A data consistency verification generation method and system for a power generation operation report

Abstract

The application discloses a kind of power generation operation report data consistency verification generation method and system, belong to intelligent report generation technical field.This method receives report template file and report generation instruction, parses report template file to construct template semantic atlas;Template semantic atlas, power generation operation index ontology and power generation operation data traceability graph are aligned, and three alignment records and alignment conflict identifiers are generated;According to three alignment records, compile report generation slot and executable check rule item, calculate candidate index correlation score and select target index, generate standard query object and obtain fact data package;Recall verification data package and generate report content, convert the conclusion to be verified into assertion triple, execute fact forward verification, traceability reverse verification and counterfactual disturbance verification, and output report or generate rollback instruction according to verification result.The application can improve the reliability and controllability of report result.

Description

Technical Field

[0001] This invention belongs to the field of intelligent report generation technology, and in particular relates to a data consistency verification and generation method and system for power generation operation reports. Background Technology

[0002] Power generation operation reports are typically used to record the operating status of power generation equipment, compile operational data, and form operational analysis conclusions. They serve as crucial evidence for dispatch management, operational review, equipment maintenance, and business analysis. With the advancement of power information technology and intelligentization, existing technologies have developed report generation methods based on report template parsing, database queries, and automatic text generation. These methods can automatically acquire relevant operational data and fill it into report templates according to user-specified conditions such as time, region, station, or equipment, thereby reducing the workload of manual data entry.

[0003] However, power generation operation reports involve numerous data sources, with complex indicator names, statistical periods, units of measurement, equipment scope, and statistical definitions. In existing automated report generation processes, systems typically focus more on whether the report can be generated successfully and whether the data can be filled into the template. However, they often lack effective process control over whether the entered data matches the template description, whether there are differences in definitions between data from different sources, and whether the generated conclusions are consistent with the actual data.

[0004] Therefore, when existing technologies are applied to generate power generation operation reports, situations may still arise where the report format is correct but the content contains hidden deviations. For example, the meanings of entered data may differ from those in the template fields, statistical definitions may not match the report descriptions, or the analysis conclusions may not match the underlying operational data. Such problems are not easily detected through simple format checks or quick manual browsing, which can reduce the credibility of the report data, affect the accuracy of operational analysis results, and adversely impact dispatch management and operational decision-making. Summary of the Invention

[0005] The purpose of this invention is to provide a data consistency verification generation method and system for power generation operation reports, in order to solve the problems that easily occur in the existing automatic generation process of power generation operation reports, such as mismatch between the entered data and the report content, inconsistency between the report conclusions and the actual operation data, and the difficulty in timely detection and handling, thereby improving the data reliability, conclusion accuracy and controllability of the automatic generation results of power generation operation reports.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] A method for generating data consistency verification of power generation operation reports includes the following steps:

[0008] S1. Receive the report template file and report generation instructions, parse the report template file to construct a template semantic graph, and determine the template content objects;

[0009] S2. Align the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph to generate a ternary alignment record and alignment conflict identifier;

[0010] S3. Based on the triple alignment record, the template content object is compiled into a report generation slot and an executable verification rule item. The executable verification rule item is used to constrain the generation of standard query objects, the recall of verification data packets, the verification of assertion triples, the verification of counterfactual disturbances and the rollback process, and to control the execution of the corresponding executable verification rule item based on the alignment conflict identifier.

[0011] S4. For each report, generate slots, calculate the candidate indicator association score based on the ternary alignment record, alignment conflict identifier and executable verification rule item, select the target indicator based on the candidate indicator association score and redundancy constraints, generate standard query objects and obtain fact data packets;

[0012] S5. Based on the executable verification rule item, recall the verification data packet and generate the report content. Convert the conclusion to be verified in the report content into an assertion triple. Perform fact forward verification, source tracing reverse verification and antifactual perturbation verification on the assertion triple.

[0013] S6. When the results of the fact forward verification, source tracing reverse verification, and antifactual disturbance verification do not meet the preset verification pass conditions, a rollback instruction is generated according to the corresponding mismatch type; when the preset verification pass conditions are met, the power generation operation report and the mapping relationship between the report content and the verification data are rendered and output.

[0014] Preferably, the parsing report template file for constructing a template semantic graph includes:

[0015] The report template file's titles, tables, field names, statistical periods, equipment objects, and indicator descriptions are structured and parsed to generate content nodes and layout nodes.

[0016] Generate semantically defined edges based on the hierarchical relationships between content nodes, header constraints, indicator definitions, and time dimensions;

[0017] Starting from the content node and ending at the report generation slot, candidate slot edges are generated. Each candidate slot edge records the slot type, dimension placeholder identifier, table header constraint chain, and candidate indicator caliber.

[0018] When the same content node corresponds to multiple slot candidate edges, the main slot candidate edge is determined based on the semantically limited edge, the table header limited chain, and the historical template mapping identifier.

[0019] A slot constraint summary is generated based on content nodes, layout nodes, semantically limited edges, and primary slot candidate edges. The slot constraint summary includes slot semantic signature, consistency verification benchmark item candidate, dimension constraint candidate, counterfactual perturbation dimension candidate, and fallback position identifier.

[0020] Preferably, aligning the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph to generate a ternary alignment record and alignment conflict identifier includes:

[0021] A slot semantic signature is generated based on the slot constraint summary. The slot semantic signature includes the indicator name, statistical object, time granularity, unit of measurement, aggregation method, and business scope.

[0022] Extract the indicator ontology signature of candidate indicators from the power generation operation indicator ontology, and extract the data traceability signature of operation data fields from the power generation operation data traceability diagram;

[0023] Perform consistency matching on slot semantic signature, indicator ontology signature and data traceability signature to generate a ternary aligned record including template content object, candidate indicator, running data field, matching benchmark item and traceability path;

[0024] When multiple ternary alignment records correspond to the same template content object and there are inconsistencies in indicator caliber, unit of measurement, time granularity, equipment object, or traceability field, an alignment conflict identifier is generated.

[0025] Preferably, the executable verification rule item is compiled from a ternary alignment record, a slot constraint summary, and an alignment conflict identifier;

[0026] The executable verification rule items include query constraints, verification data packet recall constraints, assertion triple verification constraints, counterfactual perturbation constraints, and fallback constraints;

[0027] The query constraints are used to limit the data source, statistical period, device object, indicator scope, and aggregation method in the standard query object;

[0028] The verification data packet recall constraint is used to limit the recall fields, recall sources, and recall priorities of the verification data packets;

[0029] The assertion triple check constraint is used to limit the subject, predicate, object and their consistency check benchmark items in the assertion triple;

[0030] The counterfactual perturbation constraint is used to limit the permissible perturbation dimensions, including the time dimension, device dimension, indicator dimension, or aggregation dimension, as well as the allowed perturbation range.

[0031] The rollback constraint is used to establish the correspondence between mismatch types and report template file re-parsing, data re-querying, indicator re-binding, and content re-generation;

[0032] When there are unresolved alignment conflict flags, the corresponding executable validation rule item is set to the pending parsing state, and the executable validation rule item in the pending parsing state is restricted from participating in standard query object generation, validation data retrieval, and assertion triple validation.

[0033] Preferably, the calculation of the candidate indicator correlation score includes calculation according to the following formula:

[0034] ;

[0035] In the formula: Indicates the report generation slot; Indicates candidate indicators; Indicate candidate indicators Relative to report generation slot The correlation score, Indicates the report generation slot With candidate indicators The corresponding set of unresolved alignment conflict identifiers; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true and takes the value 0 when the condition inside the parentheses is false. This indicates that the associated report generation slot will be used simultaneously. and candidate indicators The set of candidate alignment evidence; Indicates the report generation slot The corresponding set of consistency verification benchmark items; Indicates the report generation slot The corresponding set of query constraint predicates; Indicates candidate alignment evidence For consistency verification benchmark items The coverage matching degree, with a value range of 100%. ; Indicates candidate alignment evidence query constraint predicates The value that satisfies the judgment condition is either 0 or 1.

[0036] when When it is an empty set, and The value of is 0; when When it is an empty set, The value of is 0; when When it is an empty set, The value of is 1; Candidate indicators are filtered out.

[0037] Preferably, the selection of the target indicator based on the candidate indicator correlation score and redundancy constraints includes:

[0038] Will The candidate metrics were determined as a set of bindable candidate metrics. Solve the following objective function:

[0039] ;

[0040] In the formula: Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; This represents the non-negative dimensionless redundancy penalty coefficient; This represents a set of bindable candidate metrics. An unordered set of candidate indicator pairs consisting of two different candidate indicators; Indicate candidate indicators With candidate indicators The redundancy strength between them, with a value range of 100%. ;

[0041] The constraints of the objective function include: the selected candidate metrics jointly cover the necessary dimensions of the report generation slot; the selected candidate metrics can generate standard query objects that pass the query validity check; and candidate metrics with mutual exclusion relationships are not selected simultaneously.

[0042] When the objective function is maximized The candidate indicators were selected as the target indicators.

[0043] Preferably, the forward fact verification, reverse source tracing verification, and counterfactual perturbation verification include:

[0044] The conclusion to be verified is converted into an assertion triple including subject, predicate and object, and the assertion triple is matched with the indicator value, statistical period, device object and unit of measurement in the fact data packet to generate fact mismatch items;

[0045] Based on the ternary alignment record and the power generation operation data traceability diagram, reverse tracing is performed along the data source fields, calculation links and caliber conversion links corresponding to the fact data packets to generate traceability mismatch items;

[0046] Adjust the target perturbation dimension in the standard query object according to the counterfactual perturbation constraint, generate a counterfactual query object, obtain a counterfactual data packet based on the counterfactual query object, and regenerate the counterfactual conclusion based on the counterfactual data packet;

[0047] Based on the data changes between fact data packets and counterfactual data packets, and the conclusion changes between the conclusion to be verified and the counterfactual conclusion, a counterfactual mismatch item is generated;

[0048] The mismatch type is determined based on the generation of fact mismatch items, source mismatch items, and counterfactual mismatch items. A data re-query instruction is generated when a fact mismatch item is generated, an indicator rebinding instruction is generated when a source mismatch item is generated, and a content regeneration instruction is generated when a counterfactual mismatch item is generated. When two or more mismatch items are generated simultaneously, a rollback instruction corresponding to each mismatch item is generated.

[0049] Preferably, the assertion triple check includes calculating the consistency score of the assertion triple according to the following formula:

[0050] ;

[0051] In the formula: This indicates an assertion triple; Represents a data segment; Indicates assertion triples Consistency score; Expressing support for asserting triplets Supports collections of data fragments; Representation and assertion triples A set of contradictory data fragments; Representing data fragments With assertion triplet The degree of matching between the benchmark items; Representing data fragments Source quality score; Indicates assertion triples Numerical consistency factor; This indicates the necessary benchmark term verification gating factor;

[0052] , , , The range of values ​​for the preset assertion threshold is... ; The value of is 0 or 1, and the value of the empty product is 1; when asserting a triple. When all the necessary benchmark items reach the matching threshold, When asserting the triplet When there are necessary benchmark items among the corresponding necessary benchmark items that have not reached the matching threshold, ;

[0053] when Below the preset assertion threshold, or At that time, a mismatch term for the assertion triple is generated.

[0054] Preferably, the process of generating and outputting the mapping relationship between the report content and the verification data includes:

[0055] The recall fields, recall sources, and recall priorities of the verification data packets are determined according to the recall constraints of the verification data packets.

[0056] Based on target indicators, standard query objects, and ternary aligned records, data fragments containing indicator names, indicator values, statistical periods, equipment objects, units of measurement, source fields, and traceability paths are retrieved from the data sources pointed to by the power generation operation data traceability map to form a verification data package;

[0057] The fact data packet and the verification data packet are associated according to the report generation slot to generate a slot-level fact data and verification data mapping record;

[0058] Based on the report, generate slots, target indicators, fact data packages, and slot-level fact data and verification data mapping records; perform data backfilling or content generation for statistical text, tabular data, and analytical text.

[0059] When outputting the power generation operation report, the report content fragments corresponding to the generated slot, the data fragments in the verification data packet, the mapping relationship between the ternary alignment record and the verification result are output synchronously.

[0060] This invention also provides a data consistency verification and generation system for power generation operation reports, applied to the data consistency verification and generation method for power generation operation reports as described above, comprising:

[0061] The template parsing module is used to receive the report template file and the report generation instruction, parse the report template file to construct a template semantic graph, and determine the template content object;

[0062] The three-graph alignment module is used to align the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph, and generate a three-element alignment record and alignment conflict identifier;

[0063] The rule compilation module is used to compile template content objects into report generation slots and executable verification rule items based on the triple alignment records. The executable verification rule items are used to constrain the generation of standard query objects, the recall of verification data packets, the verification of assertion triples, the verification of counterfactual disturbances, and the rollback process, and control the execution of the corresponding executable verification rule items based on the alignment conflict identifier.

[0064] The indicator binding module is used to generate slots for each report, calculate the candidate indicator association score based on the ternary alignment record, alignment conflict identifier and executable verification rule item, select the target indicator based on the candidate indicator association score and redundancy constraints, generate standard query objects and obtain fact data packets;

[0065] The report generation and verification module is used to recall verification data packets based on executable verification rule items and generate report content. It converts the conclusions to be verified in the report content into assertion triples and performs forward fact verification, reverse source verification, and antifactual perturbation verification on the assertion triples.

[0066] The rollback output module is used to generate rollback instructions based on the corresponding mismatch type when the results obtained from the fact forward verification, source tracing reverse verification, and antifactual disturbance verification do not meet the preset verification pass conditions; when the preset verification pass conditions are met, it renders and outputs the power generation operation report and the mapping relationship between the report content and the verification data.

[0067] The beneficial effects of this invention are as follows: This invention constructs a template semantic graph by parsing the report template file, and aligns the template semantic graph, the power generation operation indicator ontology, and the power generation operation data traceability graph to generate ternary alignment records and alignment conflict identifiers, thus establishing a verifiable correspondence between the content objects, candidate indicators, and underlying operation data fields in the report template. Simultaneously, it compiles report generation slots and executable verification rule items based on the ternary alignment records, providing unified constraints on standard query object generation, verification data packet recall, assertion triplet verification, counterfactual disturbance verification, and rollback processing. This invention solves the problem that existing automatic power generation operation report generation processes only focus on template filling and text generation, lacking consistency control over the meaning of template fields, indicator definitions, and data sources, reducing report content deviations caused by mis-binding of indicators, incorrect statistical periods, inconsistent units of measurement, or mismatched data sources.

[0068] This invention further calculates the correlation scores of candidate indicators for each report generation slot, selects target indicators based on redundancy constraints, generates standard query objects, and obtains fact data packages. After the report content is generated, the conclusions to be verified are converted into assertion triples, and forward fact verification, reverse source tracing verification, and antifactual perturbation verification are performed. If the verification fails, a rollback instruction is generated based on the mismatch type. If the verification passes, a power generation operation report and the mapping relationship between the report content and the verification data are output. This invention can solve the problem in the prior art where the report format is correct but there are implicit inconsistencies in data, sources, and conclusions that are not easy to detect. It enables the system to automatically detect anomalies such as fact data mismatch, inconsistent data tracing, or insufficient conclusion stability, and perform targeted rollback processing, thereby improving the data reliability, conclusion accuracy, source traceability, and processing controllability of the automatically generated power generation operation report. Attached Figure Description

[0069] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method in an embodiment of the present invention; Figure 2 This is a schematic diagram of the three-class verification process for assertion triples in an embodiment of the present invention; Figure 3 This is a schematic diagram of the rollback process based on mismatch type in an embodiment of the present invention; Figure 4 This is a schematic diagram of the modular structure of the system in an embodiment of the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0071] like Figure 1 As shown, this embodiment provides a data consistency verification method for generating power generation operation reports. This method can be executed by a server, terminal processor, or computing device deployed with report generation services. This method is used to constrain and verify the consistency between the report template, operating indicators, data sources, and report conclusions during the generation of power generation operation reports. The method includes the following steps S1 to S6.

[0072] S1. Receive the report template file and report generation instructions, parse the report template file to construct a template semantic graph, and determine the template content objects;

[0073] In this embodiment, the input to S1 includes a report template file and a report generation instruction. The report template file can be a Word document, a table document, or other parsable report template file; the report generation instruction can be a natural language instruction or a parameterized instruction, used to indicate the report type, statistical time range, statistical object, or device object.

[0074] The server or terminal processor invokes the template parsing module to perform structured parsing of the titles, tables, field names, statistical periods, device objects, and indicator descriptions in the report template file, and extracts the corresponding location and layout information. For titles and body paragraphs, it reads the text content, paragraph order, title level, and location index; for tables, it reads the table number, row and column coordinates, merged cell range, horizontal header, vertical header, data cells, and remarks; for layout information, it reads the font, font size, alignment, borders, row height, column width, and cell merging rules. After parsing, text units, table units, or field units carrying business meaning are generated as content nodes, and information expressing location, level, format, and layout constraints is generated as layout nodes.

[0075] Each content node includes a node identifier, content type, original text, standardized text, location index, heading level, table identifier, row and column coordinates, and adjacent context information. Each layout node includes a layout node identifier, corresponding content node identifier, font attributes, paragraph attributes, table attributes, cell attributes, and physical location attributes. Content nodes are used to record the business semantics of the template, while layout nodes are used to record the format restoration information required for subsequent report rendering.

[0076] After generating content nodes and layout nodes, the template parsing module generates semantically limiting edges based on the hierarchical relationships, header constraints, indicator definitions, and time dimension relationships between content nodes. Specifically, hierarchical relationships are determined by title level, paragraph order, and position index; header constraints are determined by horizontal headers, vertical headers, merged cell ranges, and data cell coordinates; indicator definitions are determined by field names, indicator descriptions, units of measurement, aggregate terms, and keywords related to power generation operations; and time dimension relationships are determined by report generation instructions, template titles, time expressions in the text, and periodic information in table fields.

[0077] In one implementation, the template parsing module generates semantically constrained edges using a combination of rule parsing, dictionary matching, and semantic recognition. Rule parsing identifies title levels, table coordinates, merged cells, and paragraph order; dictionary matching identifies indicator names, equipment names, statistical periods, and units of measurement in the power generation operation scenario; and semantic recognition normalizes indicator descriptions, equipment objects, and time ranges in natural language expressions. If a semantic recognition model is used, its inputs are standardized text of content nodes, adjacent title text, table header chain text, and report generation instructions, and its outputs are content type labels, candidate indicator descriptions, candidate equipment objects, candidate time dimensions, and corresponding confidence scores. When the confidence score is lower than a preset confidence threshold, the corresponding content node is marked as an object to be verified. The semantic recognition model can be a deployed general language model or a semantic recognition model in the power generation operation domain; when using a semantic recognition model in the power generation operation domain, labeled report template text, table header chain text, and their corresponding slot types, indicator descriptions, equipment objects, and time dimensions can be used as training samples for supervised training.

[0078] Furthermore, the template parsing module generates candidate slot edges starting from content nodes and ending at report generation slots. Report generation slots represent locations in the report template file where data backfilling, text generation, or conclusion verification is required. Candidate slot edges record the slot type, dimension placeholder identifier, header constraint chain, and candidate indicator caliber. Specifically, the slot type distinguishes between statistical text slots, tabular data slots, analytical text slots, or title variable slots; the dimension placeholder identifier records time, device, region, or statistical object dimensions; the header constraint chain records the constraint relationships formed by horizontal headers, vertical headers, parent titles, or adjacent fields; and the candidate indicator caliber records the possible corresponding indicator name, unit of measurement, statistical method, and business meaning. The candidate indicator caliber is only a candidate result in the template parsing stage; the final bound indicator is determined in subsequent steps by combining ternary alignment records, alignment conflict identifiers, and executable verification rules.

[0079] When the same content node corresponds to multiple slot candidate edges, the template parsing module determines the primary slot candidate edge based on semantically limited edges, table header limited chains, and historical template mapping identifiers. Historical template mapping identifiers are derived from manually confirmed mapping results or system-confirmed mapping results saved in historical report templates, including template identifier, content node identifier, report generation slot identifier, confirmation time, and confirmation source. The priority for determining the primary slot candidate edge is as follows: historical template mapping identifier, table header limited chain completeness, number of semantically limited edges, and semantic recognition confidence. If a unique primary slot candidate edge still cannot be determined, the corresponding content node is marked as having a candidate conflict and retained for processing during subsequent three-graph alignment and verification rule compilation.

[0080] After determining the candidate edges for the main slot, the template parsing module generates template content objects based on content nodes, layout nodes, semantically defined edges, and candidate edges for the main slot. Each template content object includes a template content object identifier, associated content node identifiers, associated layout node identifiers, candidate edge identifiers for the main slot, the report chapter to which it belongs, the table or paragraph to which it belongs, the original content, standardized content, location coordinates, and layout restoration parameters.

[0081] The template parsing module generates a slot constraint summary based on the template content object. The slot constraint summary includes a slot semantic signature, consistency verification benchmark candidates, dimension constraint candidates, counterfactual disturbance dimension candidates, and a rollback position identifier. The slot semantic signature includes candidate metric names, candidate statistical objects, candidate time granularity, candidate units of measurement, candidate aggregation methods, and candidate business definitions. The consistency verification benchmark candidates record the metric name, unit of measurement, statistical period, device object, data source, and aggregation method that need to be verified later. The dimension constraint candidates record the time dimension, device dimension, region dimension, and business dimension. The counterfactual disturbance dimension candidates record the candidate dimensions that can be used for counterfactual disturbance verification later. The rollback position identifier records the original template position, content node, report generation slot, or template content object to which one can roll back in case of subsequent verification failure. The counterfactual disturbance dimension candidates are only generated as candidate constraint information in S1 and are not used for counterfactual disturbance verification in S1.

[0082] S2. Align the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph to generate a ternary alignment record and alignment conflict identifier;

[0083] In this embodiment, S2 is used to establish a correspondence between the template semantic graph and the power generation operation index ontology and the power generation operation data traceability graph. The inputs of S2 include the template semantic graph, the template content object, the slot constraint summary, the power generation operation index ontology, and the power generation operation data traceability graph; the outputs include ternary alignment records and alignment conflict identifiers.

[0084] The power generation operation indicator ontology describes the indicator system in a power generation operation scenario. The ontology can include fields such as indicator identifier, indicator name, indicator alias, indicator category, statistical object, time granularity, unit of measurement, aggregation method, business scope, applicable equipment type, indicator calculation rules, and indicator mutual exclusion relationships. The power generation operation indicator ontology can be constructed from a manually configured indicator library, an indicator dictionary in the business system, or confirmed indicator mapping relationships in historical reports.

[0085] The power generation operation data traceability diagram is used to describe the data sources and field paths involved in the data acquisition for indicators. The diagram can include data source nodes, data table nodes, operation data field nodes, calculation rule nodes, and caliber conversion nodes. These nodes are connected via source edges, calculation edges, mapping edges, or conversion edges. Operation data fields can include information such as field name, field meaning, data source, data table, unit of measurement, acquisition cycle, equipment object, queryable time range, and field quality identifier.

[0086] Specifically, the system generates a slot semantic signature based on the slot constraint summary. The slot semantic signature is used to express the business meaning of the report generation slot in the template, including the indicator name, statistical object, time granularity, unit of measurement, aggregation method, and business scope. The above fields are derived from the slot semantic signature candidates, dimension constraint candidates, table header constraint chain, candidate indicator scope, and report generation instructions formed by S1.

[0087] The system extracts the indicator ontology signatures of candidate indicators from the power generation operation indicator ontology. The indicator ontology signature expresses the standard meaning of the candidate indicator within the indicator system, and its fields correspond to the slot semantic signature, including indicator name, indicator alias, applicable scope of the statistical object, time granularity, unit of measurement, aggregation method, and business scope. The system can also standardize indicator names and aliases, including removing meaningless symbols, unifying synonyms, and standardizing unit of measurement.

[0088] The system extracts data traceability signatures for operational data fields from the power generation operation data traceability diagram. These signatures express the meaning and scope of operational data fields within the data source, including the operational data field, its meaning, the data source it belongs to, the equipment object, the time granularity, the unit of measurement, the aggregation method, the data source path, and the conversion rules. If a certain indicator needs to be calculated from multiple fields, the data traceability signature also includes the calculation path.

[0089] After obtaining the slot semantic signature, indicator ontology signature, and data traceability signature, the system performs consistency matching on these three types of signatures. This consistency matching is achieved through a combination of field-by-field matching, synonym matching, unit conversion matching, and caliber rule matching. Field-by-field matching compares whether the indicator name, statistical object, time granularity, unit of measurement, aggregation method, and business caliber are consistent; synonym matching handles indicator names with different expressions but the same meaning; unit conversion matching handles convertible units; and caliber rule matching determines whether the statistical method and calculation rules of the candidate indicator meet the business caliber requirements of the report generation slot. For each matching item, the system generates a matching benchmark item to record the matching result and matching basis for that item.

[0090] In one implementation, the system sets matching thresholds for each matching benchmark item. The indicator name and business scope can be set as required matching items, while the statistical object, time granularity, unit of measurement, and aggregation method can be set as verifiable matching items. If a required matching item does not meet the preset conditions, a valid ternary alignment record is not generated. If verifiable matching items have differences but can be eliminated through unit conversion, time granularity aggregation, or scope conversion rules, a ternary alignment record is generated and the corresponding conversion basis is recorded. If the differences cannot be eliminated, an alignment conflict flag is generated.

[0091] When the slot semantic signature, indicator ontology signature, and data traceability signature meet the preset consistency matching conditions, the system generates a ternary alignment record. Each ternary alignment record includes a template content object, a candidate indicator, an operational data field, a matching benchmark item, and a traceability path. The template content object indicates the location in the report template to be generated, backfilled, or verified; the candidate indicator indicates the indicator in the power generation operation indicator ontology that corresponds to the template content object; the operational data field indicates the data field in the power generation operation data traceability diagram that supports the data retrieval or calculation of the candidate indicator; the matching benchmark item records the matching results such as indicator name, statistical object, time granularity, unit of measurement, aggregation method, and business scope; and the traceability path records the path information from the candidate indicator to the data source field, calculation rules, and scope conversion node.

[0092] When multiple ternary alignment records correspond to the same template content object and there are inconsistencies in indicator scope, unit of measurement, time granularity, equipment object, or traceability field, the system generates an alignment conflict identifier. The alignment conflict identifier includes a conflict identifier number, the corresponding template content object, the associated ternary alignment record, the conflict type, the conflict field, the conflict reason, and whether it can be resolved. Conflict types include indicator scope conflicts, unit of measurement conflicts, time granularity conflicts, equipment object conflicts, and traceability field conflicts.

[0093] S3. Based on the triple alignment record, the template content object is compiled into a report generation slot and an executable verification rule item. The executable verification rule item is used to constrain the generation of standard query objects, the recall of verification data packets, the verification of assertion triples, the verification of counterfactual disturbances and the rollback process, and to control the execution of the corresponding executable verification rule item based on the alignment conflict identifier.

[0094] In this embodiment, S3 is used to convert the ternary alignment record generated by S2 into executable slots and rules for the report generation stage. The input of S3 includes the ternary alignment record, slot constraint summary, and alignment conflict identifier; the output includes the report generation slot and the executable verification rule item corresponding to the report generation slot.

[0095] The rule compilation module reads the ternary alignment records and, based on the template content object, candidate indicators, runtime data fields, matching benchmark items, and tracing paths recorded in the ternary alignment records, determines the report generation slot corresponding to the template content object. Each report generation slot includes a slot identifier, an associated template content object identifier, a slot type, location coordinates, candidate indicator identifiers, runtime data field identifiers, tracing path identifiers, and layout restoration parameters. Slot types include statistical text slots, tabular data slots, analytical text slots, or title variable slots.

[0096] The rule compilation module compiles executable validation rule items based on the triple alignment record, slot constraint summary, and alignment conflict identifier. Each executable validation rule item includes a rule item identifier, associated report generation slot identifier, query constraint, validation packet recall constraint, assertion triple validation constraint, counterfactual perturbation constraint, fallback constraint, and rule status. The rule status includes an executable state and a pending state.

[0097] Query constraints are used to limit the data source, statistical period, device object, indicator caliber, and aggregation method in subsequent standard queries. The parameters of the query constraints are derived from the candidate indicators, runtime data fields, and traceability paths in the ternary alignment records, as well as the slot semantic signature and dimension constraint candidates in the slot constraint summary.

[0098] The validation data packet recall constraint is used to limit the recall fields, recall sources, and recall priorities of validation data packets. Recall fields include indicator name, indicator value, statistical period, device object, unit of measurement, source field, and traceability path; recall sources include real-time operating database, historical operating database, indicator library, data governance results, or confirmed historical report data; recall priority is determined based on data source credibility, data update time, data integrity, and the degree of matching with the current report generation slot.

[0099] Assertion triple validation constraints are used to limit the subject, predicate, object, and consistency validation benchmark items in the assertion triple. The subject can correspond to a statistical object, equipment object, or report chapter object; the predicate can correspond to an indicator relationship, change relationship, or comparison relationship; the object can correspond to an indicator value, trend conclusion, or status description. Consistency validation benchmark items include indicator name, statistical period, unit of measurement, equipment object, data source, aggregation method, and business scope.

[0100] Counterfactual perturbation constraints are used to limit the permissible perturbation dimensions (time dimension, device dimension, metric caliber dimension, or aggregation dimension) and the allowed perturbation range. The parameters of counterfactual perturbation constraints are derived from the counterfactual perturbation dimension candidates, dimension constraint candidates, and metric caliber information in the slot constraint summary and ternary alignment records. In S3, counterfactual perturbation constraints only serve as constraints for generating counterfactual query objects during subsequent counterfactual perturbation verification; they are not used for perturbation verification within S3 itself.

[0101] Rollback constraints are used to establish the correspondence between mismatch types and report template file re-parsing, data re-querying, indicator rebinding, and content regeneration. The parameters of the rollback constraint are derived from the rollback position identifier in the slot constraint summary, the tracing path in the triple alignment record, and the conflict type in the alignment conflict identifier. Mismatch types include fact mismatch, tracing mismatch, counterfactual mismatch, and assertion triple mismatch.

[0102] When an alignment conflict flag remains unresolved in a report generation slot, the rule compilation module sets the corresponding executable validation rule to a pending state and restricts executable validation rule items in this pending state from participating in standard query object generation, validation data packet recall, and assertion triple validation. The pending state indicates that the rule item can only enter the executable state after subsequent conflict resolution, manual confirmation, or rollback processing.

[0103] The output of S3 includes a report generation slot and executable verification rule items. The report generation slot serves as the processing object for candidate indicator association score calculation and target indicator selection in subsequent S4; the executable verification rule items serve as the unified constraint object for subsequent standard query object generation, verification data packet recall, assertion triple verification, counterfactual perturbation verification, and rollback processing.

[0104] S4. For each report, generate slots, calculate the candidate indicator association score based on the ternary alignment record, alignment conflict identifier and executable verification rule item, select the target indicator based on the candidate indicator association score and redundancy constraints, generate standard query objects and obtain fact data packets;

[0105] In this embodiment, S4 is used to determine the target indicator from the candidate indicators in each report generation slot, and generate a standard query object based on the target indicator to obtain a fact data package. The input of S4 includes ternary alignment records, alignment conflict identifiers, and executable verification rule items; the output includes the target indicator, the standard query object, and the fact data package.

[0106] The metric binding module reads the corresponding ternary alignment record, alignment conflict flag, and executable validation rule item for each report generation slot, and calculates the candidate metric association score. For each report generation slot... and candidate indicators The candidate indicator correlation score is calculated according to the following formula:

[0107] ;

[0108] In the formula: Indicates the report generation slot; Indicates candidate indicators; Indicate candidate indicators Relative to report generation slot The correlation score, Indicates the report generation slot With candidate indicators The corresponding set of unresolved alignment conflict identifiers; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true and takes the value 0 when the condition inside the parentheses is false. This indicates that the associated report generation slot will be used simultaneously. and candidate indicators The set of candidate alignment evidence; Indicates the report generation slot The corresponding set of consistency verification benchmark items; Indicates the report generation slot The corresponding set of query constraint predicates; Indicates candidate alignment evidence For consistency verification benchmark items The coverage matching degree, with a value range of 100%. ; Indicates candidate alignment evidence query constraint predicates The value that satisfies the judgment condition is either 0 or 1.

[0109] Among them, conflict gating items Candidate metrics with unresolved alignment conflicts are excluded; the baseline coverage term measures the degree to which candidate alignment evidence covers the consistency verification baseline; and the query constraint satisfaction term determines whether a candidate metric can generate an executable standard query object. Because... The value can be 0 or 1. , ,therefore .

[0110] Alignment conflict flags generated by S2. Derived from ternary alignment records, each candidate alignment evidence This includes template content objects, candidate metrics, runtime data fields, matching benchmarks, and tracing paths. Candidate consistency verification benchmark items are derived from the assertion triplet verification constraints and slot constraint summaries in the executable verification rule items. These constraints originate from the query constraints in the executable validation rule items.

[0111] In one implementation, if the candidate alignment evidence is completely consistent with the consistency verification benchmark, then If the two have a pre-existing synonym relationship or a defined unit conversion relationship, then Take a value between 0.8 and 1; if there are executable caliber conversion rules but corrections need to be calculated, then Take a value between 0.5 and 0.8; if no match or conversion is possible, then... Synonym relationships, unit conversion relationships, and caliber conversion rules are stored in the power generation operation index ontology, executable verification rule items, or system configuration tables. If the candidate alignment evidence satisfies the query constraint predicate, then... ;otherwise, .

[0112] when When it is an empty set, and The value of is 0; when When it is an empty set, The value of is 0; when When it is an empty set, The value of is 1; Candidate indicators are filtered out.

[0113] Furthermore, The candidate metrics were determined as a set of bindable candidate metrics. Solve the following objective function:

[0114] ;

[0115] In the formula: Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; This represents the non-negative dimensionless redundancy penalty coefficient; This represents a set of bindable candidate metrics. An unordered set of candidate indicator pairs consisting of two different candidate indicators; Indicate candidate indicators With candidate indicators The redundancy strength between them, with a value range of 100. ;

[0116] The feasible region of the above objective function satisfies , , , ,and This is an unordered set of candidate indicator pairs. The first term of the objective function is used to improve the overall correlation between the selected candidate indicators and the report generation slots, and the second term is used to suppress the simultaneous selection of similar or duplicate candidate indicators; since all terms are dimensionless values, there is no issue of inconsistent dimensions.

[0117] Redundancy penalty coefficient Stored in executable verification rule entries or system configuration tables. Redundancy strength The candidate indicators can be determined based on similarity in indicator name, business scope, data source consistency, overlap in calculation rules, or historical co-occurrence relationship. Alternatively, they can be obtained through a pre-set redundant relationship table.

[0118] The constraints of the objective function include dimension coverage, query validity, and mutual exclusion. The dimension coverage condition requires that the selected candidate metrics collectively cover the necessary dimensions of the report generation slot; the query validity condition requires that the selected candidate metrics can generate a standard query object that passes query validity validation; and the mutual exclusion condition requires that candidate metrics with mutual exclusion relationships cannot be selected simultaneously. The objective metric can be one or more candidate metrics; when business rules require a single metric to be bound, additional constraints can be added. .

[0119] The objective function can be solved using an integer programming solver, heuristic search, or greedy selection. If If the objective function is an empty set or there is no feasible solution that satisfies the constraints, the corresponding report generation slot will be marked as an indicator binding failure, and an indicator rebinding instruction or a report template file re-parsing instruction will be generated based on the fallback constraints in the executable verification rule item.

[0120] After the target metrics are determined, the metric binding module generates standard query objects based on the target metrics, query constraints, and the operational data fields and traceability paths in the ternary alignment records. Each standard query object includes a query object identifier, a report generation slot identifier, a target metric identifier, a data source identifier, operational data fields, a statistical period, an equipment object, a metric caliber, an aggregation method, a unit of measurement, filtering conditions, query validity verification results, and a traceability path.

[0121] When the standard query object passes the query validity check, the data query module accesses the corresponding data source based on the standard query object to obtain the fact data package. The fact data package includes the report generation slot identifier, target indicator identifier, indicator value, statistical period, equipment object, unit of measurement, data source, source field, query time, aggregation method, and traceability path. If the standard query object fails the query validity check, a query failure identifier is generated, and rollback processing is initiated according to the rollback constraints.

[0122] like Figure 2 As shown, the report generation and verification module first extracts the conclusions to be verified from the report content and converts them into assertion triples. Then, it performs fact forward verification, source tracing reverse verification, and counterfactual perturbation verification respectively to obtain the corresponding verification results.

[0123] S5. Based on the executable verification rule item, recall the verification data packet and generate the report content. Convert the conclusion to be verified in the report content into an assertion triple. Perform fact forward verification, source tracing reverse verification and antifactual perturbation verification on the assertion triple.

[0124] In this embodiment, S5 is used to recall verification data packets based on fact data packets and executable verification rule items, generate report content, and perform consistency verification on the conclusions to be verified in the report content. The inputs of S5 include target indicators, standard query objects, fact data packets, and executable verification rule items; the outputs include report content, assertion triples, fact mismatch items, source mismatch items, and counterfactual mismatch items.

[0125] The report generation and verification module recalls verification data packets based on the verification data packet recall constraints in the executable verification rule items. Verification data packet recall constraints are used to limit the recall fields, recall sources, and recall priorities of the verification data packets. Recall fields include indicator name, indicator value, statistical period, equipment object, unit of measurement, source field, and traceability path; recall sources include the real-time operation database, historical operation database, indicator library, data governance results, or confirmed historical report data pointed to by the power generation operation data traceability map; recall priority is determined based on the reliability of the data source, data update time, field completeness, and the degree of matching with the report generation slot.

[0126] The report generation and verification module, based on target indicators, standard query objects, and ternary aligned records, retrieves data segments corresponding to the fact data packages from the data sources pointed to by the power generation operation data traceability map, forming verification data packages. The verification data packages and fact data packages are associated according to the report generation slots to verify the consistency between the generated content and the generated conclusions.

[0127] After obtaining the fact data package and verification data package, the report generation and verification module generates report content based on the report generation slot, target indicators, fact data package, and verification data package. For statistical text slots, the indicator values, statistical periods, equipment objects, and units of measurement from the fact data package are backfilled into the corresponding positions. For tabular data slots, data is backfilled according to the row and column coordinates, header constraint chains, and layout restoration parameters corresponding to the report generation slot. For analytical text slots, the fact data package, verification data package, target indicators, and report generation instructions are used as inputs, combined with preset text generation rules or a deployed large language model to generate analytical text. If a large language model is used, the model input includes the context text corresponding to the report generation slot, target indicators, fact data package, verification data package, and report generation instructions, and the model output is candidate analytical text content. The business logic layer performs consistency verification on the output content based on executable verification rule items and then determines whether to write it into the report content. When operational data, historical report data, or user input data are involved, the acquisition and processing of relevant data are based on authorized business data and comply with data security and relevant legal and regulatory requirements.

[0128] Furthermore, the report generation and verification module converts the conclusions to be verified in the report content into assertion triples. The conclusions to be verified originate from statements in statistical text, tabular data, or analytical text that contain indicator values, comparative relationships, trend judgments, or status judgments. An assertion triple includes a subject, predicate, and object. The subject is the equipment object, statistical object, report chapter object, or indicator object; the predicate is the indicator relationship, numerical relationship, change relationship, or judgment relationship; and the object is the indicator value, change trend, comparison result, or status description.

[0129] After the assertion triple transformation is completed, the report generation and verification module performs a forward fact check. The forward fact check matches the assertion triple with the indicator values, statistical periods, equipment objects, and units of measurement in the fact data package to determine whether the conclusions in the report content are supported by the fact data package. If the indicator values, statistical periods, equipment objects, or units of measurement in the assertion triple are inconsistent with the corresponding fields in the fact data package, or if there is a missing fact data fragment that can support the assertion triple, a fact mismatch item is generated.

[0130] The report generation and verification module performs reverse source verification. Based on the ternary alignment record and the power generation operation data source map, the reverse source verification traces back along the data source fields, calculation links, and caliber conversion links corresponding to the fact data packets to determine whether the source fields, calculation rules, and caliber conversion relationships of the fact data packets match the report generation slots, target indicators, and assertion triples. If the source fields, calculation links, or caliber conversion relationships are inconsistent, a source mismatch item is generated.

[0131] Furthermore, the report generation and verification module performs counterfactual perturbation verification. Based on the counterfactual perturbation constraints in the executable verification rules, the counterfactual perturbation verification adjusts the target perturbation dimension in the standard query object, generates a counterfactual query object, and obtains a counterfactual data package based on the counterfactual query object. The target perturbation dimension includes time, device, indicator caliber, or aggregation dimensions; the allowed perturbation range is limited by the counterfactual perturbation constraints. The report generation and verification module regenerates the counterfactual conclusion based on the counterfactual data package and compares the data changes between the factual data package and the counterfactual data package, as well as the conclusion changes between the conclusion to be verified and the counterfactual conclusion. If the data changes and conclusion changes do not conform to the business logic defined by the counterfactual perturbation constraints, or if the conclusion changes unreasonably under the perturbation conditions, a counterfactual mismatch item is generated.

[0132] In this embodiment, the assertion triple check also includes calculating the consistency score of the assertion triple according to the following formula.

[0133] ;

[0134] In the formula: This indicates an assertion triple; Represents a data segment; Indicates assertion triples Consistency score; Expressing support for asserting triplets Supports collections of data fragments; Representation and assertion triples A set of contradictory data fragments; Representing data fragments With assertion triplet The degree of matching between the benchmark items; Representing data fragments Source quality score; Indicates assertion triples Numerical consistency factor; This indicates the necessary benchmark term verification gating factor;

[0135] , , , The range of values ​​for the preset assertion threshold is... ; The value of is 0 or 1, and the value of the empty product is 1; The determination is based on the matching of indicator name, statistical period, equipment object, unit of measurement, aggregation method and business scope; The determination is based on the credibility of the data source, the data update time, the completeness of the fields, and the completeness of the tracing path; Used to indicate whether the values ​​in a numeric assertion triplet are consistent; for a non-numeric assertion triplet, the corresponding... The value is 1.

[0136] When asserting the triplet When all the necessary benchmark items reach the matching threshold, When asserting the triplet When there are necessary benchmark items among the corresponding necessary benchmark items that have not reached the matching threshold, Necessary benchmark items include the indicator name, statistical period, equipment object, unit of measurement, data source or business scope, etc., which are marked as mandatory verification benchmark items.

[0137] when Below the preset assertion threshold, or When this occurs, a mismatched assertion triple is generated. This mismatched assertion triple serves as supplementary evidence for factual mismatches, source mismatches, or counterfactual mismatches, or as a trigger for content regeneration.

[0138] like Figure 3 As shown, when the verification result does not meet the preset verification pass conditions, the rollback output module determines the corresponding mismatch type and generates a rollback command; when the verification result meets the preset verification pass conditions, the rollback output module outputs a power generation operation report and the mapping relationship between the report content and the verification data.

[0139] S6. When the results of the fact forward verification, source tracing reverse verification, and antifactual disturbance verification do not meet the preset verification pass conditions, a rollback instruction is generated according to the corresponding mismatch type; when the preset verification pass conditions are met, the power generation operation report and the mapping relationship between the report content and the verification data are rendered and output.

[0140] In this embodiment, S6 is used to determine whether the preset verification pass conditions are met based on the fact forward verification, source tracing reverse verification and antifactual disturbance verification results obtained in S5, and output a power generation operation report or generate a rollback command based on the judgment result.

[0141] Specifically, the rollback output module determines the mismatch type based on the generation of factual mismatches, source mismatches, and counterfactual mismatches. When a factual mismatch is generated, the rollback output module prioritizes generating a data re-query instruction, and if necessary, also generates an indicator rebinding instruction. When a source mismatch is generated, the rollback output module prioritizes generating an indicator rebinding instruction, and if necessary, also generates a data re-query instruction. When a counterfactual mismatch is generated, the rollback output module prioritizes generating a content regeneration instruction, and if necessary, also generates an indicator rebinding instruction or a data re-query instruction. When two or more mismatches are generated simultaneously, the rollback output module generates corresponding rollback instructions for each mismatch.

[0142] The data re-query instruction instructs the data query module to re-obtain the fact data package based on the original or revised standard query object; the indicator re-binding instruction instructs the indicator binding module to re-execute the candidate indicator association score calculation and target indicator selection; the content re-generation instruction instructs the report generation and verification module to regenerate the report content based on the updated fact data package, verification data package, or target indicator. If the mismatch is related to the template parsing result, report generation slot, or template content object, the rollback output module generates a report template file re-parsing instruction based on the rollback constraints.

[0143] When the results of forward fact verification, reverse source verification, and counterfactual disturbance verification meet the preset verification pass conditions, the rollback output module renders the report content as a power generation operation report. During the rendering process, the system backfills statistical text, tabular data, and analytical text into the corresponding report generation slots according to the layout restoration parameters in the template content object, and restores the layout information such as font, font size, paragraph format, table borders, row height, column width, and merged cells in the original report template file.

[0144] When outputting the power generation operation report, the system simultaneously outputs the mapping relationship between the report content and the verification data. This mapping relationship includes the correspondence between the report content fragment corresponding to the report generation slot, the data fragment in the verification data packet, the triplet alignment record, and the verification result. Specifically, the system associates the fact data packet and the verification data packet according to the report generation slot, generating a slot-level fact data to verification data mapping record; then, it binds the report content fragment, data fragment, triplet alignment record, assertion triplet, and verification result to form a traceable mapping relationship.

[0145] The output of S6 includes the power generation operation report, the mapping relationship between the report content and the verification data, and the rollback instruction generated when the verification fails. Through S1 to S6 above, the system can perform consistency verification on template semantics, indicator definitions, data sources and report conclusions during the report generation process, and perform corresponding rollback processing when mismatches are found.

[0146] In one application scenario, the above method can be embedded in an intelligent agent report generation system. This system can receive report template files and report generation instructions uploaded by users, perform document parsing, content classification, and indexing of the report template files, and obtain indicator data through an intelligent query platform, Text-to-SQL service, or business database. The document parsing results can serve as the basis for constructing a template semantic graph, the indicator knowledge base can serve as the basis for constructing a power generation operation indicator ontology, the call relationships between fields in the intelligent query platform, data middleware, and business database can serve as the basis for constructing a power generation operation data traceability graph, and the queried indicator data can serve as the data source for fact data packages. The intelligent agent report generation system can also generate analytical text based on historical reports, business rules, or the knowledge base, and upon receiving user modification instructions, use the modification records as sources for historical template mapping identifiers, historical verification results, or system configuration parameters. The above application method does not change the technical logic of tripartite alignment record generation, executable verification rule item compilation, candidate indicator selection, assertion triplet verification, and rollback processing in Embodiment 1.

[0147] like Figure 4 As shown, another embodiment of the present invention also provides a data consistency verification and generation system for power generation operation reports. This system can be deployed on a server, terminal device, or cloud computing environment and is used to execute the aforementioned data consistency verification and generation method for power generation operation reports. The system includes a template parsing module, a three-graph alignment module, a rule compilation module, an indicator binding module, a report generation and verification module, and a rollback output module. These modules are connected through data interfaces or service calls.

[0148] The template parsing module receives the report template file and the report generation instructions, parses the report template file to construct a template semantic graph, and determines the template content objects.

[0149] The three-graph alignment module is used to align the template semantic graph, the power generation operation indicator ontology, and the power generation operation data traceability graph to generate a ternary alignment record. When multiple ternary alignment records corresponding to the same template content object have inconsistencies in indicator scope, unit of measurement, time granularity, equipment object, or traceability field, an alignment conflict identifier is generated.

[0150] The rule compilation module is used to compile template content objects into report generation slots and executable verification rule items based on the triple alignment records. The executable verification rule items include query constraints, verification data packet recall constraints, assertion triple verification constraints, counterfactual perturbation constraints, and backoff constraints. When there are unresolved alignment conflict identifiers, the corresponding executable verification rule item is set to a pending parsing state, and its participation in standard query object generation, verification data packet recall, and assertion triple verification is restricted.

[0151] The indicator binding module is used to generate slots for each report, calculate the candidate indicator association score based on the ternary alignment record, alignment conflict identifier and executable verification rule item, select the target indicator based on the candidate indicator association score and redundancy constraints, generate standard query objects and obtain fact data packages.

[0152] The report generation and verification module is used to recall verification data packets based on executable verification rule items and generate report content. It converts the conclusions to be verified in the report content into assertion triples and performs fact forward verification, source tracing reverse verification, and counterfactual perturbation verification on the assertion triples to generate corresponding verification results.

[0153] The rollback output module is used to generate a rollback instruction based on the mismatch type when the verification result does not meet the preset verification pass conditions; and to render and output the power generation operation report and the mapping relationship between the report content and the verification data when the verification result meets the preset verification pass conditions.

[0154] Through the above implementation methods, the system can structurally associate the report template, operating indicators, data sources and report conclusions during the generation of power generation operation reports, and uniformly control data query, report generation, conclusion verification and rollback processing based on executable verification rules, thereby realizing the generation of data consistency verification for power generation operation reports.

[0155] In summary, this invention unifies the association between report template files, power generation operation indicators, power generation operation data traceability diagrams, and report generation conclusions. It integrates template parsing, indicator binding, data querying, content generation, consistency verification, and rollback processing into a continuous technical process. This enables the automatic generation of power generation operation reports to have traceable data sources, verifiable indicator definitions, reviewable report conclusions, and the ability to roll back in case of anomalies. This solution is applicable to report generation scenarios such as power generation operation analysis, dispatch review, equipment management, and operational statistics. It helps reduce the costs of manual verification and repeated modifications, improves the reliability, accuracy, and process controllability of report generation results, and has significant engineering application value.

[0156] In the description of this specification, references to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. The aforementioned specific features, structures, materials, or characteristics may be combined in one or more embodiments or examples in a suitable manner without contradicting each other.

[0157] The processes or methods described in this specification in conjunction with flowcharts or other means can be understood as including one or more steps, modules, segments, or parts for implementing the corresponding logical functions or processing procedures. Without affecting the necessary logical relationships between the steps, the relevant steps may be executed in the order shown, or in a substantially simultaneous manner, or in other reasonable order, depending on actual functional needs.

[0158] The above description is merely a specific embodiment of this application and is not intended to limit the scope of protection of this application. For those skilled in the art, various changes, substitutions, or combinations can be made to the technical solutions of this application without departing from the technical concept of this application; all such changes, substitutions, or combinations should be covered within the scope of protection of this application. The scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A method for generating data consistency verification data in power generation operation reports, characterized in that, Includes the following steps: Receive report template file and report generation instructions, parse report template file to construct template semantic graph, and determine template content objects; Align the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph to generate a ternary alignment record and an alignment conflict identifier; The template content object is compiled into a report generation slot and an executable verification rule item based on the ternary alignment record. The executable verification rule item is used to constrain the generation of standard query objects, the recall of verification data packets, the verification of assertion triples, the verification of counterfactual disturbances and the rollback process, and to control the execution of the corresponding executable verification rule item based on the alignment conflict identifier. For each report generation slot, the candidate indicator association score is calculated based on the ternary alignment record, alignment conflict identifier and executable verification rule item. The target indicator is selected based on the candidate indicator association score and redundancy constraints, a standard query object is generated and a fact data package is obtained. Based on the executable verification rule item, the verification data packet is recalled and the report content is generated. The conclusion to be verified in the report content is converted into assertion triples. The assertion triples are then subjected to fact forward verification, source tracing reverse verification and antifactual perturbation verification. When the results of fact forward verification, source tracing reverse verification, and antifactual disturbance verification do not meet the preset verification pass conditions, a rollback instruction is generated according to the corresponding mismatch type; when the preset verification pass conditions are met, a power generation operation report and the mapping relationship between the report content and the verification data are rendered and output.

2. The data consistency verification and generation method for a power generation operation report according to claim 1, characterized in that, The parsing report template file for constructing a template semantic graph includes: The report template file's titles, tables, field names, statistical periods, equipment objects, and indicator descriptions are structured and parsed to generate content nodes and layout nodes. Generate semantically defined edges based on the hierarchical relationships between content nodes, header constraints, indicator definitions, and time dimensions; Starting from the content node and ending at the report generation slot, candidate slot edges are generated. Each candidate slot edge records the slot type, dimension placeholder identifier, table header constraint chain, and candidate indicator caliber. When the same content node corresponds to multiple slot candidate edges, the main slot candidate edge is determined based on the semantically limited edge, the table header limited chain, and the historical template mapping identifier. A slot constraint summary is generated based on content nodes, layout nodes, semantically limited edges, and primary slot candidate edges. The slot constraint summary includes slot semantic signature, consistency verification benchmark item candidate, dimension constraint candidate, counterfactual perturbation dimension candidate, and fallback position identifier.

3. The data consistency verification and generation method for a power generation operation report according to claim 2, characterized in that, The step of aligning the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph to generate a ternary alignment record and alignment conflict identifier includes: A slot semantic signature is generated based on the slot constraint summary. The slot semantic signature includes the indicator name, statistical object, time granularity, unit of measurement, aggregation method, and business scope. Extract the indicator ontology signature of candidate indicators from the power generation operation indicator ontology, and extract the data traceability signature of operation data fields from the power generation operation data traceability diagram; Perform consistency matching on slot semantic signature, indicator ontology signature and data traceability signature to generate a ternary aligned record including template content object, candidate indicator, running data field, matching benchmark item and traceability path; When multiple ternary alignment records correspond to the same template content object and there are inconsistencies in indicator caliber, unit of measurement, time granularity, equipment object, or traceability field, an alignment conflict identifier is generated.

4. The data consistency verification and generation method for a power generation operation report according to claim 3, characterized in that, The executable verification rule item is compiled from the ternary alignment record, slot constraint summary and alignment conflict identifier; The executable verification rule items include query constraints, verification data packet recall constraints, assertion triple verification constraints, counterfactual perturbation constraints, and fallback constraints; The query constraints are used to limit the data source, statistical period, device object, indicator scope, and aggregation method in the standard query object; The verification data packet recall constraint is used to limit the recall fields, recall sources, and recall priorities of the verification data packets; The assertion triple check constraint is used to limit the subject, predicate, object and their consistency check benchmark items in the assertion triple; The counterfactual perturbation constraint is used to limit the permissible perturbation dimensions, including the time dimension, device dimension, indicator dimension, or aggregation dimension, as well as the allowed perturbation range. The rollback constraint is used to establish the correspondence between mismatch types and report template file re-parsing, data re-querying, indicator re-binding, and content re-generation; When there are unresolved alignment conflict flags, the corresponding executable validation rule item is set to the pending parsing state, and the executable validation rule item in the pending parsing state is restricted from participating in standard query object generation, validation data retrieval, and assertion triple validation.

5. The data consistency verification and generation method for a power generation operation report according to claim 4, characterized in that, The calculation of the candidate indicator correlation score includes calculation according to the following formula: ； In the formula: Indicates the report generation slot; Indicates candidate indicators; Indicate candidate indicators Relative to report generation slot The correlation score, Indicates the report generation slot With candidate indicators The corresponding set of unresolved alignment conflict identifiers; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true and takes the value 0 when the condition inside the parentheses is false. This indicates that the associated report generation slot will be used simultaneously. and candidate indicators The set of candidate alignment evidence; Indicates the report generation slot The corresponding set of consistency verification benchmark items; Indicates the report generation slot The corresponding set of query constraint predicates; Indicates candidate alignment evidence For consistency verification benchmark items The coverage matching degree, with a value range of 100%. ; Indicates candidate alignment evidence query constraint predicates The value that satisfies the judgment condition is either 0 or 1. when When it is an empty set, and The value of is 0; when When it is an empty set, The value of is 0; when When it is an empty set, The value of is 1; Candidate indicators are filtered out.

6. The data consistency verification and generation method for a power generation operation report according to claim 5, characterized in that, The selection of target indicators based on candidate indicator correlation scores and redundancy constraints includes: Will The candidate metrics were determined as a set of bindable candidate metrics. Solve the following objective function: ； In the formula: Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; Indicate candidate indicators Relative to report generation slot A binary selection variable, taking the value 0 or 1; This represents the non-negative dimensionless redundancy penalty coefficient; This represents a set of bindable candidate metrics. An unordered set of candidate indicator pairs consisting of two different candidate indicators; Indicate candidate indicators With candidate indicators The redundancy strength between them, with a value range of 100%. ; The constraints of the objective function include: the selected candidate metrics jointly cover the necessary dimensions of the report generation slot; the selected candidate metrics can generate standard query objects that pass the query validity check; and candidate metrics with mutual exclusion relationships are not selected simultaneously. When the objective function is maximized The candidate indicators were selected as the target indicators.

7. The data consistency verification and generation method for a power generation operation report according to claim 4, characterized in that, The aforementioned forward fact verification, reverse source tracing verification, and counterfactual perturbation verification include: The conclusion to be verified is converted into an assertion triple including subject, predicate and object, and the assertion triple is matched with the indicator value, statistical period, device object and unit of measurement in the fact data packet to generate fact mismatch items; Based on the ternary alignment record and the power generation operation data traceability diagram, reverse tracing is performed along the data source fields, calculation links and caliber conversion links corresponding to the fact data packets to generate traceability mismatch items; Adjust the target perturbation dimension in the standard query object according to the counterfactual perturbation constraint, generate a counterfactual query object, obtain a counterfactual data packet based on the counterfactual query object, and regenerate the counterfactual conclusion based on the counterfactual data packet; Based on the data changes between fact data packets and counterfactual data packets, and the conclusion changes between the conclusion to be verified and the counterfactual conclusion, a counterfactual mismatch item is generated; The mismatch type is determined based on the generation of fact mismatch items, source mismatch items, and counterfactual mismatch items. When a fact mismatch item is generated, a data re-query instruction is generated; when a source mismatch item is generated, an indicator rebinding instruction is generated; and when a counterfactual mismatch item is generated, a content regeneration instruction is generated. When two or more mismatch items are generated simultaneously, a rollback instruction corresponding to each mismatch item is generated.

8. The data consistency verification and generation method for a power generation operation report according to claim 7, characterized in that, The assertion triplet verification includes calculating the consistency score of the assertion triplet according to the following formula: ； In the formula: This indicates an assertion triple; Represents a data segment; Indicates assertion triples Consistency score; Expressing support for asserting triplets Supports collections of data fragments; Representation and assertion triples A set of contradictory data fragments; Representing data fragments With assertion triplet The degree of matching between the benchmark items; Representing data fragments Source quality score; Indicates assertion triples Numerical consistency factor; This indicates the necessary benchmark term verification gating factor; , , , The range of values ​​for the preset assertion threshold is... ; The value of is 0 or 1, and the value of the empty product is 1; when asserting a triple. When all the necessary benchmark items reach the matching threshold, When asserting the triplet When there are necessary benchmark items among the corresponding necessary benchmark items that have not reached the matching threshold, ; when Below the preset assertion threshold, or At that time, a mismatch term for the assertion triple is generated.

9. The data consistency verification and generation method for a power generation operation report according to claim 1, characterized in that, The process of generating and outputting the mapping relationship between the report content and the verification data includes: The recall fields, recall sources, and recall priorities of the verification data packets are determined according to the recall constraints of the verification data packets. Based on target indicators, standard query objects, and ternary aligned records, data fragments containing indicator names, indicator values, statistical periods, equipment objects, units of measurement, source fields, and traceability paths are retrieved from the data sources pointed to by the power generation operation data traceability map to form a verification data package; The fact data packet and the verification data packet are associated according to the report generation slot to generate a slot-level fact data and verification data mapping record; Based on the report, generate slots, target indicators, fact data packages, and slot-level fact data and verification data mapping records; perform data backfilling or content generation for statistical text, tabular data, and analytical text. When outputting the power generation operation report, the report content fragments corresponding to the generated slot, the data fragments in the verification data packet, the mapping relationship between the ternary alignment record and the verification result are output synchronously.

10. A data consistency verification and generation system for power generation operation reports, applied to the data consistency verification and generation method for power generation operation reports as described in any one of claims 1-9, characterized in that, include: The template parsing module is used to receive the report template file and the report generation instruction, parse the report template file to construct a template semantic graph, and determine the template content object; The three-graph alignment module is used to align the template semantic graph, the power generation operation index ontology, and the power generation operation data traceability graph, and generate a three-element alignment record and alignment conflict identifier; The rule compilation module is used to compile template content objects into report generation slots and executable verification rule items based on the triple alignment records. The executable verification rule items are used to constrain the generation of standard query objects, the recall of verification data packets, the verification of assertion triples, the verification of counterfactual disturbances, and the rollback process, and control the execution of the corresponding executable verification rule items based on the alignment conflict identifier. The indicator binding module is used to generate slots for each report, calculate the candidate indicator association score based on the ternary alignment record, alignment conflict identifier and executable verification rule item, select the target indicator based on the candidate indicator association score and redundancy constraints, generate standard query objects and obtain fact data packets; The report generation and verification module is used to recall verification data packets based on executable verification rule items and generate report content. It converts the conclusions to be verified in the report content into assertion triples and performs forward fact verification, reverse source verification, and antifactual perturbation verification on the assertion triples. The rollback output module is used to generate rollback instructions based on the corresponding mismatch type when the results obtained from the fact forward verification, source tracing reverse verification, and antifactual disturbance verification do not meet the preset verification pass conditions; when the preset verification pass conditions are met, it renders and outputs the power generation operation report and the mapping relationship between the report content and the verification data.

Images