Converter transformer internal defect detection method based on multi-physics field coupling monitoring

Abstract

The invention relates to the technical field of converter transformer state monitoring, and discloses a converter transformer internal defect detection method based on multi-physics field coupling monitoring, and the method comprises the following steps: collecting and confirming electric field, magnetic field, temperature field, stress field, sound field and flow field data of a converter transformer without internal defects under a normal working condition; taking the data as multi-physics field reference data; collecting real-time multi-physical field data of the converter transformer to be detected; performing feature extraction on the reference data and the real-time data to obtain reference features and real-time features; calculating a feature difference vector between the real-time feature and the reference feature; and comparing the feature difference vector with a preset corresponding relation library of multi-physical field signal features and internal defect types, and determining the internal defect types. Through coupling analysis and difference comparison of multi-physical field data, cross validation of multi-source information is realized, and the accuracy and reliability of defect diagnosis are improved.

Description

Technical Field

[0001] This invention relates to the field of converter transformer condition monitoring technology, specifically a method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring. Background Technology

[0002] Converter transformers are core equipment in ultra-high voltage direct current (UHVDC) transmission systems, and their operational safety and reliability directly affect the stability of the entire power grid. Converter transformers have complex internal structures and are subjected to multiple stresses from electrical, thermal, magnetic, and mechanical sources over long periods, making them prone to various internal defects. Therefore, accurate and real-time monitoring and defect diagnosis of the internal condition of converter transformers are crucial.

[0003] Existing defect detection technologies for converter transformers, such as dissolved gas analysis (DGA), electrical detection of partial discharge, ultrasonic acoustic detection, and fiber optic temperature monitoring, have played an important role in practical applications. However, these methods often rely on information from a single physical quantity for condition assessment. This dependence on singular information streams leads to significant limitations: for example, while DGA can reflect overheating or discharge faults, its information is lagging and difficult to pinpoint precisely; and while online monitoring methods such as partial discharge detection or acoustic detection offer strong real-time performance, they are highly susceptible to interference from strong electromagnetic environments or mechanical vibrations in the field.

[0004] More importantly, the evolution of defects inside converter transformers is usually the result of multi-physics coupling. For example, metal particle defects, insulation moisture defects, or metal spike defects exhibit drastically different response characteristics in electric, acoustic, temperature, or stress fields. Detection methods relying solely on a single physical field lack sufficient information dimensions, making it difficult to effectively distinguish these completely different defect types and impossible to achieve cross-verification from multiple information sources. This easily leads to misjudgments or omissions, resulting in low accuracy and reliability of defect diagnosis. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring. This method solves the problem that existing technologies rely on single-physics field information for monitoring, resulting in low differentiation of internal defect types and insufficient diagnostic reliability.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring, comprising the following steps:

[0007] On converter transformers confirmed to be free of internal defects, electric field data, magnetic field data, temperature field data, stress field data, acoustic field data, and flow field data under normal operating conditions are collected. These data together constitute multi-physics field reference data.

[0008] Real-time electric field data, magnetic field data, temperature field data, stress field data, sound field data, and flow field data are collected on the converter transformer to be tested, serving as real-time multi-physics field data.

[0009] The multiphysics benchmark data and the real-time multiphysics data are preprocessed and feature extracted to obtain benchmark features and real-time features, respectively.

[0010] Calculate the feature difference vector between the real-time features and the baseline features.

[0011] The feature difference vector is compared with a preset database of correspondences between multiphysics signal features and internal defect types.

[0012] Based on the comparison results, the type of internal defect of the converter transformer to be tested is determined.

[0013] In a specific implementation scheme, the step of collecting multi-physics reference data under normal operating conditions specifically includes: conducting impulse voltage tests, external AC withstand tests, long-term induced withstand voltage tests, and temperature rise tests on the converter transformer confirmed to be free of internal defects, and recording the electric field data, magnetic field data, temperature field data, stress field data, sound field data, and flow field data during these tests.

[0014] In one specific implementation scheme, the various data collected by this method are achieved through the arrangement of sensors:

[0015] The electric field data is acquired using an ultra-high frequency electric field sensor. In one embodiment, the ultra-high frequency electric field sensor is used to acquire the data by means of a dielectric window placed on the riser of the converter transformer, which replaces the original handhole.

[0016] The magnetic field data, temperature field data, stress field data, and acoustic field data are acquired using an optical magnetic flux leakage sensor, an optical temperature sensor, a pressure sensor, and an ultrasonic sensor, respectively. In one embodiment, all of the above sensors are arranged in the on-load tap changer area of ​​the converter transformer. Further, the optical magnetic flux leakage sensor, the optical temperature sensor, the pressure sensor, and the ultrasonic sensor arranged in the on-load tap changer area have their signals led out through an optical fiber through-hole to ensure signal transmission integrity and maintain equipment sealing.

[0017] The flow field data is acquired using a Doppler velocity sensor. In one embodiment, the Doppler velocity sensor is positioned at a 90-degree bend on the fan side of the converter transformer, specifically by replacing the original bend and installing a 90-degree bend that integrates the sensor.

[0018] In one specific implementation, the feature extraction steps include, but are not limited to: extracting the pulse amplitude, pulse repetition rate, and phase-resolved partial discharge spectrum of the partial discharge signal from the electric field data; extracting the energy, peak amplitude, and spectral distribution of the ultrasonic signal from the acoustic field data; and extracting the temperature rise rate, final steady-state temperature value, and temperature difference between different measuring points from the temperature field data.

[0019] In a specific implementation scheme, the preset correspondence library between multi-physics signal features and internal defect types is constructed through the following steps: setting at least one simulated defect inside the converter transformer; performing a pressure test on the converter transformer with the simulated defect and collecting multi-physics data under defect conditions; extracting features from the multi-physics data under defect conditions and calculating the feature difference vector between the features of the multi-physics data under defect conditions and the features of the multi-physics reference data; establishing a correspondence between the types of simulated defects and the feature difference vector to form the correspondence library.

[0020] Furthermore, the simulated defects include at least one of the following: metal particle defects, wrinkled paper dampness defects, or metal spike defects, to cover common internal defect types.

[0021] Furthermore, when constructing the corresponding relationship library, the pressure test performed on the converter transformer with the simulated defect includes a temperature rise test; and the temperature rise test duration under the defect condition is longer than the temperature rise test duration under the normal condition, in order to fully expose the evolution characteristics of the defect under continuous thermal stress.

[0022] This invention provides a method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring. It has the following beneficial effects:

[0023] 1. This invention obtains multi-dimensional coupled information by simultaneously acquiring electric field, magnetic field, temperature field, stress field, acoustic field, and flow field data inside the converter transformer. Compared with methods relying on single physical quantity detection, this method calculates the characteristic difference vector of the multi-physical fields and compares it with multi-physical field benchmark data, realizing cross-validation of multi-source information. This effectively avoids misjudgments caused by a single information source and significantly improves the accuracy and reliability of internal defect diagnosis.

[0024] 2. The core of this method lies in the synchronous acquisition of data from six physical fields and the establishment of a feature difference vector. Different types of internal defects, such as discharge, overheating, abnormal mechanical stress, or oil flow turbulence, will generate unique, coupled signal disturbances in different physical fields. By comprehensively analyzing the difference vector formed by the features of the six physical field data, this method can capture the subtle and coupled feature differences of different defects from the high-dimensional data space, thereby significantly improving the distinguishability of complex or early-stage defect types and overcoming the limitations of existing methods in defect type identification.

[0025] 3. This method collects real-time multi-physics field data, compares it instantly with baseline data under normal operating conditions, and quickly calculates feature difference vectors and compares them with a corresponding relational database, thus realizing a streamlined process for defect detection and type determination. This real-time, data difference-based diagnostic process enables the rapid capture and type determination of the internal defect state of the converter transformer, providing operators with fast, objective early warning information and reliable diagnostic basis, which is beneficial for achieving condition-based maintenance and avoiding sudden failures. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall layout of the multiphysics monitoring system of the present invention on the converter transformer;

[0027] Figure 2 This is an overall flowchart of the defect detection and diagnosis method of the present invention;

[0028] Figure 3 This is a waveform diagram of the 80kV voltage applied to the lightning grid side of the present invention;

[0029] Figure 4 This is a waveform diagram of the 80kV voltage applied to the grid side during the operation of the present invention.

[0030] Figure 5 This is a waveform diagram of the 50kV applied voltage on the overvoltage valve side of the present invention;

[0031] Figure 6 This is a schematic diagram illustrating the method of setting up metal particle-simulated defects according to the present invention;

[0032] Figure 7 This is a schematic diagram illustrating the method for simulating moisture-induced defects in crepe paper according to the present invention.

[0033] Figure 8 This is a schematic diagram illustrating the method of setting up the metal spikes to simulate defects according to the present invention;

[0034] Figure 9 This is a schematic diagram of the electric field sensor arrangement according to the present invention;

[0035] Figure 10 This is a schematic diagram of the magnetic flux leakage sensor arrangement and optical fiber through-hole of the present invention;

[0036] Figure 11 This is a schematic diagram of the temperature sensor arrangement according to the present invention;

[0037] Figure 12 This is a temperature sensor for field testing according to the present invention;

[0038] Figure 13 This is a schematic diagram of the pressure sensor arrangement according to the present invention;

[0039] Figure 14 This is a schematic diagram of the ultrasonic sensor arrangement of the present invention;

[0040] Figure 15 This is a schematic diagram of the flow velocity sensor arrangement according to the present invention;

[0041] Figure 16 This is the waveform of the long-term induced withstand voltage test of the present invention;

[0042] Figure 17 This is a schematic diagram of the metal particle defect arrangement of the present invention;

[0043] Figure 18 This is a schematic diagram illustrating the arrangement of moisture-induced defects in the crepe paper according to the present invention.

[0044] Figure 19 This is a schematic diagram of the arrangement of metal spike defects according to the present invention. Detailed Implementation

[0045] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] This invention provides an embodiment of a method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring. This embodiment first involves constructing a multi-physics field monitoring system.

[0047] See attached document Figure 1-19 This multiphysics monitoring system was built to monitor the evolution of various physical field parameters during the operation of converter transformers. It includes multiple monitoring modules, as detailed below:

[0048] The electric field monitoring module 110 includes an ultra-high frequency electric field sensor. The module is arranged such that the original manhole is replaced with a dielectric window at the riser position closest to the on-load tap changer on the converter transformer. This dielectric window is in direct contact with the transformer oil inside the converter transformer. The ultra-high frequency electric field sensor is installed through this dielectric window and used to measure the electric field. The sensor is connected to a recording device via a direct external lead.

[0049] The magnetic field monitoring module 120 includes an optical magnetic flux leakage sensor. The module is positioned such that the optical magnetic flux leakage sensor is installed at the center of the outer side of the converter transformer winding enclosure to measure changes in the magnetic field. The sensor is secured with insulating strips of cloth or cable ties. To transmit the signal, the handhole closest to the on-load tap changer is replaced with a fiber optic cable tray, allowing the optical fiber connecting the optical magnetic flux leakage sensor and the recording equipment to be led out through a fiber optic cable tray. The led-out optical fiber is secured to a nearby wooden support using insulating cable ties and then laid along the support to the recording equipment.

[0050] The temperature field monitoring module 130 includes two optical temperature sensors and one experimental temperature monitoring sensor. The two optical temperature sensors are mounted on wooden supports above and below the on-load tap changer of the converter transformer, one for measuring temperature field changes. The sensors are secured with insulating strips or cable ties. They are connected to the recording equipment by extending the sensor's optical fiber through the optical fiber conduit of the aforementioned magnetic field monitoring module 120. The experimental temperature monitoring sensor is located on the top layer of the transformer in the field.

[0051] The stress field monitoring module 140 includes a pressure sensor. This module is positioned such that the pressure sensor is mounted on a wooden bracket above the on-load tap changer of the converter transformer to measure stress changes. The sensor is secured with insulating strips of cloth or cable ties. Its connection to the recording device is the same as that of the optical temperature sensor in the temperature field monitoring module 130, i.e., an optical fiber is led out through an optical fiber conduit.

[0052] The sound field monitoring module 150 includes two ultrasonic sensors. The module is arranged such that one sensor is mounted on a wooden support above and below the on-load tap changer of the converter transformer, for measuring changes in the sound field. The sensors are secured with insulating strips of cloth or cable ties. Its connection to the recording device is the same as that of the optical temperature sensor in the temperature field monitoring module 130, i.e., an optical fiber is led out through an optical fiber through-hole.

[0053] The flow field monitoring module 160 includes a Doppler velocity sensor. The module is arranged such that the Doppler velocity sensor is installed at a 90-degree bend on the fan side of the converter transformer to measure changes in the flow field. Specifically, a pre-fabricated 90-degree bend integrating the Doppler velocity sensor is used to replace the original bend in the corresponding section of the actual transformer.

[0054] See attached document Figure 3 -Appendix Figure 5 After the multiphysics monitoring system was constructed and deployed, a series of pressure tests were conducted on the converter transformer, which was in a normal state confirmed to be free of internal defects, to collect and record multiphysics benchmark data under normal operating conditions. This benchmark data serves as a comparative reference for subsequent defect diagnosis.

[0055] This process includes an impulse voltage test to obtain a multiphysics response benchmark for the converter transformer under simulated lightning and switching transient overvoltages. Specifically, the test involves applying a full-wave voltage to the grid side of the converter transformer. A lightning impulse test of kV was conducted; a full-wave voltage of kV was applied to the neutral point on the grid side. The operating wave impact test was performed at kV; and a full-wave voltage was applied to the valve side. The kV switching wave impulse test. The lightning impulse test uses a standard full-wave lightning impulse, with a wavefront time of [missing information]. for Half-wave peak time for .

[0056] During the application of the impulse voltage, the electric field monitoring module 110 records transient partial discharge activity, the acoustic field monitoring module 150 captures transient acoustic signals generated by discharge or mechanical vibration, and the stress field monitoring module 140 records the mechanical stress response caused by transient changes in electromagnetic force. The magnetic field monitoring module 120 records leakage magnetic field fluctuations caused by changes in transient current distribution. These transient physical field data provide a benchmark for identifying the behavior of insulation defects under impulse voltage.

[0057] The process also includes an external AC withstand test to obtain a multi-physics response benchmark for the converter transformer under continuous power frequency overvoltage, in order to assess the withstand capability of its main insulation. This test specifically includes: applying a 50kV test voltage to the grid-side neutral point for 10 minutes; and applying a 30kV test voltage to the valve side for 10 minutes. During the continuous pressurization, each monitoring module continuously records data.

[0058] The process further includes a long-term induced withstand voltage test to obtain a multi-physics response benchmark of the converter transformer under induced overvoltage, in order to assess its longitudinal insulation withstand capability. In this test, the rated voltage of the equipment is set to... The highest test voltage is The long-term test voltage is The voltage values ​​satisfy the following relationship:

[0059] ;

[0060] ;

[0061] The test was conducted according to a preset pressure ramping procedure, which specifically included the following steps:

[0062] Not greater than Connect the power supply under the specified voltage.

[0063] Increase voltage to And keep it for 5 minutes;

[0064] Rise voltage to And keep it for 5 minutes;

[0065] Rise voltage to Hold for 30 seconds;

[0066] Immediately after the test, the pressure was continuously reduced to [a lower value]. Keep it on for 30 minutes;

[0067] Reduce voltage to And keep it for 5 minutes;

[0068] When the voltage drops to In the following situations, disconnect the power supply.

[0069] Long-term induced withstand voltage tests are primarily used to assess the withstand capability of longitudinal insulation components, such as inter-turn insulation, inter-layer insulation, and phase-to-phase insulation, in converter transformer windings. A high-frequency (usually a harmonic) voltage is applied to one winding of the transformer, inducing a high test voltage under unsaturated conditions. The rated voltage is... It is the effective value of the line voltage when the equipment is operating normally. As a long-term test voltage, it is mainly used to assess the reliability of transformers operating under long-term high voltage. Its holding time is relatively long, aiming to expose the aging effect of insulation. As the highest test voltage, it is typically short in duration and is used to assess the insulation strength of a transformer under instantaneous maximum stress, thereby verifying its ability to withstand transient overvoltages. The test procedure is designed with multiple voltage ramp-up and ramp-down phases to simulate the transformer's operating conditions under different voltage stresses and to observe its response at different stress levels.

[0070] For example, at the transition points where voltage increases or decreases, partial discharge activity is either activated or suppressed. During each holding phase, the various physical field sensors continuously record data, including transient changes in the electric, acoustic, and stress fields, as well as the slow drift of the temperature field. This data will provide a detailed benchmark for analyzing the dynamic response of internal defects under induced overvoltage.

[0071] The process also includes a temperature rise test to obtain a baseline of the converter transformer's thermal steady-state multiphysics response under simulated rated load conditions. This test is conducted using a short-circuit method, specifically by supplying power to the grid side, short-circuiting the valve side, and setting the tap changer to a preset appropriate position. Referring to the relevant temperature rise test records in the converter transformer's factory test report, the same test current is applied. The test duration is approximately 260 minutes to ensure the transformer reaches or approaches a thermally stable state, thereby acquiring stable temperature and flow field baseline data.

[0072] See attached document Figure 6 -Appendix Figure 8 After collecting baseline data under normal operating conditions, this embodiment artificially introduced various simulated defects on the tap changer leads of the converter transformer to reproduce typical fault scenarios, in order to study the impact of internal defects on the multi-physics distribution of the converter transformer. These simulated defects were all introduced without altering the main structure of the converter transformer, as detailed below:

[0073] Simulation of metal particle defects, used to simulate metal particle faults inside a transformer. The defect is set up as follows: one defect is placed above and one below the lead tube near the corrugated paper of the tap changer lead body. The size range of the metal particles used is... to To secure the metal particles and ensure their localized effect, the particles were first wrapped in a vacuum-sealed bag to form a particle package, which was then wrapped with insulating strips and firmly fixed to the outside of the corrugated paper of the lead tube.

[0074] The simulation of a moisture-induced defect in crepe paper is used to mimic a moisture-induced fault in the solid insulation material inside a transformer. The setup involves placing one crepe paper defect above and one below the lead wire at the top of the tap changer. Specifically, a section of pre-treated moisture-induced crepe paper is selected, wrapped around the top lead wire, and ensured to be in close contact with the lead wire to simulate a moisture-induced insulation condition.

[0075] The simulation of a metal spike defect, used to model a partial discharge fault caused by a spike effect inside a transformer, involves placing one spike above and one below the lead tube at the top of the tap changer's lead-in corrugated paper. The specific simulation method involves wrapping a thumbtack with insulating cloth, leaving only the tip exposed. This wrapped thumbtack is then fixed to the outside of the lead tube's corrugated paper, with the exposed tip facing the insulating medium, creating a localized high electric field at the spike.

[0076] See attached document Figure 2 After simulating defects inside the converter transformer, this embodiment conducts a series of pressurization tests on the defective converter transformer to collect and record multiphysics data under the condition of internal defects. This data will be used to compare with baseline data under normal operating conditions.

[0077] The process specifically includes conducting an external AC voltage withstand test on the converter transformer with internal defects, following the methods described in the aforementioned normal operating condition data acquisition.

[0078] The process also includes conducting an induced withstand voltage test on the converter transformer with internal defects, according to the induced withstand voltage test method in the aforementioned normal operating condition data acquisition.

[0079] The process further includes conducting lightning and switching overvoltage impulse tests on converter transformers with internal defects, according to the induced withstand voltage test method in the aforementioned normal operating condition data acquisition.

[0080] The process also includes conducting a temperature rise test on the converter transformer with internal defects, following the temperature rise test method described in the aforementioned normal operating condition data acquisition. A key adjustment compared to the temperature rise test under normal operating conditions is that the test duration under these defective conditions is doubled. Extending the test duration aims to fully expose the evolution characteristics of the defects under continuous thermal stress, thereby obtaining more significant changes in multiphysics field signals.

[0081] By conducting the above experiments, a complete set of data on electric field, magnetic field, temperature field, sound field, stress field, and flow field corresponding to the working conditions of converter transformers with internal defects was obtained.

[0082] After collecting multiphysics benchmark data under normal operating conditions and multiphysics data under operating conditions with internal defects, this embodiment compares and analyzes these two sets of data, and establishes the correspondence between multiphysics signal characteristics and internal defect types, thereby providing a basis for defect diagnosis.

[0083] The process first includes preprocessing the two sets of raw data collected. Preprocessing includes digital filtering of the time-series data from the electric field monitoring module 110, magnetic field monitoring module 120, temperature field monitoring module 130, stress field monitoring module 140, sound field monitoring module 150, and flow field monitoring module 160 to eliminate power frequency interference and environmental background noise.

[0084] After preprocessing, quantitative features characterizing the transformer's operating state are extracted from the two sets of data. Specifically, for the electric field monitoring data, the extracted features include the pulse amplitude, pulse repetition rate, and phase-resolved partial discharge (PRPD) spectrum of the partial discharge signal; for the acoustic field monitoring data, the extracted features include the energy, peak amplitude, and spectral distribution obtained by fast Fourier transform (FFT) of the ultrasonic signal; for the temperature field monitoring data, the extracted features include the rate of temperature rise, the final steady-state temperature, and the temperature difference between different measuring points; for the magnetic field, stress field, and flow field monitoring data, the extracted features include changes in signal amplitude, waveform shape, or harmonic components.

[0085] Feature extraction is a crucial step in connecting raw data with defect diagnosis. For electric field monitoring data, in addition to the features mentioned above, features such as the number of discharges, average discharge current, and the rise and fall times of the discharge pulses can also be extracted. PRPD maps can be further analyzed for statistical characteristics such as shape, symmetry, and ellipticity.

[0086] For sound field monitoring data, energy characteristics can be obtained by calculating the root mean square value of the signal, peak amplitude represents the instantaneous maximum sound pressure, and spectral distribution reveals the frequency components of the sound signal; for example, partial discharge typically has energy concentration in the higher frequency band. For temperature field data, the temperature rise rate... The final steady-state temperature value can be obtained by differentiating the time-temperature curve. This is the stable temperature reached after a long-term temperature rise test, and the temperature difference between different measuring points. This reflects the relative temperature difference between the hotspot and the surrounding area.

[0087] For magnetic field monitoring data, the root mean square (RMS) value, peak value, and fundamental and harmonic components obtained through FFT analysis can be extracted. For stress field monitoring data, the RMS value, peak value, vibration frequency, and energy of vibration modes obtained through FFT analysis can be extracted. For flow field monitoring data, the average flow velocity can be extracted. Flow velocity fluctuation range And the spectral characteristics of the flow velocity to identify whether there are periodic disturbances.

[0088] Next, the process calculates the characteristic changes caused by the defect. By comparing the feature values ​​extracted under the defective condition with the corresponding feature values ​​extracted from the baseline data under the normal condition, a feature difference vector is calculated. This vector quantitatively describes the combined response in all six physical fields induced by a certain type of simulated defect (e.g., a metallic particle defect).

[0089] Finally, this process establishes a correspondence between defect types and signal features. The feature difference vector calculated for each simulated defect (metal particle defect, wrinkled paper moisture defect, metal spike defect) and its corresponding defect label are stored together in a database or lookup table. This database constitutes a correspondence library between multiphysics signal features and internal defect types.

[0090] In practical applications, when using this method to inspect a converter transformer in operation, firstly, real-time multiphysics data is collected and features are extracted. Then, the feature difference vector between this data and pre-stored normal operating condition baseline data is calculated. This real-time calculated feature difference vector is matched with the defect feature difference vectors stored in the corresponding relational database. The defect label corresponding to the item with the highest matching degree is the diagnostic result of the potential internal defect of the converter transformer.

Claims

1. A method for detecting internal defects in converter transformers based on multi-physics field coupling monitoring, characterized in that, Includes the following steps: On converter transformers confirmed to be free of internal defects, electric field data, magnetic field data, temperature field data, stress field data, sound field data, and flow field data under normal operating conditions are collected as multi-physics field reference data. On the converter transformer to be tested, real-time electric field data, magnetic field data, temperature field data, stress field data, sound field data, and flow field data are collected as real-time multi-physics field data. Preprocessing and feature extraction are performed on the multiphysics benchmark data and the real-time multiphysics data to obtain benchmark features and real-time features, respectively. Calculate the feature difference vector between the real-time features and the baseline features; The feature difference vector is compared with a preset correspondence database of multiphysics signal features and internal defect types; Based on the comparison results, the type of internal defect of the converter transformer to be tested is determined.

2. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The collection of the multiphysics reference data under normal operating conditions includes: The converter transformer confirmed to be free of internal defects was subjected to impulse voltage test, external AC withstand test, long-term induced withstand voltage test and temperature rise test, and the electric field data, magnetic field data, temperature field data, stress field data, sound field data and flow field data during the test were recorded.

3. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The steps for acquiring the real-time electric field data include: The electric field data is collected by an ultra-high frequency electric field sensor arranged in the dielectric window of the riser of the converter transformer.

4. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The steps for acquiring the real-time flow field data include: The flow field data is collected using a Doppler velocity sensor positioned at a 90-degree bend on the fan side of the converter transformer.

5. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The steps for collecting the real-time magnetic field data, temperature field data, stress field data, and sound field data include: The magnetic field data, temperature field data, stress field data, and sound field data are collected by optical leakage magnetic field sensors, optical temperature sensors, pressure sensors, and ultrasonic sensors arranged in the on-load tap changer area of ​​the converter transformer.

6. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The feature extraction includes: The pulse amplitude, pulse repetition rate, and phase-resolved partial discharge spectrum of the partial discharge signal are extracted from the electric field data. Extract the energy, peak amplitude, and spectral distribution of the ultrasonic signal from the sound field data; And extract the temperature rise rate, final steady-state temperature value, and temperature difference between different measuring points from the temperature field data.

7. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 1, characterized in that, The pre-defined database of correspondences between multiphysics signal features and internal defect types is constructed through the following steps: At least one simulated defect is introduced inside the converter transformer; A pressure test was conducted on the converter transformer equipped with the simulated defect, and multi-physics field data under the defect condition were collected. Extract the features of the multiphysics data under the defective working condition, and calculate the feature difference vector between the features of the multiphysics data under the defective working condition and the features of the multiphysics reference data. Establish the correspondence between the types of simulated defects and the feature difference vectors to form the correspondence database.

8. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 7, characterized in that, The simulated defects include at least one of the following: metal particle defects, wrinkled paper dampness defects, or metal spike defects.

9. The method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 7, characterized in that, A pressure test, including a temperature rise test, was conducted on the converter transformer equipped with the simulated defect. Furthermore, the temperature rise test duration under the defective operating condition is longer than the temperature rise test duration under the normal operating condition.

10. A method for detecting internal defects in a converter transformer based on multi-physics field coupling monitoring according to claim 5, characterized in that, The optical magnetic flux leakage sensor, the optical temperature sensor, the pressure sensor, and the ultrasonic sensor all transmit signals through an optical fiber through-hole.

Images