A method for mapping relationship inside and outside of converter transformer oil tank and fault diagnosis of fused vibration acoustic print signal

Abstract

This invention discloses a method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis by integrating vibration and acoustic signals. The method includes: constructing a numerical model of the converter transformer through finite element simulation, setting key material and structural parameters, and conducting multi-field coupling simulation by combining electromagnetic field, structural field, and acoustic field coupling equations to calculate the core vibration displacement and the sound pressure level distribution inside and outside the tank, and analyzing the internal magnetic field, vibration, and noise propagation characteristics; subsequently, selecting observation points in areas with significant core vibration, analyzing the vibration response under power frequency excitation, dividing the tank area, performing Fourier transform on the signals, screening characteristic frequencies, evaluating signal correlation using the Pearson correlation coefficient, and determining the optimal measurement point scheme; finally, extracting the frequency domain signals of the optimal measurement points under normal and fault conditions, and realizing layered fault diagnosis of the transformer windings based on the Transformer-XGBoost fusion voting mechanism model, providing technical support for converter transformer condition monitoring; this invention achieves early warning, accurate identification, and degree quantification of winding faults.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring and diagnostic technology, and in particular to a method for mapping the internal and external relationships of converter transformer oil tanks and for fault diagnosis by integrating vibration and acoustic signals. Background Technology

[0002] Converter transformers are core equipment in high-voltage direct current (HVDC) transmission systems, and their operational reliability is crucial. However, under the combined effects of harmonic excitation and complex mechanical stress, their internal windings are highly susceptible to typical faults such as radial deformation, loosening of clamping force, and inter-turn short circuits. These faults directly lead to abnormal vibrations, seriously threatening power grid safety. Existing diagnostic methods based on vibration or acoustic signals generally suffer from two major drawbacks: first, they are difficult to accurately distinguish the specific type of fault; second, they cannot effectively identify the severity of the fault. This lack of diagnostic capability prevents the provision of accurate and effective decision support for on-site operation and maintenance, hindering the achievement of true early warning. Therefore, to address the problem of the single diagnostic dimension of existing methods, it is urgent to research an intelligent diagnostic method capable of hierarchical and refined identification of fault types and severity. Summary of the Invention

[0003] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a method for mapping the internal and external relationships of converter transformer oil tanks and for fault diagnosis by integrating vibration and acoustic signals. Through a technical route of multi-source signal acquisition, refined feature extraction, hierarchical fusion diagnosis, and interpretability optimization, early warning, accurate identification, and degree quantification of winding faults can be achieved.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis, which integrates vibration acoustic signals, includes the following steps: S1. Based on the finite element simulation platform, a numerical model of the converter transformer is constructed, and the material properties and key parameters of each structural component are set. S2, based on the basic control equations of electromagnetic field, structural field and sound field, conducts multi-physics field coupling analysis to calculate the core vibration displacement and the sound pressure level distribution inside and outside the oil tank; S3, combining the magnetic-structural-acoustic multi-field coupling method, systematically analyzes the internal magnetic field distribution characteristics, vibration displacement law and noise propagation characteristics of the converter transformer, and makes a comprehensive comparison; S4. Based on the vibration displacement cloud map, observation points were selected in the area of ​​significant core vibration to investigate the vibration response characteristics of the transformer during the power frequency excitation cycle. S5 divides the front of the fuel tank into several areas, performs fast Fourier transform on the vibration and acoustic time-domain signals of each zone and the signals at the observation point, selects characteristic frequency components as input variables by comparing the spectrum, and uses the Pearson correlation coefficient as the evaluation index of signal correlation. S6. Based on step S5, comprehensively evaluate the correlation coefficient between the vibration signal and the acoustic signature signal, and combine the mapping relationship between the sound field inside and outside the transformer tank and the structural vibration to determine the optimal measurement point arrangement scheme on the tank wall. S7. Based on the optimal measuring point determined in step S6, extract the vibration frequency domain signals of each measuring point under fault and normal operating conditions, and conduct comparative analysis. S8 employs a time-series-based Transformer-XGBoost fusion voting mechanism model to achieve hierarchical fault diagnosis of transformer windings.

[0005] Preferably, in step S1, during the process of constructing the numerical model of the converter transformer based on the finite element simulation platform, the model covers the key structures of the core, winding, base and oil tank; in the simulation experiment stage, corresponding magnetic field boundary conditions are applied to the above components, and then their vibration and noise related parameters are calculated, so as to realize the multi-physics coupling simulation analysis of the converter transformer.

[0006] Preferably, in step S1, the key parameters include transformer capacity, rated current, Young's modulus, Poisson's ratio, and the geometric parameters of internal components.

[0007] Preferably, in step S2, the method for calculating the core vibration displacement and the sound pressure level distribution inside and outside the oil tank includes: S201, Calculation of core vibration excitation source: Under the condition of a sinusoidal alternating magnetic field at power frequency, neglecting the displacement current effect, the governing equations describing the magnetic field—Maxwell's equations and their constitutive relations—are as follows: ; In the formula: For curl operator; The magnetic field strength; Current density; It represents the magnetic flux density; The dielectric permeability; Introducing magnetic vector position ,satisfy The governing differential equation of the magnetic field can be obtained as follows: ; In the formula: For magnetic vector position; Let be the reluctance of the medium, satisfying ; For the winding excitation current density, satisfying , The number of turns in the winding. For winding excitation current, This represents the cross-sectional area of ​​the winding. The vibration of the converter transformer core can be described as follows: the core undergoes periodic vibration in the insulating oil medium under electromagnetic excitation, and the main sources of electromagnetic excitation are Maxwell force and magnetostrictive force; its structural dynamic equation can be expressed as: ; In the formula: The mass matrix of the iron core; Here is the core displacement matrix; Here is the damping matrix; The stiffness matrix of the core can be calculated from the core material and structural dimensions. The Maxwell force matrix; The magnetostrictive force matrix; For time; The effect of core damping is ignored in the calculation, that is Therefore, the above formula simplifies to: ; In the formula: For the iron core to bear the force, ; S202, the calculation method for the sound pressure level inside and outside the fuel tank is as follows: By differentiating the core displacement over time, the core vibration velocity can be obtained; treating the air medium as an ideal fluid, the propagation of sound waves is described by the equations of motion, state, and continuity.

[0008] In the formula: For media density increment; For fluid density; , , They are iron core edges x, y, z The vibration velocity component in the direction; The sound pressure at various points in space; Speed ​​of sound; Convert sound pressure into sound pressure level It can be represented as: ; In the formula: This is the effective value of the sound pressure level; The reference sound pressure level is used.

[0009] Preferably, in step S3, the magnetic-structure-acoustic multi-field coupling method is as follows: S301, solve for the magnetic field distribution under winding excitation, and simultaneously calculate the amplitudes of Maxwell force and magnetostrictive force; S302 uses the Maxwell force and magnetostrictive force obtained from the solution as load terms to achieve coupled loading onto the structural field; S303, within the structural field solution domain, calculate the core vibration characteristic parameters of the core under load, vibration displacement and acceleration; S304 uses the structural stress distribution as the sound field excitation source to solve the key parameters of the sound field around the transformer, including sound pressure and sound pressure level, at different time points.

[0010] Preferably, in step S4, given that vibration has periodic evolution characteristics and structural symmetry, based on the results of structural field simulation analysis, an observation point is selected at the geometric center of the high vibration intensity region of the transformer to further explore the time-domain characteristic parameters of transformer vibration under power frequency excitation.

[0011] Preferably, in step S5, the front of the oil tank is equally divided into several rectangular partitions, and the geometric center point of each partition is used as the equivalent representation point. The vibration and sound field signals collected at the center point are the equivalent representation signals of the corresponding partition. After performing a fast Fourier transform on the time-domain signals of the observation point and each equivalent representation point, the intrinsic correlation characteristics between the partition vibration signals and the observation point signals are systematically explored by calculating and comparing the proportion of spectral components of the equivalent representation signals of each partition and the observation point signals. The relationship between the partition vibration signals and the observation point signals is explored.

[0012] Preferably, in step S6, based on step S5, the correlation between the monitoring point and the vibration source and acoustic signature signal is measured using the Pearson correlation coefficient. The Pearson correlation coefficient is defined as: ; In the formula: This is sample data for the iron core measuring points; This is a sample data of measuring points on the tank wall; , y and x are the sample means, respectively; Obtain the covariance for variables X and Y; and Let X and Y be the standard deviations. By combining the mapping relationship between the sound field inside and outside the transformer tank and the structural vibration, the correlation coefficient of the vibration signal and the correlation coefficient of the acoustic fingerprint signal are assigned weight coefficients and weighted. The result is defined as the sensitivity score of the vibration signal and acoustic fingerprint signal corresponding to the monitoring point. The area with the best comprehensive score value is selected, and then the optimal location for monitoring the tank wall is determined.

[0013] Preferably, in step S7, based on the optimal measuring point on the tank wall determined in step S6, vibration and acoustic frequency domain signals under normal operating conditions and various winding fault conditions are extracted respectively, and then the differences and variation laws of signal characteristics under different fault types and normal operating conditions in the frequency domain are systematically analyzed.

[0014] Preferably, in step S8, a TST-XGBoost fusion diagnostic model is constructed based on the frequency domain features extracted in step S7 under different operating conditions. The first layer of fault diagnosis focuses on classifying winding fault types, and the second layer of fault diagnosis focuses on identifying the degree of fault, thereby achieving accurate identification of winding fault types. The core calculation process is as follows: ; In the formula, Q, K, and V represent the query, key, and value matrices, respectively, and d k This is the dimension scaling factor; After obtaining the deep features extracted by TST, they are passed as input to the XGBoost ensemble learning framework for fault classification; XGBoost completes model training by optimizing the following objective function: ; In the formula: For multi-class log loss, These are genuine fault labels. This is a predicted value; ω is the regularization term, where T is the number of tree nodes, ω is the node weight, γ=0.1, and λ=1 are regularization parameters to improve the model's generalization ability. To integrate the advantages of TST and XGBoost, a weighted voting mechanism is adopted. Let the prediction probability of TST for fault category c be... XGBoost is The fusion probability is: ; In the formula: To verify the accuracy of the validation set, the final fault types were determined by... determination.

[0015] Beneficial effects of this invention: 1. This invention establishes a simulation model of a converter transformer through finite element simulation, and combines magnetic-structural-acoustic multi-physics coupling to study the distribution law of the transformer's magnetic field, vibration displacement, and noise distribution characteristics, which can provide a reference for transformer structural design, vibration reduction analysis, and fault diagnosis.

[0016] 2. This invention calculates the spectral correlation coefficients between the signals of each zone and the vibration and acoustic signature signals by dividing the surface of the oil tank into sections, thereby determining the optimal monitoring points on the transformer oil tank wall. This lays the foundation for condition monitoring and fault diagnosis of converter transformers based on vibration and acoustic signature characteristics.

[0017] 3. This invention performs layered fault diagnosis on converter transformer windings based on measurement point signals, achieving accurate identification of winding fault types. This research method can be extended to other types of power transformers.

[0018] 4. This invention achieves early warning, accurate identification and degree quantification of winding faults through a technical route of multi-source signal acquisition, refined feature extraction, hierarchical fusion diagnosis and interpretability optimization. Attached Figure Description

[0019] Figure 1 This is an overall flowchart of the present invention; Figure 2 This is a three-dimensional model diagram of the converter transformer in this invention; Figure 3 This is a diagram of the internal structure of the converter transformer in this invention, which mainly includes windings, core, and base. Figure 4 This is the magnetic field distribution of the converter transformer core in this invention; Figure 5 The vibration displacement distribution of the converter transformer core according to the present invention; Figure 6 The noise sound pressure level distribution inside the converter transformer tank of this invention; Figure 7 The sound pressure level distribution of the external noise of the converter transformer tank according to the present invention; Figure 8 This is the core observation point of the present invention; Figure 9 This is a diagram showing the front area division of the converter transformer tank according to the present invention; Figure 10 This is a heat map combining the observation points of the present invention with the correlation coefficients of the tank wall vibration and acoustic signature. Figure 11 Vibration signal spectrum diagrams extracted under different fault types and normal operating conditions; Figure 12 The results of transformer winding stratification fault diagnosis based on the optimal measuring point on the tank wall. Figure 1 ; Figure 13 The results of transformer winding stratification fault diagnosis based on the optimal measuring point on the tank wall. Figure 2 . Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: As Figure 1As shown, a method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis by integrating vibration acoustic signals includes the following steps: Step 1: Establish a simulation model of the converter transformer using finite element simulation software, and set the material properties and key parameters for each structural component; parameters include transformer capacity, rated current, Young's modulus, Poisson's ratio, and the geometric parameters of internal components; such as... Figure 2 and Figure 3 As shown, the model of the converter transformer consists of an oil tank, core, windings and base.

[0022] Step 2: For converter transformers, conduct research on the basic principles of their electromagnetic field, structural field and acoustic field, and calculate the vibration displacement of the transformer core and the magnitude of the sound pressure level inside and outside the oil tank; Step 3: Combining the magnetic-structural-acoustic multi-physics coupling, first calculate the magnetic field generated by the winding excitation, then calculate the Maxwell force and magnetostrictive force, and couple the calculated Maxwell force and magnetostrictive force as loads into the structural field. In the structural field, calculate the vibration parameters such as the core force, vibration displacement, and acceleration; in the acoustic field, calculate the acoustic field parameters such as the sound pressure inside and outside the transformer and the sound pressure level. like Figure 4 , Figure 5 , Figure 6 and Figure 7 As shown, when a 50Hz sinusoidal alternating current is applied to the transformer windings: 1. Figure 4 The magnetic flux density (T) distribution of the transformer core when a 50Hz sinusoidal alternating current is applied is shown: the two core column regions show obvious high magnetic flux density, which is the main flow path of magnetic flux, verifying the conclusion that the maximum magnetic flux density occurs in the two core columns. 2. Figure 5 The vibration displacement (μm) distribution of the transformer core structure when a 50Hz sinusoidal alternating current is applied is shown: the vibration is most intense at the upper yoke of the core, followed by the top of the core column, and the displacement distribution on the core structure is basically symmetrical. This indicates that the magnetic flux density is closely related to the magnetostriction effect of the core, and the area with larger displacement usually corresponds to the area with higher magnetic flux density.

[0023] 3. Figure 6 , Figure 7 The noise sound pressure level (dB) inside and outside the transformer core when a 50Hz sinusoidal alternating current is applied: the transformer noise is mainly distributed in the upper part of the yoke and the middle of the windings, which is consistent with... Figure 4 , Figure 5 The high degree of overlap between the upper yoke, where the vibration was most intense, and the top of the iron core column confirms the consistency between the noise source and the area of ​​intense vibration, meaning that the structural radiated noise caused by vibration mainly originates from these parts.

[0024] Step 4: Based on the displacement results, select observation points in areas of severe transformer vibration to further analyze the transformer's vibration parameters during the power frequency excitation cycle. Figure 8 This diagram illustrates the selection of observation points inside the transformer core. The observation points inside the core are used as reference measurement points to extract vibration signals.

[0025] Step 5: Divide the front of the oil tank into sections, perform Fast Fourier Transform on the time-domain signals of vibration in each section and the time-domain signals at the observation points, and select certain harmonic components as vibration parameters through comparative analysis. Use the Pearson correlation coefficient as an evaluation index for the correlation between the signals from the internal and external measurement points, reflecting the mapping relationship between the sound field and structural vibration inside and outside the transformer oil tank; Figure 9 As shown, the front of the fuel tank is divided into 50 rectangular regions. The center point of each region is taken as the equivalent point, and the signal measured at the center point is equivalent to the overall signal of its region.

[0026] Step 6: Take the average value of the correlation coefficient of the extracted vibration signal and the correlation coefficient of the acoustic fingerprint signal, and finally select the area with the higher average value as the best monitoring area of ​​the tank wall. Figure 10 The graph shows the mapping relationship between the inside and outside of the fuel tank. As can be seen from the graph, the regions with higher heat values ​​are 1, 2, 11, 20, 21, and 30. Step 7: Based on the optimal monitoring area determined in Step 6, extract the vibration signal and acoustic signature signal of each measuring point area under different fault types at 100Hz~800Hz. Then, perform noise reduction and normalization processing on the obtained signals. The extracted signal spectrum is shown below. Figure 11 As shown.

[0027] Step 8: Construct a TST-XGBoost fusion diagnostic model based on the extracted signal frequency domain features. Figure 12 The fault confusion matrix is ​​the first-level diagnosis based on the winding fault type, and its prediction accuracy is 99.45%. Figure 13 The fault confusion matrix is ​​a second-level diagnostic tool targeting the degree of winding fault, with a prediction accuracy of 98.74%. This result shows that the accuracy of layered winding fault diagnosis is 98.20%.

[0028] Example 2: The advantages of the method of the present invention compared with traditional fault diagnosis methods are as follows. Traditional fault diagnosis methods are based on the decomposition law of characteristic gases in transformer oil. They mainly rely on the proportions of characteristic gases such as H2, CH4, C2H6, C2H4, and C2H2 generated by the decomposition of insulating oil when faults such as overheating or discharge occur inside the transformer. This method selects three suitable gas ratios (such as CH4 / H2, C2H6 / CH4, and C2H4 / C2H6), and determines the fault type and nature by comparing the measured ratios with the threshold ranges specified in standards such as IEC60599. The traditional DGA three-ratio method, based on the principle of gas ratio threshold matching, has core defects such as low diagnostic accuracy, weak ability to identify mixed faults, and inability to adapt to the special operating conditions of converter transformers, making it difficult to meet the operation and maintenance needs of modern power systems. This patented TST-XGBoost hierarchical diagnostic method achieves a diagnostic accuracy of 98.20% through the fusion of vibration-acoustic multi-source signals and hierarchical diagnostic architecture design. It is significantly superior to traditional methods and existing machine learning solutions in terms of fault identification and fault degree identification, providing a reliable technical guarantee for the safe and stable operation of converter transformers.

[0029] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis by integrating vibration acoustic signals, characterized in that, Includes the following steps: S1. Based on the finite element simulation platform, a numerical model of the converter transformer is constructed, and the material properties and key parameters of each structural component are set. S2, based on the basic control equations of electromagnetic field, structural field and sound field, conducts multi-physics field coupling analysis to calculate the core vibration displacement and the sound pressure level distribution inside and outside the oil tank; S3, combining the magnetic-structural-acoustic multi-field coupling method, systematically analyzes the internal magnetic field distribution characteristics, vibration displacement law and noise propagation characteristics of the converter transformer, and makes a comprehensive comparison; S4. Based on the vibration displacement cloud map, observation points were selected in the area of ​​significant core vibration to investigate the vibration response characteristics of the transformer during the power frequency excitation cycle. S5 divides the front of the fuel tank into several areas, performs fast Fourier transform on the vibration and acoustic time-domain signals of each zone and the signals at the observation point, selects characteristic frequency components as input variables by comparing the spectrum, and uses the Pearson correlation coefficient as the evaluation index of signal correlation. S6. Based on step S5, comprehensively evaluate the correlation coefficient between the vibration signal and the acoustic signature signal, and combine the mapping relationship between the sound field inside and outside the transformer tank and the structural vibration to determine the optimal measurement point arrangement scheme on the tank wall. S7. Based on the optimal measuring point determined in step S6, extract the vibration frequency domain signals of each measuring point under fault and normal operating conditions, and conduct comparative analysis. S8 employs a time-series-based Transformer-XGBoost fusion voting mechanism model to achieve hierarchical fault diagnosis of transformer windings.

2. The method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S1, during the construction of the numerical model of the converter transformer based on the finite element simulation platform, the model covers the key structures of the core, winding, base and oil tank. In the simulation experiment stage, corresponding magnetic field boundary conditions are applied to the above components, and then their vibration and noise related parameters are calculated, so as to realize the multi-physics coupling simulation analysis of the converter transformer.

3. The method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S1, the key parameters include transformer capacity, rated current, Young's modulus, Poisson's ratio, and the geometric parameters of internal components.

4. The method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S2, the method for calculating the core vibration displacement and the sound pressure level distribution inside and outside the oil tank includes: S201, Calculation of core vibration excitation source: Under the condition of a sinusoidal alternating magnetic field at power frequency, neglecting the displacement current effect, the governing equations describing the magnetic field—Maxwell's equations and their constitutive relations—are as follows: ； In the formula: For curl operator; The magnetic field strength; Current density; It represents the magnetic flux density; The dielectric permeability; Introducing magnetic vector position ,satisfy The governing differential equation of the magnetic field can be obtained as follows: ； In the formula: The magnetic vector position; Let be the reluctance of the medium, satisfying ; For the winding excitation current density, satisfying , The number of turns in the winding. For winding excitation current, This represents the cross-sectional area of ​​the winding. The vibration of the converter transformer core can be described as follows: the core undergoes periodic vibration in the insulating oil medium under electromagnetic excitation, and the main sources of electromagnetic excitation are Maxwell force and magnetostrictive force; its structural dynamic equation can be expressed as: ； In the formula: The mass matrix of the iron core; Here is the core displacement matrix; Here is the damping matrix; The stiffness matrix of the core can be calculated from the core material and structural dimensions. The Maxwell force matrix; The magnetostrictive force matrix; For time; The effect of core damping is ignored in the calculation, that is Therefore, the above formula simplifies to: ； In the formula: For the iron core to bear the force, ; S202, the calculation method for the sound pressure level inside and outside the fuel tank is as follows: By differentiating the core displacement over time, the core vibration velocity can be obtained; treating the air medium as an ideal fluid, the propagation of sound waves is described by the equations of motion, state, and continuity. In the formula: For media density increment; For fluid density; , , They are iron core edges x, y, z The vibration velocity component in the direction; The sound pressure at various points in space; Speed ​​of sound; Convert sound pressure into sound pressure level It can be represented as: ； In the formula: This is the effective value of the sound pressure level; The reference sound pressure level is used.

5. The method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S3, the magnetic-structure-acoustic multi-field coupling method is as follows: S301, solve for the magnetic field distribution under winding excitation, and simultaneously calculate the amplitudes of Maxwell force and magnetostrictive force; S302 uses the Maxwell force and magnetostrictive force obtained from the solution as load terms to achieve coupled loading onto the structural field; S303, within the structural field solution domain, calculate the core vibration characteristic parameters of the core under load, vibration displacement and acceleration; S304 uses the structural stress distribution as the sound field excitation source to solve the key parameters of the sound field around the transformer, including sound pressure and sound pressure level, at different time points.

6. The method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S4, given that vibration has periodic evolution characteristics and structural symmetry, based on the results of structural field simulation analysis, an observation point is selected at the geometric center of the high vibration intensity region of the transformer to further explore the time-domain characteristic parameters of transformer vibration under power frequency excitation.

7. The method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S5, the front of the oil tank is equally divided into several rectangular partitions. The geometric center point of each partition is used as the equivalent representation point. The vibration and sound field signals collected at the center point are the equivalent representation signals of the corresponding partition. After performing a fast Fourier transform on the time-domain signals of the observation point and each equivalent representation point, the intrinsic correlation characteristics between the partition vibration signals and the observation point signals are systematically explored by calculating and comparing the proportion of spectral components of the equivalent representation signals of each partition and the observation point signals. The relationship between the partition vibration signals and the observation point signals is also explored.

8. The method for mapping the internal and external relationships of a converter transformer oil tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S6, based on step S5, the correlation between the monitoring point and the vibration source and acoustic signature signal is measured using the Pearson correlation coefficient. The Pearson correlation coefficient is defined as follows: ； In the formula: This is sample data for the iron core measuring points; This is a sample data of measuring points on the tank wall; , y and x are the sample means, respectively; Obtain the covariance for variables X and Y; and Let X and Y be the standard deviations. By combining the mapping relationship between the sound field inside and outside the transformer tank and the structural vibration, the correlation coefficient of the vibration signal and the correlation coefficient of the acoustic fingerprint signal are assigned weight coefficients and weighted. The result is defined as the sensitivity score of the vibration signal and acoustic fingerprint signal corresponding to the monitoring point. The area with the best comprehensive score value is selected, and then the optimal location for monitoring the tank wall is determined.

9. The method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S7, based on the optimal measuring points on the tank wall determined in step S6, vibration and acoustic frequency domain signals under normal operating conditions and various winding fault conditions are extracted respectively. Then, the differences and variation patterns of signal characteristics under different fault types and normal operating conditions in the frequency domain are systematically analyzed.

10. The method for mapping the internal and external relationships of a converter transformer tank and for fault diagnosis based on fused vibration acoustic signals according to claim 1, characterized in that, In step S8, a TST-XGBoost fusion diagnostic model is constructed based on the frequency domain features extracted in step S7 under different operating conditions. The first layer of fault diagnosis focuses on classifying winding fault types, and the second layer focuses on identifying the degree of fault, thereby achieving accurate identification of winding fault types. The core calculation process is as follows: ； In the formula, Q, K, and V represent the query, key, and value matrices, respectively, and d k This is the dimension scaling factor; After obtaining the deep features extracted by TST, they are passed as input to the XGBoost ensemble learning framework for fault classification; XGBoost completes model training by optimizing the following objective function: ； In the formula: For multi-class log loss, These are genuine fault labels. This is a predicted value; ω is the regularization term, where T is the number of tree nodes, ω is the node weight, γ=0.1, and λ=1 are regularization parameters to improve the model's generalization ability. To integrate the advantages of TST and XGBoost, a weighted voting mechanism is adopted. Let the prediction probability of TST for fault category c be... XGBoost is The fusion probability is: ； In the formula: To verify the accuracy of the validation set, the final fault types were determined by... determination.

Images