Multi-source data correction method and related equipment based on parameter constraint maps
By constructing a parameter constraint map, the accuracy and efficiency issues of multi-source data correction in power engineering were solved, realizing unified structured processing and automated correction of multi-source heterogeneous parameters, and improving the accuracy and efficiency of data verification.
Patent Information
- Application Number
- CN202611188844.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-06
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for multi-source data correction in power engineering suffer from low accuracy, serious omissions, inability to uniformly analyze and process heterogeneous multi-source data, and lack of automated correction mechanisms, resulting in parameter verification efficiency and accuracy failing to meet standardization requirements.
A multi-source data correction method based on parameter constraint graphs is adopted. By constructing parameter constraint graphs, multi-source data is acquired and preprocessed, nodes and constraint edges are constructed, deviation calculation and subgraph extraction are performed, and correction items are automatically filtered in combination with business rules to achieve local updates of the data set.
It achieves unified structured processing of multi-source heterogeneous parameters, quickly locates and quantifies parameter conflicts, automatically selects reasonable correction schemes, reduces manual workload, and improves the accuracy and efficiency of data verification.
Smart Images

Figure CN123115607A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a method and related equipment for correcting the bias of multi-source data based on parameter constraint maps. Background Technology
[0002] In practical applications of power engineering, a large amount of technical data is generated. This data comes from different sources, different units, systems, or personnel, and is produced at different stages. Inconsistencies can easily arise regarding the same or related parameters of the same equipment. For example, the rated capacity of the main transformer in the equipment technical agreement may have changed to a new value, but the design specification may retain the old value; a certain outgoing line bay in the primary system diagram may be connected to a 10kV bus, but the voltage level of that bay may be recorded as 6kV in the equipment parameter table; the current transformer ratio in the protection setting sheet may differ from the ratio in the equipment ledger; the rated current calculated based on capacity and voltage in the calculation book may differ from the rated current in the parameter table; one document may use kV as the unit, while another uses V, and discrepancies may still exist after conversion.
[0003] Existing technologies typically employ manual verification, keyword retrieval, text comparison, table field comparison, or single-rule validation to identify parameter issues in data and correct data biases. These methods have the following shortcomings:
[0004] First, it only supports comparison of a single type of data and cannot uniformly parse and process multi-source heterogeneous data such as text, tables, formulas, attached figures, engineering drawings, and structured ledgers. Multi-source parameters cannot be linked for verification, which makes it very easy to miss detections. Second, it can only complete text and literal field matching, and cannot depict complex constraint logic such as unit conversion between power equipment parameters, electrical formula coupling, equipment topology subordination, industry standard thresholds, and data version inheritance. A large number of hidden parameter conflicts cannot be identified, which greatly reduces the accuracy of verification and identification.
[0005] The aforementioned shortcomings result in low accuracy in correcting biases in multi-source data. Summary of the Invention
[0006] In view of this, the purpose of this application is to propose a multi-source data correction method and related equipment based on parameter-constrained maps to overcome all or part of the shortcomings of the prior art.
[0007] To achieve the above objectives, this application provides a multi-source data bias correction method based on a parameter constraint graph, comprising: acquiring a data set composed of data from multiple sources, and preprocessing the data set to obtain multiple sets of normalized data; constructing a parameter constraint graph based on the multiple sets of normalized data, wherein the parameter constraint graph includes multiple nodes mapped from each set of normalized data, and at least one constraint edge exists between any two nodes that are related; and calculating the bias for each set of normalized data based on the parameter constraint graph to obtain constraint data corresponding to at least one constraint edge for each set of normalized data, wherein each set of normalized data has a bias of approximately 1 / 2 ≤ ... The data set includes constraint edge deviation values and constraint satisfaction. For each constraint edge, in response to determining that the constraint edge deviation value corresponding to the constraint edge is greater than a tolerance threshold, a subgraph is extracted from the parameter constraint graph based on the constraint edge to obtain a deviation subgraph corresponding to the constraint edge. Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, at least one target correction item is determined for the deviation subgraph. For each target correction item, in response to determining that the target correction item satisfies the automatic correction condition, the data set is locally updated based on the target correction item to obtain an updated data set.
[0008] Based on the same inventive concept, this application also provides a multi-source data correction device based on a parameter constraint graph, comprising: a preprocessing module configured to acquire a data set composed of data from multiple sources and preprocess the data set to obtain multiple sets of normalized data; a construction module configured to construct a parameter constraint graph based on the multiple sets of normalized data, wherein the parameter constraint graph includes multiple nodes mapped from each set of normalized data, and at least one constraint edge exists between any two nodes that are related; and a calculation module configured to perform deviation calculation on each set of normalized data based on the parameter constraint graph to obtain constraint data corresponding to at least one constraint edge for each set of normalized data, wherein each The constraint data includes constraint edge deviation values and constraint satisfaction. An extraction module is configured to, for each constraint edge, extract a subgraph from the parameter constraint graph based on the constraint edge when the constraint edge deviation value is greater than a tolerance threshold, to obtain a deviation subgraph corresponding to the constraint edge. A determination module is configured to determine at least one target correction item corresponding to the deviation subgraph based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph. An update module is configured to, for each target correction item, locally update the data set based on the target correction item when the target correction item satisfies the automatic correction condition, to obtain an updated data set.
[0009] Based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described above when executing the computer program.
[0010] Based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to perform the method described above.
[0011] As can be seen from the above, the multi-source data correction method and related equipment based on parameter constraint graphs provided in this application include acquiring a dataset composed of data from multiple sources and preprocessing the dataset to obtain multiple sets of normalized data. This not only improves data quality but also completes the fusion and consolidation of heterogeneous parameters from multiple sources, allowing the associated parameters of the same device scattered across different data carriers to form unified structured data. Based on the multiple sets of normalized data, a parameter constraint graph is constructed, wherein the parameter constraint graph contains multiple nodes mapped from each set of normalized data. Any two nodes with an association relationship include at least one constraint edge, breaking the limitations of the original heterogeneous data being fragmented and revealing implicit parameter conflicts that are difficult to identify through text comparison and field matching. Based on the parameter constraint graph, deviation calculation is performed on each set of normalized data to obtain constraint data corresponding to at least one constraint edge for each set of normalized data. Each constraint data includes a constraint edge deviation value and a constraint satisfaction degree. Deviation calculation is performed based on a parameter constraint graph, which comprehensively records all the associated constraint relationships between devices, parameters, and data nodes. This allows for rapid location of all constraint rules matching a single set of normalized data, enabling quantitative judgment of parameter conflicts and effectively avoiding the missed detections and misjudgments caused by traditional text comparison. For each constraint edge, in response to determining that the deviation value of the constraint edge is greater than the tolerance threshold, a subgraph is extracted from the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge. This subgraph visually displays the devices, parameters, data carriers, and constraint logic involved in the conflict, clearly locating the source of the conflict and related parameters. Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, at least one target correction item is determined for the deviation subgraph. The optimal correction scheme that is compliant and reasonable is automatically selected, eliminating the need for manual comparison of multiple sets of data to determine the modification direction and reducing the workload of manual analysis. For each target correction item, in response to determining that the target correction item meets the automatic correction conditions, the data set is partially updated based on the target correction item to obtain the updated data set. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a flowchart illustrating the multi-source data correction method based on parameter constraint maps according to an embodiment of this application. Figure 2 This is a schematic diagram of the parameter constraint map structure in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a multi-source data correction device based on parameter constraint maps according to an embodiment of this application; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation
[0014] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.
[0015] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0016] As described in the background section, power engineering projects generate a large amount of technical data during practical applications such as planning and design, construction and installation, commissioning and acceptance, operation and maintenance, overhaul and renovation, and data archiving. This type of technical data typically includes design specifications, equipment technical agreements, equipment parameter tables, calculation sheets, protection setting sheets, factory test reports, handover test reports, primary system diagrams, secondary schematic diagrams, terminal block diagrams, cabinet layout diagrams, cable lists, equipment nameplate images and scanned copies, equipment ledgers, database tables, and interface data, etc.
[0017] Because different data sources are generated by different units, systems, or personnel at different stages, inconsistencies can easily arise in the same or related parameters of the same equipment. For example, the rated capacity of the main transformer in the equipment technical agreement may have been changed to a new value, but the design specification may still retain the old value; a certain outgoing line bay in the primary system diagram may be connected to a 10kV bus, but the voltage level of that bay may be recorded as 6kV in the equipment parameter table; the current transformer ratio in the protection setting sheet may be inconsistent with the ratio in the equipment ledger; the rated current calculated based on capacity and voltage in the calculation book may be inconsistent with the rated current in the parameter table; one document may use kV as the unit, while another document may use V as the unit, and discrepancies may still exist after conversion.
[0018] Existing technologies typically employ manual verification, keyword retrieval, text comparison, table field comparison, or single-rule validation to identify parameter issues in data and correct data biases. These methods have the following shortcomings: First, it only supports comparison of a single type of data and cannot uniformly parse and process multi-source heterogeneous data such as text, tables, formulas, attached figures, engineering drawings, and structured ledgers. Multi-source parameters cannot be linked for verification, which makes it very easy to miss detections. Second, it can only complete text and literal field matching, and cannot depict complex constraint logic such as unit conversion between power equipment parameters, electrical formula coupling, equipment topology subordination, industry standard thresholds, and data version inheritance. A large number of hidden parameter conflicts cannot be identified, which greatly reduces the accuracy of verification and identification.
[0019] In addition, existing technologies still have several missing mechanisms, such as: First, it is difficult to process heterogeneous data from multiple sources, such as text, tables, formulas, figures, engineering drawings, and structured data, simultaneously. Second, the lack of a parameter write-back mechanism for different data objects makes it difficult to perform unified closed-loop correction for text paragraphs, table cells, formula results, image annotations, drawing elements, and structured fields. Third, the lack of local regression verification and automatic rollback mechanism after write-back may introduce new parameter conflicts after correction. Fourth, the lack of a complete chain of evidence, from conflict detection and corrective basis to write-back location and closed-loop verification results, makes the data correction process difficult to audit and trace.
[0020] In summary, existing technical parameter verification suffers from prominent issues of missed detections and misjudgments. Furthermore, it lacks an integrated closed-loop correction solution that integrates anomaly identification, automatic correction, secondary verification, and modification tracking. It is highly dependent on manual intervention, and the efficiency of parameter correction and the accuracy of data are both insufficient to meet the standardized management and control requirements of power engineering.
[0021] In view of this, embodiments of this application propose a multi-source data correction method based on parameter constraint maps, referring to... Figure 1 This includes the following steps: Step 101: Obtain a data set consisting of data from multiple sources, and preprocess the data set to obtain multiple sets of normalized data.
[0022] In this step, data from multiple sources is acquired, such as data from different application systems. The data from multiple sources constitutes a data set, which can be a power technology data set. The power technology data set includes at least two types of data sources, such as text data, tabular data, formula data, attached figures, engineering drawings, and structured data. The following are examples of various data sources included in the power technology data collection: (1) Textual data, such as design specifications, technical specifications, equipment manuals, test reports, and acceptance reports; (2) Tabular data, such as equipment parameter tables, protection setting sheets, material lists, test data tables, and equipment lists; (3) Formula data, such as load calculation sheets, short-circuit current calculation sheets, rated current calculation sheets, and relay protection calculation sheets; (4) Attached drawings, such as equipment nameplate images, scanned images, screenshot parameter tables, and on-site photos; (5) Engineering drawing data, such as primary system diagrams, secondary schematic diagrams, terminal block diagrams, cable route diagrams, and cabinet layout diagrams; (6) Structured data, such as database tables, JSON files (JavaScript Object Notation), XML files (Extensible Markup Language), CSV files (Comma-Separated Values), configuration files, and interface data.
[0023] It should be noted that, in order to improve the efficiency of preprocessing the dataset, source metadata is added to each piece of data in the dataset before preprocessing. The source metadata corresponding to the data can be represented as follows: (1) in, Indicates the source identifier. Indicates the data source type. Indicates the name of the data. Indicates the version of the document. Indicates the data status. Indicates the generation time or effective time. Indicate the source of the data. Indicates whether the data is writable. This indicates the basic credibility of the data source. The data status can include draft, submitted for review, approved, archived, historical version, obsolete version, etc. Generally, approved and officially archived data have a higher basic credibility than draft data, temporarily exported data, or OCR (Optical Character Recognition) recognized data.
[0024] The dataset contains data from multiple units, systems, and / or stages, which may contain redundancy, incomplete fields, and inconsistent formats. Preprocessing generates multiple sets of normalized data, improving data quality and merging heterogeneous parameters from multiple sources. This allows parameters associated with the same device, scattered across different data carriers, to form unified structured data. This application can parse different data sources such as text, tables, formulas, figures, engineering drawings, and structured data, and convert the extracted parameters into unified normalized parameter records. This enables parameters of different formats, units, versions, and sources to be calculated and compared under a unified data structure.
[0025] Step 102: Based on the multiple sets of normalized data, construct a parameter constraint graph, wherein the parameter constraint graph contains multiple nodes generated by mapping each set of normalized data, and there is at least one constraint edge between any two nodes that are related.
[0026] In this step, the parameter constraint graph is a structured association graph built based on multiple sets of normalized data. It contains multiple nodes and multiple constraint edges representing verification rules. Any two nodes with an association relationship have at least one constraint edge between them. The multiple nodes include equipment object nodes, parameter entity nodes, and data evidence nodes. The multiple constraint edges include synonymous equivalence constraint edges, unit conversion constraint edges, formula calculation constraint edges, topological dependency constraint edges, standard value constraint edges, and version inheritance constraint edges. By mapping multiple sets of normalized data to standardized nodes and building constraint edges based on power business associations, it can fully support complex constraint relationships such as unit conversions, formula coupling, and topological dependency between equipment, parameters, and data. This breaks the limitations of the original heterogeneous data being fragmented, and achieves visualized storage and rapid retrieval of all parameter association relationships.
[0027] Constructing the parameter constraint graph includes: Establish equipment object nodes, which include equipment standard identifier, equipment type, equipment alias, project, bay, connection object, and topology level; Establish parameter entity nodes, which include parameter attribute standard identifier, parameter name, unit dimension, value type, allowable range, parameter risk level, and applicable equipment type; Establish data evidence nodes, which include data source identifier, data version, data status, data location anchor, resolution confidence, source credibility, and parameter anchor fingerprint; Establish constraint edges based on the relationships between device object nodes, parameter entity nodes, and data evidence nodes, and configure constraint types, constraint expressions, tolerance thresholds, edge weights, risk identifiers, effective version ranges, applicable device types, and write-back strategies for each constraint edge.
[0028] The parameter constraint map can be represented as: (2), in, A collection of device object nodes. For the set of parameter entity nodes, For data evidence nodes, For the set of constraint edges, This is a collection of metadata for the map.
[0029] Each constraint edge can be represented as: (3) in, For constraint type, For the set of parameter records associated with the constraint edges, For constraint expressions, This is the tolerance threshold. For edge weights, As a risk indicator, For the effective version range, For write-back strategy or verification strategy.
[0030] The parameter constraint graph constructed in this application not only records equipment objects and parameter values, but also expresses the constraint relationships between power parameters through various types of constraint edges. Relying on the parameter constraint graph, it fully carries the complex constraint relationships between equipment, parameters, and data, and can present implicit parameter conflicts that are difficult to identify through text comparison and field matching.
[0031] Step 103: Based on the parameter constraint map, perform deviation calculation on each set of normalized data to obtain constraint data for at least one constraint edge corresponding to each set of normalized data, wherein each constraint data includes constraint edge deviation value and constraint satisfaction degree.
[0032] In this step, due to the complex constraints among multiple data sources, text comparison alone cannot accurately quantify and determine data compliance. Therefore, deviation calculation is necessary to visually reflect the differences between parameters and constraint standards through quantitative values, objectively determining whether parameter conflicts exist. Based on the parameter constraint graph, the constraint edges associated with each group of normalized data are located and deviation calculations are performed. Quantification can be completed using various power constraint expressions bound to the parameter constraint graph, simultaneously outputting the constraint edge deviation value and constraint satisfaction degree of at least one constraint edge corresponding to each group of normalized data. The constraint edge deviation value quantifies the magnitude of parameter deviation from standard constraints, while the constraint satisfaction degree measures the degree to which the current parameter matches the corresponding constraint rule. The value directly reflects whether the parameter meets power business constraint requirements and can distinguish the severity of conflicts. Deviation calculation is performed based on the parameter constraint graph, which completely records all associated constraint relationships between equipment, parameters, and data nodes. This allows for rapid location of all constraint rules matching a single group of normalized data, achieving quantitative determination of parameter conflicts and effectively avoiding the problems of missed detections and misjudgments caused by traditional text comparison.
[0033] Step 104: For each constraint edge, in response to determining that the constraint edge deviation value corresponding to the constraint edge is greater than the tolerance threshold, subgraph extraction is performed on the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge.
[0034] In this step, if the deviation value of the constraint edge corresponding to the constraint edge is greater than the tolerance threshold, it indicates a parameter conflict. The tolerance threshold is determined based on historical experience. Constraint edges with deviation values exceeding the corresponding tolerance threshold are identified as abnormal constraint edges. Using these abnormal constraint edges as entry points, a conflict impact subgraph is extracted from the parameter constraint graph, containing associated device object nodes, associated parameter entity nodes, associated data evidence nodes, and associated constraint edges. This subgraph centrally collects the device nodes, parameter nodes, data source nodes, and all related constraint relationships associated with the conflict, discarding irrelevant graph information. It not only intuitively displays the devices, parameters, data carriers, and constraint logic involved in the conflict, clearly locating the source of the conflict and associated parameters, but also provides a complete and concise associated data carrier for subsequent screening of candidate correction items and tracing the source of problematic data, avoiding information redundancy and inefficient retrieval caused by full graph traversal.
[0035] Step 105: Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, determine at least one target correction item corresponding to the deviation subgraph.
[0036] In this step, the predetermined business constraint rules are the standardized specifications, setpoints, and input criteria uniformly implemented in the design, operation, and acceptance stages of the power industry. Based on the constraint satisfaction of the predetermined business constraint rules and the constraint edges corresponding to the deviation subgraph, at least one target correction item is determined for each deviation subgraph. The target correction item refers to a standardized parameter rectification plan that adapts to the current parameter conflict and has executable modification points and corrected values. By comprehensively selecting target correction items based on the predetermined business constraint rules and constraint satisfaction, both industry standard requirements and actual deviations from parameter conflicts can be considered. This automatically selects the compliant and reasonable optimal correction plan, eliminating the need for manual comparison of multiple documents to determine the modification direction and reducing the workload of manual analysis.
[0037] Step 106: For each target correction item, in response to determining that the target correction item meets the automatic correction condition, the data set is locally updated based on the target correction item to obtain the updated data set.
[0038] In this step, automatic correction conditions are set to differentiate between the complexity of conflicts and the risk of modification, filtering out high-risk and multi-related cross-conflicts to avoid cascading data errors caused by blind automatic modifications. Automatic correction conditions include various types of sub-correction conditions, such as risk level conditions for correction items, reliability conditions for parameter sources, expected compliance improvement conditions, new conflict prediction conditions, data write permission conditions, and parameter point fingerprint verification conditions. Before automatically correcting the parameter constraint diagram, it is determined whether the target correction item meets the automatic correction conditions. Specifically, for each target correction item, it is compared with each type of sub-correction condition to obtain the comparison result corresponding to each type of sub-correction condition. Each sub-correction condition corresponds to... The entry threshold is a pass / fail standard set in the automatic correction conditions. If the comparison results of all sub-correction conditions meet the entry threshold, the target correction item is determined to meet the automatic correction conditions, and subsequent automatic correction operations can be performed. If it is determined that a comparison result does not reach its corresponding entry threshold, the target correction item is determined to not meet the automatic correction conditions. This target correction item is marked as a pending correction item and pushed to the review end. The pending correction item carries the conflict reason, candidate parameter values, related evidence set, hit business constraint rules, and corresponding data modification location information. Local graph updates are only performed when the target correction item meets the automatic correction conditions, which can intercept high-risk, logically error-prone correction operations in advance, preventing unreasonable modifications from damaging parameter relationships.
[0039] For example, the automatic correction conditions include: the risk level of the target correction item is lower than the preset automatic write-back risk threshold; the credibility of the candidate value source of the target correction item is higher than the preset automatic write-back credibility threshold; the expected increment of constraint satisfaction corresponding to the target correction item is higher than the preset automatic write-back increment threshold; the expected number of new anomalies after the target correction item is executed is less than or equal to the preset new anomaly threshold; and the data object corresponding to the target correction item is a writable data object.
[0040] Only when the target correction item meets the automatic correction conditions is the dataset partially updated based on that item, resulting in the updated dataset. On one hand, by only updating the dataset based on the target correction item when it meets the automatic correction conditions, unreliable or high-risk correction schemes can be pre-filtered, unreasonable automated modifications avoided, unnecessary data changes reduced, and the risk of data anomalies caused by erroneous parameter adjustments lowered. On the other hand, targeted partial updates based on compliant target correction items can accurately match conflicting parameter points within multi-source data, enabling targeted parameter corrections and avoiding large-scale changes that interfere with normal data, effectively improving the accuracy of automatic correction for multi-source data.
[0041] Existing technologies can only perform literal comparison of single data points, failing to integrate multi-source heterogeneous parameters, and cannot handle complex constraint logic, resulting in severe missed conflict detection and low verification accuracy. This application unifies and integrates multi-source normalized data through parameter constraint maps, quantifies constraint deviations, and combines business rules and constraint satisfaction to select target correction items. It comprehensively considers industry standards and the degree of conflict quantification to determine the optimal correction scheme, overcoming the limitations of simple literal matching. It fully covers various implicit parameter conflicts, relying entirely on quantitative indicators and standardized business rules from conflict identification to correction scheme determination, avoiding biases caused by subjective human judgment, and fundamentally improving the accuracy of parameter conflict identification and correction determination.
[0042] It should be noted that before locally updating the parameter constraint graph based on the target correction item and the dataset to obtain the updated parameter constraint graph, the write-back position verification is performed on the parameter anchor point fingerprint corresponding to the target correction item. Before automatically correcting and writing the parameters, the points to be modified in the document are verified a second time using the parameter's unique digital fingerprint to confirm that the current area to be written is completely matched with the original area where the parameter was extracted, and there are no point misalignments caused by changes in document layout, table shifting, or paragraph disorder. The system (referring to the system equipped with the method proposed in this application) only performs automatic correction operation when the risk level of the target correction item is low, the reliability of the parameter source is high, the write-back position is confirmed to be reliable by the parameter anchor point fingerprint verification, and it is estimated that no new abnormal constraint edges will be generated after execution. In response to determining that the target correction item has a predetermined high-risk parameter, a correction item to be confirmed is generated. If the target correction item involves high-risk parameters that affect the safe operation of equipment, such as protection settings, operating time limits, current transformer ratio, voltage transformer ratio, circuit breaker breaking current, short-circuit current, main transformer capacity, and relay protection logic related parameters, then a correction item to be confirmed is generated, which includes conflict causes, candidate parameter values, evidence sets, hit constraint rules, and data modification locations, for review.
[0043] The above scheme obtains a dataset composed of data from multiple sources, and preprocesses the dataset to obtain multiple sets of normalized data. This not only improves data quality but also merges and integrates heterogeneous parameters from multiple sources, allowing parameters associated with the same device scattered across different data carriers to form unified structured data. Based on these multiple sets of normalized data, a parameter constraint graph is constructed. This graph contains multiple nodes mapped from each set of normalized data, and any two related nodes are connected by at least one constraint edge. This breaks down the limitations of the original fragmented heterogeneous data and can reveal implicit parameter conflicts that are difficult to identify through text comparison and field matching. Based on the parameter constraint graph, deviation calculations are performed on each set of normalized data to obtain constraint data for at least one constraint edge corresponding to each set of normalized data. Each constraint data includes a constraint edge deviation value and a constraint satisfaction degree. Deviation calculation is performed based on a parameter constraint graph, which comprehensively records all the associated constraint relationships between devices, parameters, and data nodes. This allows for rapid location of all constraint rules matching a single set of normalized data, enabling quantitative judgment of parameter conflicts and effectively avoiding the missed detections and misjudgments caused by traditional text comparison. For each constraint edge, in response to determining that the deviation value of the constraint edge is greater than the tolerance threshold, a subgraph is extracted from the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge. This subgraph visually displays the devices, parameters, data carriers, and constraint logic involved in the conflict, clearly locating the source of the conflict and related parameters. Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, at least one target correction item is determined for the deviation subgraph. The optimal correction scheme that is compliant and reasonable is automatically selected, eliminating the need for manual comparison of multiple sets of data to determine the modification direction and reducing the workload of manual analysis. For each target correction item, in response to determining that the target correction item meets the automatic correction conditions, the data set is partially updated based on the target correction item to obtain the updated data set.
[0044] In some embodiments, the dataset includes multiple types of data; the preprocessing of the dataset to obtain multiple sets of normalized data includes: for each type, parsing the data of that type in the dataset according to a predetermined parsing strategy corresponding to that type to obtain multiple parsed data corresponding to that type; merging all parsed data according to predetermined indicators to obtain multiple sets of merged data; and performing standardization integration processing on each set of merged data to obtain normalized data corresponding to each set of merged data.
[0045] In this embodiment, the dataset includes multiple data types, including text data, tabular data, formula data, attached figure data, engineering drawing data, and structured data. For each type, the data of that type in the dataset is parsed according to a predetermined parsing strategy, resulting in multiple parsed data for that type. The predetermined parsing strategy can be determined based on the data type; each type is pre-configured with a dedicated predetermined parsing strategy, and different types of predetermined parsing strategies are configured with differentiated parsing steps and recognition rules. For example, the following processes are performed on data corresponding to the main text data type: chapter recognition, paragraph segmentation, equipment name recognition, parameter phrase recognition, numerical unit recognition, and context window extraction; for data corresponding to the table data type: header hierarchy recognition, merged cell expansion, row and column semantic alignment, unit column recognition, parameter cell positioning, and table coordinate recording; for data corresponding to the formula data type: formula area recognition, formula symbol recognition, expression tree construction, variable definition extraction, variable-to-power parameter mapping, and formula input-output relationship binding; for data corresponding to the attached figure data type: image enhancement, text area detection, optical character recognition, annotation box positioning, and spatial proximity association between annotation text and equipment objects; for data corresponding to the engineering drawing data type: layer parsing, element recognition, block attribute extraction, equipment symbol recognition, connection line tracing, terminal number recognition, and equipment topology relationship extraction; for data corresponding to the structured data type: field path parsing, field alias mapping, field unit recognition, primary key object binding, and hierarchical object expansion.
[0046] In another example, for text-based documents, the system performs document structure parsing, heading level identification, paragraph segmentation, power equipment object identification, power parameter attribute identification, numerical and unit identification, context window extraction, and data location anchor point recording. For instance, if the text states: "The rated capacity of main transformer No. 1 is 50MVA, and the high-voltage side voltage is 110kV," the system can extract: as well as: For tabular data types, the system performs operations such as table area location, header level identification, merged cell expansion, row and column semantic alignment, unit column identification, parameter cell location, and cell coordinate binding. For example, the equipment parameter table contains the following records: Table 1 Equipment Parameter Table
[0047] The system binds the cell value to the device object, parameter attribute, and unit, and records the cell coordinates.
[0048] For formula data types, the system performs operations such as formula region location, formula symbol recognition, formula expression tree construction, variable definition extraction, variable-to-power parameter attribute mapping, formula input and output parameter binding, and formula number and location recording. For example, the formula: (4) The system can Identified as three-phase current, Identify as apparent power or rated capacity, The system identifies line voltage and establishes formula constraints for calculation. For attached image data, the system performs image denoising, tilt correction, text region detection, optical character recognition, parameter annotation box positioning, and associating annotation text with equipment objects. For example, for "Rated Voltage 10kV" in an equipment nameplate image, the system can bind this parameter to the corresponding equipment object by combining the image file name, upload directory, equipment number, or nearby annotations. For engineering drawing data, the system performs drawing layer parsing, element type recognition, text annotation extraction, block attribute extraction, equipment symbol recognition, connection line tracing, terminal number recognition, and topology connection relationship extraction. For example, if outgoing line bay 101 is identified as connected to the 10kV I section bus in a primary system diagram, the system establishes the topology connection relationship: Feeder101→Bus10kV_I. For structured data, the system performs field path parsing, field alias mapping, field unit recognition, primary key or equipment code binding, and hierarchical object expansion. For example, the field path: substation.transformer[0].ratedCapacity can be mapped to the rated capacity parameter of the device object "transformer[0]".
[0049] All parsed data is merged according to predetermined indicators, resulting in multiple sets of merged data. For example, the predetermined indicator is the standard identifier of the equipment object. Specifically, for each set of parsed data, equipment object, parameter attributes, original parameter values, original units, data location, data version, data status, and parsing confidence are extracted to obtain the extracted data corresponding to each set of parsed data. The extracted data undergoes equipment object merging, parameter attribute mapping, unit dimension normalization, version identifier binding, data location anchoring, and source confidence calculation to obtain normalized data corresponding to each set of merged data. The specific implementation methods of each processing technique are as follows: (1) Equipment Object Merging: The same equipment object may have different names in different data. For example: "No.1 Main Transformer", "1# Main Transformer", "T1 Main Transformer", "Main Transformer T1" or "No.1 Transformer". The system merges equipment objects based on equipment name, equipment type, equipment number, bay, topology location, data context and associated parameters. Equipment object merging similarity can be expressed as: (5) in, Indicates name similarity. Indicates the similarity of device types. Indicates topological similarity. Indicates contextual similarity, Indicates the similarity of associated parameters. to For weighting coefficients. When When the number of objects exceeds the preset object merging threshold, the two device objects will be merged into the same device object standard identifier.
[0050] (2) Parameter attribute mapping: The system establishes a standard identifier table for parameter attributes, for example: Table 2 Standard Identifier Table for Parameter Attributes
[0051] The system determines the standard identifier of parameter attributes based on parameter name, device type, context, and unit dimension.
[0052] (3) Unit Dimension Normalization: The system establishes a unit dimension library and unit conversion functions. For linear units, the unit conversion can be expressed as: (6) in, This represents the parameter value under normalized units. Indicates the original parameter value. Indicates the unit conversion factor. This indicates a unit offset. For example: , , .
[0053] For parameters such as "600 / 5A", the system can be broken down as follows: primary side rated current: 600A; secondary side rated current: 5A; ratio factor: 120.
[0054] For percentage parameters, such as short-circuit impedance "10.5%", the system can convert it to a per-unit value: 10.5% = 0.105.
[0055] (4) Source credibility calculation: source credibility of normalized parameter records It can be determined based on the basic credibility of the data source, the credibility of the data version, the credibility of the data status, the confidence level of the parsing, the confidence level of object merging, and the confidence level of parameter attribute mapping. For example: (7) in, To ensure the basic credibility of the data source, To ensure the reliability of the data version, To assess the reliability of the data status, To analyze the confidence level, Merge confidence for objects. Map confidence levels to parameter attributes.
[0056] (5) Parameter Anchor Fingerprint Generation: To ensure accurate write-back location, the system generates a parameter anchor fingerprint for each normalized parameter record. The parameter anchor fingerprint can be represented as: (8) in, For data source identification, For data location anchor points, For standard identification of equipment objects, For parameter attribute standard identifiers, For normalized units, This is the version of the document. The context window that represents the area surrounding the parameter value.
[0057] The parameter anchor fingerprint is used for: location verification before write-back, change tracking after write-back, binding of evidence chain for correction reports, and verification of the original position during rollback recovery.
[0058] Each group of merged data is standardized and integrated to obtain normalized data for each group. Normalized data can be represented as: (9), among which, These are the normalized parameter values. Data status. To ensure the credibility of the source, For parameter anchor fingerprint, Parameter status labels can include unvalidated, normal, suspected anomaly, conflicting, pending confirmation and correction, automatically corrected, second confirmation, and rolled back. Data location anchors include at least one of the following: page number, chapter number, paragraph number, character offset, table number, cell coordinates, formula number, image coordinate frame, drawing number, layer name, element number, block attribute name, database table name, field path, and interface field identifier. Parameter anchor fingerprints are generated based on the data source identifier, data location anchor, device object standard identifier, parameter attribute standard identifier, normalized unit, data version, and parameter context window.
[0059] This application enables the configuration of dedicated parsing strategies for various heterogeneous power technology data, including text, tables, formulas, figures, engineering drawings, and structured data, to perform precise parsing by type. It achieves unified parameter extraction across all data formats, resolving issues of scattered, missed, and incomplete parameter identification in multi-format data. Simultaneously, it converts the extracted parameters into unified, normalized parameter records. Through equipment object merging, parameter attribute standard mapping, unified unit dimension conversion, version and location information binding, source credibility quantification, and parameter anchor fingerprint generation, it standardizes original parameters with inconsistent sources, descriptions, units, and versions. This allows parameters of different formats, units, versions, and source locations to be calculated and compared horizontally under a unified data structure. It provides a unified data benchmark for constraint verification and precise location write-back, offering unified, traceable, and quantifiable underlying data support for parameter constraint map construction, parameter conflict identification, automatic correction verification, and precise write-back.
[0060] In some embodiments, the step of calculating the deviation for each set of normalized data based on the parameter constraint graph to obtain the constraint data of at least one constraint edge corresponding to each set of normalized data includes: for each constraint edge corresponding to each set of normalized data, obtaining the constraint type bound to the constraint edge based on the parameter constraint graph, and determining a predetermined constraint expression based on the constraint type; and determining the constraint data corresponding to the constraint edge based on the predetermined constraint expression.
[0061] In this embodiment, constraint edges are divided into binary edges and hyperedges connecting multiple parameter records. Binary edges connect only two normalized parameter records, suitable for scenarios requiring only two sets of parameter association constraints. Hyperedges can connect three or more normalized parameter records simultaneously, capable of supporting multi-parameter linkage constraints; formula calculation constraint edges are typically hyperedges. Constraint types are pre-defined for various electrical parameter verification scenarios, and each constraint type is matched with a unique predefined constraint expression. A binding relationship is established between constraint types, predefined constraint expressions, and constraint edges. Constraint types can include synonymous equivalence constraint types, unit conversion constraint types, formula calculation constraint types, topology dependency constraint types, standard value constraint types, and version inheritance constraint types. The specific definitions, functions, and corresponding predefined constraint expressions for each type of constraint edge are as follows: (1) Synonymous equivalence constraint type constraint edge: The synonymous equivalence constraint type constraint edge is used to ensure that the parameter values of the same equipment object and the same parameter attribute are consistent in different data sources. After converting the parameter values from different sources belonging to the same equipment object standard identifier and the same parameter attribute standard identifier to the same normalized unit, the numerical deviation is calculated. For example, the technical agreement, equipment ledger and design specification all record "rated capacity of No. 1 main transformer", and the system establishes synonymous equivalence constraint edges between these parameter records.
[0062] The predefined constraint expression corresponding to the synonym equivalence constraint type can be: (10), among which, This is the identifier for the constraint expression corresponding to a synonymous equivalence constraint. The standardized value of parameter a is the standard value after standardization. The standard value of parameter b after standardization.
[0063] (2) Constraint edge of unit conversion constraint type: Constraint edge of unit conversion constraint type is used to constrain the conversion relationship of the same physical quantity under different units. Based on the unit dimension and unit conversion function, the parameter values under different units are mapped to the same unit dimension and the conversion deviation is calculated.
[0064] The predefined constraint expression corresponding to the unit conversion constraint type can be: (11) in, Represents unit conversion functions. These are the original collected values. To and Paired raw data collection values with unit conversion associations. For example: If one source records 10kV and another source records 10000V, the units should be consistent after conversion.
[0065] (3) Constraint edges of formula calculation constraint type: Constraint edges of formula calculation constraint type are used to constrain the calculation relationship between multiple power parameters. The normalized parameter records corresponding to the formula input variables are substituted into the formula constraint expression to obtain the predicted output parameters, and the deviation between the predicted output parameters and the actual output parameters in the data source is calculated. For example, in a three-phase system, the capacity, voltage, and current can satisfy: (12), among which, Represents current. Indicates apparent power or rated capacity. This represents the line voltage.
[0066] For example, active power can be expressed as: (13), among which, The power factor.
[0067] The system calculates the constraint edges using formulas to verify whether the relevant parameters in the calculation sheet, parameter table, and technical agreement are consistent.
[0068] (4) Constraint edges of topology dependent constraint type: Constraint edges of topology dependent constraint type are used to express the relationships between equipment connections, associated bays, associated buses, terminal connections, and circuits. Based on the equipment connection relationships extracted from engineering drawings or structured topology data, the voltage level, circuit number, bay number, terminal number, connection object, or associated bus of the associated equipment objects are verified to be consistent. For example: If the outgoing bay is connected to the 10kV bus, then the voltage level of the outgoing bay should be consistent with the voltage level of the bus; if the high-voltage side of the transformer is connected to the 110kV bus, then the rated voltage of the high-voltage side of the transformer should match the voltage level of the bus; the terminal number of the secondary circuit should be consistent with the terminal number in the terminal block diagram; the starting equipment and ending equipment in the cable list should be consistent with the topology relationship in the drawings.
[0069] The predefined constraint expression corresponding to the topological dependency constraint type can be: (14) in, The attribute to be verified corresponds to the slave device. These are the standard values for attributes such as voltage level, terminal number, and circuit number of the topology reference device.
[0070] For example: if an outgoing line bay is connected to a 10kV busbar, then the voltage level of the outgoing line bay should be consistent with the voltage level of the busbar; if the high-voltage side of the transformer is connected to a 110kV busbar, then the rated voltage of the high-voltage side of the transformer should match the voltage level of the busbar; the terminal numbers of the secondary circuit should be consistent with the terminal numbers in the terminal block diagram; the starting and ending equipment in the cable list should be consistent with the topology of the drawings.
[0071] (5) Constraint edges of standard value constraint type: Constraint edges of standard value constraint type are used to constrain the allowable range or standard level under specific equipment type, parameter attribute and engineering scenario. The allowable range or standard level of the parameter is determined according to the equipment type, voltage level, parameter attribute and engineering scenario, and it is judged whether the parameter value meets the allowable range or standard level. For example, allowable range or standard level can be configured for parameters such as circuit breaker breaking current, transformer ratio, line length, rated voltage, rated current and protection action time limit.
[0072] The predefined constraint expression corresponding to the standard value constraint type can be: (15), among which, This is the lower limit threshold of the allowed value range for the parameter. This is the upper limit threshold of the allowed value range for the parameter. Standardize the values for the i-th parameter.
[0073] It can also be: (16) in, This represents the set of preset standard gears.
[0074] (6) Constraint edges of version inheritance constraint type: Constraint edges of version inheritance constraint type are used to constrain the parameter inheritance relationship between different versions of data. Based on the data version, data status, approval time, version priority and version inheritance relationship, the baseline parameter value is determined and the deviation of the parameter value to be corrected relative to the baseline parameter value is calculated. For example, when a new version of the equipment ledger or a new version of the technical agreement has been approved, but the old version of the instruction manual or old version of the drawings still retains the old parameters, the system can identify the parameters that need to be updated synchronously based on the version inheritance constraint edge.
[0075] The predefined constraint expression corresponding to the version inheritance constraint type can be: (17) in, This is the mark indicating that the new version of the documents has been approved and is now in effect. These are normalized parameter values from the old version of the data. Normalized parameter values are used to approve new versions of documentation. When the new version of documentation takes effect, the old parameters must be consistent with the new parameters. If there is a difference between the two values, the constraint is not met, and the system can identify the old parameters that need to be updated synchronously. For example, if a new version of the equipment ledger or a new version of the technical agreement has been approved, but the old version of the instruction manual or old version of the drawings still retains the old parameters, the system can identify the parameters that need to be updated synchronously based on the version inheritance constraint.
[0076] Based on each constraint edge within the parameter constraint graph, the constraint type bound to the constraint edge is read, the predefined constraint expression corresponding to that type is retrieved, and the normalized parameters associated with the constraint edge are recorded. Substitute the original residuals into the constraint residual function to calculate the original residuals, and then combine the normalized baseline function stability parameters to calculate the constraint edge deviation values. The constraint edge deviation values are then determined using the following formula: (18) in, To constrain the residual function, For the normalized reference function, To stabilize parameters and avoid the denominator being too small, This represents the constraint edge deviation value. Constraint residual function. The system directly associates the predefined constraint expressions corresponding to the aforementioned constraint edges with the normalized parameter records bound to the constraint edges. Substituting these normalized parameter records into the predefined constraint expressions for the corresponding constraint types, the resulting difference is the solution. .
[0077] Based on the constraint edge deviation value, the constraint satisfaction degree corresponding to the constraint edge is determined by the following formula. : , (19) in, This represents the tolerance threshold.
[0078] when When this happens, the system identifies the constraint edge as an abnormal constraint edge.
[0079] The constraint satisfaction of a local graph can also be determined using the following formula: (20) in, This represents the set of constraint edges in the local graph. This indicates the weight of the constraint edges. It allows for a direct and intuitive quantification of the overall compliance quality of a single device, a single bay, or a single set of engineering parameter maps.
[0080] In some embodiments, the step of extracting a subgraph from the parameter constraint graph based on the constraint edge to obtain a deviation subgraph corresponding to the constraint edge includes: using the constraint edge as a retrieval starting point, performing a neighborhood retrieval with a predetermined radius in the parameter constraint graph to obtain multiple nodes and multiple constraint edges corresponding to the constraint edge; clustering the multiple nodes and multiple constraint edges corresponding to the constraint edge to obtain multiple sets of conflicting entity clusters; performing redundancy removal on the multiple sets of conflicting entity clusters, and constructing a deviation subgraph corresponding to the constraint edge based on the multiple sets of conflicting entity clusters after redundancy removal.
[0081] In this embodiment, a neighborhood search with a defined radius is performed starting from the abnormal constraint edge to obtain multiple nodes and multiple constraint edges corresponding to the constraint edge. The defined radius is determined based on historical experience. A pre-set search radius (e.g., radius 2, 3) is used as the starting point for the search, with the constraint edge e marked as abnormal as the starting point. The search expands outward layer by layer along the connected nodes and constraint edges within the graph, only traversing related entities within a distance not exceeding the preset radius. Finally, nodes and edges with the same device object, same parameter attributes, same formula variables, same topological connection, or same version inheritance relationship as the abnormal constraint edge are collected. The search path includes three types of nodes: device object nodes, normalized parameter record nodes, data source nodes of various types, and various constraint edges connecting the nodes. Neighborhood retrieval returns scattered data containing nodes and constraint edges from multiple devices, various parameters, and multiple types of constraints. The original data lacks grouping boundaries and cannot distinguish independent conflict issues. Based on the standard identifier of the device object, the standard identifier of the parameter attribute, the constraint type, and the document version, multiple nodes and multiple constraint edges corresponding to the constraint edges are clustered to obtain multiple conflict entity clusters. This can group related entities belonging to the same device, the same parameter, and the same set of constraint logic into a single conflict entity cluster, isolating abnormal conflicts from different devices and different parameters, and avoiding interference from multiple types of anomalies in the analysis.
[0082] The conflict entity clusters obtained through clustering contain a large number of irrelevant entities that do not contribute to the abnormal constraint deviations. Therefore, a redundancy removal operation is required. The system will delete normal constraint edges, isolated evidence nodes, and ineffective version nodes that do not directly contribute to the abnormal deviations within the clusters, filter out invalid interference data, and reduce the data size of the conflict entity clusters, retaining only the core nodes and constraint edges that directly cause parameter deviations and trigger constraint anomalies. Based on the multiple conflict entity clusters after redundancy removal, a deviation subgraph corresponding to the constraint edges is constructed. The deviation subgraph can be represented as: ,(twenty one), in, Represents the set of nodes related to the conflict. This represents the set of conflict-related constraint edges.
[0083] The devices, parameters, data nodes, and constraint edges directly related to the current anomaly constraints are extracted to form independent local subgraphs. Irrelevant graph data is removed, and all related entities that directly contribute to the deviation are fully included, forming a deviation subgraph with clear boundaries and concentrated problems, without traversing the entire massive parameter constraint graph. At the same time, upstream and downstream correlation evidence of the anomaly is fully preserved, which facilitates the system's location of the root cause conflict parameter group.
[0084] In some embodiments, determining at least one target correction item corresponding to the deviation subgraph based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph includes: extracting information from the deviation subgraph to obtain at least one conflict parameter group; for each conflict parameter group, searching for at least one target predetermined business constraint rule corresponding to the conflict parameter group among the multiple predetermined business constraint rules; calling the at least one target predetermined business constraint rule to generate at least one candidate correction item corresponding to each target predetermined business constraint rule; for at least one candidate correction item corresponding to each target predetermined business constraint rule, obtaining multi-dimensional evaluation index data corresponding to the candidate correction item; determining the correction score corresponding to the candidate correction item based on the multi-dimensional evaluation index data corresponding to the candidate correction item; filtering out the target correction item corresponding to the target predetermined business constraint rule from the at least one candidate correction item based on the correction score of each candidate correction item corresponding to the target predetermined business constraint rule and the constraint satisfaction of multiple constraint edges corresponding to the conflict parameter group; and using all target correction items corresponding to the target predetermined business constraint rules as at least one target correction item corresponding to the deviation subgraph.
[0085] In this embodiment, information is extracted from the deviation subgraph to obtain at least one conflict parameter group. Normalized parameter records that directly generate deviations and contribute to abnormal constraint edges are separately collected, comprehensively preserving abnormal evidence and improving the efficiency of parameter correction and problem tracing. The conflict parameter group can include conflicting device objects, conflicting parameter attributes, a set of conflicting parameter values, parameter units, data sources, data locations, data versions, abnormal constraint types, and conflict causes. The conflict parameter group can be represented as: ,(twenty two), in, For conflicting device objects, For conflict parameter attributes, For the set of conflicting parameter values, Units for parameters For conflict resource collection, For the set of data location anchor points, This is a collection of data versions. For the set of abnormal constraint edges, Reasons for conflict include: inconsistent values for the same parameter from different data sources; inconsistent values after unit conversion; inconsistent calculation results from formulas; inconsistent device topology hierarchies; parameter values exceeding the allowable range or deviating from the standard level; inconsistencies between high-priority and low-priority versions; insufficient confidence in data location resolution; and ambiguity in device object merging. It should be noted that all types of data used in the implementation of this application can be obtained completely through multiple sources.
[0086] For each conflict parameter group, at least one target pre-defined business constraint rule corresponding to the conflict parameter group is searched among multiple pre-defined business constraint rules. This is because a single conflict parameter group may only contain some types of constraint conflicts, and only some pre-defined business constraint rules match the current conflict, while the remaining rules have no corresponding conflict and do not need to participate in the calculation. At least one target pre-defined business constraint rule is invoked to generate at least one candidate correction item corresponding to each type of target pre-defined business constraint rule. Each type of target constraint rule corresponds to an independent and compliant conflict rectification logic, and there are often multiple differentiated feasible correction paths under a single type of rule. All solutions need to be generated completely as the subsequent optimal correction item. Targeted rectification solutions can be output based on the parameters, devices, and version scenarios corresponding to the current conflict, eliminating irrelevant and redundant suggestions. Multiple compliant rectification paths can exist under the same constraint rule, and multiple candidate solutions can provide sufficient alternatives for subsequent selection of the optimal target correction item. Among them, candidate corrective items are based on conflict parameter groups and adapt to all compliant and feasible rectification operation plans generated according to the corresponding predetermined business constraint rules. Candidate corrective items may include target parameter records, candidate parameter values, candidate units, candidate value sources, corrective operation types, evidence sets, expected increments in constraint satisfaction, corrective costs, risk levels, and rollback information. A candidate corrective item can be represented as: ,(twenty three), in, Record the target parameters. Candidate parameter values, As a candidate unit, As a source of candidate values, This is a type of correction operation. For the collection of evidence, To constrain the expected increment of satisfaction, To correct the course of events, Risk level, To roll back information.
[0087] It should be noted that the pre-defined business constraint rules can include constraint rules based on source credibility, constraint rules based on weighted majority consensus, constraint rules based on forward formula calculation, constraint rules based on backward formula derivation, constraint rules based on unit conversion, constraint rules based on topological dependency, constraint rules based on standard values, and constraint rules based on version inheritance. The system generates corresponding candidate correction items based on these eight types of pre-defined business constraint rules, as follows: (1) Constraint rules based on source credibility: When the same parameter attribute of the same device object has inconsistent values in multiple data sources, the system can use the parameter value with the highest source credibility as the candidate value. For example, as shown in Table 3, Table 3. Examples of Reliability Comparison of Parameters from Multiple Data Sources
[0088] The system can generate candidate correction items to change the 40MVA correction in the design specification to 50MVA.
[0089] (2) Constraint rule based on weighted majority consensus: The system performs weighted voting on identical parameter values. The weighted votes can be expressed as: ,(twenty four), in, This indicates the source credibility of the corresponding normalized parameter record. If the weighted vote count for a certain parameter value is higher than other values and exceeds the preset confidence threshold, the system will use that parameter value as a candidate correction value.
[0090] (3) Constraint rules based on formula forward calculation: For constraint edges calculated by formula, when the source of input parameters has high credibility but the output parameters have deviation, the system calculates and outputs candidate values based on the formula forward calculation.
[0091] For example: (25) like and The source is highly credible, and the data... If the result is inconsistent with the formula calculation, then generate The candidate correction value.
[0092] (4) Constraint rules based on formula reverse derivation: When the formula output parameters and some input parameters have high credibility, and a certain input parameter is suspected to be wrong, the system can generate a candidate correction value for the input parameter through formula reverse derivation.
[0093] (5) Constraint rules based on unit conversion: When there is a fixed multiple relationship between conflicting parameters and the multiple is consistent with the unit conversion relationship, the system identifies suspected unit errors and generates unit correction items.
[0094] For example, if one source records "10000V" and another source records "10kV", and the values are the same after normalization, the system will not consider them to be conflicting. If one source records "1000V" and other highly reliable sources record "10kV", the system will generate candidate correction terms to correct "1000V" to "10kV" or "10000V".
[0095] (6) Constraint rules based on topology dependency: When the parameters of the equipment are inconsistent with the parameters of the connected objects, the system generates candidate correction items based on the topology dependency relationship. For example, if the outgoing line bay is connected to a 10kV bus, but the voltage level of the outgoing line bay is recorded as 6kV in the equipment parameter table, the system generates a candidate correction item to correct 6kV to 10kV.
[0096] (7) Constraint rules based on standard values: When a parameter value is not within the allowable range, or deviates from the standard range but is close to a certain standard range, the system can generate a standard range correction item. For example, if the rated breaking current of a circuit breaker is recorded as 30kA, and there is a neighboring 31.5kA in the standard range set, the system can generate a candidate correction item based on the equipment type, voltage level and context, and set it to be confirmed by operation or automatically corrected.
[0097] (8) Constraint rules based on version inheritance: When the same parameter in a high-priority version of the data is inconsistent with that in a low-priority version of the data, the system will use the parameter value in the high-priority version of the data as a candidate value. For example, if the rated capacity has been updated in the new version of the technical protocol, but the old value is still retained in the old version of the design specification, the system will generate a candidate correction item to synchronize the parameter in the old version of the specification with the parameter in the new version.
[0098] When generating candidate correction items for conflicting parameter groups, candidate parameter values, which serve as the core elements of the candidate correction items, can be generated through various methods. These include selecting the most reliable normalized parameter record within the conflicting parameter group, integrating parameter values from multiple data sources with a weighted majority consensus, forward calculating output parameters based on formula constraints, and reverse deriving input parameters. Alternatively, unit conversion constraints can be used to identify unit errors and generate unit correction items. Topology dependency constraints can be used to transfer bus voltage levels, circuits, bays, terminal numbers, and connection objects to generate parameters. Standard value constraints can be used to map parameters deviating from the tolerance range to the standard range, or version inheritance constraints can be used to synchronize parameters from higher-priority versions to lower-priority versions. Candidate correction items are based on candidate parameter values and integrate target parameter records, candidate units, candidate value sources, correction operation types, evidence sets, expected increments in constraint satisfaction, correction costs, risk levels, and rollback information to form a complete and executable rectification plan.
[0099] For each target-defined business constraint rule, at least one candidate corrective action is obtained. Multi-dimensional evaluation metrics data for each candidate corrective action are acquired. These metrics may include the expected increment of constraint satisfaction, source credibility, version priority, corrective action cost, risk level, expected number of new anomalies, and write-back reliability. Based on the multi-dimensional evaluation metrics data for each candidate corrective action, a corrective action score is determined. The corrective action score for each candidate corrective action is determined using the following formula: ,(26) in, Indicates candidate correction terms, This represents the expected increment in constraint satisfaction after executing the candidate corrective measures. Indicates the credibility of the candidate value source. Indicates version priority. Indicates write-back reliability. This indicates the cost of correcting the error. Indicates the risk level. This indicates the expected number of new anomalies. to These are the weighting coefficients.
[0100] Based on the correction scores of each candidate correction item corresponding to the target predetermined business constraint rule, and combined with the constraint satisfaction of all constraint edges associated with the conflict parameter group, a target correction item is selected from at least one candidate correction item corresponding to the target predetermined business constraint rule. Finally, the candidate correction item with the highest correction score and simultaneously meeting multiple judgment conditions is designated as the target correction item corresponding to the target predetermined business constraint rule. All target correction items corresponding to the target predetermined business constraint rules are used as at least one target correction item corresponding to the deviation subgraph. These multiple judgment conditions may include: correction score higher than a preset score threshold, expected increment of constraint satisfaction higher than a preset increment threshold, candidate value source credibility higher than a preset credibility threshold, risk level meeting risk gating conditions, write-back reliability higher than a preset write-back reliability threshold, and expected number of new anomalies not exceeding a preset new anomaly threshold. It should be added that two of the above six judgment conditions are related to constraint satisfaction: one is the expected increment threshold condition for constraint satisfaction, which is calculated from the original constraint satisfaction of multiple constraint edges corresponding to the conflict parameter group and is used for direct verification; the other is that the calculation process of the correction score will introduce the expected increment of constraint satisfaction as the core scoring dimension, and indirectly use constraint satisfaction related data.
[0101] This application extracts a deviation subgraph by using abnormal constraint edges as the entry point, and constructs a conflict parameter group based on the deviation subgraph. This can accurately locate conflicts to the corresponding device object, parameter attribute, data source, data location, and constraint rule, effectively reducing the false alarm rate and improving the overall efficiency of subsequent rectification and correction. At the same time, this invention can generate candidate correction items in batches based on multiple business logics such as source credibility, weighted majority consistency, formula forward and reverse calculation, unit conversion, topology dependency transmission, standard gear mapping, and version inheritance. Then, combined with the correction scoring model, it comprehensively selects the optimal target correction item based on multiple dimensions such as constraint satisfaction, credibility, risk, and reliability. This makes the entire parameter correction process have quantitative computability and complete traceability and interpretability.
[0102] In some embodiments, the step of partially updating the data set based on the target correction item to obtain an updated data set includes: finding the target data corresponding to the target correction item in the data set; writing back the parameters of the target data based on the target correction item to obtain an updated data set; and writing back the difference record and rollback handle.
[0103] In this embodiment, when the target correction item meets the automatic correction conditions, the target data corresponding to the target correction item is searched in the data set. For example, the target data includes at least one of the following: text data, table data, formula data, attached figure data, engineering drawing data, and structured data. This can accurately locate the engineering data carrier to be corrected, narrowing the scope of subsequent modifications and avoiding misoperation of irrelevant data. Based on the target correction item, the parameters of the target data are written back to obtain the updated data set, and the write-back difference record and rollback handle are also performed. Specifically, based on the target correction item, the parameters of the target data in the data set are modified using the predetermined update logic corresponding to the target data to obtain the updated data set, the write-back difference record, and the rollback handle. The write-back difference record can be a standardized change log that records information such as the transaction ID of this modification, the parameter fingerprints before and after the modification, the original value, the new value, and the operation time; the rollback handle can be an operation identifier that allows one-click restoration of data to its original state before modification based on the backup before the write-back. On the one hand, the optimal target correction item, which has undergone multi-dimensional threshold verification, is strictly used as the modification standard to ensure that the modified content is consistent with the cause of the conflict and that the correction result is compliant and reasonable. On the other hand, differentiated lossless update logic is adopted for different engineering data. Before writing back the target data based on the target correction item, the target parameter position can be locked by the parameter anchor fingerprint, the consistency of the data context window can be verified, and a backup before writing back and a unique write-back transaction identifier can be generated to retain the original snapshot and unique transaction traceability certificate for subsequent operations.
[0104] For example, the aforementioned pre-defined update logic includes: performing text replacement on target parameter values in the main text data while maintaining paragraph styles, character styles, numbering structures, and contextual semantics; performing cell updates on target parameter values in the table data while maintaining the table header structure, merged cell relationships, row and column relationships, and formula reference relationships; updating variable values, formula results, or formula footnotes in the formula data and triggering recalculation of associated formula results; updating the annotation layer on parameter labels in the attached figures data, or generating a corrective annotation layer that overwrites the original annotations; updating element text, block attributes, layer attributes, terminal numbers, or equipment labels in the engineering drawing data; and updating field values in the structured data according to the field path. Each parameter write-back generates a pre-write-back backup, a write-back difference record, a write-back transaction identifier, and a rollback handle for restoring the pre-write-back state.
[0105] It should be noted that when the target correction item meets the automatic correction conditions, the system will perform parameter write-back on the corresponding data object. Before the write-back, the system will lock the target parameter position based on the parameter anchor fingerprint, verify whether the context window of the current data object is consistent with the record before the write-back, synchronously generate a backup before the write-back, create a write-back transaction identifier and generate a rollback handle, and then process it according to the data type using the corresponding write-back method: performing text replacement on text data, performing cell update on table data, updating variables or formula results on formula data, performing annotation layer update on attached data, updating element text or block attributes on engineering drawing data, and updating field values according to field paths on structured data. After all write-back operations are completed, the system generates a standardized write-back difference record, which may include: ,(27) Where transId represents the write-back transaction identifier. This indicates writing back the anchor fingerprint of the previous parameters. This indicates the post-write-back parameter anchor fingerprint. To write back the original parameter values, To update the parameter value after writing it back, time is the change timestamp. This is an operation identifier.
[0106] Before writing back the target data, the position to be written back in the target data object can be re-verified based on the parameter anchor fingerprint. This allows for a second check to verify whether the position of the parameter to be updated matches the point in the target correction item record, identifying anomalies such as changes in data context and parameter misalignment in advance. This prevents accidental modification of irrelevant parameters and incorrect parameter placement from the source, significantly improving the accuracy of automatic write-back, reducing invalid changes and subsequent rollback operations, and ensuring the stability and reliability of automatic correction. After writing back the target data, a new parameter anchor fingerprint is generated, and a mapping relationship is established between the parameter anchor fingerprint before write-back, the parameter anchor fingerprint after write-back, and the parameter write-back difference record. By accurately matching the original and updated parameters before and after modification using a unique anchor fingerprint, and relying on the mapping relationship, the complete log corresponding to each change can be quickly located, making all parameter modification operations traceable and verifiable throughout the process. At the same time, the anchor fingerprint can serve as a unified identification basis for subsequent verification of data context, update parameter constraint graphs, and closed-loop deviation verification. Once issues such as parameter inconsistency or write-back anomalies occur, the source of the modification can be quickly located by associating the difference records. This also facilitates the generation of a complete correction report, improving the data consistency, verifiability, and troubleshooting efficiency of the automatic write-back process.
[0107] This application relies on a dual mechanism of parameter anchor fingerprint positioning and context window consistency verification to accurately assign target correction items to corresponding data objects, effectively avoiding the problem of parameter write-back location disorder. At the same time, it can automatically correct parameters for various types of engineering data, including text, tables, formulas, attached figures, engineering drawings, and structured data. It builds a complete business processing closed loop from parameter anomaly identification and optimal target correction scheme selection to automatic modification of multi-source engineering data, reducing manual rectification operations and improving the automation level and overall processing efficiency of engineering parameter correction.
[0108] In some embodiments, after writing back the target data based on the target correction item to obtain an updated data set, the method includes: preprocessing the target data to obtain updated normalized parameter data; updating the nodes and constraint edges in the parameter constraint graph corresponding to the target correction item based on the updated normalized parameter data to obtain the updated parameter constraint graph; performing local verification on the updated parameter constraint graph based on the parameter constraint graph to obtain a closed-loop verification result; in response to determining that the closed-loop verification result meets a preset pass condition, outputting the corrected target data, and generating and outputting a correction report based on the write-back difference record; in response to determining that the closed-loop verification result does not meet the preset pass condition, restoring the target data before parameter write-back based on the rollback handle, and adding a pending confirmation identifier to the conflict parameter group corresponding to the target correction item.
[0109] In this embodiment, the target data is first preprocessed to generate updated normalized parameter data in a unified format. This is to eliminate differences in the formats of multi-source engineering data, unify parameter standards, and facilitate accurate synchronization of graph-related nodes and constraint edges. This allows the parameter constraint graph to remain synchronized with the underlying entity data, preventing the graph from becoming disconnected from the actual data information. Next, based on the updated normalized parameters, the parameter constraint graph is locally updated, refreshing the corresponding nodes and constraint edges. This allows for local synchronous correction of parameter relationships without requiring a full graph reconstruction, reducing computational overhead. Specifically, based on the updated normalized parameter data, the first step is to locate all parameter nodes, device nodes, data evidence nodes, and associated constraint edges within the parameter constraint graph that are bound to the current target correction item. The old parameter values stored in the nodes are replaced, parameter fingerprints and anchor point information are updated, and invalid conflict relationships are cleared.
[0110] Subsequently, based on the parameter constraint map before the update, a local verification is performed on the updated parameter constraint map to obtain a closed-loop verification result. This automatically verifies whether the automatic correction has taken effect and whether the parameters match consistently, ensuring the accuracy and validity of the correction result. Specifically, the local influence domain corresponding to the updated parameter constraint map is determined based on the device object nodes, parameter entity nodes, and constraint edges associated with the target correction item. For each local influence domain, based on the parameter constraint map and the updated parameter constraint map, the normalized parameters before and after the local influence domain are read, and the local constraint satisfaction before and after the write-back are calculated. The constraint satisfaction increment of the two is then calculated, and the number of newly added abnormal constraint edges in the updated local influence domain is counted. Only the data fragments corresponding to the local influence domain are re-parsed, rather than a full re-parse of all data sources, thereby improving the efficiency of the closed-loop verification. The constraint satisfaction increment is then compared with a preset closed-loop increment threshold, and the number of newly added abnormal constraint edges is compared with a preset newly added abnormal threshold. If both conditions are met, the closed-loop verification is considered successful; otherwise, the verification is considered unsuccessful, thus outputting an objective and standardized closed-loop verification result. For example, a local influence domain may include: data locations that have a synonymous equivalence relationship with the target parameter record; data locations that have a unit conversion relationship with the target parameter record; data locations that have a formula input / output relationship with the target parameter record; data locations that have a device topology subordinate relationship with the target parameter record; parameter nodes that have a standard value relationship with the target parameter record; and data version nodes that have a version inheritance relationship with the target parameter record.
[0111] If the verification is successful, the target data for the corrected error can be output. Simultaneously, a complete and verifiable correction report is generated based on the write-back difference record, enabling full traceability of all modification operations. For example, the correction report includes at least one of the following: conflict parameter group identifier, abnormal constraint edge identifier, abnormal constraint type, target device object, target parameter attribute, conflict parameter value set, conflict data source, data location anchor point, data version, hit constraint rule, constraint expression, candidate correction item, target correction item, correction score, parameter value before correction, parameter value after correction, write-back difference record, write-back transaction identifier, closed-loop verification result, rollback status, and confirmation record. This application can generate a correction report containing conflict location, conflict cause, hit constraint rule, parameters before correction, parameters after correction, write-back difference, closed-loop verification result, and rollback status, facilitating review, data auditing, and quality tracking.
[0112] If the verification fails, the pre-generated rollback handle is used to undo the parameter write-back with one click, restore the original target data, and mark the conflict parameter group corresponding to the target correction item as pending confirmation for review. The entire process forms a complete closed loop of data correction, map synchronization, automatic verification, successful report generation, and automatic rollback review for failure. It not only ensures the data consistency between multi-source data and constraint maps, but also automatically verifies the effectiveness of corrections and quickly rolls back to stop losses when deviations occur, taking into account automation, data accuracy, and fault tolerance.
[0113] Calculate the local influence domain constraint satisfaction before write-back and local influence domain constraint satisfaction after write-back The constraint satisfaction increment is determined by the following formula: ,(28) If the following conditions are met: , (29), and: If (30) is true, then the closed-loop check is considered successful. This represents the closed-loop incremental threshold. This indicates the number of newly added abnormal constraint edges. This indicates the upper limit on the number of newly allowed abnormal constraint edges. It should be noted that the various thresholds mentioned in this application are determined based on historical experience. If the closed-loop verification fails, the system restores the data object before the write-back based on the rollback handle and marks the target correction item as pending confirmation.
[0114] This application delineates the local influence domain corresponding to the target correction item, and only re-analyzes and re-verifies the local influence domain associated with the target correction item, which can reduce the overhead of full recalculation. Then, it objectively verifies whether the correction has effectively eliminated parameter conflicts and has not generated new anomalies by using dual thresholds of constraint satisfaction increment and the number of newly added abnormal constraint edges. When the closed-loop verification fails, it can restore the data object before writing back based on the rollback handle, reduce the risk of data disorder caused by erroneous correction, and mark the corresponding conflict parameter group as pending confirmation. At the same time, it automatically generates a standardized correction report based on the complete retained write-back difference record, which fully records the conflict location, conflict cause, parameters before and after correction, and the whole process evidence chain, realizing full-link traceability and reviewability of automatic correction of multi-source data.
[0115] In another embodiment, after adding a pending confirmation identifier to the conflict parameter group corresponding to the target correction item, the source credibility is dynamically updated based on the closed-loop verification results. When the number or proportion of times the same source acts as an anomalous contributor in multiple conflict parameter groups exceeds a preset threshold, the source credibility of that source in the current project is reduced; when parameters from the same source pass the closed-loop verification multiple times after being written back, the source credibility of that source in the current project is increased. The updated source credibility is then used for subsequent candidate correction item generation and target correction item determination.
[0116] In another embodiment, (a) the following is an example of multi-source parameter correction for main transformer capacity: Suppose the system obtains the following data sources: For the design instruction manual, For equipment technical agreements and For equipment ledgers. The system parsed the data and obtained: (31); (32); , (33).
[0117] The system in , and Establish synonymous equivalence constraint bundles and version inheritance constraint bundles between them. Because The parameter value is 40MVA and , The parameter value of 50MVA is inconsistent, and the system identifies abnormal constraint bundle edges.
[0118] The system generates candidate correction items: , (34).
[0119] After the system calculates the correction score, It has been identified as a target correction item.
[0120] If this parameter is configured as an automatic correction parameter in the current project, the system will write back 40MVA in the design specification as 50MVA and perform closed-loop verification; if the rated capacity is configured as a high-risk parameter, the system will generate a correction item to be confirmed.
[0121] (II) The following is an example of rated current correction based on formula-calculated constraints: Assume the system extracts the following from the calculation book: capacity ,Voltage and the recorded current System recognition formula: (35) Substitute into the calculation: (36) get: , The system determined that the deviation between the recorded current of 57.7A and the calculated result of 577.35A exceeded the tolerance threshold, and therefore identified the constraint edge calculated by the formula as an abnormal constraint edge.
[0122] The system generates candidate correction items: correcting I from 57.7A to 577.35A, marking the capacity S as a suspected anomaly, and marking the voltage U as a suspected anomaly.
[0123] If the reliability of both the capacity and voltage sources exceeds the threshold, the system will use the current correction term as the target correction term. Since the rated current may be a medium-risk parameter, the system can generate correction terms to be confirmed or perform automatic correction based on the risk gating strategy.
[0124] (III) The following is an example of voltage level correction based on topology dependency constraints: The system is extracted from a single system diagram as follows: (37) , (38).
[0125] The system extracts the following from the device parameter table: , (39).
[0126] According to the topology dependency constraint, Feeder101 is connected to Bus10kV_I, therefore the voltage level of Feeder101 should be consistent with the voltage level of Bus10kV_I.
[0127] Since Bus10kV_I has a voltage rating of 10kV, while Feeder101 is recorded as having a voltage rating of 6kV, the system determines that the topology dependency constraint boundary is abnormal and generates candidate correction terms: , (40).
[0128] The system determines whether to automatically write back or generate correction items to be confirmed based on the correction score and risk gate conditions.
[0129] (iv) The following is an example of parameter correction based on unit conversion constraints: Suppose that a structured dataset records the rated voltage of a device as follows: (41) The technical agreement states: (42) The system determines this based on unit conversion relationships: (43) Therefore, the system determines that the two parameter records are consistent after unit conversion and does not generate abnormal constraint edges. If the structured data records 1000V, while the technical protocol records 10kV, the units are inconsistent after conversion, the system generates abnormal constraint edges, and generates candidate correction terms based on high-reliability data sources.
[0130] (v) The following is an example of parameter correction based on version inheritance constraints: Assume there are two versions of the equipment technical agreement: old version B1 with a rated capacity of 40MVA; and new version B2 with a rated capacity of 50MVA, both of which are approved. The system determines that version B2 has a higher priority than B1 based on version inheritance. If the design specification still references the 40MVA from the old version, the system identifies an anomaly in the version inheritance constraint edge and generates a candidate corrective action to update the 40MVA in the design specification to 50MVA. If the design specification is editable and the risk level of this parameter meets the automatic corrective action criteria, the system performs a write-back; otherwise, a corrective action to be confirmed is generated.
[0131] (vi) The following is an example of dynamic updates to source credibility: The system dynamically adjusts the data source credibility based on the closed-loop verification results. If a data source frequently acts as an anomalous contributor in multiple conflicting parameter groups, the system reduces the source credibility of that data source in the current project. The data source credibility update can be represented as: (44) in, Indicates the reliability of the updated data source. This indicates the credibility of the data source before the update. Indicates the proportion of abnormal contributions from data sources. This indicates the proportion of data sources that pass closed-loop verification after source correction. Indicates the abnormality penalty coefficient. This indicates the reward coefficient. The updated source credibility is used for subsequent candidate correction term generation and target correction term determination.
[0132] In another embodiment, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the parameter constraint graph structure in an embodiment of this application. The parameter constraint graph includes equipment object nodes, parameter entity nodes, and data evidence nodes. These three types of nodes are associated with normalized parameter records through object merging, attribute mapping, and evidence anchoring, respectively, thereby binding equipment objects, parameter attributes, parameter values, and data evidence together. The normalized parameter records are matched with constraint templates to generate a set of constraint edges. The set of constraint edges covers six major categories of constraint relationships: synonymous equivalence constraints, unit conversion constraints, formula calculation constraints, topology dependency constraints, standard value constraints, and version inheritance constraints. Consistency and compliance verification of multi-source power parameters is carried out based on these various constraints.
[0133] In another embodiment, the multi-source data parsing and parameter normalization process includes: first, accessing multi-source data such as text, tables, formulas, figures, engineering drawings, and structured data; then, performing chapter recognition, table location, formula recognition, OCR, element parsing, or field path parsing according to the data source type to extract device objects, parameter attributes, original parameter values, original units, data locations, data versions, and parsing confidence; next, performing device object merging, parameter attribute mapping, unit conversion, and version identifier binding; finally, generating normalized parameter records and calculating source confidence and parameter anchor fingerprints.
[0134] The process for identifying abnormal constraint edges and extracting deviation subgraphs includes: First, reading the constraint expression, tolerance threshold, edge weight, risk identifier, and effective version range corresponding to the constraint edge; performing matching operations on synonymous equivalence, unit conversion, formula calculation, topological dependency, standard value, and version inheritance constraints; then calculating the constraint edge deviation value and constraint satisfaction; and identifying constraint edges whose deviation exceeds the tolerance threshold as abnormal constraint edges; subsequently, using the abnormal constraint edges as starting edges, performing a limited radius search along device objects, parameter entities, and data evidence nodes; clustering and removing normal edges, isolated evidence nodes, and non-effective version nodes that do not contribute directly; and finally obtaining the deviation subgraph.
[0135] The process for generating candidate corrective items and determining target corrective items includes: First, generating candidate values based on the parameter record with the highest source credibility, weighted majority consensus results, or high-priority version data. Simultaneously, forward calculations or reverse derivations are performed in conjunction with formula calculation constraints to generate candidate corrective items. Then, parameters with unit errors, topological inconsistencies, or deviations from standard levels are identified based on unit conversion, topological subordination, and standard value constraints. Next, candidate corrective items are constructed that include target parameter records, candidate parameter values, candidate value sources, evidence sets, risk levels, and rollback information. The corrective score of the candidate corrective items is calculated by comprehensively considering the expected increment of constraint satisfaction, source credibility, version priority, and write-back reliability. Finally, the candidate corrective item with the highest score that meets the threshold, risk gating, and new anomaly restrictions is determined as the target corrective item.
[0136] The parameter write-back, closed-loop verification, and rollback process includes: First, determining whether the risk level, source credibility, write-back reliability, and expected number of new anomalies of the target correction item meet the automatic correction conditions. If the conditions are met, verifying the write-back location based on the parameter anchor fingerprint, generating a pre-write-back backup, write-back transaction identifier, and rollback handle, and performing parameter write-back for text, tables, formulas, attached figures, engineering drawings, or structured fields, and generating write-back difference records; then, determining the local influence domain based on the device object nodes, parameter entity nodes, and constraint edges associated with the target correction item, re-parseing only the data fragments corresponding to the local influence domain, updating the normalized parameter records, data evidence nodes, and constraint edge status, and calculating the local constraint satisfaction and the number of new anomalies before and after write-back to determine whether the closed-loop verification passes; if the closed-loop verification passes, outputting correction data and a report; if the verification fails, performing data recovery based on the rollback handle and entering the secondary confirmation process.
[0137] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method described.
[0138] It should be noted that some embodiments of this application have been described above. In some cases, the actions or steps described in the above embodiments can be performed in a different order than that shown in the above embodiments and the desired result can still be achieved. In addition, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0139] Based on the same inventive concept, and corresponding to any of the above embodiments, this application also provides a multi-source data correction device based on parameter constraint maps.
[0140] refer to Figure 3 The multi-source data correction device based on parameter constraint maps includes: The preprocessing module 10 is configured to acquire a data set consisting of data from multiple sources and preprocess the data set to obtain multiple sets of normalized data.
[0141] The construction module 20 is configured to construct a parameter constraint graph based on the multiple sets of normalized data, wherein the parameter constraint graph contains multiple nodes generated by mapping each set of normalized data, and there is at least one constraint edge between any two nodes that are related.
[0142] The calculation module 30 is configured to perform deviation calculation on each set of normalized data based on the parameter constraint map, and obtain constraint data of at least one constraint edge corresponding to each set of normalized data, wherein each constraint data includes constraint edge deviation value and constraint satisfaction degree.
[0143] Extraction module 40 is configured to, for each constraint edge, in response to determining that the constraint edge deviation value corresponding to the constraint edge is greater than the tolerance threshold, extract a subgraph from the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge.
[0144] The determination module 50 is configured to determine at least one target correction item corresponding to the deviation subgraph based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph.
[0145] The update module 60 is configured to, for each target correction item, in response to determining that the target correction item meets the automatic correction conditions, perform a partial update on the data set based on the target correction item to obtain an updated data set.
[0146] The aforementioned device acquires a dataset composed of data from multiple sources and preprocesses it to obtain multiple sets of normalized data. This not only improves data quality but also fuses and merges heterogeneous parameters from multiple sources, allowing parameters associated with the same device, scattered across different data carriers, to form unified structured data. Based on these multiple sets of normalized data, a parameter constraint graph is constructed. This graph contains multiple nodes mapped from each set of normalized data, with at least one constraint edge between any two related nodes. This breaks down the limitations of fragmented original heterogeneous data and can reveal implicit parameter conflicts that are difficult to identify through text comparison and field matching. Based on the parameter constraint graph, deviation calculations are performed on each set of normalized data to obtain constraint data for at least one constraint edge corresponding to each set. Each constraint data includes a constraint edge deviation value and a constraint satisfaction degree. Deviation calculation is performed based on a parameter constraint graph, which comprehensively records all the associated constraint relationships between devices, parameters, and data nodes. This allows for rapid location of all constraint rules matching a single set of normalized data, enabling quantitative judgment of parameter conflicts and effectively avoiding the missed detections and misjudgments caused by traditional text comparison. For each constraint edge, in response to determining that the deviation value of the constraint edge is greater than the tolerance threshold, a subgraph is extracted from the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge. This subgraph visually displays the devices, parameters, data carriers, and constraint logic involved in the conflict, clearly locating the source of the conflict and related parameters. Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, at least one target correction item is determined for the deviation subgraph. The optimal correction scheme that is compliant and reasonable is automatically selected, eliminating the need for manual comparison of multiple sets of data to determine the modification direction and reducing the workload of manual analysis. For each target correction item, in response to determining that the target correction item meets the automatic correction conditions, the data set is partially updated based on the target correction item to obtain the updated data set.
[0147] In some embodiments, the dataset includes multiple types of data; the preprocessing module 10 is further configured to, for each type, perform data parsing on the data of that type in the dataset according to a predetermined parsing strategy corresponding to that type, to obtain multiple parsed data corresponding to that type; merge all the parsed data according to predetermined indicators to obtain multiple sets of merged data; and perform standardization integration processing on each set of merged data to obtain normalized data corresponding to each set of merged data.
[0148] In some embodiments, the calculation module 30 is further configured to, for each constraint edge corresponding to each set of normalized data, obtain the constraint type bound to the constraint edge based on the parameter constraint graph, and determine a predetermined constraint expression based on the constraint type; and determine the constraint data corresponding to the constraint edge based on the predetermined constraint expression.
[0149] In some embodiments, the extraction module 40 is further configured to use the constraint edge as the retrieval starting point, perform a predetermined radius neighborhood retrieval in the parameter constraint graph to obtain multiple nodes and multiple constraint edges corresponding to the constraint edge; cluster the multiple nodes and multiple constraint edges corresponding to the constraint edge to obtain multiple sets of conflict entity clusters; perform redundancy removal on the multiple sets of conflict entity clusters, and construct a deviation subgraph corresponding to the constraint edge based on the multiple sets of conflict entity clusters after redundancy removal.
[0150] In some embodiments, the determining module 50 is further configured to extract information from the deviation subgraph to obtain at least one conflict parameter group; for each conflict parameter group, search for at least one target predetermined business constraint rule corresponding to the conflict parameter group among the multiple predetermined business constraint rules; call the at least one target predetermined business constraint rule to generate at least one candidate correction item corresponding to each target predetermined business constraint rule; for each candidate correction item corresponding to the at least one target predetermined business constraint rule, obtain multi-dimensional evaluation index data corresponding to the candidate correction item; determine the correction score corresponding to the candidate correction item based on the multi-dimensional evaluation index data corresponding to the candidate correction item; based on the correction score of each candidate correction item corresponding to the target predetermined business constraint rule and the constraint satisfaction of the multiple constraint edges corresponding to the conflict parameter group, select the target correction item corresponding to the target predetermined business constraint rule from the at least one candidate correction item; and use all target correction items corresponding to the target predetermined business constraint rules as at least one target correction item corresponding to the deviation subgraph.
[0151] In some embodiments, the update module 60 is further configured to: search for target data corresponding to the target correction item in the data set; write back the target data based on the target correction item to obtain the updated data set; and write back the difference record and rollback handle.
[0152] In some embodiments, a verification module is further included. This verification module is configured to: after writing back the target data based on the target correction item to obtain an updated data set, preprocess the target data to obtain updated normalized parameter data; update the nodes and constraint edges in the parameter constraint graph corresponding to the target correction item based on the updated normalized parameter data to obtain the updated parameter constraint graph; perform local verification on the updated parameter constraint graph based on the parameter constraint graph to obtain a closed-loop verification result; in response to determining that the closed-loop verification result meets a preset pass condition, output the corrected target data and generate and output a correction report based on the write-back difference record; in response to determining that the closed-loop verification result does not meet the preset pass condition, restore the target data before parameter write-back based on the rollback handle, and add a pending confirmation identifier to the conflict parameter group corresponding to the target correction item.
[0153] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.
[0154] The apparatus described above is used to implement the corresponding multi-source data correction method based on parameter constraint maps in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0155] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the multi-source data correction method based on parameter constraint maps as described in any of the above embodiments.
[0156] Figure 4 This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.
[0157] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0158] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.
[0159] The input / output interface 1030 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.
[0160] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0161] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.
[0162] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.
[0163] The electronic devices described above are used to implement the corresponding multi-source data correction methods based on parameter constraint maps in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0164] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the multi-source data correction method based on parameter constraint maps as described in any of the above embodiments.
[0165] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.
[0166] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the multi-source data correction method based on parameter constraint maps as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0167] Based on the same concept, corresponding to the methods of any of the above embodiments, this application also provides a computer program product, including computer program instructions. When the computer program instructions are run on a computer, the computer causes the computer to execute the multi-source data correction method based on parameter constraint maps as described in any of the above embodiments, which has the beneficial effects of the corresponding method embodiments, and will not be repeated here.
[0168] It is understood that before using the technical solutions of the various embodiments in this disclosure, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.
[0169] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations of this disclosed technical solution.
[0170] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.
[0171] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.
[0172] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application is limited to these examples; under the concept of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in detail for the sake of brevity.
[0173] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.
[0174] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.
[0175] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.
Claims
1. A multi-source data correction method based on parameter-constrained maps, characterized in that, include: A data set consisting of data from multiple sources is obtained, and the data set is preprocessed to obtain multiple sets of normalized data; Based on the multiple sets of normalized data, a parameter constraint graph is constructed, wherein the parameter constraint graph contains multiple nodes generated by mapping each set of normalized data, and there is at least one constraint edge between any two nodes that are related. Based on the parameter constraint map, deviation calculation is performed on each set of normalized data to obtain constraint data for at least one constraint edge corresponding to each set of normalized data, wherein each constraint data includes constraint edge deviation value and constraint satisfaction degree. For each constraint edge, in response to determining that the constraint edge deviation value corresponding to the constraint edge is greater than the tolerance threshold, a subgraph extraction is performed on the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge; Based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph, at least one target correction item corresponding to the deviation subgraph is determined. For each target correction item, in response to determining that the target correction item meets the automatic correction conditions, the data set is locally updated based on the target correction item to obtain the updated data set.
2. The method according to claim 1, characterized in that, The dataset includes various types of data; The data set is preprocessed to obtain multiple sets of normalized data, including: For each type, according to the predetermined parsing strategy corresponding to the type, the data of the type in the dataset is parsed to obtain multiple parsed data corresponding to the type; All the analyzed data are merged according to predetermined indicators to obtain multiple sets of merged data. Each group of merged data is standardized and integrated to obtain the normalized data corresponding to each group of merged data.
3. The method according to claim 1, characterized in that, Based on the parameter constraint graph, deviation calculation is performed on each set of normalized data to obtain constraint data for at least one constraint edge corresponding to each set of normalized data, including: For each constraint edge corresponding to each set of normalized data, based on the parameter constraint graph, the constraint type bound to the constraint edge is obtained, and based on the constraint type, a predetermined constraint expression is determined. Based on the predetermined constraint expression, the constraint data corresponding to the constraint edge is determined.
4. The method according to claim 1, characterized in that, The step of extracting a subgraph from the parameter constraint graph based on the constraint edges to obtain the deviation subgraph corresponding to the constraint edges includes: Using the constraint edge as the starting point, a neighborhood search with a predetermined radius is performed in the parameter constraint graph to obtain multiple nodes and multiple constraint edges corresponding to the constraint edge; Clustering is performed on multiple nodes and multiple constraint edges corresponding to the constraint edges to obtain multiple groups of conflict entity clusters; Redundancy is eliminated from the multiple conflicting entity clusters, and a deviation subgraph corresponding to the constraint edge is constructed based on the multiple conflicting entity clusters after redundancy elimination.
5. The method according to claim 1, characterized in that, The step of determining at least one target correction term corresponding to the deviation subgraph based on the constraint satisfaction of multiple predetermined business constraint rules and the multiple constraint edges corresponding to the deviation subgraph includes: Information is extracted from the deviation subgraph to obtain at least one set of conflict parameters; For each conflict parameter group, at least one target predetermined business constraint rule corresponding to the conflict parameter group is searched among the multiple predetermined business constraint rules. Invoke the at least one target pre-defined business constraint rule to generate at least one candidate correction item corresponding to each target pre-defined business constraint rule; For each target-based business constraint rule, at least one candidate correction item is obtained, and multi-dimensional evaluation index data corresponding to the candidate correction item is obtained; based on the multi-dimensional evaluation index data corresponding to the candidate correction item, the correction score corresponding to the candidate correction item is determined. Based on the correction score of each candidate correction item corresponding to the target predetermined business constraint rule and the constraint satisfaction of multiple constraint edges corresponding to the conflict parameter group, the target correction item corresponding to the target predetermined business constraint rule is selected from the at least one candidate correction item. The target correction items corresponding to all target pre-defined business constraint rules are used as at least one target correction item corresponding to the deviation subgraph.
6. The method according to claim 1, characterized in that, The step of partially updating the data set based on the target correction term to obtain the updated data set includes: In the dataset, find the target data corresponding to the target correction item; Based on the target correction term, the target data is written back with parameters to obtain the updated data set, and the difference record and rollback handle are written back.
7. The method according to claim 6, characterized in that, After writing back the parameters of the target data based on the target correction term to obtain the updated data set, the method includes: The target data is preprocessed to obtain updated normalized parameter data; Based on the updated normalized parameter data, the nodes and constraint edges corresponding to the target correction term in the parameter constraint graph are updated to obtain the updated parameter constraint graph. Based on the parameter constraint map, a local verification is performed on the updated parameter constraint map to obtain the closed-loop verification result. In response to determining that the closed-loop verification result meets the preset pass conditions, the corrected target data is output, and a correction report is generated and output based on the write-back difference record; In response to determining that the closed-loop verification result does not meet the preset pass condition, the target data before the parameter write-back is restored based on the rollback handle, and the conflict parameter group corresponding to the target correction item is marked as pending confirmation.
8. A multi-source data correction device based on parameter constraint maps, characterized in that, include: The preprocessing module is configured to acquire a dataset consisting of data from multiple sources and preprocess the dataset to obtain multiple sets of normalized data. The construction module is configured to construct a parameter constraint graph based on the multiple sets of normalized data, wherein the parameter constraint graph contains multiple nodes generated by mapping each set of normalized data, and there is at least one constraint edge between any two nodes that are related. The calculation module is configured to perform deviation calculation on each set of normalized data based on the parameter constraint graph to obtain constraint data of at least one constraint edge corresponding to each set of normalized data, wherein each constraint data includes constraint edge deviation value and constraint satisfaction degree. The extraction module is configured to, for each constraint edge, in response to determining that the constraint edge deviation value corresponding to the constraint edge is greater than the tolerance threshold, extract a subgraph from the parameter constraint graph based on the constraint edge to obtain the deviation subgraph corresponding to the constraint edge. The determination module is configured to determine at least one target correction item corresponding to the deviation subgraph based on multiple predetermined business constraint rules and the constraint satisfaction of multiple constraint edges corresponding to the deviation subgraph. The update module is configured to, for each target correction item, in response to determining that the target correction item meets the automatic correction conditions, perform a partial update on the data set based on the target correction item to obtain the updated data set.
9. The apparatus according to claim 8, characterized in that, The dataset includes multiple data types; the preprocessing module is further configured to: For each type, according to the predetermined parsing strategy corresponding to the type, the data of the type in the dataset is parsed to obtain multiple parsed data corresponding to the type; All the analyzed data are merged according to predetermined indicators to obtain multiple sets of merged data. Each group of merged data is standardized and integrated to obtain the normalized data corresponding to each group of merged data.
10. The apparatus according to claim 8, characterized in that, The computing module is further configured as follows: For each constraint edge corresponding to each set of normalized data, based on the parameter constraint graph, the constraint type bound to the constraint edge is obtained, and based on the constraint type, a predetermined constraint expression is determined. Based on the predetermined constraint expression, the constraint data corresponding to the constraint edge is determined.
11. The apparatus according to claim 8, characterized in that, The extraction module is further configured to: Using the constraint edge as the starting point, a neighborhood search with a predetermined radius is performed in the parameter constraint graph to obtain multiple nodes and multiple constraint edges corresponding to the constraint edge; Clustering is performed on multiple nodes and multiple constraint edges corresponding to the constraint edges to obtain multiple groups of conflict entity clusters; Redundancy is eliminated from the multiple conflicting entity clusters, and a deviation subgraph corresponding to the constraint edge is constructed based on the multiple conflicting entity clusters after redundancy elimination.
12. The apparatus according to claim 8, characterized in that, The determining module is further configured to: Information is extracted from the deviation subgraph to obtain at least one set of conflict parameters; For each conflict parameter group, at least one target predetermined business constraint rule corresponding to the conflict parameter group is searched among the multiple predetermined business constraint rules. Invoke the at least one target pre-defined business constraint rule to generate at least one candidate correction item corresponding to each target pre-defined business constraint rule; For each target pre-defined business constraint rule, at least one candidate correction item is obtained, and multi-dimensional evaluation index data corresponding to the candidate correction item is obtained. Based on the multi-dimensional evaluation index data corresponding to the candidate correction items, the correction score corresponding to the candidate correction items is determined. Based on the correction score of each candidate correction item corresponding to the target predetermined business constraint rule and the constraint satisfaction of multiple constraint edges corresponding to the conflict parameter group, the target correction item corresponding to the target predetermined business constraint rule is selected from the at least one candidate correction item. The target correction items corresponding to all target pre-defined business constraint rules are used as at least one target correction item corresponding to the deviation subgraph.
13. The apparatus according to claim 8, characterized in that, The update module is also configured to: In the dataset, find the target data corresponding to the target correction item; Based on the target correction term, the target data is written back with parameters to obtain the updated data set, and the difference record and rollback handle are written back.
14. The apparatus according to claim 13, characterized in that, After writing back the target data based on the target correction term to obtain the updated data set, the device further includes a verification module, which is configured as follows: The target data is preprocessed to obtain updated normalized parameter data; Based on the updated normalized parameter data, the nodes and constraint edges corresponding to the target correction term in the parameter constraint graph are updated to obtain the updated parameter constraint graph. Based on the parameter constraint map, a local verification is performed on the updated parameter constraint map to obtain the closed-loop verification result. In response to determining that the closed-loop verification result meets the preset pass conditions, the corrected target data is output, and a correction report is generated and output based on the write-back difference record; In response to determining that the closed-loop verification result does not meet the preset pass condition, the target data before the parameter write-back is restored based on the rollback handle, and the conflict parameter group corresponding to the target correction item is marked as pending confirmation.
15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.
16. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method described in any one of claims 1 to 7.
