Transformer risk detection method, device, equipment and storage medium

Abstract

This application discloses a method, apparatus, equipment, and storage medium for transformer risk detection. It establishes a three-dimensional finite element model corresponding to the transformer under test and solves for the electric field, flow field, and temperature field of the three-dimensional finite element model. Based on the electric field, flow field, and temperature field, it calculates the comprehensive risk value of each voxel. A risk heat map is generated based on the comprehensive risk value. High-risk voxels are selected from the risk heat map, and each high-risk voxel forms a first-risk candidate region. The first-risk candidate regions are comprehensively screened to determine each high-risk region. Risk location is then performed on the transformer under test according to each high-risk region. This application avoids the coarseness problem of single-field detection by solving for the electric field, flow field, and temperature field, and then combining these fields to calculate the comprehensive risk value of each voxel to determine each high-risk region. The high-risk regions are then used to locate risks on the transformer under test, thus improving detection accuracy.

Description

Technical Field

[0001] This application relates to the field of transformer risk detection technology, specifically to a transformer risk detection method, apparatus, equipment, and storage medium. Background Technology

[0002] As a core piece of equipment in the power system, the long-term, safe and stable operation of transformers is crucial to the reliability of the power grid. The internal structure of transformers is complex, and during operation, they are subjected to the coupling effects of multiple physical fields such as electricity, heat and fluid. Problems such as partial discharge, overheating and poor circulation of insulating oil often lurk inside, while traditional external detection methods are very inaccurate.

[0003] Therefore, the current mainstream solution is to provide a single risk indicator, such as judging by the overall temperature rise or average electric field strength. However, this method lacks the ability to determine the specific location of the risk in the transformer, and the detection accuracy has not been effectively improved, which still affects the normal operation of the transformer. Summary of the Invention

[0004] In view of this, this application provides a transformer risk detection method, apparatus, equipment and storage medium to solve the problem that existing methods lack the ability to determine the specific location of transformer risks, and the detection accuracy has not been effectively improved, thus still affecting the normal operation of transformers.

[0005] To achieve the above objectives, the following solution is proposed:

[0006] Firstly, a method for detecting transformer risks includes:

[0007] A three-dimensional finite element model corresponding to the transformer under test is established, and the electric field, flow field and temperature field of the three-dimensional finite element model are solved.

[0008] The comprehensive risk value of each voxel is calculated based on the electric field, flow field, and temperature field.

[0009] A risk heatmap is generated based on the comprehensive risk value;

[0010] Each high-risk voxel is selected from the risk heat map, and each high-risk voxel forms a first risk candidate region.

[0011] A comprehensive screening of each of the first risk candidate regions is conducted to determine each high-risk region;

[0012] The transformer under test is risk-identified according to each of the aforementioned high-risk areas.

[0013] Preferably, establishing the three-dimensional finite element model corresponding to the transformer under test includes:

[0014] Disassemble the various components of the transformer under test;

[0015] Determine the dimensions of each component and the assembly relationships between the components;

[0016] Based on the dimensions and assembly relationships, a finite element model is created;

[0017] The finite element model is discretized into multiple meshes, and the locations of large gradients and gradual changes of the transformer under test are determined.

[0018] Mesh refinement is performed at each large gradient location corresponding to the finite element model, while mesh coarsening is performed at each gradual change location.

[0019] Multiple preset parameters are configured on the finite element model after mesh refinement and mesh coarsening to obtain a three-dimensional finite element model.

[0020] Preferably, the calculation of the comprehensive risk value for each voxel based on the electric field, flow field, and temperature field includes:

[0021] Based on the electric field, flow field, and temperature field, calculate the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value for each voxel;

[0022] Weighting coefficients are set for the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value, respectively.

[0023] Based on the weighting coefficients, the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value are weighted and summed to obtain the comprehensive risk value.

[0024] Preferably, the step of screening high-risk voxels from the risk heatmap includes:

[0025] For each voxel in the risk heatmap, determine whether the overall risk value of that voxel is greater than a preset first threshold.

[0026] If so, then that voxel is classified as a high-risk voxel;

[0027] Using this voxel as the center, determine its neighboring voxels;

[0028] For each of the neighboring voxels, determine whether the comprehensive risk value of the neighboring voxel is not less than a preset second threshold and is less than the first threshold;

[0029] If so, then that neighboring voxel is considered a high-risk voxel.

[0030] Preferably, the step of comprehensively screening each of the first risk candidate regions to determine each high-risk region includes:

[0031] Morphological filtering is performed on each of the first risk candidate regions to obtain each of the second risk candidate regions.

[0032] For each second risk candidate region, calculate the maximum comprehensive risk value, the average comprehensive risk value of each voxel in the second risk candidate region, and the volume ratio and length-width-height ratio of the second risk candidate region in the risk heat map;

[0033] Based on the maximum comprehensive risk value, average value, volume ratio, and length-width-height ratio, calculate the comprehensive quantitative score of the second risk candidate region;

[0034] Based on the comprehensive quantitative score, each high-risk region is determined from each of the second risk candidate regions.

[0035] Preferably, the step of performing morphological filtering on each of the first risk candidate regions to obtain each of the second risk candidate regions includes:

[0036] For each of the first risk candidate regions, calculate the volume and aspect ratio of the first risk candidate region;

[0037] Calculate the minimum detectable volume based on the three-dimensional finite element model;

[0038] Determine whether the volume of the first risk candidate region is smaller than the minimum detectable volume, and / or whether the elongation ratio is greater than a preset ratio threshold;

[0039] If so, delete the first risk candidate region and use the remaining first risk candidate region as the region to be inspected.

[0040] Determine whether there are cross-material boundary phenomena in each of the areas to be inspected;

[0041] One or more first risk candidate regions with cross-material boundary phenomena are divided according to the geometric interface boundary;

[0042] Each region obtained after segmentation, along with each region to be inspected that does not exhibit cross-material boundary phenomena, are considered as second risk candidate regions.

[0043] Preferably, the method further includes calculating representative operating conditions for each of the high-risk areas, including:

[0044] For each of the high-risk areas, the mesh element corresponding to the high-risk area is determined as the first mesh element in the three-dimensional finite element model, and the non-corresponding mesh element is determined as the second mesh element.

[0045] Each of the first grid cells is marked with a first mark, and each of the second grid cells is marked with a second mark;

[0046] A three-dimensional partitioned mask is formed based on the first and second marks;

[0047] Based on the three-dimensional partition mask, the representative working conditions of the high-risk area are calculated.

[0048] Secondly, a transformer risk detection device includes:

[0049] The model building and solution module is used to build a three-dimensional finite element model corresponding to the transformer under test, and solve the electric field, flow field and temperature field of the three-dimensional finite element model.

[0050] The comprehensive risk value calculation module is used to calculate the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field.

[0051] A risk heatmap generation module is used to generate a risk heatmap based on the comprehensive risk value;

[0052] The first risk candidate region determination module is used to screen out each high-risk voxel from the risk heat map and form each first risk candidate region from each of the high-risk voxels.

[0053] The high-risk area determination module is used to comprehensively screen each of the first risk candidate areas to determine each high-risk area;

[0054] The risk location module is used to locate the risk of the transformer under test according to each of the high-risk areas.

[0055] Thirdly, a transformer risk detection device includes a memory and a processor;

[0056] The memory is used to store programs;

[0057] The processor is configured to execute the program to implement the various steps of the transformer risk detection method as described in any of the first aspects.

[0058] Fourthly, a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the transformer risk detection method as described in any of the first aspects.

[0059] As can be seen from the above technical solution, this application establishes a three-dimensional finite element model corresponding to the transformer under test, and solves the electric field, flow field, and temperature field of the three-dimensional finite element model; calculates the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field; generates a risk heat map based on the comprehensive risk value; selects each high-risk voxel from the risk heat map, and forms each first risk candidate region from each of the high-risk voxels; performs comprehensive screening on each of the first risk candidate regions to determine each high-risk region; and locates the risk of the transformer under test according to each of the high-risk regions. This application first establishes a three-dimensional finite element model corresponding to the transformer under test, eliminating the roughness problem of single detection in existing technologies, and solves for the electric field, flow field, and temperature field. Then, it calculates the comprehensive risk value of each voxel by combining the electric field, flow field, and temperature field, and generates a risk heat map based on the comprehensive risk value. This allows for the screening of high-risk voxels in the risk heat map, meaning that there is a higher probability of high risk around high-risk voxels. Therefore, in order to identify high-risk areas, candidate areas need to be formed from each high-risk voxel, and then screened. A comprehensive screening method is used to finally determine each high-risk area. In this way, the high-risk areas can be used to locate risks on the transformer under test, improving detection accuracy and facilitating the normal operation of the transformer. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0061] Figure 1 An optional flowchart of a transformer risk detection method provided in an embodiment of this application;

[0062] Figure 2 This is a schematic diagram of the structure of a transformer risk detection device provided in an embodiment of this application;

[0063] Figure 3 This is a schematic diagram of the structure of a transformer risk detection device provided in an embodiment of this application. Detailed Implementation

[0064] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] As a core piece of equipment in the power system, the long-term, safe, and stable operation of transformers is crucial to the reliability of the power grid. The internal structure of transformers is complex, and during operation, they are simultaneously subjected to the coupling effects of multiple physical fields such as electricity, heat, and fluid. The end geometry and minimum oil gap determine the concentration of the electric field. The moisture content and dissolved gas content of the insulating paper, as well as the aging of the insulating oil and insulating paper, change the dielectric constant, conductivity, viscosity, and interfacial behavior. These quantities are coupled with each other and drift with operating conditions. For example, an increase in temperature will reduce oil viscosity, change oil flow and hot spot location, and at the same time increase conductivity and polarization loss.

[0066] Traditionally, a single empirical constant is used to represent the parameters of materials such as oil and paper, with only one-dimensional or empirical corrections made for temperature in a few rare scenarios. This is acceptable near the rated steady state, but when the equipment experiences overload, cooling switching, high ambient temperature, local structural modifications, or accelerated aging, the uniform constant will lead to systematic biases: the electric field margin will be overestimated or underestimated, the location and amplitude drift of hot spots will be masked, and the risks of local stagnation and gas evolution will be difficult to quantify. As a result, problems such as partial discharge, overheating, and poor circulation of insulating oil often lurk inside, while traditional external detection methods are very inaccurate. Therefore, the current mainstream solution is to provide a single risk indicator, such as judging by the overall temperature rise or average electric field strength.

[0067] Another current solution is "full-machine multiphysics simulation + empirical material constant method," which establishes a three-dimensional geometry for the entire oil-immersed transformer, coupling electric field, heat conduction / convection, and laminar natural / forced convection fields within the same model; given the winding potential and grounding, radiator / tank wall convective heat transfer and ambient temperature, fluid inlet / natural convection conditions at the boundaries, with heat generation using copper / iron losses as the bulk heat source; engineering database tables or empirical constants are used on the material side, and the mesh is refined at the ends, air gap, and oil passages at the geometry, solved sequentially with coupling or weak coupling, using residual and temperature rise changes as convergence criteria; post-processing reads three types of "health indicators": peak electric field and The field enhancement factor is compared with the allowable electric field / creep coefficient empirical table, the hot spot temperature and temperature gradient are compared with IEC / factory limits, and the oil passage flow rate / heat transfer capacity is identified by local Re / Pr / Pe to identify low-speed recirculation and poor heat dissipation, thereby determining "where the field is strongest, where it is hottest, and where the oil flow is poor", and based on this, structural / operating condition iterations are performed on chamfering, shielding, oil passage width, radiator opening or flow rate. The advantages of this method are fast modeling, controllable computing power, and suitability for global optimization in the design and review stages. However, because the materials and interfaces use empirical values ​​or single-parameter curves, it is difficult to explicitly reflect the impact of different operating conditions on their parameters.

[0068] Therefore, based on the above analysis, several shortcomings of the current technology can be identified, including:

[0069] Parameters are disconnected from operating conditions: Material and interface parameters are mostly based on empirical constants or single empirical curves, without being linked to the actual temperature of specific zones, moisture content in paper, composition of dissolved gases, and electric field, etc., which leads to seemingly reasonable overall judgments but significant deviations in key local areas.

[0070] The identification of zones is not reproducible: high-risk areas mainly rely on engineers' experience to manually select points, lacking unified automated rules with dual threshold contiguous areas, morphological filtering and weighted sorting as the core. The results are difficult to be consistent between different personnel and different projects, making it difficult to scale up.

[0071] Lack of a mechanistic reinjection loop: The calculation results at the molecular or mesoscopic level also lack a standardized data structure that includes fields, units, interpolation methods, applicable ranges and uncertainties for reinjection. At the same time, risk assessment lacks a unified statistical caliber for the 95th percentile and a recalculation strategy, making it difficult to form a closed loop of "screening - refinement - reinjection - judgment".

[0072] Low detection accuracy: The lack of determination of the specific location of the risk in the transformer means that the detection accuracy has not been effectively improved, which will still affect the normal operation of the transformer.

[0073] To address the shortcomings of the prior art, this invention provides a transformer risk detection method. This invention can be used in numerous general-purpose or special-purpose computing device environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, distributed computing environments including any of the above devices, etc.

[0074] This method can be applied to various computer terminals or smart terminals, and its execution entity can be the processor or server of the computer terminal or smart terminal. The flowchart of the method is as follows. Figure 1 As shown, it specifically includes:

[0075] S1: Establish a three-dimensional finite element model corresponding to the transformer under test, and solve the electric field, flow field and temperature field of the three-dimensional finite element model.

[0076] This application achieves high-precision mathematical simulation of the complex internal physical environment of the transformer under test by establishing a three-dimensional model consistent with the actual geometric structure and material properties of the transformer under test. This is the basis for subsequent testing. In this process, a three-dimensional finite element model is constructed in the three-dimensional modeling software from the perspective of the whole machine. It is necessary to maintain geometric consistency with the actual object based on the dimensions and assembly relationships of the core, coil, clamps, insulating paperboard, oil passages, oil tank, radiator, etc., to ensure the correctness of key dimensions such as end gaps, corner radii, and oil passage width.

[0077] After the above settings are completed, the electric field, flow field and temperature field can be solved according to the preset order or coupling strategy, such as the order of electric field → flow field → temperature field. It is also necessary to write back the influence of temperature on physical properties and losses to form a closed loop, so as to obtain the temperature field of insulating oil and the electric field intensity field of the whole domain, and provide reliable baseline results for subsequent risk zoning.

[0078] S2: Calculate the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field.

[0079] Based on the obtained electric field, flow field, and temperature field, multiple dimensionless health indicators can be calculated point by point in the computational domain according to a unified standard. Then, by linearly weighting these dimensionless health indicators, the comprehensive risk value of each voxel can be obtained.

[0080] The electric field determines the insulation strength, the flow field shows the oil flow distribution and affects heat dissipation and impurity movement, and the temperature field reflects the key to hot spots and aging. Therefore, this application analyzes through these three fields, which overcomes the shortcomings of the single analysis in the prior art and can more realistically reflect the operating state of the voltage under test.

[0081] This application uses voxels as the basic unit for evaluation, which can achieve high spatial resolution in risk analysis and can detect localized minute defects or potential risk points that cannot be captured by traditional methods.

[0082] S3: Generate a risk heatmap based on the comprehensive risk value.

[0083] This step transforms the abstract numerical risk into a risk heat map. This allows technicians to clearly identify the global distribution, concentrated areas, and gradient changes of the internal risks of the transformer under test. Then, more detailed analysis can be carried out based on this risk heat map.

[0084] S4: Select each high-risk voxel from the risk heat map, and form each first risk candidate region from each of the high-risk voxels.

[0085] In this step, high-risk voxels can be automatically identified from a massive number of voxels by setting a threshold. Candidate regions can then be formed based on these high-risk voxels, which are considered to be potentially risky. This allows discrete risk voxels (i.e., risk points) to be aggregated into continuous regions, which conform to the morphology of defects or faults in actual engineering, such as a discharge channel or an insulation aging area, providing deterministic objects for subsequent analysis.

[0086] S5: Perform comprehensive screening on each of the first risk candidate regions to determine each high-risk region.

[0087] Understandably, the initial candidate regions may include some misjudged regions caused by model errors, computational noise, or instantaneous fluctuations. Through comprehensive screening, these "noises" can be eliminated, and the truly risky regions can be selected.

[0088] The selection process can be comprehensively based on factors such as the size, shape, and stability of the risk value of the first risk candidate area, as well as comparisons with historical data or correlations with other monitoring data.

[0089] S6: Perform risk location for the transformer under test according to each of the high-risk areas.

[0090] Finally, based on the identified high-risk areas, the specific physical location of the transformer under test can be accurately mapped to the actual location through coordinate mapping, providing a solid basis for subsequent inspection and maintenance. This can greatly save maintenance costs, ensure the normal operation of the transformer under test, and improve the reliability of the power grid.

[0091] As can be seen from the above technical solution, this application establishes a three-dimensional finite element model corresponding to the transformer under test, and solves the electric field, flow field, and temperature field of the three-dimensional finite element model; calculates the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field; generates a risk heat map based on the comprehensive risk value; selects each high-risk voxel from the risk heat map, and forms each first risk candidate region from each of the high-risk voxels; performs comprehensive screening on each of the first risk candidate regions to determine each high-risk region; and locates the risk of the transformer under test according to each of the high-risk regions. This application first establishes a three-dimensional finite element model corresponding to the transformer under test, eliminating the roughness problem of single detection in existing technologies, and solves for the electric field, flow field, and temperature field. Then, it calculates the comprehensive risk value of each voxel by combining the electric field, flow field, and temperature field, and generates a risk heat map based on the comprehensive risk value. This allows for the screening of high-risk voxels in the risk heat map, meaning that there is a higher probability of high risk around high-risk voxels. Therefore, in order to identify high-risk areas, candidate areas need to be formed from each high-risk voxel, and then screened. A comprehensive screening method is used to finally determine each high-risk area. In this way, the high-risk areas can be used to locate risks on the transformer under test, improving detection accuracy and facilitating the normal operation of the transformer.

[0092] The process of establishing a three-dimensional finite element model corresponding to the transformer under test in the method provided by this invention is described in detail below:

[0093] Disassemble the various components of the transformer under test;

[0094] Determine the dimensions of each component and the assembly relationships between the components;

[0095] Based on the dimensions and assembly relationships, a finite element model is created;

[0096] The finite element model is discretized into multiple meshes, and the locations of large gradients and gradual changes of the transformer under test are determined.

[0097] Mesh refinement is performed at each large gradient location corresponding to the finite element model, while mesh coarsening is performed at each gradual change location.

[0098] Multiple preset parameters are configured on the finite element model after mesh refinement and mesh coarsening to obtain a three-dimensional finite element model.

[0099] Specifically, after constructing the finite element model, in order to make the model more accurate, the finite element model can be discretized and then refined. This can be done by discretizing it into individual meshes and then matching each component of the transformer under test with these meshes. The refinement process involves refining the meshes at locations with large gradients and coarsening them at locations with gradual changes. Specifically, locations with large field gradients, such as corners, ends, and narrow oil passages of the transformer under test, are designated as locations with large gradients, while other locations are designated as locations with gradual changes.

[0100] This allows for a greater emphasis on different locations within the finite element model, improving the accuracy of subsequent analyses. Additionally, parameter configuration is required for the processed finite element model. For example, simplification of material and parameter settings is necessary, including setting density, viscosity, specific heat, thermal conductivity, and dielectric constant for insulating oil; setting anisotropic thermal conductivity, dielectric constant, moisture content, and related thermal parameters for insulating paperboard; and providing commonly used thermal conductivity and electrical parameters for other solid components.

[0101] Additionally, boundary conditions can be defined. The thermal boundary is determined by the convective heat transfer coefficient between the outer surface of the oil tank and the environment, as well as the ambient temperature. The radiator uses equivalent heat transfer. The coil interior is calculated as a volume heat source based on copper and iron losses. The flow boundary applies volumetric flow rate or pressure difference at the oil inlet and outlet, and the wall surface is designed to be slip-free. The electrical boundary applies potential at the high-voltage end, sets zero potential at the low-voltage end and grounding point, and sets the remaining conductors according to the actual connection.

[0102] The process of calculating the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field in this application is described in detail below.

[0103] Based on the electric field, flow field, and temperature field, calculate the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value for each voxel;

[0104] Weighting coefficients are set for the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value, respectively.

[0105] Based on the weighting coefficients, the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value are weighted and summed to obtain the comprehensive risk value.

[0106] Specifically, temperature risk value The calculation formula is:

[0107] ;

[0108] in, Use 85–95℃ (reference temperature). This represents the temperature field, which includes the temperature of each voxel. Coordinates representing voxels This indicates the maximum permissible temperature, which can be taken as 110–120℃ (permissible temperature) according to the insulation class of the whole machine. The higher the temperature, the higher the risk.

[0109] Electric field risk value The calculation formula is:

[0110] ;

[0111] in, Represents the electric field, including the electric field strength of each voxel, allowing the electric field to... Given by the “Local Structural Zoning Table”: ① Pure oil gap (thickness ≥ 1 mm) baseline 6–8 kV / mm; ② Paper-oil laminates are subject to an equivalent lower tolerance; ③ Curvature abrupt changes and sharp corners are multiplied by a safety reduction factor (0.70–0.85), and the stronger the electric field, the higher the risk.

[0112] Temperature gradient risk value The calculation formula is:

[0113] ;

[0114] in, , Represents the temperature gradient. This represents the maximum permissible temperature gradient, and the larger the gradient, the higher the risk.

[0115] Liquidity risk value The calculation formula is:

[0116] ;

[0117] in, , Indicates the critical speed threshold. This represents the velocity of a voxel within the flow field, and a higher risk only exists below a critical velocity threshold.

[0118] After solving for the four types of risk values, a weighted summation can be performed, requiring the setting of weight coefficients for each category. The calculation formula is as follows:

[0119] ;

[0120] in, , , , These are all weighting coefficients; in one example, these four weighting coefficients are as follows: It is 0.35. It is 0.35. It is 0.2. The value is 0.1. The specific weighting coefficient can be determined according to the actual situation. This embodiment does not impose any restrictions on this, but all values ​​are within the range of [0,1].

[0121] The following embodiments provide a detailed explanation of the steps for selecting high-risk voxels from the risk heatmap in this application.

[0122] For each voxel in the risk heatmap, determine whether the overall risk value of that voxel is greater than a preset first threshold.

[0123] If so, then that voxel is classified as a high-risk voxel;

[0124] Using this voxel as the center, determine its neighboring voxels;

[0125] For each of the neighboring voxels, determine whether the comprehensive risk value of the neighboring voxel is not less than a preset second threshold and is less than the first threshold;

[0126] If so, then that neighboring voxel is considered a high-risk voxel.

[0127] Specifically, this application first sets a first threshold and a second threshold. The first threshold is used to filter out high-risk voxels among the voxels. Voxels with a comprehensive risk value greater than the first threshold are considered high-risk voxels. Then, using these high-risk voxels as "seeds," several three-dimensional connected regions are generated under the 26-adjacency rule. Adjacent voxels whose comprehensive risk values ​​fall within the range of [second threshold, first threshold] are also considered high-risk voxels. These high-risk voxels can then be connected to form first-risk candidate regions, thus each high-risk voxel can form a first-risk candidate region.

[0128] Optionally, the step of comprehensively screening each of the first risk candidate regions to determine each high-risk region in this application specifically includes:

[0129] Morphological filtering is performed on each of the first risk candidate regions to obtain each of the second risk candidate regions.

[0130] For each second risk candidate region, calculate the maximum comprehensive risk value, the average comprehensive risk value of each voxel in the second risk candidate region, and the volume ratio and length-width-height ratio of the second risk candidate region in the risk heat map;

[0131] Based on the maximum comprehensive risk value, average value, volume ratio, and length-width-height ratio, calculate the comprehensive quantitative score of the second risk candidate region;

[0132] Based on the comprehensive quantitative score, each high-risk region is determined from each of the second risk candidate regions.

[0133] Specifically, when performing morphological filtering on the first risk candidate region, the filtering can be performed based on reference factors such as the volume and shape of the first risk candidate region. This is because some first risk candidate regions with special or unusual shapes are geometric anomalies, so it is not accurate to regard them as high-risk regions.

[0134] After filtering, each second risk candidate region is obtained. In addition to morphological filtering, some statistics need to be calculated to further achieve high-precision screening. The calculated statistics include the maximum comprehensive risk value in the second risk candidate region, the average value of the comprehensive risk value of each voxel, the volume ratio and length-width-height ratio of the second risk candidate region in the risk heat map. Other statistics can also be added according to the actual situation. This embodiment does not limit this.

[0135] After calculating each statistical measure, a comprehensive quantitative score is performed. The scoring formula is:

[0136] ;

[0137] in, This represents the overall risk value. This represents the average of the overall risk values ​​for each voxel. Indicates volume percentage. Indicates form penalty, , All are weights; the shape penalty refers to the reciprocal of the elongation ratio, meaning the larger the elongation ratio, the greater the shape penalty; in one example, the specific values ​​of these four weights are: 0.5 0.3 0.1 .

[0138] In determining high-risk areas from various second-risk candidate areas based on comprehensive quantitative scores, two methods can be adopted:

[0139] 1) Sort each second risk candidate region in descending order of comprehensive quantitative scores, and set the number of selections, selecting from front to back according to the number of selections.

[0140] However, if the final selected second risk candidate region has the same comprehensive quantitative score as the second risk candidate region that follows it in the ranking, then it is necessary to determine the dominant risk source of these two second risk candidate regions. The two second risk candidate regions are compared in the order of higher peak value priority, larger volume priority, and electric field dominance priority to obtain the final determined high-risk region.

[0141] The dominant risk source refers to which field dominates the second risk candidate area. For example, it may be mainly from the electric field, but a strong electric field can easily cause discharge; it may be mainly from the temperature field, but a strong temperature field can easily cause aging, or a rapid temperature change can easily generate thermal stress; or it may be mainly from the flow field, but a slow oil flow can lead to local overheating. Based on this, it can be determined that the risk in this second risk area is mainly due to this field.

[0142] In one instance, the electric field risk value in a second risk candidate region is 0.6, the temperature risk value is 0.3, the temperature gradient risk value is 0.2, and the flow risk value is 0.1. This indicates that the electric field is the dominant risk source in this second risk candidate region because it accounts for the largest proportion.

[0143] 2) For each second risk candidate region, the comprehensive quantitative score is compared with a preset third threshold. If the comprehensive quantitative score is greater than the third threshold, the second risk candidate region is designated as a high-risk region. This method will filter out all second risk candidate regions with relatively high comprehensive quantitative scores, with no limit on the number.

[0144] After identifying each high-risk area, a list of high-risk areas can be created for subsequent maintenance personnel to review. This list can include information for each high-risk area, such as its number, spatial location (centroid coordinates and outer frame), volume, maximum comprehensive risk value, average comprehensive risk value, volume percentage, dominant risk source, length-width-height ratio, shape notes, and corresponding screenshots.

[0145] To improve robustness, the contents of the above list can be stored, and a voxel set summary (such as a hash signature) of each high-risk area can be saved to enable consistency comparison and traceability between different batches and different transformers.

[0146] The process of performing morphological filtering on each of the first risk candidate regions to obtain each of the second risk candidate regions is explained in detail below:

[0147] For each of the first risk candidate regions, calculate the volume and aspect ratio of the first risk candidate region;

[0148] Calculate the minimum detectable volume based on the three-dimensional finite element model;

[0149] Determine whether the volume of the first risk candidate region is smaller than the minimum detectable volume, and / or whether the elongation ratio is greater than a preset ratio threshold;

[0150] If so, delete the first risk candidate region and use the remaining first risk candidate region as the region to be inspected.

[0151] Determine whether there are cross-material boundary phenomena in each of the areas to be inspected;

[0152] One or more first risk candidate regions with cross-material boundary phenomena are divided according to the geometric interface boundary;

[0153] Each region obtained after segmentation, along with each region to be inspected that does not exhibit cross-material boundary phenomena, are considered as second risk candidate regions.

[0154] Specifically, during morphological filtering, filtering is performed based on various aspects such as volume, region shape, and partitioning. For each first-risk candidate region, its volume and elongation ratio are calculated. At the same time, the minimum detectable volume is calculated based on the three-dimensional finite element model. If the volume of a region is equal to the minimum detectable volume, then that region is considered an isolated island. This minimum detectable volume is five ten-thousandths of the total number of voxels or three times the volume of the local insulation thickness cube. Then, first-risk candidate regions with a volume smaller than the minimum detectable volume and / or an elongation ratio greater than a preset ratio threshold are deleted because regions with a large elongation ratio belong to slit-like connected regions, and their geometry is abnormal and odd.

[0155] For the remaining first risk candidate regions, these are treated as regions to be inspected. These regions need to be checked to see if they cross material boundaries. If so, they need to be segmented along the real geometric interface to avoid merging physically discontinuous regions into one class. This way, the various second risk candidate regions can be obtained.

[0156] Furthermore, this application can also calculate representative operating conditions for each of the aforementioned high-risk areas, the specific process of which is as follows:

[0157] For each of the high-risk areas, the mesh element corresponding to the high-risk area is determined as the first mesh element in the three-dimensional finite element model, and the non-corresponding mesh element is determined as the second mesh element.

[0158] Each of the first grid cells is marked with a first mark, and each of the second grid cells is marked with a second mark;

[0159] A three-dimensional partitioned mask is formed based on the first and second marks;

[0160] Based on the three-dimensional partition mask, the representative working conditions of the high-risk area are calculated.

[0161] Specifically, the first marker is "1" and the second marker is "0". This generates a three-dimensional partition mask composed of "0" and "1". This mask is like applying a transparent layer to the three-dimensional finite element model, which only marks the high-risk areas. It can be directly called later. Fine parameters can be assigned to these high-risk areas individually, recalculated, or locally refined for analysis to improve efficiency and accuracy.

[0162] Optionally, when calculating representative operating conditions of the high-risk area based on the three-dimensional partition mask, the three-dimensional partition mask can be used as the spatial range to calculate representative operating conditions related to materials, interfaces and mass transfer within the high-risk area, and then aligned with the on-site monitoring data.

[0163] Data such as temperature, pressure, electric field strength, and flow velocity are collected for each small grid within this high-risk area. Weighted statistics are then performed based on the volume of each grid, meaning larger grids have a greater impact on the results. Temperature, pressure, and temperature gradient are represented by a volume-weighted average. The electric field strength is calculated at the 95th percentile, representing the strongest portion of the electric field. In addition to the average value, the oil flow velocity is also calculated at the 10th percentile (used to identify low-speed zones) to determine if there are heat dissipation dead zones in the slowest flow areas. These statistical values ​​comprehensively describe the actual operating conditions of the high-risk area, providing a basis for subsequent simulations and parameter updates.

[0164] In addition, data time windows and confidence markers can be used to enrich the working condition content. Finally, the working condition vector is output according to fixed fields and units as input conditions for the fifth step of molecular dynamics calculation or equivalent small-scale experiment.

[0165] For high-risk areas with large comprehensive quantitative scores, computing power can be concentrated to conduct detailed inspections. First, the representative working conditions are read out by the three-dimensional finite element model according to the three-dimensional partition mask, which includes temperature, pressure, electric field strength, temperature gradient and oil flow velocity. Then, the content of each component in the mineral oil, including basic hydrocarbon components, dissolved water, and typical non-condensable gases such as hydrogen, carbon monoxide and carbon dioxide, is obtained from the online or offline sensors closest to the high-risk area.

[0166] Then, a molecular model consistent with representative operating conditions was constructed to clarify the oil phase composition and water content, based on the temperature, pressure, and electric field strength obtained at the current moment. The force field adopted COMPASS Ⅲ / OPLS / AMBER / CHARMM, etc. (any ensemble of NPT / NVT), and molecular dynamics were performed for 5 nanoseconds under isobaric isothermal ensemble (NPT) and isovolitic isothermal ensemble (NVT), with a time step of 2 femtoseconds. Ewald was used for electrostatic interactions, and van der Waals interactions were cut off at 1.0–1.2 nanometers. Sufficient equilibrium time and plateauing of energy, temperature, and pressure were ensured before entering the production stage.

[0167] In addition, the dielectric constant is obtained by dipole moment fluctuation method or polarization autocorrelation function; under the same representative operating conditions, the solubility is obtained by Widom interpolation method or chemical potential difference method; the diffusion coefficient is calculated by the slope of the linear segment of the mean square displacement curve, and statistical analysis is performed separately for different molecular species when necessary.

[0168] To control uncertainty, at least three independent replicates are performed, the mean and the confidence interval of the 95th percentile are output, and convergence checks are performed (such as the mean square displacement linear segment determination coefficient, dipole moment autocorrelation decay integrity, chemical potential estimation variance threshold, etc.).

[0169] Ultimately, parameters such as dielectric constant, diffusion coefficient, and solubility were compiled into a "draft of zone-specific parameter packages," specifying fields, units, independent variables (such as temperature, water activity, and frequency), interpolation, and applicable ranges, for subsequent boundary recharge and boundary determination.

[0170] Furthermore, the dielectric constant, solubility of each gas, and diffusion coefficient can be compiled into a parameter package, which can be directly called and traceably managed by the finite element method model, containing three types of content:

[0171] The first category is metadata, which includes device number, partition number, timestamp, generation method "molecular simulation", number of independent repetitions, convergence index, responsible person, semantic version number and hash signature.

[0172] The second category is the data body, which includes the unit of each field, effective independent variables, value tables or coefficients, interpolation methods and extrapolation strategies, and uncertainty caliber.

[0173] The third category is applicable and prohibited domains, including temperature / frequency ranges; the upper limit of semi-empirical safe extrapolation generally does not exceed ±5% of the original training domain, and if it exceeds, it is marked as "read-only, not used for line determination".

[0174] The interpolation employs a piecewise cubic Hermitian or Archima method to maintain monotonicity, explicitly specifying units and independent variable coordinates. Physical constraints are set for properties that must be monotonic (e.g., diffusion coefficient, surface tension) and strictly positive quantities (e.g., Henry coefficient, diffusion coefficient). Point-by-point physical feasibility verification is performed after interpolation / fitting. Subsequently, a partition-specific parameter package is loaded, and the above parameters are only used to replace material properties within high-risk partitions. The re-solved field distribution and key statistics (volume-weighted average and 95th percentile) are output, along with the parameter package version and solution log, and archived for subsequent unified line determination and closed-loop iteration.

[0175] Finally, based on the above solutions, key indicators for each high-risk zone can be extracted using a fixed caliber, including: electric field strength (95th percentile and peak value) and hotspot temperature (95th percentile and peak value). All indicators are statistically weighted by volume and their uncertainties are given. Then, these key indicators are compared one by one with preset condition thresholds (including allowable electric field reduction tables, insulation class temperature limits, and values ​​given according to equipment class and safety margin). If the key indicators and their 95% confidence intervals are all within the condition thresholds, the high-risk area passes the test; if they exceed the limits or the confidence intervals cross the condition thresholds, the high-risk area needs to be re-verified.

[0176] For high-risk zones requiring review, minimum modification amounts are given according to sensitivity priority, such as: increasing the shield chamfer radius, fine-tuning the end pressure equalization structure, adjusting the oil passage width or guide vane angle, increasing the radiator opening or circulation flow, lowering the operating load or temperature rise target, etc., and clearly specifying the estimated improvement range and cost of the modifications to key indicators; then return to the overall machine angle solution steps at the beginning of this application or the subsequent zone solution steps, and enter closed-loop iteration until all high-risk areas pass or the preset iteration limit is reached.

[0177] This allows for the generation of a final report, which includes at least the following: statistical values ​​and confidence intervals for each key indicator, pass / fail conclusions, number of recalculations triggered and parameter package versions, threshold and reduction table versions used, a list of recommended minimum modifications and implementation priorities; and archived computation logs and traceable metadata (such as timestamps, grid and time step summaries, solution convergence records, etc.), which can then serve as the basis for engineering decisions and subsequent reviews.

[0178] and Figure 1 Corresponding to the method described above, this embodiment of the invention also provides a transformer risk detection device for detecting... Figure 1 In a specific implementation of the method, the transformer risk detection device provided in this embodiment of the invention can be used in a computer terminal or various mobile devices, combined with... Figure 2 This section introduces transformer risk detection devices, such as... Figure 2As shown, the device may include:

[0179] The model building and solution module 10 is used to build a three-dimensional finite element model corresponding to the transformer under test, and solve the electric field, flow field and temperature field of the three-dimensional finite element model.

[0180] The comprehensive risk value calculation module 20 is used to calculate the comprehensive risk value of each voxel based on the electric field, flow field and temperature field;

[0181] Risk heatmap generation module 30 is used to generate a risk heatmap based on the comprehensive risk value;

[0182] The first risk candidate region determination module 40 is used to screen out each high-risk voxel from the risk heat map and form each first risk candidate region from each of the high-risk voxels.

[0183] The high-risk area determination module 50 is used to comprehensively screen each of the first risk candidate areas to determine each high-risk area;

[0184] The risk location module 60 is used to locate the risk of the transformer under test according to each of the high-risk areas.

[0185] As can be seen from the above technical solution, this application establishes a three-dimensional finite element model corresponding to the transformer under test, and solves the electric field, flow field, and temperature field of the three-dimensional finite element model; calculates the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field; generates a risk heat map based on the comprehensive risk value; selects each high-risk voxel from the risk heat map, and forms each first risk candidate region from each of the high-risk voxels; performs comprehensive screening on each of the first risk candidate regions to determine each high-risk region; and locates the risk of the transformer under test according to each of the high-risk regions. This application first establishes a three-dimensional finite element model corresponding to the transformer under test, eliminating the roughness problem of single detection in existing technologies, and solves for the electric field, flow field, and temperature field. Then, it calculates the comprehensive risk value of each voxel by combining the electric field, flow field, and temperature field, and generates a risk heat map based on the comprehensive risk value. This allows for the screening of high-risk voxels in the risk heat map, meaning that there is a higher probability of high risk around high-risk voxels. Therefore, in order to identify high-risk areas, candidate areas need to be formed from each high-risk voxel, and then screened. A comprehensive screening method is used to finally determine each high-risk area. In this way, the high-risk areas can be used to locate risks on the transformer under test, improving detection accuracy and facilitating the normal operation of the transformer.

[0186] Furthermore, embodiments of this application provide a transformer risk detection device. Optionally, Figure 3The hardware structure block diagram of the transformer risk detection equipment is shown. (Refer to...) Figure 3 The hardware structure of the transformer risk detection equipment may include: at least one processor 01, at least one communication interface 02, at least one memory 03, and at least one communication bus 04.

[0187] In this embodiment, the number of processor 01, communication interface 02, memory 03 and communication bus 04 is at least one, and processor 01, communication interface 02 and memory 03 communicate with each other through communication bus 04.

[0188] Processor 01 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0189] Memory 03 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device.

[0190] The memory stores a program, which the processor can call to execute. The program is used to perform the following transformer risk detection method, including:

[0191] A three-dimensional finite element model corresponding to the transformer under test is established, and the electric field, flow field and temperature field of the three-dimensional finite element model are solved.

[0192] The comprehensive risk value of each voxel is calculated based on the electric field, flow field, and temperature field.

[0193] A risk heatmap is generated based on the comprehensive risk value;

[0194] Each high-risk voxel is selected from the risk heat map, and each high-risk voxel forms a first risk candidate region.

[0195] A comprehensive screening of each of the first risk candidate regions is conducted to determine each high-risk region;

[0196] The transformer under test is risk-identified according to each of the aforementioned high-risk areas.

[0197] Optionally, the refined and extended functions of the program can be found in the description of the transformer risk detection method in the method embodiments.

[0198] This application embodiment also provides a storage medium that can store a program suitable for execution by a processor. When the program runs, it controls the device containing the storage medium to perform the following transformer risk detection method, including:

[0199] A three-dimensional finite element model corresponding to the transformer under test is established, and the electric field, flow field and temperature field of the three-dimensional finite element model are solved.

[0200] The comprehensive risk value of each voxel is calculated based on the electric field, flow field, and temperature field.

[0201] A risk heatmap is generated based on the comprehensive risk value;

[0202] Each high-risk voxel is selected from the risk heat map, and each high-risk voxel forms a first risk candidate region.

[0203] A comprehensive screening of each of the first risk candidate regions is conducted to determine each high-risk region;

[0204] The transformer under test is risk-identified according to each of the aforementioned high-risk areas.

[0205] Specifically, the storage medium can be a computer-readable storage medium, which can be an electronic storage device such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM.

[0206] Optionally, the refined and extended functions of the program can be found in the description of the transformer risk detection method in the method embodiments.

[0207] Furthermore, the functional modules in the various embodiments of this disclosure can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a live streaming device, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this disclosure.

[0208] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0209] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0210] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting transformer risks, characterized in that, include: A three-dimensional finite element model corresponding to the transformer under test is established, and the electric field, flow field and temperature field of the three-dimensional finite element model are solved. The comprehensive risk value of each voxel is calculated based on the electric field, flow field, and temperature field. A risk heatmap is generated based on the comprehensive risk value; Each high-risk voxel is selected from the risk heat map, and each high-risk voxel forms a first risk candidate region. A comprehensive screening of each of the first risk candidate regions is conducted to determine each high-risk region; The transformer under test is risk-identified according to each of the aforementioned high-risk areas.

2. The method according to claim 1, characterized in that, The establishment of the three-dimensional finite element model corresponding to the transformer under test includes: Disassemble the various components of the transformer under test; Determine the dimensions of each component and the assembly relationships between the components; Based on the dimensions and assembly relationships, a finite element model is created; The finite element model is discretized into multiple meshes, and the locations of large gradients and gradual changes of the transformer under test are determined. Mesh refinement is performed at each large gradient location corresponding to the finite element model, while mesh coarsening is performed at each gradual change location. Multiple preset parameters are configured on the finite element model after mesh refinement and mesh coarsening to obtain a three-dimensional finite element model.

3. The method according to claim 1, characterized in that, The calculation of the comprehensive risk value for each voxel based on the electric field, flow field, and temperature field includes: Based on the electric field, flow field, and temperature field, calculate the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value for each voxel; Weighting coefficients are set for the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value, respectively. Based on the weighting coefficients, the temperature risk value, electric field risk value, temperature gradient risk value, and flow risk value are weighted and summed to obtain the comprehensive risk value.

4. The method according to claim 1, characterized in that, The process of filtering out high-risk voxels from the risk heatmap includes: For each voxel in the risk heatmap, determine whether the overall risk value of that voxel is greater than a preset first threshold. If so, then that voxel is classified as a high-risk voxel; Using this voxel as the center, determine its neighboring voxels; For each of the neighboring voxels, determine whether the comprehensive risk value of the neighboring voxel is not less than a preset second threshold and is less than the first threshold; If so, then that neighboring voxel is considered a high-risk voxel.

5. The method according to claim 1, characterized in that, The step of comprehensively screening each of the first risk candidate regions to determine each high-risk region includes: Morphological filtering is performed on each of the first risk candidate regions to obtain each of the second risk candidate regions. For each second risk candidate region, calculate the maximum comprehensive risk value, the average comprehensive risk value of each voxel in the second risk candidate region, and the volume ratio and length-width-height ratio of the second risk candidate region in the risk heat map; Based on the maximum comprehensive risk value, average value, volume ratio, and length-width-height ratio, calculate the comprehensive quantitative score of the second risk candidate region; Based on the comprehensive quantitative score, each high-risk region is determined from each of the second risk candidate regions.

6. The method according to claim 5, characterized in that, The step of performing morphological filtering on each of the first risk candidate regions to obtain each of the second risk candidate regions includes: For each of the first risk candidate regions, calculate the volume and aspect ratio of the first risk candidate region; Calculate the minimum detectable volume based on the three-dimensional finite element model; Determine whether the volume of the first risk candidate region is smaller than the minimum detectable volume, and / or whether the elongation ratio is greater than a preset ratio threshold; If so, delete the first risk candidate region and use the remaining first risk candidate region as the region to be inspected. Determine whether there are cross-material boundary phenomena in each of the areas to be inspected; One or more first risk candidate regions with cross-material boundary phenomena are divided according to the geometric interface boundary; Each region obtained after segmentation, along with each region to be inspected that does not exhibit cross-material boundary phenomena, are considered as second risk candidate regions.

7. The method according to any one of claims 1 to 6, characterized in that, It also includes calculating representative operating conditions for each of the aforementioned high-risk areas, including: For each of the high-risk areas, the mesh element corresponding to the high-risk area is determined as the first mesh element in the three-dimensional finite element model, and the non-corresponding mesh element is determined as the second mesh element. Each of the first grid cells is marked with a first mark, and each of the second grid cells is marked with a second mark; A three-dimensional partitioned mask is formed based on the first and second marks; Based on the three-dimensional partition mask, the representative working conditions of the high-risk area are calculated.

8. A transformer risk detection device, characterized in that, include: The model building and solution module is used to build a three-dimensional finite element model corresponding to the transformer under test, and solve the electric field, flow field and temperature field of the three-dimensional finite element model. The comprehensive risk value calculation module is used to calculate the comprehensive risk value of each voxel based on the electric field, flow field, and temperature field. A risk heatmap generation module is used to generate a risk heatmap based on the comprehensive risk value; The first risk candidate region determination module is used to screen out each high-risk voxel from the risk heat map and form each first risk candidate region from each of the high-risk voxels. The high-risk area determination module is used to comprehensively screen each of the first risk candidate areas to determine each high-risk area; The risk location module is used to locate the risk of the transformer under test according to each of the high-risk areas.

9. A transformer risk detection device, characterized in that, Including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the transformer risk detection method as described in any one of claims 1-7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the transformer risk detection method as described in any one of claims 1-7.

Images