A converter transformer aging assessment method based on non-uniform physical field and fault case

Abstract

This invention belongs to the field of power equipment condition assessment and intelligent diagnosis technology, and discloses a converter transformer aging assessment method based on non-uniform physical fields and fault cases. The method includes: acquiring converter transformer structural data, operating condition data, and a non-uniform parameter baseline structure; establishing an electromagnetic, thermal, and mechanical multi-physics coupling model and generating a multi-physics distribution standard package; performing non-uniform aging rate modeling and local lifetime statistics to generate an overall lifetime draft structure; performing location mapping, failure label alignment, and non-uniform coefficient reverse correction; integrating subjective and objective weights and quantifying the impact level to generate an aging index weight package; and performing level determination, parameter dictionary updating, and historical fault case database supplementation to obtain an updated non-uniform parameter baseline structure. This invention enables refined assessment of the non-uniform aging characteristics of key components, improving the accuracy and reliability of converter transformer lifetime assessment.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition assessment and intelligent data processing, specifically involving a converter transformer aging assessment method based on non-uniform physical fields and fault cases. Background Technology

[0002] Converter transformers (hereinafter referred to as "converter transformers") are core equipment in high-voltage direct current transmission systems, and their operating status directly affects the reliability and security of the entire power grid. The internal structure of converter transformers is complex, and during operation, they are simultaneously subjected to the combined effects of multiple physical fields, including electromagnetic, thermal, and mechanical fields, resulting in a significantly non-uniform aging process for their insulation materials. Therefore, accurate assessment and lifespan prediction of the aging status of key internal components of converter transformers is a crucial issue in the field of power equipment condition assessment and intelligent diagnostics.

[0003] Existing methods for aging assessment of converter transformers typically rely on segmented analysis of structural data, operational data, and local maintenance records. They often infer insulation degradation trends using single thermal or electrical stress models, combined with maintenance personnel's experiential judgment for maintenance decisions. This approach suffers from limitations such as insufficient description of the internal electromagnetic-thermal coupling behavior of the converter transformer, discontinuous updates to non-uniform parameter baseline structures, and coarse-grained reuse of historical fault cases. Existing methods often treat local monitoring information independently, manually comparing on-site anomaly locations with existing fault records during manual inspections or offline analysis. In complex service scenarios requiring simultaneous consideration of electromagnetic-thermal-mechanical multi-physics coupling modeling and operational data constraints, inconsistencies in location-level judgments and temporal sources easily arise, making it difficult to stably achieve converter transformer level determination, equipment-level archiving mapping, and parameter dictionary updates. Regarding the joint processing of how to generate an updated non-uniform parameter baseline structure through electromagnetic-thermal-mechanical multiphysics coupling modeling and non-uniform coefficient reverse correction, existing technologies generally lack a consistent judgment link between part location mapping and failure label alignment, non-uniform coefficient reverse correction, subjective and objective weight fusion and impact level quantification. It is difficult to form a continuous link in the application scenario of converter transformer aging assessment, from obtaining converter transformer structural data, operating condition data and non-uniform parameter baseline structure to generating aging index weight package and then to supplementing the historical fault case library. As a result, the updated non-uniform parameter baseline structure cannot provide closed-loop support for subsequent electromagnetic-thermal-mechanical multiphysics coupling modeling. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned shortcomings by providing a converter transformer aging assessment method based on non-uniform physical fields and fault cases. This method acquires converter transformer structural data, operating condition data, and a non-uniform parameter baseline structure; establishes an electromagnetic, thermal, and mechanical multi-physics coupling model and generates a multi-physics distribution standard package; performs non-uniform aging rate modeling and local lifetime statistics to generate an overall lifetime draft structure; performs component location mapping, failure label alignment, and non-uniform coefficient reverse correction; integrates subjective and objective weights and quantifies the impact level to generate an aging index weight package; and performs level determination, parameter dictionary updates, and historical fault case database additions to obtain an updated non-uniform parameter baseline structure. This method achieves refined assessment of the non-uniform aging characteristics of key components, improving the accuracy and reliability of converter transformer lifetime assessment. It solves the problems of insufficient description of the multi-physics coupling effect within the converter transformer, coarse granularity in the fusion of historical fault cases and real-time assessment data, inability to accurately map data, and disconnect between assessment results and physical model parameter updates, failing to form a closed-loop adaptive optimization.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An aging assessment method for converter transformers based on non-uniform physical fields and fault cases includes the following steps:

[0007] S1: Acquire converter transformer structure data, operating condition data and non-uniform parameter baseline structure, extract winding interlayer arrangement, core column arrangement and boundary correction factor fields, perform electromagnetic-thermal-mechanical multi-physics coupling modeling, spatial grid unification and coordinate alignment, boundary consistency verification, outlier annotation and numerical normalization, and generate multi-physics distribution standard package.

[0008] S2, read the electric field distribution, temperature distribution and mechanical stress distribution in the multiphysics distribution standard package according to the grid cell, perform non-uniform aging rate modeling, local lifetime integration, and combine historical load stages to perform cumulative damage statistics, regional aggregation, key part marking and global summary processing to generate an overall lifetime draft structure containing part aggregation entries.

[0009] S3, extract the key part labels in the overall life draft structure, perform part location mapping and failure label alignment, combine historical failure case library and field monitoring data to perform non-uniform coefficient reverse correction, and use the correction parameter set to drive subjective and objective weight fusion and influence level quantification to generate aging index weight package carrying level label.

[0010] S4, perform level determination, equipment-level archiving mapping, parameter dictionary update, non-uniform coefficient version registration and historical fault case library supplementation on the aging index weight package to generate the updated non-uniform parameter baseline structure.

[0011] The updated non-uniform parameter baseline structure is passed to step S1 to obtain converter transformer structure data, operating condition data, and non-uniform parameter baseline structure. Electromagnetic-thermal-mechanical multiphysics coupling modeling, spatial mesh unification and coordinate alignment, boundary consistency verification, outlier labeling, and numerical normalization are then performed to generate a multiphysics distribution standard package. This process, which is the multiphysics modeling entry point passed to step S1, forms an input link flowing back from step S4 to step S1. This allows the next round of modeling to directly load the current updated non-uniform parameter baseline structure as the latest version while reading converter transformer structure data and operating condition data. In subsequent loops, it continues to participate in non-uniform aging rate modeling, local lifetime integration and cumulative damage statistics, regional aggregation, key component marking, and global summary processing, ultimately re-entering the level determination and equipment-level archiving mapping.

[0012] Furthermore, the specific process of step S1 includes:

[0013] S1-1: Obtain converter transformer structure data, operating condition data, and non-uniform parameter baseline structure. By loading the converter transformer structure data, operating condition data, and non-uniform parameter baseline structure into the same modeling entry point, the input source for this step is formed.

[0014] S1-2 is solved by coupled modeling of electromagnetic-thermal-mechanical multiphysics fields;

[0015] Electromagnetic-thermal-mechanical multiphysics coupled modeling refers to simultaneously constructing electromagnetic field models, thermal flow field models, and mechanical stress models within the same solution framework, while maintaining the spatial interdependence of the three types of physical quantities according to the actual positional relationships of the component entities during the solution process. The above three types of models are jointly solved in the same solution process. During this process, boundary condition correction factors in the non-uniform parameter baseline structure are used to compensate for the boundary states of known high-concentration areas such as winding ends and clamp contact areas, so that the local areas exhibit stress concentration, temperature rise accumulation, and electric field distortion characteristics consistent with the actual service state. The initial multiphysics model is obtained through joint solution.

[0016] S1-3: Extract electric field distribution, temperature distribution and mechanical stress distribution from the initial multiphysics model, perform spatial grid unification and coordinate alignment processing, and form a multiphysics distribution data package through spatial grid unification and coordinate alignment processing;

[0017] The spatial grid unification and coordinate alignment process maps the electric field distribution, temperature distribution, and mechanical stress distribution to the same three-dimensional spatial grid indexing system, and uniformly registers the coordinate origin, coordinate axis direction, grid scale, and sampling step size of each distribution, forming a field distribution structure that can be compared point by point under the same grid coordinate system;

[0018] The multiphysics distribution data package uses a unified grid as an index to bind the electric field distribution, temperature distribution, and mechanical stress distribution under the same coordinates, and retains the part identifier, component name, and insulation layer type corresponding to each grid position, which is convenient for subsequent lifetime derivation stage to read point by point according to spatial location;

[0019] S1-4 involves boundary consistency verification, outlier labeling, and numerical normalization to output stable and usable data assets. Through the above boundary consistency verification, outlier labeling, and numerical normalization, a multiphysics distribution standard package is formed. The multiphysics distribution standard package is recorded as the final output field name of this step and is directly submitted to subsequent steps as input.

[0020] Furthermore, the converter transformer structure data mentioned in step S1-1 is a structured description of the target converter transformer body in terms of geometry, assembly, and insulation configuration, reflecting the relative spatial relationship between the insulation system and load-bearing components, and recording the contact surfaces, fixing points, and constraint areas between components; including the interlayer arrangement of windings and lead routing, the geometric arrangement of core columns and yokes, the positioning relationship of clamps, pressure plates, tensioning parts, and support parts, the thickness distribution of main insulation and local insulation, and the stacking order, etc.

[0021] The operating condition data is loading information related to the working status of the target converter transformer in the current or typical operating scenario, including load level, long-term average load fluctuation range, heat dissipation and cooling method, cooling flow direction of oil circuit or air circuit, ambient temperature, external mechanical fixing constraints, and historical stress traces caused by transportation, hoisting, short circuit impact, etc.

[0022] The non-uniform parameter baseline structure is a centralized representation of the initial cognitive results of the existing state of the converter transformer. The non-uniform parameter baseline structure includes a non-uniform coefficient to describe the degree of non-uniformity of the internal electric field distribution, a boundary condition correction factor to describe the locations such as the winding end, lead outlet, and pressure plate contact area, and a material aging reference curve version number to describe the trend of material degradation sensitivity.

[0023] Furthermore, the electromagnetic field model described in steps S1-2 is based on the winding position, lead direction, and insulation layer thickness information in the converter transformer structure data to arrange the applied voltage excitation boundary and dielectric distribution, forming a spatial distribution description of the internal electric field intensity; the thermal flow field model is based on the load level, cooling method, and ambient temperature information in the operating condition data to model the heat conduction, heat convection, and heat exchange of the heat-generating area and heat dissipation path, forming a spatial distribution description of the internal temperature field; the mechanical stress model is based on the constraint relationships of clamps, pressure plates, supports, and fasteners in the converter transformer structure data, and combined with the external constraint conditions and historical impact event descriptions recorded in the operating condition data to establish a distribution description of the internal mechanical stress.

[0024] Furthermore, the initial multiphysics model described in steps S1-2 is a coupled structure containing three spatial components: electric field distribution, temperature distribution, and mechanical stress distribution, which can maintain the correspondence between each component and the actual component coordinates.

[0025] Furthermore, in steps S1-3, the electric field distribution is analyzed as the electric field intensity record at each spatial location of the insulation system during the processing; the temperature distribution is analyzed as the temperature record of the winding, oil passage, adjacent area of ​​clamping parts, and insulation interlayer under steady state or specific working conditions; and the mechanical stress distribution is analyzed as the internal force record at locations such as pressure plate, tensioning parts, clamping parts, and winding end support area.

[0026] Furthermore, the boundary consistency check described in steps S1-4 is an integrity verification operation performed on the edge region, transition region, and contact interface region of the multi-physics distribution data packet. The boundary consistency check process is as follows: check whether the electric field record at the transition point from the winding end to the lead outlet is continuous; check whether there are any breaks or abrupt changes in the mechanical stress record at the contact area between the pressure plate and the winding support; check whether there are any isolated high-value points in the temperature record at the transition point from the high-temperature zone to the cooling channel. If any abnormal data points are found in the record that have obvious jumps, missing measurements, or deviate from the nominal range of the non-uniform parameter baseline structure, then the abnormal data points are marked.

[0027] Furthermore, the anomaly labeling in steps S1-4 involves adding anomaly tags and anomaly type descriptions to the locations of abnormal grids and writing them as supplementary records in the multiphysics distribution data package. The anomaly tags are used for weight adjustment or to remove interference from specific points in the subsequent aging rate modeling stage. The numerical normalization process is to perform dimensionless mapping on various numerical records in the multiphysics distribution data package after boundary consistency verification and anomaly labeling, so that the electric field distribution, temperature distribution and mechanical stress distribution are expressed in a unified scale range, which facilitates direct comparison across physical fields.

[0028] Furthermore, the multiphysics distribution standard package mentioned in steps S1-4 is a standardized data set that registers aligned electric field distribution, temperature distribution, and mechanical stress distribution under a unified coordinate system and unified dimensions. It carries anomaly labels, component names, insulation layer types, boundary condition correction factor reference identifiers, and corresponding entry indexes of non-uniform parameter baseline structures at each spatial grid location.

[0029] Furthermore, the specific process of step S2 includes:

[0030] S2-1, Import the multiphysics distribution standard package into the non-uniform aging rate modeling process to generate location-level aging rate records and form a location-level aging rate sequence.

[0031] S2-2, For the location-level aging rate sequence formed in step S2-1, perform local lifetime integration statistics to form a region-level lifetime accumulation entry; then perform cumulative damage statistics to form a local lifetime distribution description; the local lifetime distribution description is registered as a local lifetime distribution map, and the generation process of the local lifetime distribution map retains the cumulative damage statistics details for each region of interest; the local lifetime distribution map is written into the internal data buffer of the step and is ready to enter step S2-3.

[0032] S2-3, based on the local lifetime distribution map, perform regional aggregation, key component marking and global summary processing to finally form an overall lifetime draft structure. The overall lifetime draft structure is a whole-machine-level lifetime status draft formed for the target converter transformer, which uniformly records the aggregated lifetime entries, key component labels, regional aggregation relationships, historical operation segment identifiers and grid-level source references for each key component.

[0033] Furthermore, the non-uniform aging rate modeling in step S2-1 is the process of establishing aging rate descriptions for each spatial grid location inside the converter transformer. Non-uniform aging rate modeling involves reading the electric field distribution record, temperature distribution record, and mechanical stress distribution record corresponding to each grid cell, and combining this with boundary condition correction factor references to weight the stress concentration, heat accumulation, and electric field distortion at locations such as winding ends, lead outlets, and clamp contact areas. Additionally, grid cells marked as abnormal points in the previous step are weighted down or removed according to anomaly labels, generating location-level aging rate records.

[0034] Furthermore, the local life integral statistics in step S2-1 aggregates the location-level aging rate records of multiple grid cells belonging to the same insulation region according to their spatial proximity, and then combines them with the historical operating load stage, cooling condition stage and known maintenance intervention stage of the insulation region to form a region-level life cumulative entry.

[0035] Cumulative damage statistics are based on the above-mentioned regional-level cumulative lifespan entries. They are performed in segments for operation phases with long time spans or significant load fluctuations, recording the contribution percentage of each segment to the lifespan consumption of the same region, thus providing a description of the local lifespan distribution.

[0036] Furthermore, the specific process of step S4 includes:

[0037] S4-1, Perform a level determination on the aging index weight package to obtain the aging status level record corresponding to the overall machine status;

[0038] Specifically, each key component label and its corresponding level label in the aging index weight package are analyzed item by item;

[0039] After analyzing each item, all key part labels are grouped to form a set of part-level labels;

[0040] Based on the set of part-level labels, and according to the distribution range and concentration of each label throughout the machine, as well as whether it involves windings, lead wire outlets, pressure plate contact areas, and clamp support areas, a whole-machine level label is formed.

[0041] Furthermore, the item-by-item parsing process includes: reading the component aggregation entry corresponding to the key component label, reversing the position of the component aggregation entry in the overall life draft structure, and confirming the structural hierarchy and assembly connection relationship of the key component in the whole machine; reading the corrected non-uniformity coefficient and influence level description corresponding to the key component label, and determining whether the influence level of the key component in the aging index weight package reaches the level that needs to be listed separately; reading the field monitoring data reference pointer and historical fault case library reference pointer associated with the key component label, and setting a mark for key component labels with long-term repeated abnormal records or clear defect records.

[0042] The above determination process generates an aging status level record. The aging status level record is the status archive entry data used for subsequent processing in this step. It contains the correspondence between the part level label set and the whole machine level label, and maintains the reference relationship between the part level label and the whole machine level label as a traceable index so that it can be directly bound in subsequent archiving mapping.

[0043] S4-2 After obtaining the aging status level record, perform equipment-level archiving mapping processing to form an aging assessment result archive.

[0044] Equipment-level archiving mapping is used to directly bind aging status level records to the engineering management objects of converter transformers, and form an aging assessment result archive that can be archived for a long time.

[0045] Specifically, the set of part level labels in the aging status level record is read, and each part level label is bound one-to-one with the aforementioned key part label. The physical installation location of the key part in the converter transformer structure, the corresponding insulation layer type, the description of the load-bearing or support component, the description of the assembly transition surface, the description of the thermal channel, and the description of the stress concentration contact surface are recorded. In this way, the abstract level label is applied to the specific component unit that can be identified in the on-site maintenance process.

[0046] Furthermore, for the same key component label, the corrected non-uniform coefficient in the aging index weight package is called, and the corrected non-uniform coefficient is recorded in the file along with the physical description of the key component. Additionally, field monitoring data reference pointers and historical fault case library reference pointers are added, so that the record of the key component in the file carries the perspectives of operation monitoring, fault experience, and physical field modeling.

[0047] Based on this, this step archives the overall machine level label, which is then written into the overall machine level entry in the aging assessment result file, forming an overall machine level entry. A two-way reference relationship is established between the overall machine level entry and the part level label, allowing for direct retrieval of related parts in subsequent maintenance management systems or condition assessment systems.

[0048] The output of the aforementioned equipment-level archiving mapping forms the aging assessment result archive. This archive is an intermediate archiving product of this step, recording components such as part-level labels, overall machine-level labels, corrected non-uniformity coefficients, descriptions of the physical installation locations of key components, field monitoring data reference pointers, and historical fault case library reference pointers. Structurally, the aging assessment result archive is rooted at the converter, with entries for key components linked downwards. It also retains index pointers to the overall lifespan draft structure, allowing for backtracking to the lifespan derivation process of the previous main step when needed.

[0049] S4-3 After generating the aging assessment result file, update the parameter dictionary to solidify the part-level status data in the aging assessment result file into searchable parameter entries.

[0050] Parameter dictionary update refers to writing the descriptive fields related to key parts in the aging assessment result file into the long-term parameter dictionary; the parameter dictionary is a centralized descriptive set used by this invention to record the operating status characteristics, component relationship characteristics, boundary condition characteristics and change records of converter transformers.

[0051] During the parameter dictionary update process, a parameter entry is created for each key component. This entry includes a key component label, the corresponding corrected non-uniformity coefficient, the corresponding component level label, the overall machine level label to which the key component belongs, a pointer to the relevant field monitoring data, a pointer to the relevant historical fault case library, and an archiving time stamp. Each parameter entry is assigned a current version number and a source identifier when written to the parameter dictionary, distinguishing whether it originates from the current aging assessment result archive or from existing data. In this way, the parameter dictionary update operation solidifies the component-level status data in the aging assessment result archive into searchable parameter entries.

[0052] S4-4, Register the version of the non-uniform coefficients based on the completed parameter dictionary update.

[0053] Non-uniformity coefficient version registration refers to the process of forming a continuous version sequence of corrected non-uniformity coefficients over time, used to record the evolution of the non-uniformity coefficient of a key component across multiple assessments. Specifically, for each key component label, the corrected non-uniformity coefficient recorded in the aging assessment result file is read, and the non-uniformity coefficient of the same key component label in the previous version is also read from the parameter dictionary. If a difference is found between the corrected non-uniformity coefficient and the previous version record, a new version entry is generated in the non-uniformity coefficient version registration, and the key component label, the corrected non-uniformity coefficient, the reference pointer of the previous version record, the explanation of the source of change, the corresponding component level label, and the corresponding whole machine level label are registered as the same entry.

[0054] Understandably, non-uniform coefficient version registration is not simply an overwrite, but rather an accumulation of time-series version links based on the parameter dictionary, allowing the values ​​of non-uniform coefficients for the same critical component to be traced back to historical stages. This version link maintains a pointer-level association with the aging assessment result archive within this step, enabling subsequent calls to locate the monitoring and fault context from the version record.

[0055] S4-5, Perform the process of supplementing the historical fault case database.

[0056] The historical fault case database replenishment process involves adding new information generated during the current assessment process back to the historical fault case database. In this invention, the historical fault case database is defined as a centralized knowledge resource that records the failure location labels, maintenance records, decommissioning descriptions, and defect-causing environments of converter transformers or converter transformers of the same series.

[0057] During the supplementary data entry process, for newly emerging critical component labels or critical component labels identified as high-concern components in this assessment, the entry content of the critical component label in the aging assessment result file is first read. Then, it is determined whether there is already a failure component label record for the critical component in the historical failure case database. If not, the critical component label, the corresponding component level label, the corresponding whole machine level label, the current corrected non-uniformity coefficient, the field monitoring data reference pointer, and the entry pointer of the critical component in the aging assessment result file are written as new case entries into the historical failure case database. If it has already appeared, the case entry corresponding to the critical component label is appended with the time stamp of this assessment, the component level label, and the corrected non-uniformity coefficient. The original record is not overwritten, but rather a multi-stage record is added.

[0058] Through the above supplementary recording process, the historical failure case library obtained a state expression method consistent with this assessment, so that the part location mapping and failure label alignment in the subsequent main steps can directly reference these updated entries without having to re-establish the mapping rules.

[0059] After completing equipment-level archive mapping, parameter dictionary updates, non-uniform coefficient version registration, and historical fault case library supplementation, the updated non-uniform parameter baseline structure is output. The updated non-uniform parameter baseline structure is the final output field name of this step. It is defined as a set of baseline parameters that can be directly loaded into subsequent modeling processes, obtained by fusing the corrected non-uniform coefficients, part-level labels, whole-machine-level labels, parameter dictionary update results, non-uniform coefficient version registration results, and the latest supplemented entries in the historical fault case library generated in this assessment. The updated non-uniform parameter baseline structure not only describes the values ​​of the corrected non-uniform coefficients of each key part at the current assessment time, but also carries the corresponding time stamp and version information, and can be traced back to the archived entries in the aging assessment result file through internal pointers.

[0060] The updated non-uniform parameter baseline structure is passed to step S1 to obtain converter transformer structure data, operating condition data, and non-uniform parameter baseline structure. This process involves electromagnetic-thermal-mechanical multiphysics coupling modeling, spatial mesh unification and coordinate alignment, boundary consistency verification, outlier labeling, and numerical normalization to generate a multiphysics distribution standard package. This process is the multiphysics modeling entry point passed to step S1. This transfer forms an input link flowing back from step S4 to step S1, enabling the next round of modeling to directly load the updated non-uniform parameter baseline structure as the latest version while reading converter transformer structure data and operating condition data. In subsequent loops, it continues to participate in non-uniform aging rate modeling, local lifetime integration and cumulative damage statistics, regional aggregation, key component marking, and global summary processing, ultimately re-entering the level determination and equipment-level archiving mapping.

[0061] Beneficial effects:

[0062] Compared with the prior art, the present invention has the following significant beneficial technical effects:

[0063] 1. By generating a multiphysics distribution standard package, the converter transformer structural data, operating condition data, and non-uniform parameter baseline structure are unified into the same spatial grid after electromagnetic-thermal-mechanical multiphysics coupling modeling. After boundary consistency verification, outlier labeling, and numerical normalization, the electric field distribution, temperature distribution, and mechanical stress distribution can be directly called by non-uniform aging rate modeling, local lifetime integration, and cumulative damage statistics in the same coordinate system. This avoids limiting the converter transformer aging assessment to independent judgment scenarios based on a single physical field or manual inspection records. It enables non-uniform aging rate modeling and regional aggregation, key part marking, and global summary processing to be carried out around the same multiphysics distribution standard package and output the overall lifetime draft structure.

[0064] 2. By generating an aging index weight package, the overall life draft structure is no longer just a static description of the insulation life prediction result. Instead, it is synchronously aligned with the failure location labels in the historical fault case library, the abnormal observations of field monitoring data, and the local field distribution parameters of the multi-physics field distribution standard package under the same key location label. After non-uniformity coefficient inverse correction, a non-uniform correction parameter set is obtained. Then, after subjective and objective weight fusion and impact level quantification, an aging index weight package that can be directly used for grade determination is output. This transforms the converter transformer aging assessment process from a scattered judgment based solely on human experience to generating verifiable location-level impact level labels within the same link of location mapping and failure label alignment, non-uniformity coefficient inverse correction, and impact level quantification. This supports equipment-level archiving mapping and aging status level recording.

[0065] 3. By generating an updated non-uniform parameter baseline structure, the aging index weight package is solidified into an aging assessment result archive after level determination and equipment-level archiving mapping. It is then used for parameter dictionary updates, non-uniform coefficient version registration, and historical fault case library supplementation. This allows the non-uniform coefficients corrected at the part level, aging status level records, and the archived overall system level to be written into the updated non-uniform parameter baseline structure within the same process. This serves as input for electromagnetic-thermal-mechanical multiphysics coupling modeling, forming a closed loop from the multiphysics distribution standard package, overall life draft structure, aging index weight package to the updated non-uniform parameter baseline structure. This enables the converter transformer status assessment to continue to be called in subsequent loops and enter the chain of non-uniform aging rate modeling, local life integration and cumulative damage statistics, and subjective and objective weight fusion and impact level quantification. This ensures continuous and traceable data consistency between the entry point of acquiring converter transformer structural data, operating condition data, and the non-uniform parameter baseline structure, and the exit point of historical fault case library supplementation. Attached Figure Description

[0066] The present invention will be further described below with reference to the accompanying drawings:

[0067] Figure 1 This is a flowchart illustrating a converter state assessment method based on component importance and data fusion provided in an embodiment of the present invention. Detailed Implementation

[0068] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0069] The following detailed description, with reference to the accompanying drawings and specific embodiments, illustrates a converter transformer state assessment method based on component importance and data fusion provided by the present invention.

[0070] Example 1:

[0071] Reference manual attached Figure 1 The diagram shows a flowchart of a converter state assessment method based on component importance and data fusion provided by an embodiment of the present invention.

[0072] In this embodiment, a converter transformer aging assessment method based on non-uniform physical fields and fault cases includes the following steps:

[0073] S1 acquires converter transformer structure data, operating condition data, and non-uniform parameter baseline structure, extracts winding interlayer arrangement, core column arrangement, and boundary correction factor fields, performs electromagnetic-thermal-mechanical multiphysics coupling modeling, spatial grid unification and coordinate alignment, boundary consistency verification, outlier annotation, and numerical normalization processing, and generates a multiphysics distribution standard package.

[0074] In one possible implementation, step S1 specifically includes:

[0075] S1-1: Obtain converter transformer structure data, operating condition data, and non-uniform parameter baseline structure. By loading the converter transformer structure data, operating condition data, and non-uniform parameter baseline structure into the same modeling entry point, the input source for this step is formed.

[0076] Converter transformer structural data is a structured description of the target converter transformer body in terms of geometry, assembly, and insulation configuration. This data includes the interlayer arrangement of windings and lead routing, the geometric arrangement of core columns and yokes, the positioning relationships of clamps, pressure plates, tensioning components, and supports, the thickness distribution and stacking order of main and local insulation, and other information. The converter transformer structural data reflects the relative spatial relationship between the insulation system and load-bearing components, and records the contact surfaces, fixing points, and constraint areas between components.

[0077] Operating condition data refers to loading information related to the working status of the target converter transformer under current or typical operating scenarios. This operating condition data includes load level, long-term average load fluctuation range, heat dissipation and cooling methods, cooling flow direction of oil or air circuits, ambient temperature, external mechanical constraints, and historical stress traces caused by transportation, hoisting, short-circuit impacts, etc.

[0078] The non-uniform parameter baseline structure is a centralized representation of the initial cognitive results of the existing state of the converter transformer. The non-uniform parameter baseline structure includes non-uniform coefficients to describe the degree of non-uniformity of the internal electric field distribution, boundary condition correction factors to describe the locations such as winding ends, lead outlets, and pressure plate contact areas, and material aging reference curve version numbers to describe the trend of material degradation sensitivity.

[0079] S1-2 is solved by coupled modeling of electromagnetic-thermal-mechanical multiphysics fields;

[0080] Electromagnetic-thermal-mechanical multiphysics coupled modeling refers to simultaneously constructing an electromagnetic field model, a thermal flow field model, and a mechanical stress model within the same solution framework, while maintaining the spatial interdependence of the three types of physical quantities according to the actual positional relationship of the component entities during the solution process.

[0081] The electromagnetic field model, based on the winding position, lead routing, and insulation thickness information in the converter transformer structure data, arranges the applied voltage excitation boundary and dielectric distribution to form a spatial distribution description of the internal electric field intensity. The thermal flow field model, based on the load level, cooling method, and ambient temperature information in the operating condition data, models the heat conduction, heat convection, and heat exchange of the heat-generating area and heat dissipation path to form a spatial distribution description of the internal temperature field. The mechanical stress model, based on the constraint relationships of clamps, pressure plates, supports, and fasteners in the converter transformer structure data, and combined with the external constraint conditions and historical impact event descriptions recorded in the operating condition data, establishes a distribution description of the internal mechanical stress.

[0082] The three types of models mentioned above are solved jointly in the same solution process. During this process, the boundary condition correction factor in the non-uniform parameter baseline structure is used to compensate for the boundary state of known high concentration areas such as the winding end and clamp contact area, so that the local area exhibits stress concentration, temperature rise accumulation and electric field distortion characteristics consistent with the actual service state. The initial multiphysics model is obtained through joint solution. The initial multiphysics model is a coupled result structure containing three spatial components: electric field distribution, temperature distribution and mechanical stress distribution, and maintains the correspondence between each component and the actual component coordinates.

[0083] S1-3: Electric field distribution, temperature distribution, and mechanical stress distribution are extracted from the initial multiphysics model and processed through spatial grid unification and coordinate alignment. This process maps the electric field distribution, temperature distribution, and mechanical stress distribution to the same three-dimensional spatial grid indexing system, and uniformly registers the coordinate origin, coordinate axis direction, grid scale, and sampling step size of each distribution, forming a field distribution structure that can be compared point-by-point within the same grid coordinate system. During this process, the electric field distribution is resolved as electric field intensity records at various spatial locations of the insulation system; the temperature distribution is resolved as temperature records of the winding, oil passages, adjacent areas of clamps, and insulation interlayers under steady-state or specific operating conditions; and the mechanical stress distribution is resolved as internal force records at locations such as pressure plates, tensioners, clamps, and winding end support areas. Through spatial grid unification and coordinate alignment, a multiphysics distribution data package is formed. The multiphysics distribution data package uses a unified grid as an index to bind the electric field distribution, temperature distribution, and mechanical stress distribution under the same coordinates, and retains the part identifier, component name, and insulation layer type corresponding to each grid location, which facilitates point-by-point reading according to spatial location in the subsequent lifetime derivation stage.

[0084] S1-4 performs boundary consistency checks, outlier labeling, and numerical normalization to output stable and usable data assets. Boundary consistency checks are integrity verification operations performed on the edge regions, transition regions, and contact interface regions of the multiphysics distribution data package. Specifically, it checks whether the electric field record at the transition from the winding end to the lead outlet is continuous; it checks whether there are breaks or abrupt changes in the mechanical stress record at the contact area between the pressure plate and the winding support; and it checks whether there are isolated high-value points in the temperature record at the transition from the high-temperature zone to the cooling channel. If any abnormal data points with obvious jumps, missing measurements, or deviations from the nominal range of the non-uniform parameter baseline structure are found in the records, these data points are labeled as outliers. Outlier labeling involves adding anomaly tags and anomaly type descriptions to the abnormal mesh locations and writing them as supplementary records in the multiphysics distribution data package. The anomaly tags are used for weight adjustment or removal of interference at specific points in the subsequent aging rate modeling stage. Numerical normalization, following boundary consistency checks and outlier labeling, performs dimensionless mapping on various numerical records in the multiphysics distribution data package. This allows the electric field distribution, temperature distribution, and mechanical stress distribution to be expressed within a unified scale, facilitating direct comparison across physical fields. Through these boundary consistency checks, outlier labeling, and numerical normalization processes, a multiphysics distribution standard package is formed. This standard package is a standardized data set of aligned electric field, temperature, and mechanical stress distributions registered in a unified coordinate system and with unified dimensions. Each spatial grid location carries an outlier label, component name, insulation type, boundary condition correction factor reference identifier, and corresponding entry index for the non-uniform parameter baseline structure. This multiphysics distribution standard package is recorded as the final output field name of this step and directly submitted to subsequent steps as input. The multiphysics distribution standard package is passed to subsequent steps to drive processing steps such as non-uniform aging rate modeling, local lifetime integration, and cumulative damage statistics. Non-uniform aging rate modeling is part of the calculation process corresponding to the next step, S2. In other words, the multiphysics distribution standard package, once generated, serves as the input for step S2, providing data for the subsequent derivation of location-level aging rates, local lifetime distributions, and the overall lifetime draft structure. This forms a top-down continuous input chain in the overall process of the commutator aging assessment method.

[0085] The technical effects of this step can be summarized as follows: By jointly solving the converter variable structure data, operating condition data and non-uniform parameter baseline structure under the same electromagnetic-thermal-mechanical multiphysics coupled modeling framework, and performing spatial grid unification and coordinate alignment, boundary consistency verification, outlier annotation and numerical normalization on the results, a multiphysics distribution standard package is output, forming a unified field distribution basic data asset that can be directly input into the subsequent step S2.

[0086] S2, read the electric field distribution, temperature distribution and mechanical stress distribution in the multiphysics distribution standard package according to the grid cell, perform non-uniform aging rate modeling, local lifetime integration, and combine historical load stages to perform cumulative damage statistics, regional aggregation, key part marking and global summary processing to generate an overall lifetime draft structure containing part aggregation entries.

[0087] The multiphysics distribution standard package is a standardized data set that records the electric field distribution, temperature distribution, and mechanical stress distribution under a unified grid coordinate system. Each grid cell carries spatial coordinates, component name, insulation layer type, boundary condition correction factor reference identifier, and anomaly label.

[0088] In one possible implementation, step S2 specifically includes:

[0089] S2-1, Import the multiphysics distribution standard package into the non-uniform aging rate modeling process to generate location-level aging rate records and form a location-level aging rate sequence.

[0090] Non-uniform aging rate modeling is the process of establishing aging rate descriptions for each spatial grid location within a converter transformer. Non-uniform aging rate modeling involves reading the electric field distribution record, temperature distribution record, and mechanical stress distribution record corresponding to each grid cell, and combining this with boundary condition correction factors to weight the stress concentration, heat accumulation, and electric field distortion at locations such as winding ends, lead outlets, and clamp contact areas. Furthermore, grid cells marked as anomalous points in the previous step are weighted down or removed based on anomaly labels, generating location-level aging rate records.

[0091] The location-level aging rate record is a set of time-series entries of aging rates registered grid by grid, used to reflect the degradation tendency of the insulating medium or load-bearing component corresponding to the grid cell under actual service conditions. Through the above processing, a location-level aging rate sequence can be understood to be formed. The location-level aging rate sequence is regarded as the calculation product of this section of processing and will be directly called in the subsequent local life integration and cumulative damage statistics stages.

[0092] S2-2, For the location-level aging rate sequence formed in step S2-1, perform local lifetime integration statistics to form regional lifetime accumulation entries; then perform cumulative damage statistics to form a local lifetime distribution description.

[0093] Local lifetime integration refers to the aggregation of location-level aging rate records of several grid cells belonging to the same structural part within the same insulation region, to obtain the remaining lifetime characterization of the insulation region under current and historical operating conditions. In one possible implementation, the same insulation region may be a winding interlayer insulation region, a wire end insulation encapsulation region, a pressure plate contact pad region, a transition region between clamps and winding supports, etc. These regions are uniquely indicated in the multiphysics distribution standard package through component name and insulation layer type fields.

[0094] Local life integral statistics aggregates the location-level aging rate records of multiple grid cells belonging to the same insulation region according to their spatial proximity, and then combines them with the historical operating load stage, cooling condition stage and known maintenance intervention stage of the insulation region to form a region-level life cumulative entry.

[0095] Cumulative damage statistics are based on the above-mentioned regional-level cumulative lifespan entries. They are performed in segments for operation phases with long time spans or significant load fluctuations, recording the contribution percentage of each segment to the lifespan consumption of the same region, thus providing a description of the local lifespan distribution.

[0096] The local lifetime distribution description is registered as a local lifetime distribution map. The local lifetime distribution map is a panoramic dataset of lifetime decay presented at the regional level, carrying records such as region identifiers, grid set pointers, historical operational segment identifiers, and cumulative lifetime entries. The generation process of the local lifetime distribution map also retains detailed cumulative damage statistics for each region of interest. This detail describes which operational phases contributed significantly to lifetime loss in that region, used for subsequent identification of critical components. At this point, understandably, the local lifetime distribution map is written into the internal data buffer of this step and is ready for region aggregation, critical component marking, and global summary processing.

[0097] S2-3 performs regional aggregation, key component marking, and global summary processing based on the local lifetime distribution map.

[0098] Region aggregation merges multiple insulation regions within the same structural level to form a location-level aging profile. Specifically, multiple insulation regions belonging to the same winding end, multiple insulation regions belonging to the same lead outlet path, and multiple load-bearing contact regions belonging to the same clamping area of ​​the pressure plate and tensioner are aggregated according to their spatial continuity and assembly relationship to form location aggregation entries. These location aggregation entries not only include the cumulative lifetime entries from the local lifetime distribution map but also carry component names and insulation layer type fields copied from the multiphysics distribution standard package, allowing the location aggregation entry to directly pinpoint the physical components and geometric locations within the converter transformer.

[0099] Critical component marking is a highlighting process performed based on the aforementioned component aggregation entries. It identifies components exhibiting rapid attenuation characteristics in their cumulative lifespan entries, components subjected to prolonged high loads during operation, and stress-sensitive components such as the ends of cooling paths or transition points of support components, generating critical component labels for these components. The critical component label is a component-level indicator field containing the component name, insulation type, assembly positioning description, support or clamping relationship description, and a pointer to the corresponding component aggregation entry. Using critical component labels, subsequent steps can quickly map the component to historical maintenance records, failure reports, and online monitoring records.

[0100] Global aggregation processing is a system-wide unified sorting operation performed after regional aggregation and key component marking. Specifically, all component aggregation entries and their key component labels are arranged according to the structural hierarchy of the converter transformer, establishing an index structure from the whole system to components and from components to regions. Within the same structure, reverse pointers to local lifetime distribution maps and location-level aging rate sequences are retained, forming a traceable data reference link. The result of global aggregation processing is the overall lifetime draft structure, which is a system-wide lifetime status draft for the target converter transformer. It uniformly records the aggregated lifetime entries, key component labels, regional aggregation relationships, historical operation segment identifiers, and grid-level source references for each key component.

[0101] After this step is completed, the overall lifetime draft structure is registered as an output field name and directly passed to subsequent main steps. Specifically, the overall lifetime draft structure is input into the part location mapping and failure label alignment processing in subsequent steps, driving the invocation of the historical failure case library, the splicing of field monitoring data, and the reverse correction of non-uniformity coefficients in the next main step. Part location mapping and failure label alignment fall under the processing scope of the subsequent main step, i.e., step S3, and depend on the key part labels, part aggregation entries, and regional aggregation relationships within the overall lifetime draft structure. Simultaneously, after being consumed by subsequent steps, the overall lifetime draft structure will continue to influence the sorting logic of subjective and objective weight fusion and influence level quantification processing, and indirectly participate in the formation of the final aging index weight package; this aging index weight package will serve as the pre-input for the next stage of level determination and equipment-level archiving mapping, and then enter the processing flow of step S4.

[0102] The technical effects of this step can be summarized as follows: By performing non-uniform aging rate modeling, local lifetime integration, and cumulative damage statistics on the multi-physics distribution standard package, and based on this, regional aggregation, key part marking, and global summary processing are carried out to obtain the overall lifetime draft structure. This allows the lifetime consumption behavior of the converter transformer in different components, different insulation regions, and different assembly transition positions to be uniformly converged into referenceable data results at the part level and the whole machine level. This provides clear input for subsequent steps such as reverse correction based on fault cases, fusion of subjective and objective weights, and aging state classification.

[0103] S3, extract the key part labels in the overall life draft structure, perform part location mapping and failure label alignment, combine historical failure case library and field monitoring data to perform non-uniform coefficient reverse correction, and use the correction parameter set to drive subjective and objective weight fusion and influence level quantification to generate aging index weight package carrying level label.

[0104] This step directly builds upon the overall lifetime draft structure formed in step S2, and simultaneously integrates the historical fault case library, field monitoring data, and multiphysics distribution standard packages. The overall lifetime draft structure is the system-level lifetime status draft obtained in step S2, containing component aggregation entries, key component labels, cumulative lifetime entries corresponding to local lifetime distribution maps, regional aggregation relationships, and grid-level source references. It is used to describe the lifetime consumption degree and spatial orientation relationships of each key component within the converter transformer.

[0105] The historical fault case database is a collection of operational fault records collected for target converter transformers and converter transformers of the same type or series. The records in the historical fault case database include information such as the identification of the part that had an anomaly, maintenance records, decommissioning descriptions, disassembly and inspection photos, the environment in which the defect occurred, and the operational stage identification when the defect occurred. Each record is also labeled with a failure part tag.

[0106] The on-site monitoring data is a collection of online data acquired during operation, including oil chromatography analysis records, partial discharge monitoring records, acoustic inspection records, dielectric loss angle inspection records, etc. It can be cross-linked with the historical fault case database through timestamps and location identifiers.

[0107] The multiphysics distribution standard package is derived from the modeling and calculation results of step S1.

[0108] In one possible implementation, step S3 specifically includes:

[0109] S3-1, align the execution part location mapping and the failure label to obtain the fault alignment mapping relationship;

[0110] Specifically, the aforementioned overall lifespan draft structure is used as input, and through part location mapping and failure label alignment processing, a data structure that points to specific maintenance experience records is obtained.

[0111] Location mapping refers to the process of reading key component tags one by one within the overall lifespan draft structure and comparing the component name, insulation type, assembly positioning description, and support or clamping relationship description described by the key component tag with the corresponding failure component tags in the historical failure case database at both the spatial location and naming levels. This comparison process first identifies identifiable structural components such as the converter transformer winding end, lead outlet, transition position between the pressure plate and winding support, and clamping area based on the assembly positioning description of the key component tag. Then, based on the insulation type and support or clamping relationship description, it determines whether a corresponding failure component tag record exists in the historical failure case database. If multiple records for the same physical location or assembly relationship are found in the historical failure case database, they are sorted according to the time sequence of the maintenance records and the operating stage identifier when the defect occurred, and the record closest to the current operating stage is selected as the primary reference record. This primary reference record is then bound to the corresponding key component tag, generating a component alignment entry. In this step, the location alignment entry is recorded as a structured record object, which records the key location label, the failure location label, the maintenance record, the operation stage identifier, and the index of the local life distribution entries related to the location.

[0112] The fault alignment mapping relationship is obtained through the above-mentioned location mapping and failure tag alignment processing. This fault alignment mapping relationship is not output separately in this step, but is used as the direct input for the downstream non-uniform coefficient reverse correction processing.

[0113] S3-2, perform reverse correction of non-uniform coefficients to form a non-uniform correction parameter set;

[0114] Specifically, the labels of key parts, labels of failed parts, field monitoring data, and local field distribution parameters in the multiphysics distribution standard package are extracted from the fault alignment mapping relationship, and the non-uniformity coefficient is reversed.

[0115] The non-uniformity coefficient reverse correction refers to the situation where the same critical part has three types of information at the same time: the life consumption performance identified in the overall life draft structure, the failure part labels and maintenance records accumulated in the historical failure case library, and the abnormal monitoring records of the part in the corresponding time period of the field monitoring data. Combined with the electric field distribution, temperature distribution and mechanical stress distribution data of the grid coordinates of the part in the multiphysics distribution standard package, non-uniformity correction parameters are generated to correct the original physical field description. The non-uniformity coefficient is a set of coefficients that describe the degree of unevenness of the spatial distribution of the physical field in the part, and is used to express the degree of field distribution distortion in the actual service process.

[0116] The process of inverse correction of non-uniformity coefficients includes the correspondence and constraints of three types of data. First, according to the key part label pointed to by the part alignment entry, the grid cells consistent with the key part label in the multiphysics distribution standard package are extracted to obtain the local field distribution parameters of the key part in terms of electric field, temperature, and mechanical stress. Second, according to the maintenance records, decommissioning descriptions, and defect generation environments associated with the failure part label, the abnormal description fields in these records are compared item by item with the aforementioned local field distribution parameters to determine whether there is a significant non-uniformity concentration trend. Third, according to the abnormal records of the field monitoring data, it is confirmed whether the part exhibits long-term continuous heat accumulation, stress concentration, or discharge symptoms consistent with the local field distribution parameters during the corresponding operating stage, and if consistency exists, the original non-uniformity coefficient is corrected and registered. The result of the correction registration forms a non-uniformity correction parameter set, which records the corrected non-uniformity coefficient of each key part, the source description of the field monitoring data that triggered the correction, the corresponding failure part label reference pointer, and the associated local field distribution parameter index. The non-uniformity correction parameter set is temporarily reserved as an intermediate product of calibration and submitted to the subsequent subjective and objective weight fusion and impact level quantification processing.

[0117] S3-3, based on the non-uniform correction parameter set, performs subjective and objective weight fusion and influence level quantification to obtain the field name aging index weight package.

[0118] The subjective and objective weight fusion process integrates subjective weights derived from the long-term field experience of maintenance experts with objective weights calculated using the entropy weight method. Subjective weights, provided by maintenance experts, establish a priority order for key components. This order reflects the maintenance experts' level of attention to these key components during the converter transformer's lifespan, the allocation of routine maintenance resources, and the priority of repairs. Objective weights are calculated using the entropy weight method. In this invention, the entropy weight method refers to a set of weight coefficients derived from the dispersion analysis of the fluctuation amplitude, duration, and abnormal repetition degree of field monitoring data. This set of weight coefficients is no longer expressed as mathematical symbols but rather as a description of the objective level of attention given to each key component.

[0119] When integrating subjective and objective weights, subjective and objective weights are read synchronously, and the subjective and objective weights of the same key part are weighted and synthesized according to the corrected non-uniform coefficients recorded in the non-uniform correction parameter set, so that the key part can obtain a single influence level description.

[0120] Impact level quantification is performed after the weighted synthesis described above. Impact level quantification involves dividing the synthesized impact level description into several level tiers, each corresponding to a directly referenceable level label. The level label is an archiveable field that describes the current aging importance of a critical component and maintains a one-to-one correspondence with the critical component's critical component label.

[0121] By integrating subjective and objective weights and quantifying the impact level, a set of level labels for the entire machine is obtained. This set of level labels, along with the basis for its generation, the corresponding key component labels, and the index entries for the corresponding non-uniform correction parameter set, are written into the aging index weight package.

[0122] Understandably, the aging index weight package is recorded as the output field name of this step and is directly passed to the next main step, namely step S4. Structurally, the aging index weight package integrates key component labels, corrected non-uniform coefficients, subjective and objective weight fusion results, level labels obtained from impact level quantification, reference pointers to field monitoring data, reference pointers to the historical fault case library, and the mapping relationship with component aggregation entries in the overall lifespan draft structure. This aging index weight package, as an output field name, is sent to the subsequent main steps for level determination, equipment-level archiving mapping, parameter dictionary update, non-uniform coefficient version registration, and historical fault case library supplementation. The subsequent step S4, by calling the level labels and corrected non-uniform coefficient fields in the aging index weight package, provides an aging state level record and further forms an aging assessment result archive. It then registers the updated non-uniform parameter baseline structure, enabling the converter transformer aging assessment method of this invention to transition from the early multiphysics modeling stage and lifespan derivation stage to the archive archiving and baseline update stage, forming a cyclically updated closed loop.

[0123] The technical effects of this step can be summarized as follows: By mapping the execution parts of the overall lifespan draft structure and aligning failure labels, key part labels are bound point-by-point to historical failure case libraries and on-site monitoring data; through non-uniform coefficient reverse correction processing, a non-uniform correction parameter set is formed to achieve dynamic correction of the unevenness of local field distribution; through the fusion of subjective and objective weights and the quantification of impact levels, an aging index weight package is generated, providing a directly callable data foundation for the level determination and archiving of the next main step, and the status assessment results can be included in a consistent level label system and enter into the long-term management process.

[0124] In one possible implementation, the key component labels in the overall lifespan draft structure are expanded one by one, and a component candidate set is constructed according to their component name, insulation layer type, assembly positioning description, and support or clamping relationship description. Simultaneously, the failure component labels and their accompanying maintenance records, decommissioning descriptions, and defect generation environments are read from the historical failure case library, and a failure candidate set is constructed according to the timestamp and operation stage identifier. Then, oil chromatography analysis records, partial discharge monitoring records, acoustic inspection records, and dielectric loss angle inspection records associated with the components are extracted from the field monitoring data, and a one-to-one reference relationship is established between them and the grid coordinates of the multiphysics distribution standard package.

[0125] To achieve part location mapping and failure label alignment, bidirectional similarity calculation and connectivity smoothing are first performed within a unified part label space and fault label space. To this end, an initial matching score from part to fault is defined, calculated through a combination of embedding alignment and topological nearest neighbor, and corrected according to the coherence of the monitoring signals after time alignment, yielding an intermediate quantity for alignment determination. Next, adjacency normalization is implemented on the part-fault bipartite graph to suppress the randomness of isolated high-matching pairs and enhance cluster consistency, thereby outputting a stable representation of the alignment relationship. The two calculation steps are presented sequentially: Equation 1 gives the initial matching score, and Equation 2 gives the smooth and consistent representation under graph structure constraints.

[0126] Formula 1

[0127] in:

[0128] This is the initial matching score vector from location to fault, with a value range of [value range missing]. arrive ;

[0129] , These are non-negative weight coefficients, given by the matching strategy table in the parameter dictionary;

[0130] The similarity of the embedding of parts and faults is obtained by extracting key part labels from the overall life draft structure and extracting failure part labels from the historical failure case library, and then performing name matching and assembly semantic mapping.

[0131] This is a cross-correlation function used to align location-related time series in field monitoring data;

[0132] The sequence of event pulses from partial discharge monitoring records is extracted from on-site monitoring data and then normalized in amplitude.

[0133] The key component change sequences were extracted from field monitoring data and recorded by oil chromatography analysis, and then the amplitude was normalized.

[0134] : Normalized mapping function;

[0135] The mapping from data source to metric to variable is as follows:

[0136] The partial discharge event sequence extracted from the on-site monitoring data is denoted as follows: The key component sequences of oil chromatography were extracted from the field monitoring data and denoted as follows: The two together form Then proceed to Formula 1 to finally generate... Based on the dependency relationship, Formula 1 yields... This serves as an intermediate quantity for subsequent calculations and alignment filtering, and also as one of the inputs for the next step. To suppress noise matching and reflect the topological constraints of the component structure, an adjacency matrix of the part-fault bipartite graph is constructed, and adjacency normalization and one message passing are performed to obtain a smoothing score:

[0137] Formula 2

[0138] in:

[0139] This is the alignment score vector after graph structure smoothing;

[0140] To add a unit self-loop to the location-failure bipartite graph, the adjacency matrix is ​​derived from the connectivity between the location aggregation entries in the overall lifetime draft structure and the failure location labels in the historical failure case library.

[0141] for A diagonal matrix of degrees;

[0142] : Adjacency normalization operator, used to suppress the bias of height nodes and perform one message passing;

[0143] The mapping from data source to metric to variable is as follows:

[0144] Extracting regional aggregation relationships from the overall lifespan draft structure and generating connected pairs with maintenance records from the historical failure case library, constructing... ;Depend on generate Then, use the formula obtained from formula 1. Substituting into formula 2, we get During operation, for Set a minimum connectivity threshold and a time consistency threshold (the thresholds are read from the parameter dictionary). Eliminate alignment pairs that do not meet the consistency requirements of assembly positioning description and operation phase identifier through logical consistency scoring. The retained alignment pairs are written into the fault alignment mapping relationship and stored as a fault alignment mapping package. The output field name is directly input as the output of this section and is used as the inverse correction of the non-uniformity coefficient in the next step.

[0145] Following the aforementioned fault alignment mapping package, this section proceeds to the non-uniformity coefficient reverse correction process. First, in the unified grid coordinate system of the multiphysics distribution standard package, local field distribution parameters of electric field, temperature, and mechanical stress are read according to the key locations marked in the fault alignment mapping package, and field index vectors are constructed accordingly. Second, within the same time window, anomaly observation vectors for these locations are extracted from the field monitoring data, and comparable operational anomaly characteristics are obtained through standardization and detrending processing. Then, the two types of vectors are aligned to the same location label to form a location-level residual representation, which drives the calculation of the correction amount. For this purpose, the location-level residual is first defined as the calibration target:

[0146] Formula 3

[0147] in:

[0148] This is the location-level residual vector;

[0149] The weighting matrix for each physical field component is given by the parameter dictionary and can be switched according to the operating conditions;

[0150] It is a spliced ​​vector of local field distribution parameters from the multiphysics distribution standard package, including electric field distribution, temperature distribution and mechanical stress distribution;

[0151] The abnormal observation feature vector is derived from field monitoring data, including the intensity of partial discharge events, the intensity of changes in key components of oil chromatography, and the amplitude of acoustic inspection anomalies.

[0152] The mapping from data source to metric to variable is as follows:

[0153] The local field distribution parameters of the part are extracted from the multiphysics distribution standard package and then normalized and spliced ​​as follows: Anomaly observation features are extracted from on-site monitoring data and then normalized and spliced ​​together as follows: The field component weights are extracted from the parameter dictionary using a strategy that yields the desired results. The matching score obtained from Formula 1 was used in the next step to filter part-fault pairs, and now it is obtained from Formula 2. Used as part alignment weight for selection , The co-located sample set is then incorporated into the construction of Formula 3. After obtaining the residual representation, a quadratic cost is established, and a corrected non-uniform coefficient is introduced as a variable to be solved, thus forming a convex optimization problem and obtaining the part-level correction:

[0154] Formula 4

[0155] in:

[0156] The solution is the corrected non-uniform coefficient vector;

[0157] : The operator that minimizes the value of the independent variable in the objective function;

[0158] To solve for the variable, we have the location-level increment of the non-uniform coefficient;

[0159] The sensitivity mapping matrix is ​​obtained by linear approximation of the unit perturbation response of the grid near this location to the electric field distribution, temperature distribution, and mechanical stress distribution using the multiphysics distribution standard package.

[0160] Positive regularization coefficients are read from the regularization strategy table in the parameter dictionary;

[0161] It is a norm 2;

[0162] The mapping from data source to metric to variable is as follows:

[0163] Constructed from co-located site samples filtered by fault alignment mapping package ; obtained from Formula 3 Substitute directly into the first term of Formula 4; obtain the regularization strength from the parameter dictionary. The solution obtained It is registered as a core field of the non-uniform correction parameter set along with the location label. In the runtime chain, it is first obtained from Equation 1. Generated by Formula 2 And then filter the samples at the same address, and then form in Formula 3 Finally, the solution is obtained from formula 4. Will contain The records of the reference pointers are written into the non-uniform correction parameter set, and the output field name serves as the input for the next step of subjective and objective weight fusion and influence level quantification of the output of this section.

[0164] Following the non-uniform correction parameter set, this section performs subjective and objective weight fusion and impact level quantification, outputting a consistent level label set for the entire machine and forming an aging index weight package. In this processing chain, the subjective weight vector provided by the operation and maintenance expert and the objective weight vector calculated by the entropy weight method are read first. Then, the corrected non-uniform coefficient obtained from Formula 4, together with the aforementioned aligned weights, are used as fusion factors in the weighted normalization mapping. To improve the sensitivity to high-confidence and high-correction-amplitude regions, a soft maximum mapping with a temperature parameter is introduced to complete the normalization score. Subsequently, a first-order linear programming is constructed to select a single level label on the discrete level label set, ensuring that each region corresponds to only one level result and retaining bidirectional references with the region label. The specific two steps are as follows. First, temperature-inclusive fusion normalization is performed:

[0165] Formula 5

[0166] in:

[0167] This is the fused, part-level normalized score vector;

[0168] : Normalized mapping function;

[0169] The temperature parameter is given by the temperature and threshold strategy table in the parameter dictionary.

[0170] The coefficient is a non-negative fusion coefficient.

[0171] This is a subjective weight vector, formed by the weight configuration of operations and maintenance experts;

[0172] The objective weight vector is calculated from field monitoring data using the entropy weight method.

[0173] Select the operator for the location, and apply formula 4. Map body part labels to corresponding locations;

[0174] The mapping from data source to metric to variable is as follows:

[0175] The weight priority of the extracted parts is obtained by the weight configuration of the operation and maintenance experts. The dispersion of the indicators extracted from the field monitoring data was obtained using the entropy weight method. The corrected non-uniformity coefficients are read from the non-uniformity correction parameter set and then processed. Mapping Read from the parameter dictionary Among them, Formula 3 yields... It has been used to solve Formula 4. Now obtained from Formula 4 Proceed directly to the third term of Equation 5. After normalization, to make a unique choice on the discrete rank set, construct the assignment form of the linear programming problem to solve for the label selection vector:

[0176] Formula 6

[0177]

[0178]

[0179]

[0180] in:

[0181] A selection vector for discrete level labels;

[0182] : The operator that maximizes the value of the independent variable in the objective function;

[0183] Candidates for decision vectors;

[0184] : Probabilistic simplex set;

[0185] Let it be the set of its poles;

[0186] It is a vector of all ones;

[0187] The objective function is linear.

[0188] It is the transpose symbol;

[0189] For the notation of "subject to", optimize constraint symbols;

[0190] The mapping from data source to metric to variable is as follows:

[0191] Generated from Formula 5 Directly use Formula 6 as the target coefficient to obtain... Constraints are defined by the set of level labels, specifying their dimensions and feasible regions. This is derived from Equation 5. The linear target term is explicitly substituted into Formula 6. The results are obtained for all key component labels. After that, with The non-zero component locations are mapped to specific level labels, and together with the location labels, corrected non-uniformity coefficients, subjective and objective weights, and reference pointers, they are compiled into an aging index weight package. This output field name is then used in step S4 for level determination, equipment-level archiving mapping, parameter dictionary updates, non-uniformity coefficient version registration, and historical fault case library supplementation. Simultaneously, the aging index weight package retains the mapping relationship with the location aggregation entries in the overall lifespan draft structure, allowing direct referencing of the source entries in step S2 when forming aging status level records and aging assessment result archives in step S4. In summary, this technical effect involves a three-stage link of alignment, calibration, and fusion, forming a verifiable intermediate quantity sequence from location-fault alignment score to corrected non-uniformity coefficients and then to discrete level labels. The aging index weight package serves as the steady-state output of the structured product, supporting consistent cross-step invocation and closed-loop backfeedback.

[0192] S4, perform level determination, equipment-level archiving mapping, parameter dictionary update, non-uniform coefficient version registration, and historical fault case library supplementation on the aging index weight package to generate an updated non-uniform parameter baseline structure.

[0193] The aging index weight package is a structured data set generated in step S3, containing key component labels, corrected non-uniformity coefficients, impact level descriptions after the fusion of subjective and objective weights, level labels obtained after impact level quantification, field monitoring data reference pointers, historical fault case library reference pointers, and mapping relationships between the component aggregation entries and the overall lifespan draft structure. Specifically, key component labels characterize the specific physical location, assembly relationship, and stress concentration area within the converter transformer; the corrected non-uniformity coefficients describe the spatial unevenness of the electric field distribution, temperature distribution, and mechanical stress distribution at that physical location; and the level labels are directly referenceable fields formed by dividing the impact level descriptions after the fusion of subjective and objective weights into grades.

[0194] In one possible implementation, step S4 specifically includes:

[0195] S4-1 performs a level determination on the aging index weight package to obtain the aging status level record corresponding to the overall machine status.

[0196] Specifically, each key component label and its corresponding level label in the aging index weight package are analyzed item by item.

[0197] The item-by-item analysis process includes: reading the component aggregation entry corresponding to the critical component label, reversing the position of the component aggregation entry in the overall life draft structure, and confirming the structural hierarchy and assembly connection relationship of the critical component in the whole machine; reading the corrected non-uniformity coefficient and influence level description corresponding to the critical component label, and determining whether the influence level of the critical component in the aging index weight package reaches the level that needs to be listed separately; reading the field monitoring data reference pointer and historical fault case library reference pointer associated with the critical component label, and setting a mark for critical component labels with long-term repeated abnormal records or clear defect records.

[0198] After the above item-by-item analysis, all key component labels are grouped to form a component-level label set. The component-level label set describes the components inside the converter transformer according to their level labels, and explains the clustering results for each category by recording the corrected non-uniformity coefficient and the influence level description.

[0199] Next, based on the set of component-level labels, and according to the distribution range and concentration of each label across the entire machine, as well as whether it involves high-concern areas such as windings, lead outlets, pressure plate contact areas, and clamp support areas, a whole-machine-level label is formed. The whole-machine-level label is a hierarchical description of the overall state of the converter transformer and is registered as a whole-machine-level summary field.

[0200] The above determination process generates an aging status level record. The aging status level record is the status archive entry data used for subsequent processing in this step. It contains the correspondence between the part level label set and the whole machine level label, and maintains the reference relationship between the part level label and the whole machine level label as a traceable index so that it can be directly bound in subsequent archiving mapping.

[0201] S4-2 After obtaining the aging status level record, perform equipment-level archiving mapping processing to form an aging assessment result archive.

[0202] Equipment-level archiving mapping is used to directly bind aging status level records to the engineering management objects of converter transformers, and form an aging assessment result archive that can be archived for a long time.

[0203] Specifically, the set of part level labels in the aging status level record is read, and each part level label is bound one-to-one with the aforementioned key part label. The physical installation location of the key part in the converter transformer structure, the corresponding insulation layer type, the description of the load-bearing or support component, the description of the assembly transition surface, the description of the thermal channel, and the description of the stress concentration contact surface are recorded. In this way, the abstract level label is applied to the specific component unit that can be identified in the on-site maintenance process.

[0204] Furthermore, for the same key component label, the corrected non-uniform coefficient in the aging index weight package is called, and the corrected non-uniform coefficient is recorded in the file along with the physical description of the key component. Additionally, field monitoring data reference pointers and historical fault case library reference pointers are added, so that the record of the key component in the file carries the perspectives of operation monitoring, fault experience, and physical field modeling.

[0205] Based on this, this step archives the overall machine level label, which is then written into the overall machine level entry in the aging assessment result file, forming an overall machine level entry. A two-way reference relationship is established between the overall machine level entry and the part level label, allowing for direct retrieval of related parts in subsequent maintenance management systems or condition assessment systems.

[0206] The output of the aforementioned equipment-level archiving mapping forms the aging assessment result archive. This archive is an intermediate archiving product of this step, recording components such as part-level labels, overall machine-level labels, corrected non-uniformity coefficients, descriptions of the physical installation locations of key components, field monitoring data reference pointers, and historical fault case library reference pointers. Structurally, the aging assessment result archive is rooted at the converter, with entries for key components linked downwards. It also retains index pointers to the overall lifespan draft structure, allowing for backtracking to the lifespan derivation process of the previous main step when needed.

[0207] S4-3 After generating the aging assessment result file, update the parameter dictionary to solidify the part-level status data in the aging assessment result file into searchable parameter entries.

[0208] Parameter dictionary update refers to writing the descriptive fields related to key parts in the aging assessment result file into the long-term parameter dictionary; the parameter dictionary is a centralized descriptive set used by this invention to record the operating status characteristics, component relationship characteristics, boundary condition characteristics and change records of converter transformers.

[0209] During the parameter dictionary update process, a parameter entry is created for each key component. This entry includes a key component label, the corresponding corrected non-uniformity coefficient, the corresponding component level label, the overall machine level label to which the key component belongs, a pointer to the relevant field monitoring data, a pointer to the relevant historical fault case library, and an archiving time stamp. Each parameter entry is assigned a current version number and a source identifier when written to the parameter dictionary, distinguishing whether it originates from the current aging assessment result archive or from existing data. In this way, the parameter dictionary update operation solidifies the component-level status data in the aging assessment result archive into searchable parameter entries.

[0210] S4-4, Register the version of the non-uniform coefficients based on the completed parameter dictionary update.

[0211] Non-uniformity coefficient version registration refers to the process of forming a continuous version sequence of corrected non-uniformity coefficients over time, used to record the evolution of the non-uniformity coefficient of a key component across multiple assessments. Specifically, for each key component label, the corrected non-uniformity coefficient recorded in the aging assessment result file is read, and the non-uniformity coefficient of the same key component label in the previous version is also read from the parameter dictionary. If a difference is found between the corrected non-uniformity coefficient and the previous version record, a new version entry is generated in the non-uniformity coefficient version registration, and the key component label, the corrected non-uniformity coefficient, the reference pointer of the previous version record, the explanation of the source of change, the corresponding component level label, and the corresponding whole machine level label are registered as the same entry.

[0212] Understandably, non-uniform coefficient version registration is not simply an overwrite, but rather an accumulation of time-series version links based on the parameter dictionary, allowing the values ​​of non-uniform coefficients for the same critical component to be traced back to historical stages. This version link maintains a pointer-level association with the aging assessment result archive within this step, enabling subsequent calls to locate the monitoring and fault context from the version record.

[0213] S4-5, Perform the process of supplementing the historical fault case database.

[0214] The historical fault case database replenishment process involves adding new information generated during the current assessment process back to the historical fault case database. In this invention, the historical fault case database is defined as a centralized knowledge resource that records the failure location labels, maintenance records, decommissioning descriptions, and defect-causing environments of converter transformers or converter transformers of the same series.

[0215] During the supplementary data entry process, for newly emerging critical component labels or critical component labels identified as high-concern components in this assessment, the entry content of the critical component label in the aging assessment result file is first read. Then, it is determined whether there is already a failure component label record for the critical component in the historical failure case database. If not, the critical component label, the corresponding component level label, the corresponding whole machine level label, the current corrected non-uniformity coefficient, the field monitoring data reference pointer, and the entry pointer of the critical component in the aging assessment result file are written as new case entries into the historical failure case database. If it has already appeared, the case entry corresponding to the critical component label is appended with the time stamp of this assessment, the component level label, and the corrected non-uniformity coefficient. The original record is not overwritten, but rather a multi-stage record is added.

[0216] Through the above supplementary recording process, the historical failure case library obtained a state expression method consistent with this assessment, so that the part location mapping and failure label alignment in the subsequent main steps can directly reference these updated entries without having to re-establish the mapping rules.

[0217] After completing equipment-level archive mapping, parameter dictionary updates, non-uniform coefficient version registration, and historical fault case library supplementation, the updated non-uniform parameter baseline structure is output. The updated non-uniform parameter baseline structure is the final output field name of this step. It is defined as a set of baseline parameters that can be directly loaded into subsequent modeling processes, obtained by fusing the corrected non-uniform coefficients, part-level labels, whole-machine-level labels, parameter dictionary update results, non-uniform coefficient version registration results, and the latest supplemented entries in the historical fault case library generated in this assessment. The updated non-uniform parameter baseline structure not only describes the values ​​of the corrected non-uniform coefficients of each key part at the current assessment time, but also carries the corresponding time stamp and version information, and can be traced back to the archived entries in the aging assessment result file through internal pointers.

[0218] In steps S4-5, the updated non-uniform parameter baseline structure is passed to step S1 to obtain converter transformer structure data, operating condition data, and the non-uniform parameter baseline structure. This process involves electromagnetic-thermal-mechanical multiphysics coupling modeling, spatial mesh unification and coordinate alignment, boundary consistency verification, outlier labeling, and numerical normalization to generate a multiphysics distribution standard package. This process is the multiphysics modeling entry point passed to step S1. This transfer forms an input link flowing back from step S4 to step S1, enabling the next round of modeling to directly load the updated non-uniform parameter baseline structure as the latest version while reading converter transformer structure data and operating condition data. In subsequent loops, it continues to participate in non-uniform aging rate modeling, local lifetime integration and cumulative damage statistics, regional aggregation, key component marking, and global summary processing, ultimately re-entering the level determination and equipment-level archiving mapping.

[0219] The technical effects of this step can be summarized as follows: by generating aging status level records through level determination and completing equipment-level archiving mapping, an aging assessment result archive is obtained; through parameter dictionary updates, non-uniform coefficient version registration, and historical fault case library supplementation, an updated non-uniform parameter baseline structure is output; the updated non-uniform parameter baseline structure is directly sent back to the multiphysics modeling entry point of step S1, so that the converter transformer aging assessment method forms a closed loop that can be updated cyclically on the same engineering object.

[0220] In conjunction with the above embodiments, the present invention:

[0221] (1) Based on the acquired converter structure data, operating condition data and non-uniform parameter baseline structure, electromagnetic-thermal-mechanical multi-physics coupling modeling, spatial grid unification and coordinate alignment, boundary consistency verification, outlier labeling and numerical normalization are performed to generate a multi-physics distribution standard package. The multi-physics distribution standard package simultaneously expresses electric field distribution, temperature distribution and mechanical stress distribution under a unified grid coordinate system, and maintains a corresponding relationship with the non-uniform parameter baseline structure. It is used to provide directly referenced basic data for subsequent non-uniform aging rate modeling and local life integration and cumulative damage statistics.

[0222] (2) Perform location mapping and failure label alignment, non-uniform coefficient reverse correction, and subjective and objective weight fusion and impact level quantification to generate aging index weight package; Location mapping and failure label alignment establish a location-level correspondence between the key location labels in the overall life draft structure and the failure location labels in the historical failure case library; Non-uniform coefficient reverse correction compares the local field distribution parameters of the multi-physics distribution standard package with the field monitoring data and generates a non-uniform correction parameter set; Subjective and objective weight fusion and impact level quantification unifies the non-uniform correction parameter set with the subjective weight of operation and maintenance experts and the objective weight of entropy weight method into the aging index weight package to achieve impact level quantification for key locations.

[0223] (3) Perform level determination, equipment-level archiving mapping, parameter dictionary update, non-uniform coefficient version registration, and historical fault case library supplementation processing to generate the updated non-uniform parameter baseline structure; after completing the aging state level record and aging assessment result archive, the updated non-uniform parameter baseline structure is written back to the non-uniform parameter baseline structure and enters the parameter dictionary update and non-uniform coefficient version registration, while triggering the historical fault case library supplementation processing, so that the updated non-uniform parameter baseline structure becomes the input for electromagnetic-thermal-mechanical multi-physics coupling modeling again, thereby forming a closed-loop link between acquiring converter transformer structure data, operating condition data, and non-uniform parameter baseline structure to generating the updated non-uniform parameter baseline structure.

Claims

1. A method for evaluating the aging of commutator transformers based on non-uniform physical fields and fault cases, characterized in that, The method includes the following steps: Acquire converter transformer structure data, operating condition data and non-uniform parameter baseline structure, extract winding interlayer arrangement, core column arrangement and boundary correction factor fields, perform electromagnetic-thermal-mechanical multiphysics coupling modeling, spatial grid unification and coordinate alignment, boundary consistency verification, outlier annotation and numerical normalization, and generate multiphysics distribution standard package. S2, read the electric field distribution, temperature distribution and mechanical stress distribution in the multiphysics distribution standard package according to the grid cell, perform non-uniform aging rate modeling, local lifetime integration, and combine historical load stages to perform cumulative damage statistics, regional aggregation, key part marking and global summary processing to generate an overall lifetime draft structure containing part aggregation entries. S3, extract the key part labels in the overall life draft structure, perform part location mapping and failure label alignment, combine historical failure case library and field monitoring data to perform non-uniform coefficient reverse correction, and use the correction parameter set to drive subjective and objective weight fusion and influence level quantification to generate aging index weight package carrying level label. S4, perform level determination, equipment-level archiving mapping, parameter dictionary update, non-uniform coefficient version registration and historical fault case library supplementation on the aging index weight package to generate the updated non-uniform parameter baseline structure. The updated non-uniform parameter baseline structure is passed to the multiphysics modeling entry point in step S1, forming an input link that flows back from step S4 to step S1.

2. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 1, characterized in that, The specific process of step S1 includes: S1-1: Obtain converter transformer structure data, operating condition data, and non-uniform parameter baseline structure. By loading the converter transformer structure data, operating condition data, and non-uniform parameter baseline structure into the same modeling entry point, the input source for this step is formed. S1-2, the initial multiphysics model is obtained by coupled solution through electromagnetic-thermal-mechanical multiphysics coupling modeling; S1-3: Extract electric field distribution, temperature distribution and mechanical stress distribution from the initial multiphysics model, perform spatial grid unification and coordinate alignment processing, and form a multiphysics distribution data package through spatial grid unification and coordinate alignment processing; S1-4 performs boundary consistency verification, outlier labeling, and numerical normalization to output stable and usable data assets.

3. The converter transformer aging assessment method based on non-uniform physical field and fault case according to claim 2, characterized in that, The converter transformer structure data mentioned in step S1-1 is a structured description of the target converter transformer body in terms of geometry, assembly and insulation configuration, reflecting the relative spatial relationship between the insulation system and load-bearing components, and recording the contact surfaces, fixing points and constraint areas between components; This includes the interlayer arrangement and lead routing of the windings, the geometric arrangement of the core columns and yokes, the positioning relationship of clamps, pressure plates, tensioners and supports, and the thickness distribution and stacking sequence information of the main insulation and local insulation. The operating condition data is loading information related to the working status of the target converter transformer in the current or typical operating scenario, including load level, long-term average load fluctuation range, heat dissipation and cooling method, cooling flow direction of oil circuit or air circuit, ambient temperature, external mechanical fixing constraints, and historical stress traces caused by transportation, hoisting, and short circuit impact. The non-uniform parameter baseline structure is a centralized representation of the initial cognitive results of the existing state of the converter transformer. The non-uniform parameter baseline structure includes a non-uniform coefficient to describe the degree of non-uniformity of the internal electric field distribution, a boundary condition correction factor to describe the location of the winding end, lead outlet, and pressure plate contact area, and a material aging reference curve version number to describe the trend of material degradation sensitivity.

4. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 1, characterized in that, The electromagnetic field model described in step S1-2 is based on the winding position, lead direction and insulation layer thickness information in the converter transformer structure data to arrange the applied voltage excitation boundary and dielectric distribution, forming a spatial distribution description of the internal electric field intensity; The thermal flow field model is based on the load level, cooling method and ambient temperature information in the operating condition data to model the heat conduction, heat convection and heat exchange of the heat-generating area and heat dissipation path, forming a spatial distribution description of the internal temperature field; the mechanical stress model is based on the constraint relationship of clamps, pressure plates, supports and fasteners in the converter variable structure data, and combined with the external constraint conditions and historical impact event descriptions recorded in the operating condition data to establish a distribution description of the internal mechanical stress. The initial multiphysics model is a coupled structure that includes three spatial components: electric field distribution, temperature distribution, and mechanical stress distribution. It can maintain the correspondence between each component and the actual component coordinates.

5. The aging assessment method for converter transformers based on non-uniform physical fields and fault cases according to claim 1, characterized in that, In steps S1-3, the electric field distribution is analyzed as the electric field intensity record at each spatial location of the insulation system during the processing; the temperature distribution is analyzed as the temperature record of the winding, oil passage, adjacent area of ​​clamping parts and insulation interlayer under steady state or specific working conditions; the mechanical stress distribution is analyzed as the internal force record at locations such as pressure plate, tensioning parts, clamping parts, and winding end support area.

6. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 2, characterized in that, The boundary consistency check described in steps S1-4 is an integrity verification operation performed on the edge region, transition region, and contact interface region of the multi-physics distribution data packet. The boundary consistency check process is as follows: check whether the electric field record at the transition point from the winding end to the lead outlet is continuous; check whether there are any breaks or abrupt changes in the mechanical stress record at the contact area between the pressure plate and the winding support; and check whether there are any isolated high-value points in the temperature record at the transition point from the high-temperature zone to the cooling channel. If any abnormal data points are found in the records that have obvious jumps, missing measurements, or deviate from the nominal range of the non-uniform parameter baseline structure, then the data points are marked as abnormal points. Anomaly labeling involves adding anomaly tags and anomaly type descriptions to abnormal mesh locations and writing them as supplementary records within the multiphysics distribution data package. Anomaly tags are used for weight adjustment or to remove interference from specific points during the subsequent aging rate modeling stage. Numerical normalization is performed after boundary consistency verification and anomaly labeling. It involves dimensionless mapping of various numerical records in the multiphysics distribution data package, so that electric field distribution, temperature distribution, and mechanical stress distribution are expressed on a unified scale, facilitating direct comparison across physical fields.

7. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 2, characterized in that, The specific process of step S2 includes: S2-1, Import the multiphysics distribution standard package into the non-uniform aging rate modeling process to generate location-level aging rate records and form a location-level aging rate sequence. S2-2, For the location-level aging rate sequence formed in step S2-1, perform local lifetime integration statistics to form a region-level lifetime accumulation entry; then perform cumulative damage statistics to form a local lifetime distribution description; the local lifetime distribution description is registered as a local lifetime distribution map, and the generation process of the local lifetime distribution map retains the cumulative damage statistics details for each region of interest; the local lifetime distribution map is written into the internal data buffer of the step and is ready to enter step S2-3. S2-3, based on the local lifetime distribution map, perform regional aggregation, key component marking and global summary processing to finally form an overall lifetime draft structure. The overall lifetime draft structure is a whole-machine-level lifetime status draft formed for the target converter transformer, which uniformly records the aggregated lifetime entries, key component labels, regional aggregation relationships, historical operation segment identifiers and grid-level source references for each key component.

8. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 7, characterized in that, The non-uniform aging rate modeling in step S2-1 is a process of establishing an aging rate description for each spatial grid position inside the converter transformer. By reading the electric field distribution record, temperature distribution record and mechanical stress distribution record corresponding to each grid cell, and combining the boundary condition correction factor reference label, the stress concentration, heat accumulation and electric field distortion of parts such as winding ends, lead outlets and clamp pressing areas are weighted and processed. The grid cells marked as abnormal points in the previous step are reduced or removed according to the abnormal label to generate a position-level aging rate record. The local life integral statistics are to aggregate the location-level aging rate records of multiple grid units belonging to the same insulation region according to their spatial proximity, and then combine them with the historical operating load stage, cooling condition stage and known maintenance intervention stage of the insulation region to form a region-level life accumulation entry. Cumulative damage statistics are based on the above-mentioned regional-level cumulative lifespan entries. They are performed in segments for operation phases with long time spans or significant load fluctuations, recording the contribution percentage of each segment to the lifespan consumption of the same region, thus providing a description of the local lifespan distribution.

9. The aging assessment method for converter transformers based on non-uniform physical fields and fault cases according to claim 7, characterized in that, The specific process of step S4 includes: S4-1, Perform a level determination on the aging index weight package to obtain the aging status level record corresponding to the overall machine status; S4-2 After obtaining the aging status level record, perform equipment-level archiving mapping processing to form an aging assessment result archive; S4-3 After generating the aging assessment result file, update the parameter dictionary and solidify the part-level status data in the aging assessment result file into searchable parameter entries. S4-4, Register the version of the non-uniform coefficients based on the completed parameter dictionary update; S4-5, perform historical fault case database supplementation processing, and output the updated non-uniform parameter baseline structure; pass the updated non-uniform parameter baseline structure to the multiphysics modeling entry point of step S1, forming an input link from step S4 back to step S1.

10. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 9, characterized in that, The specific process for determining the grade of the aging indicator weight package includes: Each key component label and its corresponding level label in the aging index weight package will be analyzed item by item. After analyzing each item, all key part labels are grouped to form a set of part-level labels; Based on the set of part-level labels, and according to the distribution range and concentration of each label throughout the machine, as well as whether it involves windings, lead wire outlets, pressure plate contact areas, and clamp support areas, a whole-machine level label is formed.

11. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 9, characterized in that, The device-level archiving mapping described in step S4-2 is used to directly bind the aging status level record with the engineering management object of the converter transformer, and form an aging assessment result archive that can be archived for a long time. Specifically, the set of part level labels in the aging status level record is read, and each part level label is bound one-to-one with the aforementioned key part label. The physical installation location of the key part in the converter transformer structure, the corresponding insulation layer type, the description of the load-bearing or support component, the description of the assembly transition surface, the description of the thermal channel, and the description of the stress concentration contact surface are recorded, so that the abstract level label is applied to the specific component unit that can be identified in the on-site maintenance process. For the same key component label, the corrected non-uniform coefficient in the aging index weight package is called, and the corrected non-uniform coefficient is recorded in the file along with the physical description of the key component. The field monitoring data reference pointer and the historical fault case library reference pointer are added, so that the record of the key component in the file carries the operation monitoring perspective, the fault experience perspective and the physical field modeling perspective. Based on this, this step archives the whole machine grade label, which is then written into the whole machine level entry in the aging assessment result file to form the whole machine grade entry.

12. The converter transformer aging assessment method based on non-uniform physical field and failure case according to claim 9, characterized in that, The non-uniformity coefficient version registration mentioned in step S4-4 refers to the process of forming a continuous version sequence of the corrected non-uniformity coefficient in the time dimension, which is used to record the evolution of the non-uniformity coefficient of the key part between multiple assessments; specifically, for each key part label, the corrected non-uniformity coefficient recorded in the aging assessment result file is read, and the non-uniformity coefficient of the same key part label recorded in the previous version is read in the parameter dictionary. If a difference is found between the corrected non-uniform coefficient and the previous version record, a new version entry will be generated in the non-uniform coefficient version registration. The key part label, the corrected non-uniform coefficient, the reference pointer of the previous version record, the change source description, the corresponding part level label, and the corresponding whole machine level label will be registered as the same entry.

Images