A method and system for detecting a turn-to-turn short circuit of a converter transformer based on acoustic characteristics

Abstract

The application discloses a kind of based on acoustic feature converter interturn short circuit detection method and system, method includes the following steps: constructing converter three-dimensional simulation model, setting electromagnetic-thermal-force-acoustic multi-physical field coupling, extracting temperature, equivalent stress and sound pressure level and normalizing processing, determine peak region and divide peak connected region;The spatial consistency factor between peak connected region is calculated, and acoustic sensitive area is identified;The acoustic thermal force joint sensitivity in acoustic sensitive area is calculated, to maximize total acoustic thermal force joint sensitivity as target, solve optimal monitoring point arrangement scheme;With optimal monitoring point arrangement scheme, monitoring point arrangement is carried out, the acoustic characteristic vector under different working conditions and actual working condition is calculated, and the actual state of converter under actual working condition is determined.The present application realizes the high-precision detection, accurate positioning of converter interturn short circuit fault by multi-physical field information, combined with acoustic sensitive area identification, monitoring point optimization arrangement and working condition matching analysis.

Description

Technical Field

[0001] This invention belongs to the field of power equipment fault technology, specifically relating to a method and system for detecting inter-turn short circuits in converter transformers based on acoustic characteristics. Background Technology

[0002] In ultra-high voltage direct current (UHVDC) transmission projects, the converter transformer, as the core equipment connecting the AC and DC sides, directly affects the stability and reliability of the entire DC system. The windings, as key load-bearing components subjected to electromagnetic, thermal, and mechanical forces, are highly susceptible to inter-turn short circuits under the long-term influence of various factors such as AC / DC mixed excitation, harmonic current superposition, DC bias, and frequent load fluctuations. Once an inter-turn short circuit occurs, the fault current rises abruptly within milliseconds, causing a rapid concentration of local electromagnetic forces and a rapid accumulation of Joule heat. Simultaneously, it leads to enhanced local structural vibration of the windings, resulting in significant changes in the acoustic response of the transformer casing and tank. Current research and engineering practice indicate that changes in acoustic radiation caused by inter-turn short circuits are among the earliest, most easily externally detectable, and highly correlated with internal faults in transformers. During a short-circuit fault, the enhanced pulsation of internal electromagnetic forces causes structural vibrations in the windings, core, and tank. This vibrational energy is radiated into the external air in the form of sound waves, forming a sound pressure distribution with specific directionality and peak characteristics. Therefore, the acoustic response provides an important external observation window for internal short-circuit faults in converter transformers.

[0003] However, current analysis and diagnosis of inter-turn short circuit conditions in converter transformers still mainly rely on traditional indicators such as oil temperature, vibration, and casing noise. These external quantities cannot accurately reflect the concentration area of ​​electromagnetic force and the location of local short circuits within the windings. Existing acoustic monitoring methods generally suffer from the following problems: acoustic signals are lagging and lack mechanistic correspondence. Traditional diagnostic methods based on noise changes mainly observe whether the noise amplitude increases, but fail to consider the electromagnetic-thermal-mechanical coupling mechanism caused by short circuits, lacking a correspondence between acoustic changes and internal short circuit mechanisms. Furthermore, they do not consider the differences in sound fields at different short circuit locations. When a short circuit occurs in different parts of the winding (upper, middle, and lower sections), the peak position of the electromagnetic force, the structural vibration modes, and the acoustic radiation characteristics of the oil tank are all different. External single-point acoustic monitoring cannot capture these differences, making it difficult to locate the short circuit. There is a lack of acoustic feature analysis methods based on physical field simulation. Although some existing studies use electromagnetic or thermal field simulation analysis, very few incorporate the sound field into a unified modeling system, and even fewer systematically analyze the mechanistic relationship between peak sound pressure, directivity changes, and structural vibration after an inter-turn short circuit. Monitoring point placement is based on experience, and the acoustic sensitive area is not clearly defined. Traditional acoustic monitoring usually places points at fixed locations without optimization based on peak sound pressure distribution or concentrated sound source energy areas, resulting in low short-circuit feature signal capture rate and unstable diagnostic results.

[0004] Currently, existing power equipment monitoring methods include: fixed-location monitoring methods, which deploy acoustic, vibration, and temperature probes on the outer surface of the transformer tank or at fixed structural locations, and conduct standardized monitoring according to a uniform deployment method; however, this method does not consider the differences in the distribution of electromagnetic fields inside the transformer and the sensitivity of different short-circuit locations; fixed measuring points may be far from areas of concentrated short-circuit current or peak mechanical stress, resulting in weak correlation between the acquired signals and the actual short-circuit state, and failing to support accurate assessment. Diagnostic methods based on external characteristics identify the short-circuit state by observing external features such as increased noise, abnormal vibration modes, or changes in gas in the oil, combined with limited physical quantities; however, appearance or external manifestations are not synchronized with changes in the internal electromagnetic-thermal-mechanical field, and there are often no obvious external characteristics in the early stages of a short circuit; external noise, load fluctuations, and other factors interfere with the accuracy of judgment, easily leading to missed detections or misjudgments.

[0005] For example, patent application CN117763907A discloses a simulation method for inter-turn short circuits in converter transformers based on bidirectional field-circuit coupling. This method establishes a bidirectional coupling model between the electromagnetic field and the external circuit to simulate and analyze the current distribution, electromagnetic parameter changes, and electrical response characteristics under inter-turn short circuit conditions, thereby achieving electromagnetic behavior modeling and evaluation of converter transformers under inter-turn short circuit conditions. This method mainly focuses on the impact of short circuits on electromagnetic parameters and electrical characteristics, without further addressing the structural mechanical response and acoustic radiation characteristics caused by short circuits, nor establishing a correspondence between acoustic response and short circuit location. Patent application CN117875082A discloses a simulation method and system for inter-turn short circuit faults in dry-type air-core reactors. This method, based on a reactor structural model, simulates the electromagnetic and thermal effects under short circuit conditions by setting equivalent parameters for inter-turn short circuits, thereby achieving analysis of the inter-turn short circuit fault characteristics of dry-type air-core reactors. This technical solution is applicable to dry-type reactor structures, focusing on the simulation of electromagnetic and thermal effects of short-circuit faults. It does not consider the complex structural vibration and sound field propagation problems in oil-immersed converter transformers, nor does it address the application of acoustic characteristics in short-circuit diagnosis. Patent application CN119692112A provides a method and system for constructing a simulation model of ultrasonic propagation in transformer partial discharge. This method constructs a propagation model of ultrasonic waves in the transformer's internal medium and structure, analyzes the propagation path and attenuation characteristics of ultrasonic signals generated by partial discharge, and thus achieves simulation research on the acoustic detection mechanism of partial discharge. This method uses partial discharge as the sound source and focuses on the propagation law of ultrasonic waves, without addressing the acoustic response mechanism caused by electromagnetic forces and structural vibrations under electromagnetic fault conditions such as inter-turn short circuits.

[0006] The shortcomings of existing technologies include: (1) The monitoring logic lacks a physical field basis and does not consider the electromagnetic-thermal-mechanical coupling distribution characteristics. The existing point layout is not planned based on the internal electromagnetic force concentration area, hot spot sensitive area or strong sound field directional area, resulting in a low capture rate of key fault features and an inability to support accurate short circuit feature identification and reliable diagnosis. (2) The monitoring location is random or fixed, ignoring the differences in winding structure and the influence of operating conditions, resulting in poor universality and low consistency. The distribution of multiple physical fields varies significantly under different winding types, load levels and DC bias conditions, and fixed or empirical monitoring methods cannot adapt to these differences; the acquisition results lack consistency and it is difficult to form a standardized and reproducible short circuit monitoring and evaluation process. Summary of the Invention

[0007] To address the technical bottlenecks in existing acoustic diagnostic methods, such as the lack of a real excitation source and the difficulty in quantitatively characterizing the differences between multiple fields at different short-circuit locations, this invention provides a method and system for detecting inter-turn short circuits in circulating transformers based on acoustic features. The method includes the following steps: constructing a three-dimensional simulation model of the converter transformer; setting up electromagnetic-thermal-mechanical-acoustic multi-physics coupling; extracting and normalizing temperature, equivalent stress, and sound pressure level; determining peak regions and dividing peak connected regions; calculating the spatial consistency factor between peak connected regions and identifying acoustically sensitive areas; calculating the combined acoustic-thermal-mechanical sensitivity within the acoustically sensitive areas; solving for the optimal monitoring point layout scheme with the goal of maximizing the total combined acoustic-thermal-mechanical sensitivity; arranging monitoring points according to the optimal scheme; calculating acoustic feature vectors under different operating conditions and actual operating conditions; and determining the actual state of the converter transformer under actual operating conditions. This invention achieves high-precision detection and accurate location of inter-turn short-circuit faults in converter transformers by combining multi-physics information, acoustically sensitive area identification, optimized monitoring point layout, and operating condition matching analysis.

[0008] The present invention adopts the following technical solution: The first aspect of the present invention provides a method for detecting inter-turn short circuits in converter transformers based on acoustic characteristics, comprising: S1. Construct a three-dimensional simulation model of the converter transformer, mesh the three-dimensional simulation model of the converter transformer, set electromagnetic-thermal-mechanical-acoustic multi-physics coupling, and set boundary conditions; S2. After the multiphysics solution is completed, extract the temperature, equivalent stress and sound pressure level at each spatial location along the preset path; S3. Normalize the collected temperature, equivalent stress, and sound pressure level to determine the peak regions of different physical quantities and divide them into multiple peak connected regions; calculate the spatial weighting center of the peak connected region for each physical quantity, and determine the spatial consistency factor between the peak connected regions of different physical quantities by combining the predefined spatial deviation distance; when the peak connected region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, the region is determined to be the acoustically sensitive area of ​​the converter transformer. S4. Calculate and normalize the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determine the multi-field gradient coordination factor, and calculate the acoustic-thermal-mechanical joint sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor; initialize the monitoring point layout scheme, and solve for the optimal monitoring point layout scheme with the goal of maximizing the total acoustic-thermal-mechanical joint sensitivity of all monitoring points. S5. Layout the monitoring points according to the optimal monitoring point layout scheme, obtain the sound pressure level data under different working conditions through simulation, calculate the corresponding sound pressure level peak, increment and gradient to form the acoustic feature vector; collect the sound pressure, temperature and equivalent stress data under actual working conditions in real time, determine the corresponding actual acoustic feature vector; calculate the working condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different working conditions, and determine the actual state of the converter transformer.

[0009] Preferably, the process of dividing the peak connected region in S3 is as follows: For any physical quantity at each spatial location, if each physical quantity at the corresponding spatial location is not less than the predefined judgment threshold of the corresponding physical quantity, the corresponding spatial location is taken as the peak region of the corresponding physical quantity, and the peak regions of temperature, stress and sound pressure level are extracted respectively. For each physical quantity's peak region, if the spatial distance between any two peak points in the corresponding peak region is not greater than the predefined neighborhood radius, it indicates that the corresponding peak points are connected. The corresponding peak points are then divided into the same connected region, generating multiple peak connected regions for the corresponding physical quantity.

[0010] Preferably, the process of calculating the spatial weighted center of the peak connected region for each physical quantity in S3 is as follows: For any peak connected region of a physical quantity, the normalized value of the physical quantity at each peak point in the corresponding peak connected region is multiplied by the corresponding spatial coordinate and summed to form the numerator. The normalized physical quantities at all peak points in the corresponding peak connected region are summed to form the denominator. The spatial weighted center of the peak connected region of the corresponding physical quantity is then calculated.

[0011] Preferably, the process of determining the spatial consistency factor between the peak connected regions of different physical quantities in S3 is as follows: From the peak connected regions of different physical quantities, select any peak connected region and match it into a peak connected region group to generate multiple peak connected region groups. For each peak connected region group, calculate the sum of squared distances between the spatial weighted centers of the peak connected regions of different physical quantities and sum them as the numerator. Use the square of 3 times the predefined spatial deviation distance as the denominator to calculate the spatial exponential coefficient. Use the negative of the spatial exponential coefficient as the exponent to calculate the value of the predefined exponential function as the spatial consistency factor of the corresponding peak connected region group.

[0012] Preferably, the process of determining the acoustically sensitive area in S3 is as follows: For each peak connected region group, calculate the sum of squared distances between the spatial weighted centers of the peak connected regions for different physical quantities, take the average and take the square root, and calculate the spatial matching distance of the corresponding peak connected region group. For each peak connectivity region group, when the corresponding spatial matching distance is not greater than the predefined spatial deviation distance, the temperature peak connectivity region, stress peak connectivity region, and sound pressure level peak connectivity region of the corresponding peak connectivity region group are considered to be spatially consistent and are regarded as an effective multi-field peak matching region; if there are multiple multi-field peak matching regions, the multi-field peak matching region with a spatial consistency factor greater than the predetermined consistency threshold is selected as the corresponding acoustic sensitive region.

[0013] Preferably, the process of determining the multi-field gradient coordination factor in S4 is as follows: For each physical quantity at each spatial location within the acoustically sensitive area, the difference between corresponding physical quantities at adjacent spatial locations is used as the numerator, and the distance between adjacent spatial locations is used as the denominator. The gradient of the corresponding physical quantity at the corresponding spatial location is calculated and normalized. The multi-field gradient coordination factor at the corresponding spatial location is calculated by multiplying the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at the corresponding spatial location as the numerator, and taking the maximum value of the product of the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at all spatial locations as the denominator.

[0014] Preferably, the process of calculating the combined acoustic, thermal, and mechanical sensitivity at different spatial locations in S4 is as follows: Multiply the multi-field gradient coordination factor at the corresponding spatial location by the predefined gradient coupling correction coefficient and add 1 to obtain the multi-field coordination coefficient; multiply the normalized temperature, equivalent stress, and sound pressure level at the corresponding spatial location as the base, calculate the value of the predefined power function, and multiply the corresponding value by the multi-field coordination coefficient and the multi-field spatial consistency factor of the peak connected region group to which the corresponding spatial location belongs in turn to obtain the corresponding acoustic-thermal-mechanical joint sensitivity.

[0015] Preferably, the process of solving for the optimal monitoring point layout scheme in S4 is as follows: All spatial locations within the acoustically sensitive area are used as candidate locations to generate a candidate location set. Under the constraint of the upper limit of candidate monitoring points, candidate locations are selected from the candidate location set as candidate monitoring points, and a monitoring point layout scheme is randomly generated. The upper limit constraint of candidate monitoring points is that the number of candidate monitoring points in the monitoring point layout scheme is not greater than the predetermined upper limit of the number of monitoring points. The combined acoustic, thermal, and mechanical sensitivities at all candidate monitoring points in the monitoring point layout scheme are summed and used as the objective function for monitoring point optimization. Under the set constraints, the objective function for monitoring point optimization is maximized to solve for the optimal monitoring point layout scheme.

[0016] Preferably, the constraints include: monitoring point sensitivity constraints, spatial spacing constraints, and information redundancy constraints; Among them, the sensitivity constraint of the monitoring point is that the combined acoustic, thermal and mechanical sensitivity of each candidate monitoring point in the monitoring point layout scheme is not less than the predetermined sensitivity threshold; the spatial distance constraint is that the spatial distance between any two candidate monitoring points in the monitoring point layout scheme is not less than the predetermined minimum layout distance; the information redundancy constraint is that the difference in the combined acoustic, thermal and mechanical sensitivity at any two candidate monitoring points with a spatial distance less than the predetermined neighborhood distance threshold is not less than the predefined lower limit of difference.

[0017] Preferably, the process of constructing the acoustic feature vector in S5 is as follows: The monitoring points are arranged according to the optimal monitoring point layout scheme. The sound pressure level data of each monitoring point under different working conditions are collected through multi-physics field coupling simulation. For any working condition, the maximum sound pressure level among all monitoring points under the corresponding working condition is taken as the peak sound pressure level. For each monitoring point in the optimal monitoring point layout scheme, calculate the difference in sound pressure level at the corresponding monitoring point under different operating conditions and normal operating conditions, and use it as the sound pressure level increment; calculate the sound pressure level gradient at each monitoring point in the optimal monitoring point layout scheme. For any operating condition, the sound pressure level obtained from the simulation of each monitoring point in the optimal monitoring point layout scheme is spliced ​​together to obtain the acoustic feature vector under the corresponding operating condition.

[0018] Preferably, the process of determining the actual acoustic feature vector in S5 is as follows: The collected sound pressure data is preprocessed to obtain the preprocessed sound pressure signal; the time window is divided based on the predefined time window length; for any time window, the root mean square of the preprocessed sound pressure signal at each monitoring point within the corresponding time window is calculated as the effective sound pressure value at the corresponding monitoring point; for any monitoring point, the effective sound pressure value at the corresponding monitoring point is converted into a sound pressure level based on the predefined reference sound pressure. The sound pressure levels of all monitoring points within the corresponding time window are spliced ​​together to form the actual monitored sound pressure vector. The actual sound pressure level peak, sound pressure level increment, and sound pressure level gradient are calculated respectively, and then spliced ​​together to obtain the actual acoustic feature vector.

[0019] Preferably, the process of determining the actual state of the converter transformer in S5 is as follows: For any operating condition, calculate the distance between the actual sound pressure level increment at all monitoring points and the sound pressure level increment under the corresponding operating condition, as the sound pressure level increment term; calculate the distance between the actual sound pressure level gradient at all monitoring points and the sound pressure level gradient under the corresponding operating condition, as the sound pressure level gradient term; take the absolute value of the difference between the actual sound pressure level peak and the sound pressure level peak under the corresponding operating condition as the sound pressure level peak term; based on predefined weighting coefficients, weight the sound pressure level increment term, the sound pressure level gradient term, and the sound pressure level peak term to obtain the operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under the corresponding operating condition. The operating condition with the smallest matching distance to the actual acoustic feature vector is taken as the matching operating condition; if the matching operating condition is a normal operating condition and the maximum value of the actual sound pressure level increment in all monitoring points is not greater than the predefined increment threshold, then the converter transformer is in normal condition. If the matching condition is normal and the maximum value of the actual sound pressure level increment at all monitoring points is greater than the predefined increment threshold, then the converter transformer is in an uncertain state. If the matching condition is not a normal operating condition, the converter transformer is in a fault state, and the fault condition corresponding to the matching condition is taken as the fault type of the converter transformer.

[0020] Preferably, if the converter transformer is in a fault state, for each monitoring point, the temperature, equivalent stress and sound pressure level at the corresponding monitoring point are weighted by a predefined physical quantity weighting coefficient to construct a multi-field consistency discrimination function at the corresponding monitoring point; the monitoring point with the largest multi-field consistency discrimination value is taken as the fault occurrence area.

[0021] A second aspect of the present invention provides an acoustic feature-based inter-turn short-circuit detection system for converter transformers, employing an acoustic feature-based inter-turn short-circuit detection method for converter transformers, comprising: The simulation model building module constructs a three-dimensional simulation model of the converter transformer, performs mesh generation on the three-dimensional simulation model of the converter transformer, sets electromagnetic-thermal-mechanical-acoustic multiphysics coupling, and sets boundary conditions; After the multiphysics solution is completed, the simulation data acquisition module extracts the temperature, equivalent stress and sound pressure level at each spatial location along a preset path. The acoustic sensitive area determination module normalizes the collected temperature, equivalent stress, and sound pressure level, identifies the peak regions of different physical quantities, and divides them into multiple peak connected regions. It calculates the spatial weighting center of the peak connected region for each physical quantity and, combined with a predefined spatial deviation distance, determines the spatial consistency factor between the peak connected regions of different physical quantities. When the peak connected region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, the region is determined to be the acoustic sensitive area of ​​the converter transformer. The monitoring point layout optimization module calculates and normalizes the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determines the multi-field gradient coordination factor, and calculates the combined acoustic-thermal-mechanical sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor. It initializes the monitoring point layout scheme and solves for the optimal monitoring point layout scheme with the goal of maximizing the total combined acoustic-thermal-mechanical sensitivity of all monitoring points. The transformer short-circuit detection module arranges monitoring points according to the optimal monitoring point layout scheme. It acquires sound pressure level data under different operating conditions through simulation, calculates the corresponding sound pressure level peak, increment, and gradient to form an acoustic feature vector. It also collects sound pressure, temperature, and equivalent stress data under actual operating conditions in real time to determine the corresponding actual acoustic feature vector. Finally, it calculates the operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different operating conditions to determine the actual state of the converter transformer.

[0022] A third aspect of the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of a converter transformer inter-turn short-circuit detection method based on acoustic features.

[0023] A fourth aspect of the invention provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of a method for detecting inter-turn short circuits in a converter transformer based on acoustic features.

[0024] Compared with the prior art, the beneficial effects of the present invention include at least the following: 1. This invention constructs a multi-physics coupling model of converter transformers, integrating electromagnetic, thermal, mechanical, and acoustic fields. It comprehensively analyzes multi-source information such as temperature, equivalent stress, and sound pressure level, moving beyond fault diagnosis methods limited to single acoustic parameters. Through peak connectivity region extraction, spatial weighting center calculation, and spatial consistency factor construction, it achieves cross-validation and collaborative identification of multiple abnormal regions. This effectively eliminates misjudgments caused by local noise, environmental interference, and single-physics anomalies, improving the correspondence between fault characteristics and actual fault locations. Consequently, it significantly enhances the accuracy, stability, and reliability of inter-turn short-circuit fault detection in converter transformers.

[0025] 2. Based on the acoustic sensitive area identification results, this invention further introduces multi-field gradient coordination factors, spatial consistency factors, and acoustic-thermal joint sensitivity evaluation to optimize the monitoring point layout. Under constraints such as the number of monitoring points, spatial spacing, and information redundancy, the optimal monitoring point layout scheme is determined with the goal of maximizing joint sensitivity, ensuring that monitoring points preferentially cover fault-sensitive areas and areas with significant multi-field coupling changes. Compared to empirical or uniform point distribution methods, this invention can acquire more effective fault information under limited monitoring resources, improving the coverage and fault response capabilities of the monitoring system, while reducing monitoring costs and data redundancy.

[0026] 3. This invention utilizes multiphysics simulation to establish an acoustic feature library under different operating conditions, and combines this with actually collected sound pressure, temperature, and stress data to construct actual acoustic feature vectors. Accurate differentiation between normal, fault, and uncertain states is achieved through operating condition matching distance. Furthermore, after a fault is detected, a multi-field consistency discriminant function is constructed using temperature, stress, and sound pressure level to perform location analysis of the fault area. This not only identifies the fault type but also determines the fault location. This provides more accurate data support for the operation, maintenance, and fault diagnosis of converter transformers, improving equipment operating safety and maintenance efficiency. Attached Figure Description

[0027] Figure 1 A flowchart of a method for detecting inter-turn short circuits in a circulating current transformer based on acoustic features, provided by the present invention; Figure 2 This is a schematic diagram of the converter transformer simulation model in an embodiment of the present invention; Figure 3 This is a schematic diagram of the mesh division results in an embodiment of the present invention; Figure 4 This is a schematic diagram of the data extraction path in an embodiment of the present invention; Figure 5 This is a temperature distribution diagram of the upper right and middle right short-circuit positions of the converter transformer in an embodiment of the present invention. Figure 6 This is a temperature distribution diagram of the upper right and middle right short-circuit detection paths of the converter transformer in an embodiment of the present invention; Figure 7 This is a stress distribution diagram at the corresponding positions of the upper right and middle right short circuits of the converter transformer in an embodiment of the present invention; Figure 8 This is a stress distribution diagram of the upper right and middle right short-circuit detection paths of the converter transformer in an embodiment of the present invention; Figure 9 This is a sound pressure level distribution diagram at the corresponding positions of the upper right and middle right short circuits of the converter transformer in an embodiment of the present invention; Figure 10This is a sound pressure level distribution diagram of the upper right and middle right short-circuit detection paths of the converter transformer in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0029] Example 1 Embodiment 1 of the present invention provides a method for detecting inter-turn short circuits in a circulating current transformer based on acoustic characteristics, see reference. Figure 1 This includes the following steps: S1. Construct a three-dimensional simulation model of the converter transformer, mesh the three-dimensional simulation model of the converter transformer, set electromagnetic-thermal-mechanical-acoustic multi-physics coupling, and set boundary conditions.

[0030] A three-dimensional simulation model of the converter transformer was constructed. The three-dimensional simulation model of the converter transformer was meshed and electromagnetic-thermal-mechanical-acoustic multi-field coupling was set.

[0031] The three-dimensional simulation model of the converter transformer established in this embodiment covers key structures such as the core, grid-side winding, valve-side winding, insulating oil, clamps, and the air domain of the tank, and can completely characterize the propagation path of inter-turn short circuits in electromagnetic, thermal, mechanical, and acoustic physical fields. To improve computational efficiency while ensuring calculation accuracy, some small structures (such as fasteners and small chamfers) have been appropriately simplified. The voltage regulating winding has limited coupling effect on the main magnetic circuit under actual operation and has a relatively small impact on key short-circuit parameters (such as winding losses and stress response). Therefore, this embodiment does not include the voltage regulating winding in the final simulation scope, focusing instead on the response characteristics of the dominant part of the short-circuit coupling physical field. The basic physical properties of the core, conductors, insulating oil, and structural materials are shown in Table 1, and the simulation model is shown in Figure 2.

[0032] Table 1 Material physical parameters of various parts in the simulation model

[0033] To ensure the accuracy and stability of the multiphysics coupling simulation results for inter-turn short circuits, this embodiment adopts a partitioned meshing strategy for the simulation model based on the differences in physical characteristics of different structural regions within the converter transformer. Specifically, for areas prone to inter-turn short circuits, contact areas between the winding and insulation structure, and critical areas with drastic changes in electromagnetic force, temperature rise, and structural stress, a local mesh refinement method is used for discretization to improve the ability to distinguish local field gradient changes. The minimum mesh element size within the critical region is controlled within the millimeter range to meet the computational accuracy requirements of multiphysics coupling solutions. For structural regions far from the inter-turn short circuit area and with relatively gentle field changes, a progressively thicker meshing strategy is adopted, allowing the mesh size to transition smoothly along the spatial direction. This effectively reduces the overall mesh size while ensuring computational stability, thus balancing simulation accuracy and computational efficiency. Through this combination of partitioned refinement and gradual meshing, the multiphysics response in the critical inter-turn short circuit region can be finely characterized, while avoiding the waste of computational resources caused by an excessively dense global mesh. The meshing results are shown in Figure 3.

[0034] As a preferred implementation method, four physical domains—electromagnetic field, temperature field, structural mechanical field, and acoustic field—are established in the three-dimensional simulation model of the converter transformer; the specific process is as follows: In an electromagnetic field, the rated operating current is applied to the three-dimensional simulation model of the converter transformer, and the equivalent resistance of the inter-turn short circuit is added at a predetermined short-circuit location. The potential distribution and magnetic field distribution under current variation are determined through the magnetic field control equation; the specific formula is as follows: ; in, It is a magnetic vector potential; Permeability; The applied current density will change when a short-circuit current is applied. Indicates the current density inside the conductor; The conductivity of the conductor; Potential; Expressing the calculation of curl; For gradient operators; The electric field strength is determined based on the current density inside the conductor; the Joule heat power generated per unit volume is determined by multiplying the electric field strength by the current density inside the conductor, which serves as the heat source for the thermal field; specifically expressed as: ; In the formula, The Joule heat power generated per unit volume; Electric field strength; The curl of the magnetic vector potential is calculated to obtain the magnetic flux density; using the Lorentz force formula, combined with the current density inside the conductor and the magnetic flux density, the electromagnetic force density is calculated as the volume load term of the mechanical field; specifically expressed as: ; In the formula, Electromagnetic force density (from electromagnetic field); It represents the magnetic flux density; In the temperature field, Joule heat and iron loss are used as volumetric heat sources. A transient heat conduction control equation for the temperature field is established, and the Joule heat output from the electromagnetic field is applied to the grid cells of the winding and core. The temperature field distribution is obtained by solving the transient heat conduction control equation. The specific formula for the transient heat conduction control equation is as follows: ; in, Density; The heat capacity is at constant pressure; T is temperature; t is time. Thermal conductivity; As a volumetric heat source, in this embodiment it comes from Joule heating and iron loss; Iron loss; For divergence operators; By setting convective heat transfer boundary conditions at the interface between solids and insulating oil (such as the outer surface of the winding, the contact surface between the core surface and the transformer oil), the temperature rise distribution characteristics at various points in the winding, core, and insulating oil are obtained; the specific expression of the convective heat transfer boundary conditions is as follows: ; In the formula; The convective heat transfer coefficient; The solid surface temperature; The ambient fluid temperature; In a mechanical field, thermal stress is calculated based on the temperature field distribution and the coefficient of thermal expansion. A geometric equation is used to construct the relationship between total stress and displacement. Subtracting the thermal stress from the total strain yields the relationship between elastic strain and displacement. The specific formula is as follows: ; The first term is a geometric equation; For total strain; It is a displacement vector; For elastic strain; Thermal strain; The coefficient of thermal expansion; For reference temperature; It is a second-order unit tensor; Based on the stress-strain relationship, and combined with the expression for the relationship between elastic strain and displacement, the expression for the relationship between stress tensor and displacement is determined; specifically, it is expressed as: ; In the formula, This is the fourth-order elastic stiffness tensor; It is the elastic strain tensor; During the simulation, structural mechanics follows the linear elastic constitutive relation (i.e., stress-strain relationship) and the structural equilibrium equations. By setting the structural equilibrium equations and loading the thermal stress, electromagnetic force density, and the relationship between stress tensor and displacement, the displacement vector distribution is output. Based on the displacement vector, the stress tensor, and the relationship between stress tensor and displacement, the stress tensor is determined, and the equivalent stress distribution is calculated. The structural equilibrium equations are as follows: ; in, For stress tensor; It is a displacement vector; For divergence operators; In an acoustic field, structural vibration velocity is applied as the sound source boundary, and the Helmholtz equation is used to describe the propagation of pressure waves; the frequency domain sound field governing equation is: ; in, Sound pressure level; For wave number, Angular frequency, Speed ​​of sound; A coupling boundary between structural vibration and the sound field is set at the contact surfaces of the windings, core, inner wall of the tank, and transformer oil. That is, the structural vibration velocity is used as the boundary velocity and loaded into the acoustic domain, outputting the sound pressure level distribution in the internal and external air domains of the tank. The specific expression of the boundary velocity is as follows: ; in, The density of the medium; The displacement vector is the structural displacement. For the boundary normal.

[0035] S2. After the multiphysics solution is completed, the temperature, equivalent stress and sound pressure level at each spatial location are extracted along the preset path to obtain the basic data reflecting the multiphysics response under the inter-turn short circuit condition.

[0036] After the model is solved, a fixed data extraction path is set to quantitatively compare the physical field response under different fault conditions (e.g., Figure 4 (As shown). This path allows for the acquisition of spatial distribution data for sound pressure level, temperature field, and stress field.

[0037] In this embodiment, the temperature rise, structural stress concentration, and acoustic response caused by the inter-turn short circuit exhibit obvious distribution characteristics along the winding axis; the winding axis is one of the directions where the coupling effects of short-circuit electromagnetic force, thermal stress, and structural vibration are most concentrated; selecting the winding center axial path can avoid the interference of edge structures and local geometric discontinuities on the multiphysics results, which is beneficial for obtaining representative field quantity variation laws. A fixed data extraction path is set (e.g., Figure 4 As shown (in the diagram), the path is set along the axial direction of the converter transformer winding, located within the plane of symmetry of the winding's central axis. The starting point is in the lower region of the winding, and the ending point extends to the upper region, traversing the entire effective height range of the winding. In the 3D model, this path is represented as a linear path along the winding height, used to continuously extract physical quantities such as temperature, stress, and sound pressure at different spatial locations. By simultaneously sampling the temperature field, structural mechanical field, and acoustic field results along this fixed data extraction path, a continuous distribution curve of the multi-physics response varying with spatial location can be obtained. This curve is used for subsequent multi-field consistency analysis, acoustic sensitive area identification, and acoustic feature comparison analysis.

[0038] S3. Normalize the collected data. Based on the normalized data, determine the peak regions of different physical quantities and divide them into multiple peak connected regions. Calculate the spatial weighting center of the peak connected region for each physical quantity. Combined with the predefined spatial deviation distance, determine the spatial consistency factor between the peak connected regions of different physical quantities. When the peak connected region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, determine that the region is the acoustically sensitive area of ​​the converter transformer.

[0039] To address the lack of clear physical evidence for acoustic monitoring locations, this embodiment does not analyze the acoustic field in isolation. Instead, it constructs a multi-field consistency discriminant function based on the joint response characteristics of the temperature field, structural stress field, and acoustic field. The multi-field consistency discriminant function comprehensively considers the following factors: the location of local temperature rise peaks and their spatial gradient changes in the temperature field; the equivalent stress peaks and stress concentration regions in the structural stress field; and the distribution of sound pressure level peaks and their directivity characteristics in the acoustic field. This function is used to comprehensively evaluate the degree of consistency in the responses of the temperature field, structural stress field, and acoustic field at the same spatial location.

[0040] The temperature, equivalent stress, and sound pressure level at each spatial location along the preset path are normalized; specifically, this is expressed as follows: ; in, The normalized temperature at spatial location x; Indicates the spatial position x on the preset path; This represents the temperature at spatial location x. The normalized equivalent stress at spatial location x; This represents the equivalent stress at spatial location x; The normalized sound pressure level at spatial location x; This represents the sound pressure level at spatial location x; , These represent the maximum and minimum temperatures along the preset path, respectively. , These represent the maximum and minimum values ​​of the equivalent stress along the preset path, respectively. , These represent the maximum and minimum sound pressure levels along the preset path, respectively. For any physical quantity at each spatial location, if each physical quantity at that location is not less than a predefined threshold, the corresponding spatial location is taken as the peak region of the corresponding physical quantity, and the peak regions of temperature, stress, and sound pressure level are extracted respectively; specifically expressed as: ; in, , , These are the peak regions of temperature, stress, and sound pressure level, respectively; that is, the set of peak points. , , The predefined threshold values ​​for temperature, stress, and sound pressure level range from 0.7 to 0.9. In temperature, stress, and acoustic fields, peak regions may not be single points but multiple continuous regions. Therefore, the peak regions of each physical quantity are first divided into several connected sub-regions, and the weighting center of each peak sub-region is calculated. For each physical quantity's peak region, if the spatial distance between any two peak points in the corresponding peak region is not greater than a predefined neighborhood radius, it indicates that the corresponding peak points are connected. The corresponding peak points are then divided into the same connected region, generating multiple peak connected regions for the corresponding physical quantity. Specifically: ; in, This represents the connected region of the m-th temperature peak. This represents the connected region of the nth stress peak. Let be the peak connected region of the qth sound pressure level; m, n, and q are the number of peak connected regions of temperature, stress, and sound pressure level, respectively. For the peak connected regions of temperature, stress, and sound pressure level, their weighting centers are defined respectively. For the peak connected region of any physical quantity, the normalized physical quantity value at each peak point in the corresponding peak connected region is multiplied by the corresponding spatial coordinate and summed to obtain the numerator. The normalized physical quantities at all peak points in the corresponding peak connected region are summed to obtain the denominator. The spatial weighting center of the peak connected region of the corresponding physical quantity is then calculated. Specifically, it is expressed as follows: ; in, This represents the connected region of the a-th temperature peak. This represents the connected region of the b-th stress peak. This represents the connected region of the c-th sound pressure level peak. , , These are the spatial weighting centers of the corresponding peak connected regions; Construct a spatial consistency factor among the peak regions of the three fields; arbitrarily select a peak connected region from the peak connected regions of different physical quantities, match them into a peak connected region group, and generate multiple peak connected region groups; for each peak connected region group, calculate the sum of squared distances between the spatial weighted centers of the peak connected regions of different physical quantities and sum them as the numerator, use the square of 3 times the predefined spatial deviation distance as the denominator, and calculate the spatial exponent coefficient; use the negative of the spatial exponent coefficient as the exponent, calculate the value of the predefined exponent function as the spatial consistency factor of the corresponding peak connected region group; specifically expressed as: ; in, The spatial consistency factor is the connection region between the temperature peak connectivity region a, the stress peak connectivity region b, and the sound pressure level peak connectivity region c. Allowable spatial deviation distance; This represents the Euclidean distance; the closer the peak temperature, peak stress, and peak sound pressure level are to each other... The closer to 1; when the spatial distance between the three is large, Approaching 0; In this embodiment, the spatial consistency factor among the three peak regions is not simply determined by whether a certain point simultaneously meets the threshold. Instead, the peak regions of the three physical fields—temperature, stress, and sound pressure—are matched as spatial objects, and the spatial coupling degree of the three types of peak regions is quantitatively described by an exponential decay function. This method can adapt to the actual situation where peak regions are multi-point, multi-block, and not completely overlapping, and is more consistent with the complex multi-physics propagation characteristics inside the converter transformer.

[0041] As a preferred implementation, for each peak connected region group, the sum of squared distances between the spatial weighted centers of the peak connected regions for different physical quantities is calculated, and the average is taken to obtain the square root of the sum of squared distances. This is then used to calculate the spatial matching distance for the corresponding peak connected region group; specifically, it is expressed as follows: ; in, The spatial matching distance between temperature peak connected region a, stress peak connected region b, and sound pressure level peak connected region c; For each peak connected region group, when the corresponding spatial matching distance is not greater than the predefined spatial deviation distance, i.e. At that time, the temperature peak connectivity region, stress peak connectivity region, and sound pressure level peak connectivity region of the corresponding peak connectivity region group are considered to have spatial consistency and are thus considered as a valid multi-field peak matching region; if multiple multi-field peak matching regions exist, then a region with a spatial consistency factor greater than a predetermined consistency threshold is selected. The multi-field peak matching region is used as the corresponding acoustic sensitive area.

[0042] This embodiment does not simply calculate the distance between all peak points pairwise, but first merges the peak points into peak regions, and then performs spatial combination matching of peak regions of different physical fields.

[0043] S4. Calculate and normalize the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determine the multi-field gradient coordination factor, and calculate the acoustic-thermal-mechanical joint sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor; initialize the monitoring point layout scheme, and solve for the optimal monitoring point layout scheme with the goal of maximizing the total acoustic-thermal-mechanical joint sensitivity of all monitoring points.

[0044] Based on the identification of acoustically sensitive areas, this invention further constructs an acoustic sensitivity distribution map to characterize the sensitivity of different regions to the acoustic response of inter-turn short circuits. This acoustic sensitivity distribution map comprehensively considers temperature gradient, stress gradient, and sound pressure change intensity, and introduces the following optimization mechanisms: a multi-physics gradient superposition mechanism to highlight key areas where multiple fields change drastically simultaneously; spatial correlation constraints to avoid excessive concentration of monitoring points or information redundancy; and an information redundancy suppression mechanism to eliminate information regions with weak or repetitive multi-field responses.

[0045] An acoustic sensitivity distribution map is a spatial distribution result calculated based on the combined effects of temperature gradient, stress gradient, and sound pressure variation intensity at different spatial locations. Its purpose is to characterize the sensitivity of a location to the acoustic response in the same frequency band.

[0046] As a preferred implementation method, the specific process for calculating the combined acoustic-thermal-mechanical sensitivity at each spatial location within the acoustically sensitive region is as follows: For each physical quantity at each spatial location within the acoustically sensitive region, the difference between corresponding physical quantities at adjacent spatial locations is used as the numerator, and the distance between adjacent spatial locations is used as the denominator to calculate the gradient of the corresponding physical quantity at the corresponding spatial location; specifically expressed as: ; In the formula, , and These are the i-th spatial positions within the acoustically sensitive area. The gradient of temperature, equivalent stress, and sound pressure level; Let i be the spatial location within the acoustically sensitive area. Temperature at that location; Let i be the spatial location within the acoustically sensitive area. Equivalent stress at the location; Let i be the spatial location within the acoustically sensitive area. The sound pressure level at that location; Let be the i-th spatial location within the acoustically sensitive region; in this embodiment, the gradient at the last spatial location within the acoustically sensitive region is equal to the gradient at the previous spatial location; Based on the maximum and minimum gradient values ​​of the corresponding physical quantities within the acoustically sensitive region, the gradients are normalized to their maximum and minimum values ​​to obtain... , , , , are the normalized gradients at spatial location x; For any spatial location within the acoustically sensitive region, the numerator is the product of the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at that location, and the denominator is the maximum product of the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at all spatial locations. The multi-field gradient coordination factor at the corresponding spatial location is then calculated; specifically, it is expressed as: ; in, For multi-field gradient coordination factors; , and These are the spatial locations within the acoustically sensitive area. The normalized gradients of temperature, equivalent stress, and sound pressure level; After identifying the peak matching combinations, a combined acoustic-thermal-mechanical sensitivity is further constructed. The multi-field gradient coordination factor at the corresponding spatial location is multiplied by a predefined gradient coupling correction coefficient and then incremented by 1 to obtain the multi-field coordination coefficient. The normalized temperature, equivalent stress, and sound pressure level at the corresponding spatial location are multiplied together as the base to calculate the value of a predefined power function. This value is then multiplied sequentially by the multi-field coordination coefficient and the multi-field spatial consistency factor of the peak connectivity region group to which the corresponding spatial location belongs, yielding the corresponding combined acoustic-thermal-mechanical sensitivity. Specifically, it is expressed as follows: ; in, Indicates spatial location The combined acoustic and thermal sensitivity; , , These are the normalized temperature, equivalent stress, and sound pressure level at spatial location x, respectively. This is the multi-field spatial consistency factor of the peak connectivity region group to which this spatial location belongs; These are predefined gradient coupling correction coefficients; Indicates the exponent is The power function; In this formula, It is used to reflect the degree of common enhancement of three physical quantities, namely temperature, stress, and sound pressure, at the same location; Used to reflect the spatial consistency of the three peak regions; It is used to reflect whether three types of physical fields undergo drastic changes simultaneously at that location; In this embodiment, a region is only identified as a highly acoustically sensitive region when it simultaneously meets the three conditions of strong numerical value, spatial consistency, and drastic change, thus avoiding misjudgment based solely on a single sound pressure peak value or simple weighting.

[0047] As an alternative implementation, the gradient of each physical quantity at each spatial location can also be expressed as: Gradient calculation is performed based on the three-dimensional spatial representation of spatial location, specifically as follows: ; ; ; In the formula, , and These represent the gradients of temperature, equivalent stress, and sound pressure level at spatial location x, respectively. , and These represent the partial derivatives of the temperature at spatial location x with respect to the x, y, and z axes, respectively. , and These represent the partial derivatives of the equivalent stress at spatial location x with respect to the x, y, and z axes, respectively. , and These represent the partial derivatives of the sound pressure level at spatial location x with respect to the x, y, and z axes, respectively. In practical applications, the discrete differences in the three-dimensional coordinates of adjacent spatial locations are used instead of the partial derivatives for gradient calculation.

[0048] Based on the above mechanism, an optimal monitoring point layout scheme for inter-turn short-circuit acoustic monitoring is automatically generated, enabling monitoring points to preferentially cover acoustically sensitive areas, thereby improving the capture efficiency and stability of short-circuit acoustic features. The optimal monitoring point layout scheme is generated based on the acoustic-thermal joint sensitivity distribution map. Its core principle is to prioritize locations with high acoustic-thermal joint sensitivity, reasonable spatial coverage, and low information redundancy as monitoring points.

[0049] In a preferred implementation, all spatial locations within the acoustically sensitive area are selected as candidate locations to generate a candidate location set. Under the constraint of an upper limit on the number of candidate monitoring points, candidate locations are selected from the candidate location set as candidate monitoring points, and a monitoring point layout scheme is randomly generated. Specifically, the upper limit constraint on the number of candidate monitoring points in the layout scheme is that the number of candidate monitoring points does not exceed a predetermined upper limit for the number of monitoring points. ; In the formula, The number of candidate monitoring points in the monitoring point layout plan; This represents the maximum number of pre-defined monitoring points. The objective function for optimizing monitoring points is constructed with the goal of maximizing the total acoustic sensitivity of the monitoring point layout scheme. Constraints on monitoring point sensitivity, spatial spacing, and information redundancy are constructed as constraints. Under these constraints, an intelligent optimization algorithm is used to solve the objective function for optimizing monitoring points and generate the optimal monitoring point layout scheme. The intelligent optimization algorithm includes, but is not limited to, particle swarm optimization, genetic algorithm, ant colony optimization, and their variations.

[0050] As a preferred implementation, the combined acoustic, thermal, and mechanical sensitivities at all candidate monitoring points in the monitoring point layout scheme are summed and used as the objective function for monitoring point optimization; specifically expressed as: ; In the formula, Optimize the objective function for monitoring point layout scheme X; Candidate monitoring points in monitoring point layout scheme X The combined acoustic and thermal sensitivity at the location; The number of candidate monitoring points in the monitoring point layout plan; The constraints are set as follows: monitoring point sensitivity constraint, spatial spacing constraint, and information redundancy constraint. Specifically, the monitoring point sensitivity constraint ensures that the combined acoustic, thermal, and mechanical sensitivity of each candidate monitoring point in the monitoring point layout scheme is not less than a predetermined sensitivity threshold; the spatial spacing constraint ensures that the spatial distance between any two candidate monitoring points in the monitoring point layout scheme is not less than a predetermined minimum spacing between monitoring points; and the information redundancy constraint ensures that the difference in the combined acoustic, thermal, and mechanical sensitivity at any two candidate monitoring points whose spatial distance is less than a predetermined neighborhood distance threshold is not less than a predefined lower limit for difference. Specifically, this is expressed as follows: ; In the formula, the first clause is the sensitivity constraint of the monitoring point; The predetermined sensitivity threshold is generally set to [value missing]. , The first is the maximum combined acoustic-thermal-mechanical sensitivity among all candidate locations; the second is the spatial spacing constraint. For any two candidate monitoring points in monitoring point layout scheme X and For spatial distance calculation, Euclidean distance is used in this embodiment. The first rule is the minimum spacing between monitoring points; the second rule is the information redundancy constraint. If the spatial distance between two candidate monitoring points is close and the difference in the joint acoustic-thermal sensitivity is small, it means that the acoustic response similarity between the two candidate monitoring points is high. In this case, the candidate monitoring point with higher joint acoustic-thermal sensitivity is retained, and the candidate monitoring point with lower sensitivity is removed. Candidate monitoring points The combined acoustic and thermal sensitivity; A predefined lower bound for the difference; This is the predetermined neighborhood distance threshold.

[0051] S5. The optimal monitoring point layout scheme is used to arrange monitoring points. Sound pressure level data under different operating conditions are obtained through simulation. The corresponding sound pressure level peak, increment and gradient are calculated to form an acoustic feature vector. Sound pressure, temperature and equivalent stress data under actual operating conditions are collected in real time to determine the corresponding actual acoustic feature vector. The operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different operating conditions is calculated, and the actual state of the converter transformer is determined.

[0052] After obtaining the optimized acoustic monitoring point layout scheme, the present invention further extracts acoustic features under different short-circuit locations and short-circuit degrees based on the sound pressure level data collected or simulated by each monitoring point, and uses them to assist in the determination of short-circuit locations.

[0053] Monitoring points were arranged according to the optimal monitoring point layout scheme. Under different typical short-circuit locations and short-circuit degrees, sound pressure level data of each monitoring point was collected through multi-physics coupling simulation under different operating conditions. For any given operating condition, the maximum sound pressure level among all monitoring points under that condition was taken as the peak sound pressure level. These operating conditions included normal operating condition, upper short-circuit condition, middle short-circuit condition, lower short-circuit condition, and two-point short-circuit condition. The specific representation of the peak sound pressure level is as follows: ; In the formula, The peak sound pressure level under operating condition k; Monitoring points in the optimal monitoring point layout scheme The sound pressure level under operating condition k; For each monitoring point in the optimal monitoring point layout scheme, the difference in sound pressure level at the corresponding monitoring point under different operating conditions and normal operating conditions is calculated as the sound pressure level increment; specifically expressed as: ; In the formula, For the i-th monitoring point The sound pressure level increment under operating condition k is 0 under normal operating conditions. For the i-th monitoring point under operating condition k The sound pressure level; The corresponding monitoring point under normal operating conditions The sound pressure level; Calculate the sound pressure level gradient at each monitoring point in the optimal monitoring point layout scheme; specifically expressed as: ; In the formula, For the i-th monitoring point The sound pressure level gradient under operating condition k; For any operating condition, the sound pressure levels obtained from the simulation of each monitoring point in the optimal monitoring point layout scheme are spliced ​​together to obtain the acoustic feature vector for the corresponding operating condition; specifically, it is expressed as: ; in, is the acoustic feature vector under operating condition k; N is the number of monitoring points in the optimal monitoring point layout scheme.

[0054] In this embodiment, according to the aforementioned optimal monitoring point layout scheme, acoustic sensors are deployed at externally installable locations on the outer wall of the converter transformer tank, near the clamps, or corresponding to acoustically sensitive areas. Each acoustic sensor is used to collect the time-domain sound pressure signal at the corresponding monitoring point. During the actual operation of the converter transformer, the acoustic sensors deployed at the optimal monitoring point locations collect sound pressure data in real time, and the actual monitored sound pressure level is obtained after filtering, frequency band extraction, and sound pressure level calculation.

[0055] As a preferred implementation method, under actual working conditions, sensors are deployed according to the optimal monitoring point layout scheme to collect sound pressure data at each monitoring point in real time. Temperature data and equivalent stress data; among them, The Nth monitoring point at time t The raw sound pressure signal collected at the location; To reduce the impact of environmental noise, load fluctuations, and background vibrations on acoustic assessment, the raw sound pressure signal is preprocessed, including DC component removal, bandpass filtering, abnormal pulse removal, and time window segmentation, to obtain the preprocessed sound pressure signal. Based on a predefined time window length, the time window is divided; for any time window, the root mean square of the preprocessed sound pressure signal at each monitoring point within the corresponding time window is calculated as the effective sound pressure value at the corresponding monitoring point; specifically expressed as: ; In the formula, Let be the effective sound pressure value at the i-th monitoring point; The length of the predefined time window; This is the initial acquisition time within the time window; The preprocessed sound pressure signal at the i-th monitoring point at time t; For any monitoring point, based on a predefined reference sound pressure, the effective sound pressure value at the corresponding monitoring point is converted into a sound pressure level; specifically expressed as: ; in, Let i be the sound pressure level at the i-th monitoring point; It is a logarithmic function with base 10; For reference sound pressure level; The sound pressure levels of all monitoring points within the corresponding time window are concatenated to form the actual monitored sound pressure vector. This vector is used for subsequent similarity matching with the simulated acoustic feature template; based on the actual monitored sound pressure vector, the actual sound pressure level peak, sound pressure level increment, and sound pressure level gradient are calculated respectively, and then concatenated to obtain the actual acoustic feature vector. ; During actual monitoring, the real-time acquired acoustic feature vectors are matched with the acoustic feature vectors under various operating conditions. For any operating condition, the distance between the actual sound pressure level increment at all monitoring points and the sound pressure level increment under the corresponding operating condition is calculated as the sound pressure level increment term. The distance between the actual sound pressure level gradient at all monitoring points and the sound pressure level gradient under the corresponding operating condition is calculated as the sound pressure level gradient term. The absolute value of the difference between the peak sound pressure level and the peak sound pressure level under the corresponding operating condition is taken as the sound pressure level peak term. Based on predefined weighting coefficients, the sound pressure level increment term, sound pressure level gradient term, and sound pressure level peak term are weighted to obtain the operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under the corresponding operating condition. Specifically, it is expressed as follows: ; in, The operating condition matching distance is the actual acoustic feature vector and the acoustic feature vector under operating condition k. This is the actual sound pressure level increment vector. This represents the actual sound pressure level increment at the Nth monitoring point; represents the sound pressure level increment vector under operating condition k; This is the actual sound pressure level gradient vector. This represents the actual sound pressure level gradient at the (N-1)th monitoring point; Let be the sound pressure level gradient vector under operating condition k; This represents the actual peak sound pressure level. , , These are the weighting coefficients for the sound pressure level increment term, sound pressure level gradient term, and sound pressure level peak term, respectively.

[0056] When the working condition matching distance corresponding to a certain working condition is the smallest, it means that the actual monitored acoustic mode is closest to that working condition. If the working condition is not a normal working condition, the short circuit position or short circuit degree corresponding to the template can be used as an auxiliary judgment result.

[0057] As an optional implementation, the preprocessed sound pressure signal can also be converted into a sound pressure level as follows: During the operation of a converter transformer, to improve the stability of the acoustic characteristics of inter-turn short circuits, characteristic frequency bands related to transformer vibration and electromagnetic forces can be selected, such as the frequency band near the second harmonic of the power frequency, to extract the sound pressure signal in the frequency domain. Correspondingly, the frequency band sound pressure level of the i-th monitoring point within the characteristic frequency band Ω can be expressed as: ; in, Ω represents the frequency domain amplitude of the preprocessed sound pressure signal at the i-th monitoring point, which is the amplitude of the frequency domain signal obtained by Fourier transforming the preprocessed sound pressure signal at that point; Ω is a predefined characteristic frequency band.

[0058] As a preferred implementation, the operating condition with the smallest matching distance to the actual acoustic feature vector is selected as the matching operating condition; specifically, it is expressed as follows: ; In the formula, To match the operating conditions; If the matching condition is normal and the maximum value of the actual sound pressure level increment at all monitoring points is not greater than the predefined increment threshold, then the converter transformer is considered to be in a normal state, i.e., a fault-free state. If the matching condition is normal and the maximum value of the actual sound pressure level increment at all monitoring points is greater than the predefined increment threshold, the converter transformer is considered to be in an uncertain state and further manual verification is triggered. If the matching condition is not a normal operating condition, the converter transformer is in a fault state, and the fault condition corresponding to the matching condition is taken as the fault type of the converter transformer. For each monitoring point, the temperature, equivalent stress, and sound pressure level at the corresponding monitoring point are weighted by predefined physical quantity weighting coefficients to construct a multi-field consistency discrimination function for the corresponding monitoring point; specifically expressed as: ; in, The value of the multi-field consistency discriminant function at monitoring point x; , and These represent the temperature, equivalent stress, and sound pressure level at monitoring point x, respectively. , , These are the weighting coefficients for physical quantities such as temperature, stress, and sound pressure level, respectively. Their values ​​are all within the interval (0,1) and satisfy the following conditions: In this embodiment, the following can be adopted: ; Simultaneously, multi-field consistency functions can be used for auxiliary localization; the location of the monitoring point with the largest multi-field consistency discriminant value is taken as the fault occurrence area; specifically expressed as: ; In the formula, The fault location, i.e. the region with the highest multi-field consistency discrimination value, can be regarded as the region where short-circuit faults are most likely to occur.

[0059] The degree of short circuit can be judged based on the increase in sound pressure level, peak amplitude, and the range of the sensitive area. For example, a small increase in sound pressure level and a relatively localized peak area indicate a mild short circuit; a significant increase in sound pressure level and an expansion of the high sound pressure area indicate a more severe short circuit; the appearance of multiple sound pressure peak areas may correspond to a two-point short circuit or a multi-point short circuit.

[0060] Based on the optimized acoustic monitoring point layout scheme, a comparative analysis of acoustic characteristics under different inter-turn short-circuit locations and short-circuit degrees was conducted. Analysis of the sound pressure level variation patterns and spatial distribution differences provides a basis for determining the location of inter-turn short circuits and for operation and maintenance decisions.

[0061] To analyze the impact of short-circuit location changes on the multiphysics response, this implementation case establishes three typical short-circuit scenarios: upper short circuit, middle short circuit, and two-point short circuit. The short-circuit parameter settings are shown in Table 2.

[0062] Table 2 Short-circuit parameter settings

[0063] See Figure 5 and Figure 6 This invention compares and analyzes the differences in temperature distribution under the operating conditions of converter transformers based on three-dimensional thermal field calculation data of three typical operating conditions (normal operation, upper right short circuit, and double short circuit of upper right and middle right). The results show that under no short circuit condition, the overall temperature level of the converter transformer is relatively stable, with a maximum temperature of approximately 78℃ and a uniform temperature distribution, consistent with normal operation characteristics. When an upper right short circuit occurs, although the average temperature only rises slightly, the temperature of local hot spots increases significantly, reaching a maximum of approximately 190℃, and the standard deviation of the temperature distribution increases significantly, indicating that thermal stress begins to concentrate in localized areas. Under the condition of simultaneous upper right and middle right short circuits, the overall average temperature of the system decreases, but the maximum temperature still exceeds 170℃, indicating a dangerous state of "overall temperature reduction but more prominent local extreme overheating," which is more concealed. The hot spot temperatures under both fault conditions exceed the tolerance limit of the oil-immersed insulation system, posing risks of material thermal decay, oil aging, and insulation breakdown.

[0064] See Figure 7 and Figure 8Based on three-dimensional stress field calculation data for three operating conditions (normal operation, upper right short circuit, and upper right-middle double short circuit), the mechanical stress distribution of the converter transformer under different electromagnetic loads was compared and analyzed. The results show that under no short circuit condition, the overall equivalent stress level of each component is relatively low, with stress mainly concentrated in structural transitions, material interfaces, and geometric abrupt change areas. However, the maximum stress values ​​are all within the safety threshold of the structural design, demonstrating good overall mechanical stability. When an upper right short circuit occurs, the stress field distribution changes significantly, with a substantial increase in local stress peak values. In some areas, the stress exceeds 5 MPa, and the stress distribution gradient increases significantly, indicating a strong local concentration trend of electromagnetic force caused by the short circuit current. Under the simultaneous upper right and middle right short circuit condition, the stress concentration phenomenon is further intensified, the number of high-stress areas increases, and the stress peak value remains at a high level similar to that of a single-point short circuit. This indicates that the double short circuit makes the electromagnetic force path more complex, increasing the non-uniformity and suddenness of the local load on the structure.

[0065] As can be seen from the three operating conditions, the electromagnetic force inside the converter transformer is significantly enhanced and highly concentrated under short-circuit conditions, resulting in greater mechanical impact on the windings and their supporting structures. Especially in the upper right and middle right regions, the high stress peaks that appear consecutively in different short-circuit scenarios indicate that these locations are weak points in terms of mechanical stress, and there is a potential risk of insulation deformation, winding instability under pressure, and fatigue damage to the supporting structure.

[0066] See Figure 9 , Figure 10 Based on the three-dimensional calculation data of the sound pressure level on the surface of the converter transformer under three operating conditions, the acoustic field distribution pattern shows significant differences. Overall, the sound pressure level is non-uniformly distributed on the transformer surface, mainly affected by electromagnetic force excitation, structural vibration response, and local geometric features.

[0067] Under normal operating conditions, the overall sound pressure level is relatively stable, typically ranging from 144 to 151 dB. The sound field distribution is continuous, with only slight local enhancement at the edges of the heat sink and structural corners, indicating that the structural vibration excitation is relatively stable and the noise source is evenly distributed.

[0068] When the fault condition (upper right short circuit) is entered, the sound pressure level distribution changes significantly: the maximum sound pressure level increases to a high level of approximately 153–160 dB, with a noticeable concentration of sound field peaks, especially in the panel area near the dense short-circuit current zone. The sound pressure level gradient and spatial non-uniformity are significantly enhanced, indicating that the electromagnetic force pulsation caused by the short circuit leads to an increase in the local vibration intensity of the winding and core, exacerbating local sound radiation.

[0069] Under the dual short-circuit conditions at the upper right and middle right, the sound pressure level distribution further exhibits a "hotspot enhancement" trend: although the sound pressure in some areas decreases, multiple panel locations show local sound pressure peaks exceeding 155 dB. The overall sound pressure level distribution is significantly more dispersed than under normal conditions, and local noise sources exhibit stronger directivity and concentration. This indicates that the dual short circuit makes the electromagnetic excitation frequency components more complex and the superposition of local vibration modes more pronounced, resulting in a significant local enhancement of the sound field.

[0070] Overall, the sound pressure level increased under both short-circuit conditions, and the local peak value significantly exceeded that under normal operating conditions, indicating that the structure generates stronger vibration energy radiation under fault excitation.

[0071] Overall, the sound pressure level increased under both short-circuit conditions, and the local peak value significantly exceeded that under normal operating conditions, indicating that the structure generates stronger vibration energy radiation under fault excitation.

[0072] Based on the results from the temperature, stress, and sound pressure fields, this embodiment summarizes the following patterns: 1. Significant hot spot consistency: Under different short-circuit locations, the hot spots of temperature rise, the peak stress region, and the peak sound pressure level all highly overlap.

[0073] 2. The chain-like enhancement effect is prominent: the response sequence of electromagnetic distortion → temperature rise → stress → sound pressure is completely consistent with that of this invention, and the short circuit acts as the source to drive the continuous enhancement of multiple fields.

[0074] 3. Response pattern changes with short circuit location: The peak positions and distribution patterns of multiple fields migrate with the short circuit location, providing a basis for understanding the short circuit failure mechanism.

[0075] Ultimately, this invention achieves the full reproduction of the process of inter-turn short circuits in converter transformers in multiple physical fields including electromagnetic, thermal, mechanical, and acoustic domains, providing a quantitative reference for short circuit location determination, identification of structural weak links, and short circuit protection design.

[0076] Multiphysics simulation analysis reveals that the converter transformer exhibits significant acoustic non-uniformity under different inter-turn short-circuit locations. Severe short-circuit regions (such as the upper right and middle right sections) correspond to areas of intense electromagnetic force concentration and rapid Joule heat accumulation, resulting in a strong increase in structural vibration energy and thus forming a significant sound pressure enhancement zone in the sound field. Simulation results show that the local hotspot temperature in these regions can reach 170–190℃, with peak equivalent stress exceeding 5 MPa. Simultaneously, the peak sound pressure significantly increases from 144–151 dB during normal operation to 153–160 dB, indicating that short-circuit faults have a high amplification effect on acoustic characteristics. These acoustic anomalies are generated by a chain coupling effect driven by short-circuit current distortion: electromagnetic force concentration → increased temperature rise → sudden increase in structural stress → enhanced vibration → enhanced sound radiation. Under the combined influence of multiple factors such as oil convection, material thermal properties, winding support stiffness, and enclosure acoustic radiation characteristics, the sound pressure distribution caused by different short-circuit locations shows significant differences. Due to the complex magnetic flux path and sudden increase in electromagnetic force at the upper and middle short-circuit points, sound pressure peaks and strong directional sound radiation are easily formed; while in the region with a more uniform electromagnetic field, sound pressure changes are more affected by structural modes, presenting more discrete and directional acoustic modes.

[0077] Furthermore, simulations reveal a high degree of spatial consistency between the temperature rise hotspots, stress concentration zones, and sound pressure peak regions, verifying the physical nature of the short-circuit acoustic response being dominated by an internal four-field chain enhancement effect. This trend reflects the typical step-by-step amplification mechanism of "current distortion → thermal runaway → force concentration → sound radiation" after an inter-turn short circuit.

[0078] In summary, this invention achieves quantitative reproduction of short-circuit acoustic characteristics through multiphysics coupling simulation, providing a reliable basis for short-circuit sound pressure peak identification, acoustic sensitive area localization, and acoustic radiation pattern analysis. Its research results can effectively improve the acoustic diagnostic accuracy of inter-turn short circuits in converter transformers, providing important technical support for acoustic monitoring point optimization, early fault warning, and short-circuit resistant structure design. It also has significant engineering application value for the operation assurance and life assessment of ultra-high voltage direct current systems.

[0079] Example 2 Embodiment 2 of the present invention provides an acoustic feature-based inter-turn short-circuit detection system for converter transformers, which performs the acoustic feature-based inter-turn short-circuit detection method for converter transformers described in Embodiment 1, including: The simulation model building module constructs a three-dimensional simulation model of the converter transformer, performs mesh generation on the three-dimensional simulation model of the converter transformer, sets electromagnetic-thermal-mechanical-acoustic multiphysics coupling, and sets boundary conditions; After the multiphysics solution is completed, the simulation data acquisition module extracts the temperature, equivalent stress and sound pressure level at each spatial location along a preset path. The acoustic sensitive area determination module normalizes the collected temperature, equivalent stress, and sound pressure level, identifies the peak regions of different physical quantities, and divides them into multiple peak connected regions. It calculates the spatial weighting center of the peak connected region for each physical quantity and, combined with a predefined spatial deviation distance, determines the spatial consistency factor between the peak connected regions of different physical quantities. When the peak connected region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, the region is determined to be the acoustic sensitive area of ​​the converter transformer. The monitoring point layout optimization module calculates and normalizes the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determines the multi-field gradient coordination factor, and calculates the combined acoustic-thermal-mechanical sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor. It initializes the monitoring point layout scheme and solves for the optimal monitoring point layout scheme with the goal of maximizing the total combined acoustic-thermal-mechanical sensitivity of all monitoring points. The transformer short-circuit detection module arranges monitoring points according to the optimal monitoring point layout scheme. It acquires sound pressure level data under different operating conditions through simulation, calculates the corresponding sound pressure level peak, increment, and gradient to form an acoustic feature vector. It also collects sound pressure, temperature, and equivalent stress data under actual operating conditions in real time to determine the corresponding actual acoustic feature vector. Finally, it calculates the operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different operating conditions to determine the actual state of the converter transformer.

[0080] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0081] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0082] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0083] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for detecting inter-turn short circuits in converter transformers based on acoustic characteristics, characterized in that, include: S1. Construct a three-dimensional simulation model of the converter transformer, mesh the three-dimensional simulation model of the converter transformer, set electromagnetic-thermal-mechanical-acoustic multi-physics coupling, and set boundary conditions; S2. After the multiphysics solution is completed, extract the temperature, equivalent stress and sound pressure level at each spatial location along the preset path; S3. Normalize the collected temperature, equivalent stress and sound pressure level, determine the peak regions of different physical quantities and divide them into multiple peak connected regions. Calculate the spatial weighting center of the peak connectivity region for each physical quantity, and determine the spatial consistency factor between the peak connectivity regions of different physical quantities by combining the predefined spatial deviation distance; when the peak connectivity region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, the region is determined to be the acoustically sensitive area of ​​the converter transformer. S4. Calculate and normalize the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determine the multi-field gradient coordination factor, and calculate the acoustic-thermal-mechanical joint sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor; initialize the monitoring point layout scheme, and solve for the optimal monitoring point layout scheme with the goal of maximizing the total acoustic-thermal-mechanical joint sensitivity of all monitoring points. S5. Layout the monitoring points according to the optimal monitoring point layout scheme, obtain the sound pressure level data under different working conditions through simulation, calculate the corresponding sound pressure level peak, increment and gradient to form the acoustic feature vector; collect the sound pressure, temperature and equivalent stress data under actual working conditions in real time, determine the corresponding actual acoustic feature vector; calculate the working condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different working conditions, and determine the actual state of the converter transformer.

2. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of dividing the peak connected region in S3 is as follows: For any physical quantity at each spatial location, if each physical quantity at the corresponding spatial location is not less than the predefined judgment threshold of the corresponding physical quantity, the corresponding spatial location is taken as the peak region of the corresponding physical quantity, and the peak regions of temperature, stress and sound pressure level are extracted respectively. For each physical quantity's peak region, if the spatial distance between any two peak points in the corresponding peak region is not greater than the predefined neighborhood radius, it indicates that the corresponding peak points are connected. The corresponding peak points are then divided into the same connected region, generating multiple peak connected regions for the corresponding physical quantity.

3. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of calculating the spatial weighted center of the peak connected region for each physical quantity in S3 is as follows: For any peak connected region of a physical quantity, the normalized value of the physical quantity at each peak point in the corresponding peak connected region is multiplied by the corresponding spatial coordinate and summed to form the numerator. The normalized physical quantities at all peak points in the corresponding peak connected region are summed to form the denominator. The spatial weighted center of the peak connected region of the corresponding physical quantity is then calculated.

4. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of determining the spatial consistency factor between the peak connected regions of different physical quantities in S3 is as follows: From the peak connected regions of different physical quantities, select any peak connected region and match it into a peak connected region group to generate multiple peak connected region groups. For each peak connected region group, calculate the sum of squared distances between the spatial weighted centers of the peak connected regions of different physical quantities and sum them as the numerator. Use the square of 3 times the predefined spatial deviation distance as the denominator to calculate the spatial exponential coefficient. Use the negative of the spatial exponential coefficient as the exponent to calculate the value of the predefined exponential function as the spatial consistency factor of the corresponding peak connected region group.

5. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 4, characterized in that: The process for determining the acoustically sensitive area in S3 is as follows: For each peak connected region group, calculate the sum of squared distances between the spatial weighted centers of the peak connected regions for different physical quantities, take the average and take the square root, and calculate the spatial matching distance of the corresponding peak connected region group. For each peak connectivity region group, when the corresponding spatial matching distance is not greater than the predefined spatial deviation distance, the temperature peak connectivity region, stress peak connectivity region, and sound pressure level peak connectivity region of the corresponding peak connectivity region group are considered to be spatially consistent and thus constitute an effective multi-field peak matching region. If multiple multi-field peak matching regions exist, the multi-field peak matching region with a spatial consistency factor greater than a predetermined consistency threshold is selected as the corresponding acoustic sensitive region.

6. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of determining the multi-field gradient coordination factor in S4 is as follows: For each physical quantity at each spatial location within the acoustically sensitive area, the difference between corresponding physical quantities at adjacent spatial locations is used as the numerator, and the distance between adjacent spatial locations is used as the denominator. The gradient of the corresponding physical quantity at the corresponding spatial location is calculated and normalized. The multi-field gradient coordination factor at the corresponding spatial location is calculated by multiplying the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at the corresponding spatial location as the numerator, and taking the maximum value of the product of the absolute values ​​of the normalized gradients of temperature, equivalent stress, and sound pressure level at all spatial locations as the denominator.

7. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of calculating the combined acoustic-thermal-mechanical sensitivity at different spatial locations in S4 is as follows: Multiply the multi-field gradient coordination factor at the corresponding spatial location by the predefined gradient coupling correction coefficient and add 1 to obtain the multi-field coordination coefficient; multiply the normalized temperature, equivalent stress, and sound pressure level at the corresponding spatial location as the base, calculate the value of the predefined power function, and multiply the corresponding value by the multi-field coordination coefficient and the multi-field spatial consistency factor of the peak connected region group to which the corresponding spatial location belongs in turn to obtain the corresponding acoustic-thermal-mechanical joint sensitivity.

8. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of finding the optimal monitoring point layout scheme in S4 is as follows: All spatial locations within the acoustically sensitive area are used as candidate locations to generate a candidate location set. Under the constraint of the upper limit of candidate monitoring points, candidate locations are selected from the candidate location set as candidate monitoring points, and a monitoring point layout scheme is randomly generated. The upper limit constraint of candidate monitoring points is that the number of candidate monitoring points in the monitoring point layout scheme is not greater than the predetermined upper limit of the number of monitoring points. The combined acoustic, thermal, and mechanical sensitivities at all candidate monitoring points in the monitoring point layout scheme are summed and used as the objective function for monitoring point optimization. Under the set constraints, the objective function for monitoring point optimization is maximized to solve for the optimal monitoring point layout scheme.

9. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 8, characterized in that: The constraints include: monitoring point sensitivity constraints, spatial spacing constraints, and information redundancy constraints; Among them, the sensitivity constraint of the monitoring point is that the combined acoustic, thermal and mechanical sensitivity of each candidate monitoring point in the monitoring point layout scheme is not less than the predetermined sensitivity threshold; the spatial distance constraint is that the spatial distance between any two candidate monitoring points in the monitoring point layout scheme is not less than the predetermined minimum layout distance; the information redundancy constraint is that the difference in the combined acoustic, thermal and mechanical sensitivity at any two candidate monitoring points with a spatial distance less than the predetermined neighborhood distance threshold is not less than the predefined lower limit of difference.

10. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of constructing acoustic feature vectors in S5 is as follows: The monitoring points are arranged according to the optimal monitoring point layout scheme. The sound pressure level data of each monitoring point under different working conditions are collected through multi-physics field coupling simulation. For any working condition, the maximum sound pressure level among all monitoring points under the corresponding working condition is taken as the peak sound pressure level. For each monitoring point in the optimal monitoring point layout scheme, calculate the difference in sound pressure level at the corresponding monitoring point under different operating conditions and normal operating conditions, and use it as the sound pressure level increment; Calculate the sound pressure level gradient at each monitoring point in the optimal monitoring point layout scheme; For any operating condition, the sound pressure level obtained from the simulation of each monitoring point in the optimal monitoring point layout scheme is spliced ​​together to obtain the acoustic feature vector under the corresponding operating condition.

11. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process of determining the actual acoustic feature vector in S5 is as follows: The collected sound pressure data is preprocessed to obtain the preprocessed sound pressure signal; the time window is divided based on the predefined time window length; for any time window, the root mean square of the preprocessed sound pressure signal at each monitoring point within the corresponding time window is calculated as the effective sound pressure value at the corresponding monitoring point; for any monitoring point, the effective sound pressure value at the corresponding monitoring point is converted into a sound pressure level based on the predefined reference sound pressure. The sound pressure levels of all monitoring points within the corresponding time window are spliced ​​together to form the actual monitored sound pressure vector. The actual sound pressure level peak, sound pressure level increment, and sound pressure level gradient are calculated respectively, and then spliced ​​together to obtain the actual acoustic feature vector.

12. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 1, characterized in that: The process for determining the actual state of the converter transformer in S5 is as follows: For any operating condition, calculate the distance between the actual sound pressure level increment of all monitoring points and the sound pressure level increment under the corresponding operating condition, and use it as the sound pressure level increment term; calculate the distance between the actual sound pressure level gradient of all monitoring points and the sound pressure level gradient under the corresponding operating condition, and use it as the sound pressure level gradient term. The absolute value of the difference between the actual peak sound pressure level and the peak sound pressure level under the corresponding working condition is taken as the peak sound pressure level term. Based on the predefined weighting coefficient, the sound pressure level increment term, the sound pressure level gradient term, and the peak sound pressure level term are weighted to obtain the working condition matching distance between the actual acoustic feature vector and the acoustic feature vector under the corresponding working condition. The operating condition with the smallest matching distance to the actual acoustic feature vector is taken as the matching operating condition; if the matching operating condition is a normal operating condition and the maximum value of the actual sound pressure level increment in all monitoring points is not greater than the predefined increment threshold, then the converter transformer is in normal condition. If the matching condition is normal and the maximum value of the actual sound pressure level increment at all monitoring points is greater than the predefined increment threshold, then the converter transformer is in an uncertain state. If the matching condition is not a normal operating condition, the converter transformer is in a fault state, and the fault condition corresponding to the matching condition is taken as the fault type of the converter transformer.

13. The method for detecting inter-turn short circuits in a converter transformer based on acoustic characteristics according to claim 12, characterized in that: If the converter transformer is in a fault state, for each monitoring point, the temperature, equivalent stress and sound pressure level at the corresponding monitoring point are weighted by a predefined physical quantity weighting coefficient to construct a multi-field consistency discrimination function at the corresponding monitoring point. The location of the monitoring point with the highest consistency discrimination value across multiple fields is taken as the fault occurrence area.

14. A converter transformer inter-turn short-circuit detection system based on acoustic characteristics, using the method described in any one of claims 1-13, characterized in that, include: The simulation model building module constructs a three-dimensional simulation model of the converter transformer, performs mesh generation on the three-dimensional simulation model of the converter transformer, sets electromagnetic-thermal-mechanical-acoustic multiphysics coupling, and sets boundary conditions; After the multiphysics solution is completed, the simulation data acquisition module extracts the temperature, equivalent stress and sound pressure level at each spatial location along a preset path. The acoustic sensitive area determination module normalizes the collected temperature, equivalent stress and sound pressure level, determines the peak regions of different physical quantities and divides them into multiple peak connected regions. Calculate the spatial weighting center of the peak connectivity region for each physical quantity, and determine the spatial consistency factor between the peak connectivity regions of different physical quantities by combining the predefined spatial deviation distance; when the peak connectivity region meets the preset consistency criteria in terms of spatial location and spatial consistency factor, the region is determined to be the acoustically sensitive area of ​​the converter transformer. The monitoring point layout optimization module calculates and normalizes the spatial gradients of temperature, equivalent stress, and sound pressure level within the acoustically sensitive area, determines the multi-field gradient coordination factor, and calculates the combined acoustic-thermal-mechanical sensitivity at different spatial locations by combining the normalized physical quantity values ​​and the spatial consistency factor. It initializes the monitoring point layout scheme and solves for the optimal monitoring point layout scheme with the goal of maximizing the total combined acoustic-thermal-mechanical sensitivity of all monitoring points. The transformer short-circuit detection module arranges monitoring points according to the optimal monitoring point layout scheme. It acquires sound pressure level data under different operating conditions through simulation, calculates the corresponding sound pressure level peak, increment, and gradient to form an acoustic feature vector. It also collects sound pressure, temperature, and equivalent stress data under actual operating conditions in real time to determine the corresponding actual acoustic feature vector. Finally, it calculates the operating condition matching distance between the actual acoustic feature vector and the acoustic feature vector under different operating conditions to determine the actual state of the converter transformer.

15. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-13.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-13.

Images