Method, medium and device for searching power distribution network multi-element business fuzzy semantics

Abstract

The application discloses a search method, medium and device for power distribution network multi-element business fuzzy semantics, and belongs to the technical field of power distribution network operation and maintenance. In the application, the fuzzy semantic level of power distribution network query is determined in the determined search range, and the corresponding search link is selected. The recall candidate result of the candidate document associated with the power distribution network query is screened, and the search result meeting the relevance of the power distribution network query business and the physical rationality is obtained. The search range of the query can be accurately positioned, and the processing strength of the query fuzzy semantics can be matched. The search efficiency is improved, the search accuracy is ensured, and the problems of extensive search range limitation and mixed search results of a large amount of irrelevant noise in the current power distribution network query business search are solved.

Description

Technical Field

[0001] This invention relates to a retrieval method, medium, and device for fuzzy semantics of diverse services in distribution networks, belonging to the field of distribution network operation and maintenance technology. Background Technology

[0002] With the advancement of digital transformation of power distribution networks, massive heterogeneous data resources have emerged, encompassing equipment operation, fault handling, and production maintenance. These data have various business scenarios, including fault assessment, ledger query, work order retrieval, and operational traceability. The data focus, search scope, and result judgment criteria differ significantly across these different business scenarios. However, existing retrieval systems typically fail to differentiate between business types before searching, resulting in a broad and limited search scope.

[0003] Furthermore, frontline operations and maintenance personnel often use colloquial expressions, abbreviated terms, and incomplete descriptions in actual interactions, resulting in a high degree of semantic ambiguity in queries due to missing entities and insufficient context. Faced with complex and diverse business needs and highly fuzzy queries, if the existing system continues to use a uniform "one-size-fits-all" retrieval path, it is highly susceptible to problems such as missing key information, shifted candidate results, or the introduction of a large amount of irrelevant noise. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a retrieval method, medium and device for fuzzy semantics of multiple services in the distribution network. It can accurately locate the retrieval range of the query and match the processing intensity of the fuzzy semantics of the query, thereby improving the retrieval efficiency while ensuring the retrieval accuracy. It solves the problems of the current distribution network query service retrieval being limited and coarse, and the retrieval results being mixed with a large amount of irrelevant noise.

[0005] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution:

[0006] The first aspect of this invention provides a retrieval method for fuzzy semantics of multiple services in a distribution network, comprising:

[0007] Based on the input distribution network query data, determine the retrieval scope of the distribution network query;

[0008] Fuzzy semantic quantization processing is performed on the distribution network query data to determine the fuzzy semantic level of the distribution network query, including:

[0009] Calculate the overall fuzziness of the distribution network query based on the distribution network query data;

[0010] The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows:

[0011] like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic;

[0012] like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics;

[0013] like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics;

[0014] in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold;

[0015] ;

[0016] ;

[0017] in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism;

[0018] Based on the fuzzy semantic level, select the corresponding retrieval link within the retrieval scope of the power distribution network query;

[0019] Based on the retrieval link, determine the recall candidate results of candidate documents associated with the power distribution network query;

[0020] The recalled candidate results are reordered to obtain retrieval results that meet the business relevance and physical rationality requirements of the power distribution network query.

[0021] Furthermore, determining the retrieval scope of the distribution network query based on the input distribution network query data includes:

[0022] The data queryed from the power distribution network is classified to obtain the business type of the power distribution network query.

[0023] Based on the business type of the distribution network query, determine the time range, distribution network topology space range, and text semantic document range of the distribution network query;

[0024] The retrieval scope of the distribution network query is determined based on the time range, topological space range, and semantic document range of the distribution network query.

[0025] Furthermore, determining the retrieval scope of the distribution network query based on the time range, topological space range, and semantic document range of the distribution network query includes:

[0026] The following formula is used to represent the retrieval scope of the distribution network query:

[0027] ;

[0028] in, Indicates distribution network query The scope of the search; Indicates distribution network query Time range Indicates distribution network query The spatial range of the distribution network topology, Indicates distribution network query The text semantic document scope, Indicates the business type of the power distribution network query; Indicates intersection;

[0029] Distribution network query The business type is represented as:

[0030] ;

[0031] in, This represents a function that takes the maximum value of the independent variable. Indicates distribution network query Business type, Indicates distribution network query Belongs to candidate business type The probability of; Represents a collection of diverse business types. , This indicates reasoning-related business. This indicates query-type services. This indicates pending business items.

[0032] Furthermore, the formula for calculating the comprehensive fuzziness of the distribution network query based on the distribution network query data is as follows:

[0033] ;

[0034] in, Indicates distribution network query Intentional entropy Indicates distribution network query Contextual completeness; , , , The non-negative weighting coefficients representing the overall ambiguity, and ;

[0035] ;

[0036] in, Indicates distribution network query Belongs to candidate business type The probability, It represents a collection of diverse business types.

[0037] Furthermore, the step of selecting the corresponding retrieval link in the retrieval range of the distribution network query based on the fuzzy semantic level includes: if the fuzzy semantic level of the distribution network query is level two fuzzy semantic or level three fuzzy semantic, then a hybrid recall link is adopted.

[0038] The hybrid recall link includes a basic hybrid recall link that balances semantic matching and graph association with low computational complexity, and an extended hybrid recall link that considers keyword matching degree and time consistency.

[0039] Furthermore, determining the recall candidate results of candidate documents associated with the distribution network query based on the retrieval link includes: calculating the recall score of the distribution network query for the candidate documents using the basic hybrid recall link, with the formula as follows:

[0040] ;

[0041] in, This indicates that the basic hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Vector semantic similarity; Indicates distribution network query Related node subgraphs Indicates candidate documents The associated node subgraph; Indicates distribution network query Related node subgraphs and candidate documents The degree of structural overlap of the associated node subgraphs in the distribution network map; , Denotes the non-negative weighting coefficient of the mixed recall, and ;

[0042] Based on the recall score, the candidate documents associated with the distribution network query are sorted in descending order to obtain the recall candidate results.

[0043] Furthermore, the step of determining the recall candidate results of candidate documents associated with the distribution network query based on the retrieval link also includes: calculating the recall score of the distribution network query for the candidate documents using an extended hybrid recall link, with the formula as follows:

[0044] ;

[0045] in, This indicates that the extended hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Vector semantic similarity; Indicates distribution network query Related node subgraphs Indicates candidate documents The associated node subgraph; Indicates distribution network query Related node subgraphs and candidate documents The degree of structural overlap of the associated node subgraphs in the distribution network map; Indicates distribution network query With candidate documents Keyword matching similarity, Indicates distribution network query With candidate documents Temporal consistency similarity; , , and Denotes the non-negative weighting coefficient of the mixed recall, and ;

[0046] Based on the recall score, the candidate documents associated with the distribution network query are sorted in descending order to obtain the recall candidate results.

[0047] Furthermore, the reordering of the recalled candidate results to obtain retrieval results that satisfy the relevance and physical rationality of the distribution network query business includes:

[0048] The re-ranking score of candidate documents in the recall results is calculated using the following formula:

[0049] ;

[0050] in, Indicates distribution network query About candidate documents The reordering score; Indicates distribution network query About candidate documents Recall rating, Indicates the business type corresponding to the distribution network query. Next candidate document The weight of business importance; Indicates distribution network query Corresponding equipment With candidate documents Associated devices Topological distance attenuation relationship under the premise of satisfying basic physical rationality; Indicates distribution network query About candidate documents Rule review factor;

[0051] ;

[0052] in, Indicates equipment and The distance between the distribution network topologies, Represents positive real numbers; This represents an exponential function with the natural constant e as its base.

[0053] ;

[0054] in, This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. express and An indicator of whether there is an electrical connection between them. Indicates distribution network query The corresponding voltage level, Indicates candidate documents The corresponding voltage level, Indicates candidate documents timestamp, Indicates distribution network query Allowed time range;

[0055] Candidate documents are reordered from highest to lowest according to the reordering score to form the final recall candidate results;

[0056] The top-ranked candidate documents in the final recall results will be used as the search results.

[0057] A second aspect of the present invention also provides a retrieval device for fuzzy semantics of multiple services in a distribution network, comprising:

[0058] The retrieval range constraint module is used to determine the retrieval range of the distribution network query based on the input distribution network query data;

[0059] The fuzzy semantic grading module is used to perform fuzzy semantic quantization processing on distribution network query data to determine the fuzzy semantic level of the distribution network query, including:

[0060] Calculate the overall fuzziness of the distribution network query based on the distribution network query data;

[0061] The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows:

[0062] like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic;

[0063] like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics;

[0064] like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics;

[0065] in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold;

[0066] ;

[0067] ;

[0068] in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism;

[0069] The retrieval link selection module is used to select the corresponding retrieval link in the retrieval range of the power distribution network query based on the fuzzy semantic level.

[0070] The hybrid recall module is used to determine the recall candidate results of candidate documents associated with the power distribution network query based on the retrieval link;

[0071] The filtering module is used to reorder the recalled candidate results to obtain retrieval results that meet the business relevance and physical rationality of the power distribution network query.

[0072] A third aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the retrieval method for fuzzy semantics of multiple services in the power distribution network as described above.

[0073] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0074] 1. This invention determines the fuzzy semantic level of the distribution network query within a defined search scope and selects the corresponding search link. By filtering the recall candidate results of candidate documents associated with the distribution network query, it obtains search results that meet the business relevance and physical rationality of the distribution network query. It can accurately locate the search scope of the query and match the processing intensity of the fuzzy semantics of the query, thereby improving search efficiency while ensuring search accuracy. It solves the problems of the current distribution network query business having a broad search scope and search results mixed with a large amount of irrelevant noise.

[0075] 2. This invention selects different retrieval links based on different levels of fuzzy semantics, and introduces a hybrid recall mechanism in second-level or third-level fuzzy semantic processing scenarios to achieve synergy between keyword matching, vector semantic similarity and topological graph association capabilities, taking into account both recall speed and recall coverage, and achieving fast recall.

[0076] 3. In the process of re-ranking the recalled candidate results, this invention introduces the topological distance attenuation relationship between the corresponding equipment in the distribution network query and the associated equipment of the candidate document, as well as the rule review factor. By implementing hard constraint review and filtering on candidate results that are electrically disconnected, have mismatched voltage levels, or inconsistent time ranges, this invention not only considers the semantic association strength, but also the document priority, the topological proximity relationship of the equipment, and physical constraints. This can significantly reduce the false recall rate and false ranking rate of results that are "literally relevant but physically unrelated", thus achieving quasi-re-ranking. Attached Figure Description

[0077] Figure 1 This is a flowchart of a retrieval method for fuzzy semantics of multiple services in a power distribution network, provided in an embodiment of the present invention.

[0078] Figure 2 This is a flowchart of determining the retrieval range for a power distribution network query, provided in an embodiment of the present invention;

[0079] Figure 3This is a schematic diagram of the structure of the retrieval device for fuzzy semantics of multiple services in the power distribution network provided in an embodiment of the present invention. Detailed Implementation

[0080] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0081] Example 1

[0082] like Figure 1 As shown, this embodiment provides a retrieval method for fuzzy semantics of multiple services in the distribution network, including:

[0083] Step 1: Based on the input distribution network query data, determine the retrieval scope of the distribution network query; such as... Figure 2 As shown, specifically:

[0084] The data queryed from the power distribution network is classified to obtain the business type of the power distribution network query.

[0085] Based on the business type of the distribution network query, determine the time range, distribution network topology space range, and text semantic document range of the distribution network query;

[0086] Based on the time range, topological space range, and semantic document range of the distribution network query, the retrieval scope of the distribution network query is determined, including:

[0087] The following formula is used to represent the retrieval scope of the distribution network query:

[0088] ;

[0089] in, Indicates distribution network query The scope of the search; Indicates distribution network query Time range Indicates distribution network query The spatial range of the distribution network topology, Indicates distribution network query The text semantic document scope, Indicates the business type of the power distribution network query; Indicates intersection;

[0090] Distribution network query The business type is represented as:

[0091] ;

[0092] in, This represents a function that takes the maximum value of the independent variable. Indicates distribution network query Business type, Indicates distribution network query Belongs to candidate business type The probability of; Represents a collection of diverse business types. , This indicates reasoning-based business logic, suitable for scenarios involving tracing the causes of phenomena, analyzing faults, and explaining anomalies. This indicates query-type services, suitable for scenarios involving ledger retrieval, work order status retrieval, and operation record retrieval. This indicates pending business information, suitable for scenarios where the intent is incomplete, requires completion, or is an open-ended search.

[0093] It should be noted that, in this embodiment, Using Bayes' theorem, the determination can be made comprehensively based on the equipment entities, fault phenomenon words, status query words, work order instruction words, time constraint words, and contextual expression features in the distribution network query.

[0094] When the query is "What are the reasons for the frequent alarms of a certain transformer?", this query represents a scenario where a phenomenon is being investigated for an explanation. ; The time window is limited to the period from 24 hours before the anomaly occurred to 2 hours after the anomaly occurred. The scope is limited to the feeder where the target transformer is located and the upstream and downstream electrical nodes. Limited to alarm logs, fault waveforms, protection action records, and abnormal event chain documents;

[0095] When the query is "work order status for a certain device", the query is considered a precise query scenario. ; Limited to the lifecycle period of the work order corresponding to the equipment. Limited to the target device itself and its directly affiliated distribution area. Limited to equipment ledgers, work order transfer records, and structured inspection forms;

[0096] When the query is "Why is the power tripping so frequently in this area lately?", the query is vague and lacks complete intent. ; The restrictions have been relaxed to nearly a month. The region representation is mapped to multiple candidate feeders. The scope has been broadened to include summary-type runtime documents and comprehensive overview documents, in order to provide sufficient background for subsequent completion of the larger model intent;

[0097] Different business types obtain retrieval boundaries that match their business objectives before entering the recall stage, thereby avoiding large-scale entry of data from irrelevant time periods, irrelevant topological regions, and irrelevant document semantic spaces into the subsequent recall chain, reducing candidate space redundancy, and improving the targeting of retrieval.

[0098] Step 2: Perform fuzzy semantic quantization on the distribution network query data to determine the fuzzy semantic level of the distribution network query; specifically:

[0099] The comprehensive fuzziness of the distribution network query is calculated based on the distribution network query data, using the following formula:

[0100] ;

[0101] in, Indicates distribution network query The overall fuzziness; Indicates distribution network query Intentional entropy Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism, Indicates distribution network query Contextual completeness; , , , The non-negative weighting coefficients representing the overall ambiguity, and ;

[0102] ;

[0103] The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows:

[0104] like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic;

[0105] like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics;

[0106] like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics;

[0107] in, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold;

[0108] ;

[0109] ;

[0110] in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification;

[0111] This embodiment introduces... and This allows for dynamic adjustment of the hierarchical boundaries based on the clarity of the query's scope, the clarity of the entity, and the degree of colloquialism, thereby avoiding the processing intensity mismatch problem caused by using fixed thresholds.

[0112] Step 3: Select the corresponding retrieval link in the retrieval scope of the distribution network query according to the fuzzy semantic level; specifically: if the fuzzy semantic level of the distribution network query is level two or level three fuzzy semantic, then a hybrid recall link is used.

[0113] It should be noted that, in this embodiment, since the business intent of the distribution network query in the first-level fuzzy semantics is relatively clear, the key entities are relatively clear, and the context is relatively complete, a fast retrieval method combining keyword inverted index and entity exact matching is adopted for the first-level fuzzy semantics to achieve high-efficiency result retrieval with low processing cost.

[0114] Among them, the hybrid recall link includes a basic hybrid recall link that takes into account both semantic matching and graph association with low computational complexity, and an extended hybrid recall link that considers keyword matching degree and time consistency.

[0115] Step 4: Based on the retrieval link, determine the recall candidate results of candidate documents associated with the distribution network query; specifically:

[0116] If the recall uses a basic hybrid recall link, the recall score for the distribution network query regarding the candidate document is calculated using the following formula: ;

[0117] in, This indicates that the basic hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Vector semantic similarity; Indicates distribution network query Related node subgraphs Indicates candidate documents The associated node subgraph; Indicates distribution network query Related node subgraphs and candidate documents The degree of structural overlap of the associated node subgraphs in the distribution network map; , Denotes the non-negative weighting coefficient of the mixed recall, and ;

[0118] If the recall uses an extended hybrid recall link, the recall score for the distribution network query on the candidate document is calculated using the following formula: ;

[0119] in, This indicates that the extended hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Keyword matching similarity, Indicates distribution network query With candidate documents Temporal consistency similarity; , , and Denotes the non-negative weighting coefficient of the mixed recall, and ;

[0120] Based on the recall score, the candidate documents associated with the distribution network query are sorted in descending order to obtain the recall candidate results.

[0121] Step 5: Reorder the recalled candidate results to obtain retrieval results that meet the business relevance and physical rationality requirements of the distribution network query; specifically:

[0122] The re-ranking score of candidate documents in the recall results is calculated using the following formula:

[0123] ;

[0124] in, Indicates distribution network query About candidate documents The reordering score, Indicates distribution network query About candidate documents Recall rating;

[0125] Indicates the business type corresponding to the distribution network query. Next candidate document The business importance weight, for example:

[0126] when Increase the weight of alarm logs, fault waveforms, protection action records, and abnormal event chain documents; when Increase the weighting of equipment ledgers, work order status, maintenance records, and inspection records; when Increase the weight of summary-type running documents, recently frequently associated documents, and comprehensive overview documents;

[0127] If you are searching for "the cause of abnormal voltage at the end of a feeder", then... The trip alarm log of a circuit breaker upstream of the same feeder was recalled. Because this alarm log is key evidence for fault deduction, it received an extremely high business importance weight. Furthermore, it is electrically connected to the query point and is close in distance, ultimately achieving a very high re-ranking score and being accurately pushed to the front;

[0128] Indicates distribution network query Corresponding equipment With candidate documents Associated devices Topological distance attenuation relationship under the premise of satisfying basic physical rationality; Indicates distribution network query About candidate documents Rule review factor;

[0129] ;

[0130] in, Indicates equipment and The distance between the distribution network topologies, Represents positive real numbers; This represents an exponential function with the natural constant e as its base.

[0131] ;

[0132] in, This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. express and An indicator of whether there is an electrical connection between them. Indicates distribution network query The corresponding voltage level, Indicates candidate documents The corresponding voltage level, Indicates candidate documents timestamp, Indicates distribution network query Allowed time range;

[0133] It should be noted that, Applying topological distance decay to candidate results that satisfy basic physical rationality is a soft decay reordering; Hard filtering is applied to results of electrical disconnection, voltage level mismatch, and time range inconsistency, which is a hard constraint review filtering. The two are not simply superimposed general ranking optimization, but form a synergistic relationship of "soft decay reordering and hard constraint review filtering".

[0134] For example, if you search for "cause of overload of transformer A", in traditional search, the "transformer B high temperature alarm log" in another region will be incorrectly recalled based on the high similarity of vector semantics "transformer, overheat / overload", and this log will be incorrectly sorted at the high position.

[0135] In this embodiment, transformer A and transformer B belong to completely different feeders, are mutually insulated, and have no electrical connection. Substituting the rule review factor directly leads to , The correlation instantly resets to zero; thus, the spurious correlation result, which was "literally highly correlated but physically unrelated," was precisely eliminated during the reordering review stage.

[0136] The candidate documents in the recall candidate results are reordered from high to low according to the reordering score to form the final recall candidate results;

[0137] The top-ranked candidate documents in the final recall results will be used as the search results.

[0138] Example 2

[0139] like Figure 3 As shown, this embodiment provides a retrieval device for fuzzy semantics of multiple services in the power distribution network, including:

[0140] The retrieval range constraint module is used to determine the retrieval range of the distribution network query based on the input distribution network query data;

[0141] The fuzzy semantic grading module is used to perform fuzzy semantic quantization processing on distribution network query data to determine the fuzzy semantic level of the distribution network query, including:

[0142] Calculate the overall fuzziness of the distribution network query based on the distribution network query data;

[0143] The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows:

[0144] like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic;

[0145] like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics;

[0146] like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics;

[0147] in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold;

[0148] ;

[0149] ;

[0150] in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism;

[0151] The retrieval link selection module is used to select the corresponding retrieval link in the retrieval range of the power distribution network query based on the fuzzy semantic level.

[0152] The hybrid recall module is used to determine the recall candidate results of candidate documents associated with the power distribution network query based on the retrieval link;

[0153] The filtering module is used to reorder the recalled candidate results to obtain retrieval results that meet the business relevance and physical rationality of the power distribution network query.

[0154] The specific functions of each module described above are explained in the relevant content of the method in Embodiment 1, and will not be repeated here.

[0155] Example 3

[0156] This embodiment provides a computer-readable storage medium on which a computer program is stored. When executed by a processor, the computer program implements the following method steps:

[0157] Based on the input distribution network query data, determine the retrieval scope of the distribution network query;

[0158] Fuzzy semantic quantization processing is performed on the distribution network query data to determine the fuzzy semantic level of the distribution network query, including:

[0159] Calculate the overall fuzziness of the distribution network query based on the distribution network query data;

[0160] The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows:

[0161] like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic;

[0162] like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics;

[0163] like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics;

[0164] in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold;

[0165] ;

[0166] ;

[0167] in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism;

[0168] Based on the fuzzy semantic level, select the corresponding retrieval link within the retrieval scope of the power distribution network query;

[0169] Based on the retrieval link, determine the recall candidate results of candidate documents associated with the power distribution network query;

[0170] The recalled candidate results are reordered to obtain retrieval results that meet the business relevance and physical rationality requirements of the power distribution network query.

[0171] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0172] This application is described with reference to flowchart illustrations of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each step in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the process. Figure 1 One or more processes or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0173] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 The function specified in one or more processes.

[0174] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 Steps of a specified function in one or more processes.

[0175] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A retrieval method for fuzzy semantics of multiple services in distribution networks, characterized in that, include: Based on the input distribution network query data, determine the retrieval scope of the distribution network query; Fuzzy semantic quantization processing is performed on the distribution network query data to determine the fuzzy semantic level of the distribution network query, including: Calculate the overall fuzziness of the distribution network query based on the distribution network query data; The formula for calculating the overall fuzziness of a distribution network query is as follows: ； in, Indicates distribution network query Intentional entropy Indicates distribution network query Contextual completeness; , , , The non-negative weighting coefficients representing the overall ambiguity, and ; ； in, Indicates distribution network query Belongs to candidate business type The probability, Represents a collection of diverse business types; The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows: like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic; like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics; like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics; in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold; ； ； in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism; Based on the fuzzy semantic level, select the corresponding retrieval link within the retrieval scope of the power distribution network query; Based on the retrieval link, determine the recall candidate results of candidate documents associated with the power distribution network query; The recalled candidate results are reordered to obtain retrieval results that meet the requirements of distribution network query business relevance and physical rationality, including: The re-ranking score of candidate documents in the recall results is calculated using the following formula: ； in, Indicates distribution network query About candidate documents The reordering score; Indicates distribution network query About candidate documents Recall rating, Indicates the business type corresponding to the distribution network query. Next candidate document The weight of business importance; Indicates distribution network query Corresponding equipment With candidate documents Associated devices Topological distance attenuation relationship under the premise of satisfying basic physical rationality; Indicates distribution network query About candidate documents Rule review factor; ； in, Indicates equipment and The distance between the distribution network topologies, Represents positive real numbers; This represents an exponential function with the natural constant e as its base. ； in, This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. express and An indicator of whether there is an electrical connection between them. Indicates distribution network query The corresponding voltage level, Indicates candidate documents The corresponding voltage level, Indicates candidate documents timestamp, Indicates distribution network query Allowed time range; Candidate documents are reordered from highest to lowest according to the reordering score to form the final recall candidate results; The top-ranked candidate documents in the final recall results will be used as the search results.

2. The retrieval method for fuzzy semantics of multiple services in distribution networks according to claim 1, characterized in that, The determination of the retrieval scope for the distribution network query based on the input distribution network query data includes: The data queryed from the power distribution network is classified to obtain the business type of the power distribution network query. Based on the business type of the distribution network query, determine the time range, distribution network topology space range, and text semantic document range of the distribution network query; The retrieval scope of the distribution network query is determined based on the time range, topological space range, and semantic document range of the distribution network query.

3. The retrieval method for fuzzy semantics of multiple services in distribution networks according to claim 2, characterized in that, The process of determining the retrieval scope of the distribution network query based on the time range, topological space range, and semantic document range of the distribution network query includes: The following formula is used to represent the retrieval scope of the distribution network query: ； in, Indicates distribution network query The scope of the search; Indicates distribution network query Time range Indicates distribution network query The spatial range of the distribution network topology, Indicates distribution network query The text semantic document scope, Indicates the business type of the power distribution network query; Indicates intersection; Distribution network query The business type is represented as: ； in, This represents a function that takes the maximum value of the independent variable. Indicates distribution network query Business type, Indicates distribution network query Belongs to candidate business type The probability of; Represents a collection of diverse business types. , This indicates reasoning-related business. This indicates query-type services. This indicates pending business items.

4. The retrieval method for fuzzy semantics of multiple services in distribution networks according to claim 1, characterized in that, The step of selecting the corresponding retrieval link in the retrieval range of the power distribution network query according to the fuzzy semantic level includes: if the fuzzy semantic level of the power distribution network query is level two fuzzy semantic or level three fuzzy semantic, then a hybrid recall link is adopted. The hybrid recall link includes a basic hybrid recall link that balances semantic matching and graph association with low computational complexity, and an extended hybrid recall link that considers keyword matching degree and time consistency.

5. The retrieval method for fuzzy semantics of multiple services in distribution networks according to claim 4, characterized in that, The step of determining the recall candidate results of candidate documents associated with the distribution network query based on the retrieval link includes: calculating the recall score of the distribution network query for the candidate documents using the basic hybrid recall link, with the formula as follows: ； in, This indicates that the basic hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Vector semantic similarity; Indicates distribution network query Related node subgraphs Indicates candidate documents The associated node subgraph; Indicates distribution network query Related node subgraphs and candidate documents The degree of structural overlap of the associated node subgraphs in the distribution network map; , Denotes the non-negative weighting coefficient of the mixed recall, and ; Based on the recall score, the candidate documents associated with the distribution network query are sorted in descending order to obtain the recall candidate results.

6. The retrieval method for fuzzy semantics of multiple services in distribution networks according to claim 4, characterized in that, The step of determining the recall candidate results of candidate documents associated with the distribution network query based on the retrieval link further includes: calculating the recall score of the distribution network query for the candidate documents using an extended hybrid recall link, with the formula as follows: ； in, This indicates that the extended hybrid recall link is used to calculate the distribution network query. About candidate documents Recall rating; Indicates distribution network query With candidate documents Vector semantic similarity; Indicates distribution network query Related node subgraphs Indicates candidate documents The associated node subgraph; Indicates distribution network query Related node subgraphs and candidate documents The degree of structural overlap of the associated node subgraphs in the distribution network map; Indicates distribution network query With candidate documents Keyword matching similarity, Indicates distribution network query With candidate documents Temporal consistency similarity; , , and Denotes the non-negative weighting coefficient of the mixed recall, and ; Based on the recall score, the candidate documents associated with the distribution network query are sorted in descending order to obtain the recall candidate results.

7. A retrieval device for fuzzy semantics of diverse services in a power distribution network, characterized in that, include: The retrieval range constraint module is used to determine the retrieval range of the distribution network query based on the input distribution network query data; The fuzzy semantic grading module is used to perform fuzzy semantic quantization processing on distribution network query data to determine the fuzzy semantic level of the distribution network query, including: Calculate the overall fuzziness of the distribution network query based on the distribution network query data; The formula for calculating the overall fuzziness of a distribution network query is as follows: ； in, Indicates distribution network query Intentional entropy Indicates distribution network query Contextual completeness; , , , The non-negative weighting coefficients representing the overall ambiguity, and ; ； in, Indicates distribution network query Belongs to candidate business type The probability, Represents a collection of diverse business types; The overall fuzziness is divided to obtain the fuzzy semantic level of the distribution network query, which is represented as follows: like Then the distribution network query The fuzzy semantic level is level one fuzzy semantic; like Then the distribution network query The fuzzy semantic level is level two fuzzy semantics; like Then the distribution network query The fuzzy semantic level is level three fuzzy semantics; in, Indicates distribution network query The overall fuzziness, This indicates the lower limit of the adaptive grading threshold. Indicates the upper limit of the adaptive grading threshold; ； ； in, and This indicates the preset basic classification threshold; , , and This represents the non-negative adjustment coefficient. Indicates distribution network query Scope clarification, Indicates distribution network query The integrity of the entity, Indicates distribution network query The degree of colloquialism; The retrieval link selection module is used to select the corresponding retrieval link in the retrieval range of the power distribution network query based on the fuzzy semantic level. The hybrid recall module is used to determine the recall candidate results of candidate documents associated with the power distribution network query based on the retrieval link; The filtering module is used to reorder the recalled candidate results to obtain retrieval results that meet the business relevance and physical rationality requirements of the distribution network query, including: The re-ranking score of candidate documents in the recall results is calculated using the following formula: ； in, Indicates distribution network query About candidate documents The reordering score; Indicates distribution network query About candidate documents Recall rating, Indicates the business type corresponding to the distribution network query. Next candidate document The weight of business importance; Indicates distribution network query Corresponding equipment With candidate documents Associated devices Topological distance attenuation relationship under the premise of satisfying basic physical rationality; Indicates distribution network query About candidate documents Rule review factor; ； in, Indicates equipment and The distance between the distribution network topologies, Represents positive real numbers; This represents an exponential function with the natural constant e as its base. ； in, This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. express and An indicator of whether there is an electrical connection between them. Indicates distribution network query The corresponding voltage level, Indicates candidate documents The corresponding voltage level, Indicates candidate documents timestamp, Indicates distribution network query Allowed time range; Candidate documents are reordered from highest to lowest according to the reordering score to form the final recall candidate results; The top-ranked candidate documents in the final recall results will be used as the search results.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the retrieval method for fuzzy semantics of multiple services in the power distribution network as described in any one of claims 1 to 6.

Images