Methods, systems, equipment and storage media for resolving conflicts related to power grid dispatch quotas
By constructing a quota query slot template and semantic extraction from a large language model, structured quota rule units are generated, which solves the problems of accuracy and conflict resolution in quota clause identification in power grid dispatching. This improves the accuracy and interpretability of quota selection, reduces the risk of model illusion, and meets the safety and management needs of power grid dispatching.
Patent Information
- Application Number
- CN202611383947.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-08
- Publication Date
- 2026-11-10
AI Technical Summary
In the current power grid dispatching, the sources of quota clauses are diverse and the forms of expression are not uniform, making it difficult to accurately identify the truly applicable quotas in the current dispatching scenario. This leads to overly strict control, reduced equipment utilization, or increased operational risks. Furthermore, large language models are prone to illusion risks in quota judgment and lack traceability in conflict resolution.
By constructing a quota query slot template, using a large language model for semantic extraction, generating structured quota rule units, constructing a quota conflict graph, performing safety-first and evidence-first resolution, determining the final effective quota, combining the power grid topology and operating mode to perform scenario applicability scoring, and outputting an interpretable final result.
It improves the accuracy of quota selection and the efficiency of scheduling verification, reduces the influence of model illusion, enhances the verifiability and interpretability of results, reduces the workload of manual table lookup, and meets the actual needs of power grid security constraints and scheduling management.
Smart Images

Figure CN123073984A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power grid dispatching and operation technology, and in particular to a method, system, device and storage medium for resolving power grid dispatching quota conflicts. Background Technology
[0002] During power grid dispatching and operation, various constraints are typically imposed on objects such as lines, main transformers, busbars, cross sections, stability control sections, and regional mutual aid channels. These constraints include thermal stability limits, transient stability limits, static safety limits, maintenance mode limits, temporary limits, and seasonal limits. These limits may originate from dispatching procedures, operation mode arrangements, stability calculation reports, maintenance tickets, equipment manufacturer data, or from temporary notices, shift handover records, or dispatching instructions issued by the operations department. The numerous limit clauses, diverse sources, and inconsistent expressions constitute one of the fundamental data sets for dispatching operation safety verification and decision support.
[0003] In actual operations, the same controlled object may correspond to multiple limit values under different operating modes, time periods, directions, and safety constraints. For example, the same line may have a long-term permissible power flow limit under normal conditions, a reduced operating limit under high-temperature weather, and a more stringent control limit when adjacent lines are under maintenance; the same cross-section may have different limits in the outgoing and incoming directions; the same equipment may also have annual procedure limits, monthly mode limits, and temporary notification limits simultaneously. If the truly applicable limits in the current scheduling scenario cannot be accurately identified, it can easily lead to overly strict control, reduced equipment utilization, or overly broad limit selection, increasing operational risks.
[0004] Existing scheduling quota management methods typically rely on manual review of procedures, forms, method sheets, and historical notifications, or on rule-based retrieval primarily using device names and keywords. While these methods work when structured fields are complete and quota clauses are well-defined, they often struggle with accurate recall and precise matching for quota clauses described in natural language, aliases for the same device, multi-granularity cross-section expressions, temporary quota validity periods, conditional quotas, and directional constraints. Especially when multiple quota clauses simultaneously apply to the current scenario but with inconsistent quota values, existing methods often lack a mechanism for unified modeling of conflict causes, applicable conditions, priorities, and security constraint strength.
[0005] In recent years, large language models have been able to semantically understand natural language scheduling instructions, operating mode descriptions, and quota clauses, and can be used to assist in extracting information such as controlled objects, operating conditions, quota values, and effective times. However, directly using large language models for quota judgment has the following shortcomings: 1) The model may generate conclusions not supported by the original clauses, posing a risk of illusion; 2) When the model lacks constraints on grid topology, operating mode, and safety rules, it is difficult to determine whether the clauses are truly applicable to the current scenario; 3) The model lacks traceable and structured evidence for resolving conflicts between multiple candidate quotas, making it difficult to meet the requirements of scheduling operations for security, interpretability, and verifiability. Existing methods lack a mechanism for unified modeling of natural language quota clauses, scheduling scenario constraints, grid topology relationships, and multi-rule conflict relationships.
[0006] Therefore, there is an urgent need to propose a dedicated processing scheme for resolving grid dispatching quota conflicts. Without replacing the existing dispatch automation system and safety verification system, this scheme should utilize a large language model to complete the semantic understanding and evidence organization of quota clauses. By combining structured quota units, scenario applicability scoring, conflict graph construction, and priority resolution strategies, the scheme can determine the final effective constraint quota under the current dispatching scenario, thereby improving the accuracy, interpretability, and efficiency of quota selection and dispatching review.
[0007] In view of this, the present invention is hereby proposed. Summary of the Invention
[0008] The purpose of this invention is to provide a method, system, device, and storage medium for resolving conflicts in power grid dispatching quotas, which can improve the accuracy, interpretability, and efficiency of quota selection and dispatching review.
[0009] The objective of this invention is achieved through the following technical solution: A method for resolving conflicts related to power grid dispatch quotas includes: Retrieve scheduling scenario information and generate a quota query context; Construct a quota query slot template; the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level; Based on the aforementioned limit query context, several candidate limit terms are retrieved from limit term data from multiple sources; Based on the large language model, semantic extraction is performed on each candidate limit clause through the limit query slot template to obtain the structured limit rule unit corresponding to each candidate limit clause. For each structured quota rule unit, calculate its applicability score to the current scheduling scenario; Select structured quota rule units with an applicability score higher than the threshold as nodes. If two structured quota rule units conflict, establish a conflict edge to form a quota conflict graph. For each connected subgraph in the limit conflict graph, security priority and evidence priority resolution are performed: if all the structured limit rule units contained therein are applicable and there are no mutually exclusive conditions, then the structured limit rule unit with the smallest limit value is selected as the final effective rule, and its limit value is used as the final effective limit. At the same time, the large language model generates a readable interpretation based on the clause evidence fragments in the structured limit rule unit. For each connected subgraph in the quota conflict graph, the final output result is generated by combining the final effective rule, the final effective quota, and the structured quota rule unit in the connected subgraph.
[0010] A power grid dispatch quota conflict resolution system, used to implement the aforementioned method, includes: The quota query context generation unit is used to retrieve scheduling scenario information and generate quota query context; The quota query slot template construction unit is used to construct the quota query slot template; wherein, the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level; The candidate limit clause retrieval unit is used to retrieve several candidate limit clauses from limit clause data from multiple sources, in conjunction with the limit query context. The semantic extraction unit based on the large language model is used to extract the semantics of each candidate quota clause based on the large language model and the quota query slot template to obtain the structured quota rule unit corresponding to each candidate quota clause. The applicability score calculation unit is used to calculate the applicability score of each structured quota rule unit in relation to the current scheduling scenario. The quota conflict graph construction unit is used to select structured quota rule units with an applicability score higher than the threshold as nodes. If two structured quota rule units conflict, a conflict edge is established to form a quota conflict graph. The conflict resolution unit is used to perform security-first and evidence-first resolution on each connected subgraph in the limit conflict graph: if all the structured limit rule units included are applicable and there are no mutually exclusive conditions, the structured limit rule unit with the smallest limit value is selected as the final effective rule and its limit value is used as the final effective limit. At the same time, the large language model generates a readable interpretation based on the clause evidence fragments in the structured limit rule unit.
[0011] The final output generation unit is used to generate the final output result for each connected subgraph in the quota conflict graph by combining the final effective rule, the final effective quota, and the structured quota rule unit in the connected subgraph.
[0012] A processing device includes: one or more processors; and a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.
[0013] A readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method.
[0014] As can be seen from the technical solutions provided by the present invention above, 1) the present invention uses a large language model for semantic slot extraction and evidence organization of quota clauses, without directly outputting scheduling quota conclusions, which can reduce the influence of model illusion on scheduling quota judgment and improve the verifiability of the results; 2) it converts natural language quota clauses into structured quota rule units, so that key information such as device object, quota value, effective time, applicable conditions, direction and priority can participate in subsequent verification and conflict resolution in a unified manner, improving the level of structured utilization of quota knowledge; 3) by calculating scenario applicability, it can avoid mismatches caused by keyword similarity alone, and improve the matching accuracy between candidate quota clauses and the current scheduling scenario; 4) it constructs a quota conflict graph, which combines multiple 5) Conflicts between candidate limits are explicitly expressed, making the limit conflict identification process clearer and easier for dispatchers to review; 6) Conflict resolution is performed based on the connected subgraph of the limit conflict graph, and the final effective limit is determined, which can take into account both power grid security constraints and actual dispatch management, and reduce the workload of manual table lookup and manual comparison; 7) The main control rules, the rules to be resolved, the conflict type and the original evidence fragments are output at the same time as the final limit, which can form a complete chain of evidence and improve the traceability and interpretability of dispatch limit management results; 8) Focusing on the conflict resolution link in the limit management process, it can be connected as an auxiliary module of the existing dispatch automation system, safety verification system or limit knowledge base without replacing the original system, and has good engineering implementation. Attached Figure Description
[0015] 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.
[0016] Figure 1 This is a flowchart of a power grid dispatch quota conflict resolution method provided in an embodiment of the present invention.
[0017] Figure 2 This is a schematic diagram of a power grid dispatch quota conflict resolution method provided in an embodiment of the present invention.
[0018] Figure 3 This is a schematic diagram of a power grid dispatch quota conflict resolution system provided in an embodiment of the present invention.
[0019] Figure 4 This is a schematic diagram of a processing device provided in an embodiment of the present invention. Detailed Implementation
[0020] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.
[0021] First, the following explanations are provided for the terms that may be used in this article: The terms “comprising,” “including,” “containing,” “having,” or other similar semantic descriptions should be interpreted as non-exclusive inclusion. For example, “comprising a technical feature element (such as raw material, component, ingredient, carrier, dosage form, material, size, part, component, mechanism, device, step, process, method, reaction conditions, processing conditions, parameter, algorithm, signal, data, product or article of manufacture, etc.)” should be interpreted as including not only the expressly listed technical feature element, but also other technical feature elements that are not expressly listed and are well-known in the art.
[0022] The term "composed of" excludes any technical features not expressly listed. When used in a claim, it closes the claim to exclude all technical features other than those expressly listed, except for associated conventional impurities. If the term appears only in a clause of a claim, it limits the claim to the elements expressly listed in that clause; elements recited in other clauses are not excluded from the overall claim.
[0023] The following provides a detailed description of a method, system, device, and storage medium for resolving power grid dispatch quota conflicts provided by this invention. Contents not described in detail in the embodiments of this invention are prior art known to those skilled in the art. Where specific conditions are not specified in the embodiments of this invention, they are performed according to conventional conditions in the art or conditions recommended by the manufacturer. Where the manufacturer of the instruments used in the embodiments of this invention is not specified, they are all conventional products that can be purchased commercially.
[0024] Example 1 like Figure 1 The diagram shows a flowchart of a power grid dispatch quota conflict resolution method provided by an embodiment of the present invention, which mainly includes the following steps: Step 1: Obtain scheduling scenario information and generate a quota query context.
[0025] In this embodiment of the invention, current scheduling operation scenario data is obtained, including but not limited to: one or more of the following: controlled object name, object type, operation mode, maintenance equipment, power flow direction, planned time period, real-time power flow, meteorological conditions, N-1 check status, scheduling instruction text, and alarm information; the current scheduling scenario information is processed by object standardization, time normalization, etc., to generate a quota query context.
[0026] In this embodiment of the invention, N-1 refers to the safety verification after any single operating element in the power system exits. Here, the N-1 verification status indicates whether this type of safety verification is enabled in the current dispatch scenario.
[0027] Step 2: Construct a quota query slot template.
[0028] In this embodiment of the invention, the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level.
[0029] In this embodiment of the invention, object granularity refers to different control levels such as lines, transmission channels, and stable sections; object-level mapping is the process of mapping different object granularities to a unified control boundary.
[0030] Step 3: Recall candidate limit clauses and form a candidate set.
[0031] In this embodiment of the invention, several candidate quota terms are retrieved from the multi-source quota term data in conjunction with the quota query context. Specifically: based on the quota query context, candidate quota terms are retrieved from multi-source quota term data (e.g., quota knowledge base, operation procedure library, mode sheet library, maintenance plan library, temporary notice library, and historical shift handover record library); during the retrieval process, the symbol matching score, vector semantic retrieval score, topological proximity score, time correlation score, and source reliability score of each quota term are calculated respectively, and the candidate recall score of each quota term data is obtained by combining all the above scores; the quota term data with a candidate recall score not less than a preset threshold are selected as candidate quota terms.
[0032] Step 4: Convert candidate clauses into structured limit rule units.
[0033] In this embodiment of the invention, based on a large language model, semantic extraction is performed on each candidate quota clause through the quota query slot template to obtain the structured quota rule unit corresponding to each candidate quota clause.
[0034] In this embodiment of the invention, the large language model can only output structured results within the slot template, and each structured field is bound to evidence fragments in the original scene text.
[0035] In addition, during the semantic extraction process, unit conversion is performed for limit values with inconsistent units; and object-level mapping is performed for clauses with inconsistent object granularity, such as mapping "line", "transmission channel" and "stable section" to a unified control boundary level.
[0036] Step 5: Calculate the scenario applicability score of the candidate quota rules.
[0037] In this embodiment of the invention, for each structured quota rule unit, its applicability score with the current scheduling scenario is calculated.
[0038] In this embodiment of the invention, for each structured quota rule unit, its applicability score for the current scheduling scenario is calculated, taking into account object consistency, time validity, direction consistency, operation mode matching, and condition completeness. If any strong constraint field is not satisfied, such as time being expired, direction being opposite, or objects not belonging to the same control boundary, the applicability score is reduced to zero or marked as unusable.
[0039] Specifically: For each structured quota rule unit, an initial applicability score is calculated by comprehensively considering object consistency, time validity, directional consistency, operational mode matching, topology reachability, security check type matching, and condition integrity. Set a strong constraint discriminant function B. i If time validity, direction consistency, and object consistency are all satisfied, then B i =1; if any condition is not met, then B i =0; using the strong constraint discriminant function B i Initial applicability score Make corrections to obtain the final applicability score. .
[0040] In this embodiment of the invention, object consistency is compared by the name of the controlled object or control boundary; time validity is compared by the validity period of the rule and the planned time period; direction consistency is compared by the power flow direction; operation mode matching is compared by the operation mode and maintenance conditions; topology reachability is compared by the corresponding topology relationship; safety check type matching is compared by the corresponding check status; and the condition integrity judgment rule is evaluated by the satisfaction of the applicable conditions in the current scenario.
[0041] Step 6: Construct a limit conflict graph and identify conflict types.
[0042] In this embodiment of the invention, structured quota rule units with an applicability score higher than the threshold are selected as nodes. If two structured quota rule units conflict, a conflict edge is established to form a quota conflict graph.
[0043] Specifically: If two structured quota rules overlap in the controlled object name or control boundary, time window, power flow direction, and quota type, and are inconsistent in the quota value, applicable conditions, or source priority, then they conflict, and a conflict edge is established. Conflict types include, but are not limited to: numerical conflicts within the same object, granularity conflicts between superior and subordinate objects, conflicts between temporary and regular quotas, directional quota conflicts, time overlap conflicts, condition inclusion conflicts, and source version conflicts.
[0044] In this embodiment of the invention, the time window refers to the effective time interval determined by the rule's effective time and expiration time, and is compared with the current planned time period. The quota type is a field in the structured quota rule unit, used to indicate the quota constraint type corresponding to the rule.
[0045] Step 7: Resolve conflicts based on the rules of security priority and evidence priority.
[0046] In this embodiment of the invention, for each connected subgraph in the quota conflict graph, a safety-first and evidence-first resolution process is performed: if all the included structured quota rule units are applicable and there are no mutually exclusive conditions, the structured quota rule unit with the smallest quota value is selected as the final effective rule, and its quota value is used as the final effective quota. Simultaneously, the large language model generates a readable interpretation based on the clause evidence fragments in the structured quota rule unit. Here, "no mutually exclusive conditions" means that the applicable conditions of multiple rules can be simultaneously met in the current scheduling scenario. For example, "normal operation mode" and "one-circuit maintenance, single-circuit operation" are operating mode conditions that cannot be satisfied simultaneously, and therefore cannot be used as simultaneously applicable rules for minimum quota comparison.
[0047] Specifically: During the resolution process, the large language model is used to generate readable interpretations based on evidence fragments, but does not directly determine the limit value. If multiple rules are applicable and there are no mutually exclusive conditions, the limit with stronger security constraints is selected first, that is, the stricter limit value is selected under the same unit and direction; if temporary limits and regular limits are applicable at the same time, the temporary limit is used first during the validity period of the temporary limit; if the upper-level section limit and the lower-level equipment limit are constrained at the same time, they are both retained according to the control boundary relationship, and the constraint that is triggered earlier is marked as the main control limit.
[0048] In this embodiment of the invention, conventional limits, temporary limits, upper-level section limits, and lower-level equipment limits are all common terms in the art. Conventional limits and temporary limits mainly reflect differences in applicable conditions and validity periods, while upper-level section limits and lower-level equipment limits mainly reflect differences in the controlled object and control boundary level. Specifically: conventional limits refer to limits applicable under long-term or regular operating conditions; temporary limits refer to dispatch notification limits with a clear temporary validity period; upper-level section limits refer to limits acting on sections or larger control boundaries, and lower-level equipment limits refer to limits acting on specific lines or equipment within the control boundary.
[0049] Step 8: Output the final effective limit and traceability chain.
[0050] In this embodiment of the invention, for each connected subgraph in the limit conflict graph, the final output result is generated by combining the final effective rule, the final effective limit, and the structured limit rule unit in the connected subgraph.
[0051] Specifically: Combining the final effective rule, the final effective limit, and the structured limit rule units in the connected subgraph, the final output result is composed of the controlled object name, applicable time, power flow direction, limit type, final effective rule, set of rules to be resolved, set of conflict types, set of reasons for resolution, and set of original evidence fragments corresponding to the final effective rule. Among them, the set of rules to be resolved includes other structured limit rule units in the connected subgraph except for the final effective rule; the set of conflict types includes the conflict types between the final effective rule and other structured limit rule units in the connected subgraph; the applicable time is the actual planning or query period in which the final rule applies in the current scheduling scenario; the set of reasons for resolution is formed based on the rule conflict relationship and resolution result, specifically by the large language model generating readable interpretations based on the corresponding clause evidence fragments; the set of original evidence fragments comes from the clause evidence fragments in each structured limit rule unit.
[0052] In addition, the final output includes a manual review flag and a confidence level. The manual review flag includes two options: "Requires Manual Confirmation" and "Automatic Approval." For results with a confidence level below a preset value or missing evidence, the output will show a "Requires Manual Confirmation" status and list the fields requiring confirmation. The confidence level is the same as the confidence level in the structured quota rule unit corresponding to the final effective rule. Through this output, schedulers can quickly view the final quota and its source basis, while retaining the review path for candidate clauses.
[0053] The above-mentioned solutions provided by the embodiments of the present invention have the following main advantages: 1. Using a large language model for semantic slot extraction and evidence organization of quota clauses, instead of directly outputting scheduling quota conclusions, can reduce the influence of model illusion on scheduling quota judgment and improve the verifiability of results.
[0054] 2. Convert natural language quota terms into structured quota rule units so that key information such as device object, quota value, effective time, applicable conditions, power flow direction and priority can be uniformly involved in subsequent verification and conflict resolution, thereby improving the structured utilization level of quota knowledge.
[0055] 3. By calculating scenario applicability through factors such as object consistency, time validity, direction consistency, and operation mode matching, it can avoid mismatches caused by keyword similarity alone and improve the matching accuracy between candidate quota terms and the current scheduling scenario.
[0056] 4. Construct a limit conflict diagram to explicitly express numerical conflicts, granular conflicts, temporary and regular conflicts, time overlap conflicts, and source version conflicts among multiple candidate limits, making the limit conflict identification process clearer and facilitating review by schedulers.
[0057] 5. Conflict resolution is performed on the connected subgraph of the limit conflict graph, and the final effective limit is determined by comprehensively considering the strength of security constraints, the validity period of temporary limits, the level of control boundaries, the fineness of conditions, and the support of evidence. This can take into account both power grid security constraints and actual dispatch management, and reduce the workload of manual table lookup and manual comparison.
[0058] 6. When outputting the final limit, the system outputs the master control rule, the rules to be resolved, the conflict type, and fragments of original evidence, forming a complete chain of evidence and improving the traceability and interpretability of the scheduling limit management results.
[0059] 7. Focusing on the conflict resolution process in the quota management process, it can be integrated as an auxiliary module of existing scheduling automation system, security verification system or quota knowledge base without replacing the original system, and has good engineering implementation.
[0060] To more clearly demonstrate the technical solution and its effects provided by this invention, the method provided by the embodiments of this invention is described in detail below with reference to specific examples. The station names, times, power flow, and quota values in the following specific examples are only used to illustrate the implementation process of this invention and do not constitute a limitation on actual power grid operating parameters. In this embodiment, the inputs are current dispatch operation scenario data, multi-source quota clause text, and related power grid topology and operation mode information; the outputs are the final effective quota under the current dispatch scenario, the resolved rule, the conflict type, the resolution reason, and the traceable evidence chain. Figure 2 The diagram illustrates the process of resolving power grid dispatch quota conflicts in this embodiment.
[0061] 1. Obtain scheduling scenario information and generate a quota query context.
[0062] To obtain the current scheduling scenario information, let the data set D of the scheduling scenario information be: ; Where, d i Let represent the i-th type of scheduling scenario information, and n be the total number of data categories in the scheduling scenario information.
[0063] The data set D of the scheduling scenario information includes one or more of the following: controlled object name, object type, operating mode, maintenance equipment, power flow direction, planned time period, real-time power flow, weather conditions, N-1 check status, scheduling instruction text, and alarm information. The scheduling scenario data set D is processed by object standardization, time normalization, direction normalization, and unit unification to obtain the quota query context: ; in, Representation of object standardization and time normalization processing, This indicates that the direction is normalized and the units are uniformly processed.
[0064] Quota query context Q is the name of the controlled object. Object type Operating mode Equipment maintenance Trends Planned time period Real-time trends Meteorological conditions N-1 verification status Dispatch instruction text and alarm information One or more of the above, when containing all of the above information, can be represented as: .
[0065] For example, if a dispatcher inputs a limit query request as: "From 20:00 to 24:00 on August 15, 2024, the I line of the 220kV A-B double-circuit transmission channel in a certain area is scheduled for maintenance. The current power flow from A to B is approximately 430MW. It is necessary to confirm the controllable power flow limit in this direction.", then the dispatcher will simultaneously obtain the maintenance plan, real-time cross-sectional power flow, grid topology, operation mode sheet, and weather warning information, and identify: o q =A 220kV power transmission corridor from Substation A to Substation B in a certain region; g q =Power transmission channel; m q =One maintenance cycle, single-cycle operation; x q =Station A, Line I under maintenance; d q =Station A → Station B; t q =2024-08-15 20:00 to 2024-08-15 24:00; p q=430MW; w q =High temperatures; n q =N-1 verification enabled. The above fields together constitute the quota query context for the current scheduling scenario.
[0066] II. Construction of constrained semantic slots based on large language models.
[0067] The scheduling scenario text and candidate quota terms are input into the semantic extraction module of the large language model. To avoid the large language model generating fields without original textual basis, quota slot templates are predefined: ; Among them, s o Indicates the name of the controlled object, s g Indicates object granularity, s k Indicates the limit type, s l Indicates the limit value, s u Indicates the unit of measurement, s c Indicates the applicable conditions, s b Indicates the effective date, s e Indicates the expiration time, s d Indicates the direction of the trend, s p Indicates the source type, s v This indicates a fragment of evidence related to the terms of the agreement, s h This indicates the confidence level. Large language models can only output structured results within the specified slot template, and each field must be bound to a fragment of the original clause evidence.
[0068] 3. Recall candidate quota clauses and form a candidate set.
[0069] Based on the quota query context Q, candidate quota clauses are retrieved from the quota knowledge base, operating procedure database, mode sheet database, maintenance plan database, temporary notice database, and historical shift handover record database. The retrieval process comprehensively employs symbol matching, vector semantic retrieval, and topological proximity filtering. For candidate clause c... i Its candidate recall score can be expressed as: ; ; Among them, M sym M represents the symbol matching score. vec M represents the vector semantic retrieval score. topo M represents the topological nearest neighbor score. time M represents the time-related score. srcThis represents the source reliability score. For the four candidate clauses mentioned above, R1=0.86, R2=0.89, R3=0.95, and R4=0.98. Clause four has the highest recall score because it is highly consistent with the current scenario in terms of target audience, trend direction, time window, and temporary notification source.
[0070] Clauses whose recall scores meet the threshold are selected to form the candidate quota clause set: .
[0071] For example, when the threshold θ R When the value is 0.80, all of the above clauses are included in the candidate limit clause set. .
[0072] For example, the system recalled the following candidate limit clauses: Clause 1 is "Under normal operation, the long-term power flow limit of the 220kV A-B double-circuit transmission channel in a certain area is 600MW in the direction from A to B"; Clause 2 is "During the peak summer season, the control limit of the A-B transmission channel in the direction from A to B is adjusted to 540MW"; Clause 3 is "When any circuit of the A-B double-circuit transmission channel is under maintenance, the power flow control value in the direction from A to B during the operation of the other circuit shall not exceed 480MW"; Clause 4 is "From 18:00 on August 15, 2024 to 06:00 on August 16, 2024, due to the impact of high temperature load and important supply guarantee requirements, the temporary power flow limit of the A-B channel in the direction from A to B shall be implemented at 450MW".
[0073] Let the set of candidate quota clause texts be: .
[0074] For any candidate clause c i Slot results are generated using a constrained large language model semantic extraction function: .
[0075] For example, for Clause 3, the following can be extracted: s o =A 220kV power transmission line from station A to station B in a certain area, s g =Transmission channel, s k =Trend ceiling, s l =480, s u =MW,s c =When one line is under maintenance and the other line is in operation, s b =Maintenance start time, s e = Maintenance completion time, s d =Station A → Station B, s p =Maintenance method limit clause, s v=“During any line maintenance...the power flow control value shall not exceed 480MW”, s h =0.96. The extraction results together constitute the structured slot description of Clause 3.
[0076] For clause four, the following can be extracted: s o =A 220kV power transmission line from station A to station B in a certain area, s g =Transmission channel, s k =Trend ceiling, s l =450, s u =MW,s c =High-temperature load and critical supply requirements, s b =2024-08-15 18:00, s e =2024-08-16 06:00, s d =Station A → Station B, s p =Temporary scheduling notification, s v =“From 18:00 on August 15, 2024 to 06:00 on August 16, 2024… the temporary power flow quota will be implemented at 450MW,” s h =0.97.
[0077] To ensure the traceability of extraction results, the following binding relationship is established between slots and evidence fragments: .
[0078] Fourth, convert candidate clauses into structured limit rule units.
[0079] Set of candidate quota terms For each candidate quota clause, a structured quota rule unit is generated using the slot extraction results obtained in step two. For any candidate clause c i Its corresponding structured quota rule unit It can be represented as: .
[0080] Preferably, a structured quota rule unit It can be represented as: ; The corresponding representations for each item are: controlled object name, object granularity, limit type, limit value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level.
[0081] For example, Clause 3 can be converted to r3=<220kV substation A to substation B transmission channel in a certain region, transmission channel, maintenance start time, maintenance end time, substation A → substation B, power flow limit, 480 MW, one maintenance and single-circuit operation, maintenance method limit, evidence fragment v3, 0.96>. Clause 4 can be converted to r4=<220kV substation A to substation B transmission channel in a certain region, transmission channel, 2024-08-15 18:00, 2024-08-16 06:00, substation A → substation B, power flow limit, 450 MW, high temperature load and important supply guarantee, temporary dispatch notice, evidence fragment v4, 0.97>.
[0082] For limit values with inconsistent units, perform unit conversion; for clauses with inconsistent object granularity, perform object hierarchy mapping. Unit conversion can be expressed as: ; in, This represents a unit conversion function used to convert the original units of measurement. The limit value below Convert to a unified unit of measurement The limit value below . Indicates its original unit of measurement, This indicates the unified unit of measurement.
[0083] For example, if a clause states "the power flow limit is 0.48GW", the system will convert it to 480MW; if a clause describes "Line II of Station A", and the current query object is "Transmission channel from Station A to Station B", the system will map the line to the corresponding transmission channel control boundary based on the topology.
[0084] V. Calculate the scenario applicability score of the candidate quota rules.
[0085] For each structured quota rule unit r i The applicability score for the current scheduling scenario Q is calculated. The applicability score comprehensively considers object consistency, time validity, direction consistency, operation mode matching, topology reachability, security check type matching, and condition integrity.
[0086] ; Among them, the above There are 7 weighting coefficients, which satisfy: .
[0087] To avoid incorrect retention of fields that do not meet strong constraints, let the strong constraint discrimination function be B. i If the strong constraints such as time, power flow direction, and control boundary are all satisfied, then B i =1; if any strong constraint field is not satisfied, then Bi =0. The corrected scenario applicability score is: .
[0088] For example, in the current scheduling scenario of "Station A to Station B corridor, Station A to Station B direction, one maintenance cycle, 20:00 to 24:00, high-temperature operation", A1 is calculated. ′ =0.62, A2 ′ =0.82, A3 ′ =0.94, A4 ′ =0.97. Among them, r1 is the long-term limit under normal conditions, which has low applicability; r2 is the limit for peak summer demand; r3 covers maintenance methods; r4 covers the current time window, object direction, and temporary operation requirements, so it has the highest applicability.
[0089] VI. Construct a limit conflict diagram and identify conflict types.
[0090] Candidate quota rules with modified scene adaptability scores higher than the threshold are selected as nodes to construct a quota conflict graph: .
[0091] For any two candidate quota rules r i and r j If the two overlap in the name of the controlled object or control boundary, time window, power flow direction, and quota type, and are inconsistent in the quota value, applicable conditions, or source priority, then a conflict edge is established: ; in, The conflict type indicates the conflict edge, including one or more of the following: numerical conflict of the same object, granularity conflict of superior and subordinate objects, conflict between temporary and regular quotas, directional quota conflict, time overlap conflict, conditional inclusion conflict, and source version conflict.
[0092] In this embodiment, r1, r2, r3, and r4 all involve the same controlled object name or control boundary, and all act in the direction from station A to station B, but their limit values are 600MW, 540MW, 480MW, and 450MW, respectively. Therefore, the system establishes a conflict edge: e 12 This indicates a numerical conflict between the regular quota and the peak quota; e 13 This indicates a conflict in mode conditions between the normal mode limit and the maintenance mode limit; e 14 This indicates a conflict in the source and values between the regular and temporary limits; e 24 This indicates a temporary coverage conflict between peak-hour limits and temporary limits; e 34 This indicates a conflict in numerical values and source priority between the maintenance method limit and the temporary limit.
[0093] VII. Conflict resolution based on the rules of security priority and evidence priority.
[0094] Perform conflict resolution on each connected subgraph in the quota conflict graph G. Suppose that the quota conflict graph G can be decomposed into several connected subgraphs: .
[0095] For any connected subgraph G k The corresponding candidate rule set for: .
[0096] In this embodiment, r1, r2, r3, and r4 belong to the same connected subgraph G1. Security-first and evidence-first resolution is performed on each candidate rule in this connected subgraph. The large language model is only used to generate readable interpretations based on evidence fragments and does not directly determine the final limit value. For any rule r... i Definition of resolution score: ; ; Among them, U i E represents the source priority score. i Indicates that the evidence supports the score, H i The score for safety strictness is represented by T. i Indicates the timeliness score, C i This indicates the score for the granularity of the conditions or the inclusion relationship of the conditions. There are 6 weighting coefficients.
[0097] The aforementioned resolution score is used to comprehensively prioritize candidate rules within the connected subgraph based on their scenario applicability, source priority, evidentiary support, security strictness, timeliness, and conditional granularity. It is a quantitative expression of "security priority and evidence priority" and is not directly used as the final limit value. Subsequently, when all rules are applicable and there are no mutually exclusive conditions, the final upper limit will be determined according to the security priority principle.
[0098] For upper limit types of limits, the smaller the limit value, the stronger the security constraint, satisfying: .
[0099] If multiple rules apply and there are no mutually exclusive conditions, the final effective limit is determined according to the principle of safety priority. For upper limit types, the final effective limit is... for: .
[0100] In this embodiment, the current set of valid rules is R1. eff={r2,r3,r4}, the valid rule set R1 eff This is a subset of rules that are actually effective in the current scheduling scenario after determining the applicability and conditions of the scenario. In this example, r1 is the normal long-term limit, which does not completely match the current return-to-maintenance scenario. Therefore, the effective rule set is R1. eff ={r2,r3,r4}. Since r4's validity period covers the current query time and its limit value is the most stringent, L1 is determined. * =450MW, final effective rule r * =r4. Further, calculate the remaining margin between the current power flow and the final limit: .
[0101] For example, the current real-time power flow p q =430MW, final quota L1 * =450MW, then the remaining margin is 20MW, indicating that the channel has not yet exceeded the limit, but is approaching the temporary control boundary. The dispatchers should be reminded to pay attention to subsequent power flow changes.
[0102] 8. Output the final effective limit and traceability chain.
[0103] Based on the final effective rules and final effective limits obtained in step seven, the system outputs the limit conflict resolution results. Let the final output result be: ;
[0104] Among them, L * Indicates the final effective limit, O * T represents the name of the controlled object. * Indicates the applicable time, D * Indicates the trend direction, K * Indicates the limit type, R * Represents the master rule (the final effective rule), R - H represents the set of rules to be resolved, Z represents the set of conflict types, and V represents the set of reasons for resolution.
[0105] In this embodiment, the output is: L * =450MW; O * =A 220kV power transmission corridor from Substation A to Substation B in a certain area; T * =2024-08-15 20:00 to 2024-08-15 24:00; D * =Station A → Station B; K * =Trend ceiling; R * =r4;R -={r1,r2,r3}. Here, r1 is eliminated because the normal mode limit does not fully match the current maintenance mode; r2 is eliminated because the peak-load limit is covered by the currently valid temporary dispatch notification; r3 is eliminated because although the maintenance mode limit applies, the temporary limit has a higher priority and stricter limits within the validity period of the temporary notification.
[0106] To ensure the traceability of the final results, the system constructs a limit-based chain. : .
[0107] In this embodiment, the chain of reference includes, in sequence: the quota query context. Candidate Clause Set , effective rule set R', limit conflict graph G, master rule (final effective rule) r * Final limit L * And evidence fragment V * Among them, V * The output should state "From 18:00 on August 15, 2024 to 06:00 on August 16, 2024... the temporary power flow limit will be implemented at 450MW". If the confidence level of the result is lower than the preset threshold or there is missing evidence, the output should state "Manual confirmation required" and list the fields requiring confirmation; otherwise, the output should state "Automatic approval".
[0108] Through the above output, dispatchers can view the final effective limit, master control rules, resolved rules, conflict types, resolution reasons, and original evidence fragments in the current scenario. They can also trace back the source terms, conflict rules, and resolution decision-making process of the limit along the basis chain, thereby improving the accuracy, interpretability, and verifiability of dispatch limit management.
[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of the above embodiments can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.), including several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0110] Example 2 This invention also provides a power grid dispatch quota conflict resolution system, which is mainly used to implement the methods provided in the foregoing embodiments, such as... Figure 3 As shown, the system mainly includes: The quota query context generation unit is used to retrieve scheduling scenario information and generate quota query context; The quota query slot template construction unit is used to construct the quota query slot template; wherein, the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level; The candidate limit clause retrieval unit is used to retrieve several candidate limit clauses from limit clause data from multiple sources, in conjunction with the limit query context. The semantic extraction unit based on the large language model is used to extract the semantics of each candidate quota clause based on the large language model and the quota query slot template to obtain the structured quota rule unit corresponding to each candidate quota clause. The applicability score calculation unit is used to calculate the applicability score of each structured quota rule unit in relation to the current scheduling scenario. The quota conflict graph construction unit is used to select structured quota rule units with an applicability score higher than the threshold as nodes. If two structured quota rule units conflict, a conflict edge is established to form a quota conflict graph. The conflict resolution unit is used to perform security-first and evidence-first resolution on each connected subgraph in the limit conflict graph: if all the structured limit rule units included are applicable and there are no mutually exclusive conditions, the structured limit rule unit with the smallest limit value is selected as the final effective rule and its limit value is used as the final effective limit. At the same time, the large language model generates a readable interpretation based on the clause evidence fragments in the structured limit rule unit. The final output generation unit is used to generate the final output result for each connected subgraph in the quota conflict graph by combining the final effective rule, the final effective quota, and the structured quota rule unit in the connected subgraph.
[0111] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above.
[0112] Example 3 The present invention also provides a processing device, such as Figure 4 As shown, it mainly includes: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the foregoing embodiments.
[0113] Furthermore, the processing device also includes at least one input device and at least one output device; in the processing device, the processor, memory, input device, and output device are connected via a bus.
[0114] In this embodiment of the invention, the specific types of the memory, input device, and output device are not limited; for example: Input devices can be touchscreens, image acquisition devices, physical buttons, or mice, etc. The output device can be a display terminal; The memory can be random access memory (RAM) or non-volatile memory, such as disk storage.
[0115] Example 4 The present invention also provides a readable storage medium storing a computer program that, when executed by a processor, implements the method provided in the foregoing embodiments.
[0116] In this embodiment of the invention, the readable storage medium is a computer-readable storage medium and can be disposed in the aforementioned processing device, for example, as a memory in the processing device. Furthermore, the readable storage medium can also be any medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.
[0117] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims. The information disclosed in the background section is intended only to enhance the understanding of the overall background technology of the present invention and should not be construed as an admission or implication in any way that such information constitutes prior art known to those skilled in the art.
Claims
1. A method for resolving conflicts related to power grid dispatch quotas, characterized in that, include: Retrieve scheduling scenario information and generate a quota query context; Construct a quota query slot template; the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level; Based on the aforementioned limit query context, several candidate limit terms are retrieved from limit term data from multiple sources; Based on the large language model, semantic extraction is performed on each candidate limit clause through the limit query slot template to obtain the structured limit rule unit corresponding to each candidate limit clause. For each structured quota rule unit, calculate its applicability score to the current scheduling scenario; Select structured quota rule units with applicability scores higher than the threshold as nodes. If two structured quota rule units conflict, establish a conflict edge to form a quota conflict graph. For each connected subgraph in the limit conflict graph, security priority and evidence priority resolution are performed: if all the structured limit rule units contained therein are applicable and there are no mutually exclusive conditions, then the structured limit rule unit with the smallest limit value is selected as the final effective rule, and its limit value is used as the final effective limit. At the same time, the large language model generates a readable interpretation based on the clause evidence fragments in the structured limit rule unit. For each connected subgraph in the quota conflict graph, the final output result is generated by combining the final effective rule, the final effective quota, and the structured quota rule unit in the connected subgraph.
2. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, The step of generating a quota query context using the current scheduling scenario information includes: The current scheduling scenario information includes one or more of the following: controlled object name, object type, operating mode, maintenance equipment, power flow direction, planned time period, real-time power flow, weather conditions, N-1 check status, scheduling instruction text, and alarm information; The current scheduling scenario information is processed by object standardization and time normalization to generate a quota query context.
3. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, The process of retrieving several candidate limit terms from multi-source limit term data, based on the limit query context, includes: Combining the context of the quota query, the symbol matching score, vector semantic retrieval score, topological proximity score, time correlation score, and source reliability score of each quota clause are calculated separately. The candidate recall score of each quota clause data is obtained by combining all the above scores. Select the data of the limit terms with a recall score not less than the preset threshold as candidate limit terms.
4. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, Semantic extraction of each candidate limit clause using the limit query slot template includes: For limit values with inconsistent units, perform unit conversion; for clauses with inconsistent object granularity, perform object hierarchy mapping.
5. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, The calculation of the applicability score of each structured quota rule unit to the current scheduling scenario includes: For each structured quota rule unit, an initial applicability score is calculated by considering object consistency, time validity, directional consistency, operational mode matching, topology reachability, security check type matching, and condition integrity. ; Set a strong constraint discriminant function B i If the time, current direction, and control boundary segment are all satisfied, then B i =1; if any condition is not met, then B i =0; Using the strongly constrained discriminant function B i Initial applicability score Make corrections to obtain the final applicability score. .
6. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, If two structured quota rule units conflict, establishing a conflict edge includes: If two structured quota rules overlap in terms of controlled objects or control boundaries, time windows, power flow direction, and quota type, and are inconsistent in terms of quota values, applicable conditions, or source priority, then they are in conflict, and a conflict edge is established.
7. The method for resolving power grid dispatch quota conflicts according to claim 1, characterized in that, The process of generating the final output result for each connected subgraph in the limit conflict graph, combining the final effective rule, the final effective limit, and the structured limit rule unit in the connected subgraph, includes: Combining the final effective rule, the final effective limit, and the structured limit rule unit in the connected subgraph, the final output result is composed of the controlled object name, applicable time, power flow direction, limit type, final effective rule, set of rules to be resolved, set of conflict types, set of reasons for resolution, and set of original evidence fragments corresponding to the final effective rule. The set of rules to be resolved includes other structured quota rule units in the connected subgraph besides the final effective rule, and the set of conflict types includes the conflict types between the final effective rule and other structured quota rule units in the connected subgraph.
8. A power grid dispatching quota conflict resolution system, characterized in that, To implement the method according to any one of claims 1 to 7, comprising: The quota query context generation unit is used to retrieve scheduling scenario information and generate quota query context; The quota query slot template construction unit is used to construct the quota query slot template; wherein, the quota query slot template includes: controlled object name, object granularity, quota value, unit of measurement, applicable conditions, effective time, expiration time, flow direction, source type, clause evidence fragment, and confidence level; The candidate limit clause retrieval unit is used to retrieve several candidate limit clauses from limit clause data from multiple sources, in conjunction with the limit query context. The semantic extraction unit based on the large language model is used to extract the semantics of each candidate quota clause based on the large language model and the quota query slot template to obtain the structured quota rule unit corresponding to each candidate quota clause. The applicability score calculation unit is used to calculate the applicability score of each structured quota rule unit in relation to the current scheduling scenario. The quota conflict graph construction unit is used to select structured quota rule units with an applicability score higher than the threshold as nodes. If two structured quota rule units conflict, a conflict edge is established to form a quota conflict graph. The conflict resolution unit is used to perform security-first and evidence-first resolution on each connected subgraph in the limit conflict graph: if all the structured limit rule units included are applicable and there are no mutually exclusive conditions, the structured limit rule unit with the smallest limit value is selected as the final effective rule and its limit value is used as the final effective limit. At the same time, the large language model generates a readable interpretation based on the clause evidence fragments in the structured limit rule unit. The final output generation unit is used to generate the final output result for each connected subgraph in the quota conflict graph by combining the final effective rule, the final effective quota, and the structured quota rule unit in the connected subgraph.
9. A processing device, characterized in that, include: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1 to 7.
10. A readable storage medium storing a computer program, characterized in that, When a computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.
