration and maintenance demonstration, and the data statistics are chaotic. The visual degree is still difficult to meet the timeliness and effectiveness of the existing intelligent management science, and improvement is urgently needed. SUMMARY

[0006] The purpose of the present application is to solve the problems existing in the prior art, and a transformer bushing safety monitoring system based on artificial intelligence is provided.

[0007] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0008] The transformer bushing safety monitoring system based on artificial intelligence mainly consists of a fault diagnosis expert system and a transformer oil chromatographic online analysis system,

[0009] The first part of the fault diagnosis expert system includes:

[0010] The transformer bushing electrical knowledge base stores historical data on transformer bushings and diagnostic reports from experts based on that data. It integrates data on possible faults, causes, diagnostic methods, and characteristics caused by transformer oil into an experience data package for easy access.

[0011] SQL Server databases, including a comprehensive transformer information database, a routine electrical test database, and a dissolved gas composition database in oil, provide a data storage and feedback platform;

[0012] The inference engine stores mathematical reasoning algorithms, analyzes the composition data of dissolved gas in transformer oil, analyzes the real-time concentration value and concentration change rate, and compares and calculates with historical data to obtain fault data that matches the experience data package, thereby obtaining fault diagnosis.

[0013] The human-computer interaction interface uses data tables and image formats to monitor data and can call up real-time data and related historical data at any time to obtain visualized fault reports.

[0014] The controller is used to connect various components and servers, and to forward fault reports to on-site clients, providing early warning information and timely maintenance.

[0015] The second part of the online transformer oil chromatography analysis system includes:

[0016] The gas source module includes a gas production unit, a gas storage unit, a purification unit, a pressure control and alarm unit, etc., and is used to provide the carrier gas that drives the gas flow.

[0017] The oil-gas separation module includes an oil sample circulation and acquisition unit, an oil sample quantitative unit, an oil sample processing and return unit, and a degassing and gas collection unit, which are used to separate dissolved gases from transformer oil.

[0018] The gas detection module includes a gas separation unit, a constant temperature and constant current control unit, and a gas detection unit, which is used to collect and remove nitrogen and oxygen, purify various fault gases, and detect and analyze them.

[0019] The chromatography data acquisition and processing module includes a data acquisition unit, a field control and processing unit, a communication control unit, and a communication unit, etc., which are used to acquire and analyze data and output signals. If data and alarm signals are not directly uploaded through the communication unit, a computer and monitoring software can be installed in the main control room to directly monitor the operation of the equipment in the main control room.

[0020] Auxiliary units include those placed at transformer interfaces, oil pipes, and communication and power cables.

[0021] Preferably, the working process of the transformer oil chromatography online analysis system is as follows: First, the oil sample acquisition unit circulates the oil circuit to remove stagnant oil in the connecting pipeline, and then the oil sample is quantified; the oil-gas separation unit quickly separates the dissolved gas in the oil and delivers it to the quantitative tube of the six-way valve for automatic injection; under the propulsion of the carrier gas, the sample gas passes through the chromatographic column for separation and sequentially enters the gas detection unit; the data acquisition unit completes the conversion and acquisition of AD data, the embedded processing unit stores, calculates and analyzes the acquired data, and uploads the data to the data processing server (installed in the main control room) through the Ethernet interface; finally, the XS-DGA-03 monitoring and early warning software performs data processing and fault analysis.

[0022] Furthermore, the monitoring and early warning software categorizes faults into multiple levels based on gas concentration values: "General Fault," "Severe Fault," and "Critical Fault." "Critical Fault" requires immediate on-site handling. "Severe Fault" is analyzed in depth based on time constraints, combined with other electrical inspection reports and expert judgments. "General Fault" is summarized as a preventative measure against "Severe Fault" and "Critical Fault," forming a regular diagnostic report for archiving and future use, facilitating long-term data processing and fault analysis.

[0023] Preferably, the mathematical reasoning algorithm flow in the inference engine is as follows:

[0024] Based on the correlation between the frequency and concentration of various gases, a gas concentration distribution map with a certain order is formed. By connecting any two of the multiple concentration values ​​to form a triangle, the area of ​​each triangle is counted and compared to obtain the difference between the concentration of two gases and the concentration of other gases. Thus, the fault type can be quickly determined by observation.

[0025] Furthermore, the mathematical reasoning algorithm in the inference engine also includes the following process: By comparing the areas of each triangle, the ratio of any two sets of data from S1 to S6 requires a total of C(6,2) = 15 calculations, corresponding to the performance ratio of each triangle area in various fault types. For example, in the "oil overheating" fault, S5 > S1, so the performance ratio is: S5 / S1 > 1, while for others S2 / S1≈S3 / S1≈S4 / S1≈S6 / S1≈0; the real-time tested area ratios are then compared with the inherent performance ratios of each fault type. By comparing the data, the fault type can be obtained immediately. For the faults "partial discharge in oil-paper insulation" and "arc in oil and paper", the sum of S1-S6 should also be referenced, which is the fixed sum of the areas of each triangle of each fault. The real-time sum of S1-S6 is compared with the fixed sum. If the real-time sum / fixed sum = 0.7-1.3, it can be considered that the "corresponding fault has been matched". By comparing the above two ratios, the fault type can be quickly diagnosed.

[0026] Furthermore, the mathematical reasoning algorithm in the inference engine also includes the following process: simultaneously, based on the time statistics of the concentrations of each of the two gases and the visualization effect of the curves changing with time, i.e., the St curves, it is possible to see from the coordinate system which one or more areas of each St curve are at high values ​​and which one or more areas are at low values ​​over time. This allows for real-time monitoring and early warning of gas concentration changes before some faults occur, thus providing system support for the implementation of fault early warning mechanisms.

[0027] More importantly, the method for constructing the gas concentration distribution map is as follows:

[0028] 1) Based on the empirical data in the database, the concentration values ​​of a certain gas that occurs during a certain type of fault in the transformer bushing are sorted out and averaged. The average value is multiplied by 3 to obtain the limit value of the gas concentration. The limit values ​​of six gases, H2, C2H2, CH4, C2H4, CO2 and CO, are calculated in sequence. The six limit values ​​are multiplied by a different ratio κ1-κ6 to enlarge or reduce to the same value. With the origin O as the six polar coordinate axes, the six polar coordinate axes are distributed in an equal-angle ring to construct a six-axis polar coordinate system. The endpoints correspond to the limit values, and adjacent endpoints are connected to form a hexagon.

[0029] 2) The real-time concentration value obtained from the test is multiplied by one of the corresponding ratios κ1-κ6, and then multiplied by 3 to obtain the marked value in the coordinate.

[0030] 3) Automatically connect the marked values ​​of adjacent polar coordinate axes. A triangle is constructed between the two marked values ​​and the origin O. Output the graph. Calculate the area of ​​each triangle in each of the six polar coordinate systems using graph recognition. Let S1 be the area of ​​the triangle between the H2 and C2H2 polar coordinate axes, S2 be the area between the H2 and CO polar coordinate axes, and S3, S4, S5, and S6 be the areas between the other two polar coordinate axes (CO, CO2, C2H4, CH4, and C2H2). This yields a gas concentration distribution map that can record concentrations in real time. Figure 2 As shown;

[0031] 4) Based on the test time sequence, the area of ​​each triangle related to the six gases changes over time. After linear fitting, the St curve for a period of time is obtained. This allows for real-time observation of whether the state of the oil in the transformer bushing is abnormal, as well as real-time observation of the gas concentration change trend when a certain fault occurs and for a period of time before it. This provides the possibility of summarizing the rules for subsequent fault early warning. Therefore, the St curve of this invention provides a new approach for real-time fault monitoring.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. In view of the shortcomings of the low visualization processing level of existing intelligent transformer online monitoring systems, this invention, based on the three-ratio method in existing visualization processing and the experience of gas composition analysis during actual transformer bushing faults, found that the possibility of a single fault gas is almost non-existent when a fault occurs, and at least two gases are often generated simultaneously when a fault occurs; and according to the composition analysis report, there is a certain established pattern between the frequency and concentration of the nine gases when a fault occurs.

[0034] 2. This invention forms a gas concentration distribution map in a certain order based on the correlation between two gases, and forms triangles by connecting the concentration values ​​of each pair. By comparing the areas of each triangle, the difference between the concentrations of two gases and the concentrations of other gases can be obtained, thus allowing for quick observation to determine the type of fault.

[0035] 3. The gas concentration distribution map of the present invention is a novel visualization interface. Its six-axis polar coordinate system can collect real-time detected gas concentration data and obtain a label value. The label values ​​of adjacent polar coordinate axes form a triangle. By calculating the area of ​​each triangle in the six-axis polar coordinate system through image recognition, a gas concentration distribution map that can record the concentration in real time is obtained.

[0036] 4. This invention obtains the St curve over a period of time by statistically analyzing the area of ​​each triangle related to six gases over time according to the test time sequence and performing linear fitting. This allows for real-time observation of whether the state of the oil inside the transformer bushing is abnormal, as well as real-time observation of the gas concentration change trend when a certain fault occurs and for a period of time before it, providing the possibility of summarizing the patterns for subsequent fault early warning.

[0037] 5. By combining gas concentration distribution maps and St curves for diagnosis, the visualization of various fault diagnoses is greatly improved. Specifically, the interface of the two displays the pre-assigned characteristics of each fault (which can be displayed in the coordinate system of the gas concentration distribution map and St curve), which greatly improves the speed of fault analysis. Furthermore, the data from a period of time before the fault is analyzed to obtain the pattern of fault occurrence, providing a new approach for real-time fault monitoring. Attached Figure Description

[0038] Figure 1 This is an architecture diagram of the transformer bushing safety monitoring system based on artificial intelligence proposed in this invention;

[0039] Figure 2 This is a comparison chart of gas analysis data from the artificial intelligence-based transformer bushing safety monitoring system proposed in this invention.

[0040] Figure 3 for Figure 2The curve of the interference area of ​​the detected gas concentration versus time obtained from the gas analysis comparison graph (St graph). Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0042] I. System Architecture

[0043] The transformer bushing safety monitoring system based on artificial intelligence mainly consists of two parts: a fault diagnosis expert system and an online transformer oil chromatography analysis system.

[0044] The first part of the fault diagnosis expert system includes:

[0045] The transformer bushing electrical knowledge base stores historical data on transformer bushings and diagnostic reports from experts based on that data. It integrates data on possible faults, causes, diagnostic methods, and characteristics caused by transformer oil into an experience data package for easy access.

[0046] SQL Server databases, including a comprehensive transformer information database, a routine electrical test database, and a dissolved gas composition database in oil, provide a data storage and feedback platform;

[0047] The inference engine stores mathematical reasoning algorithms, analyzes the composition data of dissolved gas in transformer oil, analyzes the real-time concentration value and concentration change rate, and compares and calculates with historical data to obtain fault data that matches the experience data package, thereby obtaining fault diagnosis.

[0048] The human-computer interaction interface uses data tables and image formats to monitor data and can call up real-time data and related historical data at any time to obtain visualized fault reports.

[0049] The controller is used to connect various components and servers, and to forward fault reports to on-site clients, providing early warning information and timely maintenance.

[0050] The second part of the online transformer oil chromatography analysis system includes:

[0051] The gas source module includes a gas production unit, a gas storage unit, a purification unit, a pressure control and alarm unit, etc., and is used to provide the carrier gas that drives the gas flow.

[0052] The oil-gas separation module includes an oil sample circulation and acquisition unit, an oil sample quantitative unit, an oil sample processing and return unit, and a degassing and gas collection unit, which are used to separate dissolved gases from transformer oil.

[0053] The gas detection module includes a gas separation unit, a constant temperature and constant current control unit, and a gas detection unit, which is used to collect and remove nitrogen and oxygen, purify various fault gases, and detect and analyze them.

[0054] The chromatography data acquisition and processing module includes a data acquisition unit, a field control and processing unit, a communication control unit, and a communication unit, etc., which are used to acquire and analyze data and output signals. If data and alarm signals are not directly uploaded through the communication unit, a computer and monitoring software can be installed in the main control room to directly monitor the operation of the equipment in the main control room.

[0055] Auxiliary units include those placed at transformer interfaces, oil pipes, and communication and power cables.

[0056] Specifically, the working process of the online transformer oil chromatography analysis system is as follows:

[0057] First, the oil sampling unit circulates the oil circuit to remove stagnant oil in the connecting pipelines, and then the oil sample is quantified. The oil-gas separation unit quickly separates dissolved gases from the oil and delivers them to the quantitative tube of the six-way valve for automatic injection. Driven by the carrier gas, the sample gas passes through the chromatographic column for separation and sequentially enters the gas detection unit. The data acquisition unit completes the conversion and acquisition of AD data, the embedded processing unit stores, calculates and analyzes the acquired data, and uploads the data to the data processing server (installed in the main control room) via the Ethernet interface. Finally, the XS-DGA-03 monitoring and early warning software performs data processing and fault analysis.

[0058] Furthermore, the monitoring and early warning software categorizes faults into multiple levels based on gas concentration values: "General Fault," "Severe Fault," and "Critical Fault."

[0059] "Critical faults" require immediate on-site handling;

[0060] "Serious faults" are analyzed in depth based on time constraints, other electrical inspection reports, and expert judgment.

[0061] "General failures" are summarized as preventative measures against "serious failures" and "critical failures," forming regular diagnostic reports for archiving and future use, facilitating long-term data processing and failure analysis.

[0062] II. Algorithm improvements for visualization processing:

[0063] 1) Based on the three-ratio method in existing visualization processing and the experience of gas composition analysis during actual transformer bushing faults, this invention found that the possibility of a single fault gas occurring during a fault is almost non-existent. At least two gases are often produced simultaneously during a fault. Furthermore, according to the composition analysis report, there is a certain established pattern between the frequency and concentration of the nine gases during a fault.

[0064] 2) Based on this analysis, this invention forms a gas concentration distribution map in a certain order according to the correlation between the frequency and concentration of various gases. By connecting any two of the concentration values ​​to form triangles, and comparing the areas of each triangle, the difference between the concentrations of two gases and their relative concentrations of other gases can be obtained. Therefore, the fault type can be quickly determined through observation. Figure 2 ;

[0065] 3) By comparing the areas of each triangle, the ratio of any two sets of data from S1 to S6 requires a total of C(6,2)=15 calculations, which corresponds to the performance ratio of the area of ​​each triangle in various fault types.

[0066] For example, in the "oil overheating" fault, S5 > S1, then the performance ratio is: S5 / S1 > 1, while for others S2 / S1≈S3 / S1≈S4 / S1≈S6 / S1≈0. By comparing the area ratios of each real-time test with the inherent performance ratios of each fault type, the fault type can be obtained immediately.

[0067] The inherent ratios of the faults “partial discharge in oil-paper insulation” and “electric arc in oil and paper” are similar. Therefore, the sum of S1-S6, that is, the fixed sum of the areas of each triangle of each fault, should also be taken into account. The real-time sum of S1-S6 detected in real time is compared with the fixed sum. If the real-time sum / fixed sum = 0.7-1.3, it can be identified as “the corresponding fault has been matched”.

[0068] By comparing the two ratios above, the diagnosis of the fault type can be quickly determined, as shown in Table 1.

[0069] Table 1. Transformer Bushing Fault Analysis Based on Gas Composition

[0070]

[0071] It should be noted that a fault arc occurs when conductors at different potentials, such as phase conductors to ground, experience insulation failure without forming a metallic connection, resulting in a short circuit through the arc. Partial discharge occurs when local insulation deteriorates, resulting in excessively high local electric field strength (e.g., at a sharp point), or when air on the insulation surface is ionized, potentially generating an arc. However, this arc does not extend to other conductors, hence the term "partial" discharge. Partial discharge can be corona discharge from air ionization or flashover discharge along the insulation surface, but it does not extend to other conductors. When S1≈S3>S5, it is necessary to analyze the difference in the total fault gas content. That is, the total fault gas concentration generated by partial discharge faults in oil-paper insulation is significantly lower than the total fault gas concentration generated by arc faults in oil and paper (compared to the fault gas range of historical data). Therefore, the display area of ​​the St curve should also show the various fault gas concentration values ​​and the total fault gas concentration values ​​generated by different types of faults, so that human experts can quickly make a judgment and the intelligent expert system can quickly determine the fault type after comparison and calculation.

[0072] 4) Simultaneously, based on the time-varying statistics of the concentrations of each of the two gases and the visualization of the St curves (i.e., the area of ​​each St curve at a high value and the area of ​​each St curve at a low value) over time, it is possible to monitor and provide real-time warnings of gas concentration changes before some faults occur. This provides system support for the implementation of fault early warning mechanisms. (Refer to...) Figure 3 .

[0073] III. Construction of Gas Concentration Distribution Diagram:

[0074] The method for constructing a gas concentration distribution map is as follows:

[0075] 1) Based on the empirical data in the database, the concentration values ​​of a certain gas that occurs during a certain type of fault in the transformer bushing are sorted out and averaged. The average value is multiplied by 3 to obtain the limit value of the gas concentration. The limit values ​​of six gases, H2, C2H2, CH4, C2H4, CO2 and CO, are calculated in sequence. The six limit values ​​are multiplied by a different ratio κ1-κ6 to enlarge or reduce to the same value. With the origin O as the six polar coordinate axes, the six polar coordinate axes are distributed in an equal-angle ring to construct a six-axis polar coordinate system. The endpoints correspond to the limit values, and adjacent endpoints are connected to form a hexagon.

[0076] 2) The real-time concentration value obtained from the test is multiplied by one of the corresponding ratios κ1-κ6, and then multiplied by 3 to obtain the marked value in the coordinate.

[0077] 3) Automatically connect the marked values ​​of adjacent polar coordinate axes. A triangle is constructed between the two marked values ​​and the origin O. Output the graph. Calculate the area of ​​each triangle in each of the six polar coordinate systems using graph recognition. Let S1 be the area of ​​the triangle between the H2 and C2H2 polar coordinate axes, S2 be the area between the H2 and CO polar coordinate axes, and S3, S4, S5, and S6 be the areas between the other two polar coordinate axes (CO, CO2, C2H4, CH4, and C2H2). This yields a gas concentration distribution map that can record concentrations in real time. Figure 2 As shown;

[0078] 4) Based on the test time sequence, the area of ​​each triangle related to the six gases is statistically analyzed over time. A linear fit yields the St curve for a given period, allowing real-time observation of whether the oil condition inside the transformer bushing is abnormal. It also allows real-time observation of the gas concentration trends during and before a specific fault occurs, providing a possibility for summarizing patterns for subsequent fault early warning. Figure 3 The figure shows the dynamic curve changes during the occurrence of the fault "spark discharge in oil" and the 24 hours prior. S1 shows a sudden increase, while S2-S5 fluctuate around 0 (actually the electrical signal fluctuation of the gas detection limit). Before the fault occurred, S6 showed large fluctuations, indicating that the CH4 concentration fluctuated to some extent. When the fault occurred, S6 tended to flatten out, proving that a small amount of arc was generated during the discharge, but it did not cause actual short-circuit heat release. It did not actually evolve into other faults (such as "spark discharge in oil"). This is consistent with the possibility and complexity of fault types and their transformations in actual testing. Therefore, the St curve of this invention provides a new approach for real-time fault monitoring.

[0079] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based transformer bushing safety monitoring system, mainly composed of a fault diagnosis expert system and a transformer oil chromatography online analysis system, characterized in that, The fault diagnosis expert system includes: The transformer bushing electrical knowledge base stores historical data on transformer bushings and diagnostic reports from experts based on that data. It integrates data on possible faults, causes, diagnostic methods, and characteristics caused by transformer oil into an experience data package for easy access. SQL Server databases, including a comprehensive transformer information database, a routine electrical test database, and a dissolved gas composition database in oil, provide a data storage and feedback platform; The inference engine stores mathematical reasoning algorithms, analyzes the composition data of dissolved gas in transformer oil, analyzes the real-time concentration value and concentration change rate, and compares and calculates with historical data to obtain fault data that matches the experience data package, thereby obtaining fault diagnosis. The human-computer interaction interface uses data tables and image formats to monitor data and can call up real-time data and related historical data at any time to obtain visualized fault reports. The controller is used to connect various components and servers, and to forward fault reports to on-site clients, providing early warning information and timely maintenance. The online transformer oil chromatography analysis system includes: The gas source module includes a gas production unit, a gas storage unit, a purification unit, a pressure control and alarm unit, and is used to provide carrier gas to drive the gas flow; The oil-gas separation module includes an oil sample circulation and acquisition unit, an oil sample quantitative unit, an oil sample processing and return unit, and a degassing and gas collection unit, which are used to separate dissolved gases from transformer oil. The gas detection module includes a gas separation unit, a constant temperature and constant current control unit, and a gas detection unit, which is used to collect and remove nitrogen and oxygen, purify various fault gases, and detect and analyze them. The chromatography data acquisition and processing module includes a data acquisition unit, a field control and processing unit, a communication control unit, and a communication unit, which are used to acquire analytical data and output signals. Auxiliary units, including those located at the transformer interface, oil pipes, and communication and power cables; The mathematical reasoning algorithm flow in the inference engine is as follows: Based on the correlation between the frequency and concentration of various gases, a gas concentration distribution map with a certain order is formed. By connecting any two of the multiple concentration values ​​to form a triangle, the area of ​​each triangle is counted and compared to obtain the difference between the concentration of two gases and the concentration of other gases. Thus, the fault type can be quickly determined by observation. The mathematical reasoning algorithm in the inference engine also includes the following process: By comparing the areas of each triangle, the ratio of any two sets of data from S1 to S6 requires a total of C(6,2) = 15 calculations, corresponding to the performance ratio of each triangle area in various fault types. The real-time tested area ratios are compared with the inherent performance ratios of each fault type. The sum of S1 to S6, i.e., the fixed sum of the areas of each triangle in each fault, is also referenced. The real-time sum of S1 to S6 is compared with the fixed sum. A real-time sum / fixed sum of 0.7-1.3 can be considered as "corresponding fault matched state". Through the above two ratio comparison calculations, the diagnosis of the fault type is quickly determined. The mathematical reasoning algorithm in the inference engine also includes the following process: based on the time statistics of the concentration of each of the two gases and the curve of change over time, i.e. the visualization effect of the St curve, observe from the coordinate system which one or more areas of each St curve are at high value and which one or more areas are at low value, and monitor and warn of gas concentration changes before some faults occur in real time. The method for constructing the gas concentration distribution map is as follows: 1) Based on the empirical data in the database, the concentration values ​​of a certain gas that occurs during a certain type of fault in the transformer bushing are sorted out and averaged. The average value is multiplied by 3 to obtain the limit value of the gas concentration. The limit values ​​of six gases, H2, C2H2, CH4, C2H4, CO2 and CO, are calculated in sequence. The six limit values ​​are multiplied by a different ratio κ1-κ6 to enlarge or reduce to the same value. With the origin O as the six polar coordinate axes, the six polar coordinate axes are distributed in an equal-angle ring to construct a six-axis polar coordinate system. The endpoints correspond to the limit values, and adjacent endpoints are connected to form a hexagon. 2) The real-time concentration value obtained from the test is multiplied by one of the corresponding ratios κ1-κ6, and then multiplied by 3 to obtain the marked value in the coordinate. 3) Automatically connect the marked values ​​of adjacent polar coordinate axes, construct a triangle with the origin O between the two marked values, output the graph, and calculate the area of ​​each triangle in each six-axis polar coordinate system through graph recognition. Set the area of ​​the triangle between the H2 polar coordinate axis and the C2H2 polar coordinate axis as S1, the area of ​​the triangle between the H2 polar coordinate axis and the CO polar coordinate axis as S2, and the areas of the triangles between the other two polar coordinate axes of CO, CO2, C2H4, CH4 and C2H2 as S3, S4, S5 and S6 respectively. Record the gas concentration distribution graph in real time. 4) Based on the test time sequence, statistically analyze the changes in the area of ​​each triangle related to the six gases over time, obtain the St curve for a period of time after linear fitting, observe in real time whether the state of the oil in the transformer bushing is abnormal, and observe in real time the trend of gas concentration changes during and before a certain fault occurs.

2. The transformer bushing safety monitoring system based on artificial intelligence according to claim 1, characterized in that, The working process of the transformer oil chromatography online analysis system is as follows: First, the oil sample acquisition unit circulates the oil circuit to remove stagnant oil in the connecting pipeline, and then the oil sample is quantified; the oil-gas separation module quickly separates the dissolved gas in the oil and delivers it to the quantitative tube of the six-way valve for automatic injection; under the propulsion of the carrier gas, the sample gas passes through the chromatographic column for separation and sequentially enters the gas detection unit; the data acquisition unit completes the conversion and acquisition of AD data, the embedded processing unit stores, calculates and analyzes the acquired data, and uploads the data to the data processing server through the Ethernet interface; finally, the monitoring and early warning software performs data processing and fault analysis.

3. The transformer bushing safety monitoring system based on artificial intelligence according to claim 2, characterized in that, The monitoring and early warning software classifies faults into multiple levels based on gas concentration values: "general faults," "serious faults," and "critical faults." "Critical faults" require immediate on-site handling. "Serious faults" are analyzed in depth based on time constraints, combined with other electrical inspection reports and expert judgments. "General faults" are summarized as preventative measures against "serious faults" and "critical faults," forming a regular diagnostic report for archiving and future use, facilitating long-term data processing and fault analysis.

Images