A point table generation method, device, equipment and medium

Abstract

This application discloses a point table generation method, apparatus, device, and medium, relating to the field of power grid technology. The method includes: acquiring first-type and second-type multimodal data of a target power station; performing entity recognition on the first-type and second-type multimodal data to obtain first-type entity data and second-type entity data; performing multimodal feature fusion on the first-type and second-type multimodal data to obtain a first-type target feature vector and a second-type target feature vector; constructing a knowledge graph based on the first-type and second-type entity data, and obtaining logical rules based on the knowledge graph; determining candidate matching signals based on the first-type and second-type target feature vectors; verifying the candidate matching signals based on the logical rules to obtain matching verification results; and performing confidence analysis on the matching verification results to obtain a target point table. This method enables intelligent generation of point tables.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a method, apparatus, equipment and medium for generating power grid points. Background Technology

[0002] Against the backdrop of the rapid development of smart grids and substation automation, the master-slave station interactive information point table serves as the core data hub connecting field equipment and the dispatch and monitoring system. Its accuracy is directly related to the real-time monitoring, fault analysis, and safety control of the power grid.

[0003] Currently, the generation and verification of data point tables still heavily rely on manual methods. When constructing new substations, designers must manually fill in thousands to tens of thousands of data point table entries in Excel or specialized tools, based on technical documents such as SCD configuration files, design drawings, and equipment lists. During the renovation of existing substations or periodic verifications, verification personnel must manually compare the dispatch master site table with the substation signal list, judging the correctness of signal mapping relationships through visual inspection and expert experience. While some improved methods have introduced simple script tools or keyword-based matching, they essentially remain within the scope of manual identification and operation.

[0004] Therefore, there is an urgent need for a method to achieve intelligent generation of point tables. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for generating point tables, which enables intelligent generation of point tables.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, this application provides a method for generating a point table, including:

[0008] Acquire the first type of multimodal data of the target power station on the dispatch side, and the second type of multimodal data on the power plant side;

[0009] Entity recognition is performed on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data;

[0010] Multimodal feature fusion is performed on the first type of multimodal data to obtain the first target feature vector, and multimodal feature fusion is performed on the second type of multimodal data to obtain the second target feature vector;

[0011] Based on the first entity data and the second entity data, a knowledge graph is constructed, and logical rules are obtained based on the knowledge graph;

[0012] Based on the first target feature vector and the second target feature vector, candidate matching signals are determined;

[0013] Based on logical rules, candidate matching signals are verified to obtain matching verification results;

[0014] Confidence analysis was performed on the matching verification results to obtain the target point table.

[0015] In one embodiment, determining candidate matching signals based on a first target feature vector and a second target feature vector includes:

[0016] The first target feature vector and the second target feature vector are combined into a feature pair to be matched;

[0017] At least two independent semantic mapping methods are used to calculate the matching degree of the feature pairs to be matched, so as to obtain at least two matching degree data.

[0018] Calculate the similarity between at least two matching data points;

[0019] Based on similarity, the final matching score is determined from at least two matching score data.

[0020] Based on the final matching degree data, candidate matching signals corresponding to the first type of multimodal data are determined from the second type of multimodal data.

[0021] In one embodiment, logical rules are derived based on a knowledge graph, including:

[0022] Extracting relationships and attribute constraints between entities from knowledge graphs;

[0023] Based on the relationship and attribute constraints, logical rules are generated to verify the rationality of signal matching; among them, the logical rules include at least one of the following: device level attribution rules, device type and standard signal type correspondence rules, signal logical linkage rules, and terminology standardization verification rules.

[0024] In one embodiment, candidate matching signals are verified based on logical rules to obtain matching verification results, including:

[0025] Based on the equipment hierarchy affiliation rules, verify whether the equipment associated with the dispatch side and the plant side signals in the candidate matching signals has the same topology hierarchy affiliation;

[0026] Based on the correspondence rules between equipment type and standard signal type, verify whether the types of signals from the dispatch side and the plant side in the candidate matching signals match the corresponding standardized signal templates;

[0027] Based on the logical linkage rules between signals, verify whether there is a preset logical conflict in the combination of dispatch side and plant side signals in the candidate matching signals;

[0028] Based on the terminology standardization verification rules, we verify whether the path features contained in the signal description in the candidate matching signal conform to the standard terminology specification.

[0029] In one embodiment, a confidence analysis is performed on the matching verification results to obtain a target point table, including:

[0030] For each signal pair in the matching verification results, the comprehensive confidence score of the signal pair is calculated based on the verification status, semantic compliance, historical frequency of occurrence and device correlation of the signal pair.

[0031] The confidence level of each signal pair is determined based on the overall confidence score.

[0032] A target point table is generated based on the confidence level of each signal pair.

[0033] In one embodiment, the confidence levels include high, medium, and low. Based on the confidence levels of each signal pair, a target point table is generated, including:

[0034] Add the signal pairs with high confidence levels to the target point table;

[0035] Mark the signal pairs with a confidence level of 0 as items to be reviewed;

[0036] Signal pairs with low confidence levels are marked as suspected invalid items;

[0037] The items pending review and those suspected of being invalid will be reviewed, and the target point table will be updated based on the review results.

[0038] In one embodiment, a knowledge graph is constructed based on first entity data and second entity data, including:

[0039] Based on the first entity data and the second entity data, extract the device entity, signal entity, and attribute information;

[0040] Based on the power grid topology, determine the hierarchical relationships between equipment entities;

[0041] Based on the device function type, determine the mapping relationship between device entities and signal entities;

[0042] Based on historical operational data, determine the logical relationships between signal entities;

[0043] A knowledge graph is constructed based on device entities, signal entities, attribute information, hierarchical relationships, mapping relationships, and logical relationships.

[0044] Secondly, this application provides a dot table generation apparatus, comprising:

[0045] The acquisition module is used to acquire the first type of multimodal data of the target power station on the dispatch side, and the second type of multimodal data on the power station side;

[0046] The recognition module is used to perform entity recognition on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data.

[0047] The fusion module is used to perform multimodal feature fusion on the first type of multimodal data to obtain a first target feature vector, and to perform multimodal feature fusion on the second type of multimodal data to obtain a second target feature vector;

[0048] The building module is used to construct a knowledge graph based on the first entity data and the second entity data, and to obtain logical rules based on the knowledge graph.

[0049] The matching module is used to determine candidate matching signals based on the first target feature vector and the second target feature vector;

[0050] The verification module is used to verify candidate matching signals based on logical rules and obtain matching verification results.

[0051] The generation module is used to perform confidence analysis on the matching verification results and obtain the target point table.

[0052] Thirdly, this application provides a computing device, including a memory and a processor;

[0053] The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of the first aspects.

[0054] Fourthly, this application provides a computer-readable storage medium for storing a computer program for performing the method as described in any one of the first aspects.

[0055] Fifthly, this application provides a computer program product comprising one or more computer instructions, wherein when the computer instructions are executed by a computer, the computer performs the method as described in any one of the first aspects.

[0056] As can be seen from the above technical solution, this application has at least the following beneficial effects:

[0057] In this application, by acquiring first-type multimodal data of the target power station on the dispatch side and second-type multimodal data on the substation side, sufficient data is provided for thorough and accurate analysis. Furthermore, entity recognition is performed on the first-type and second-type multimodal data to obtain first entity data corresponding to the first-type multimodal data and second entity data corresponding to the second-type multimodal data, laying the foundation for signal matching. Multimodal feature fusion is performed on the first-type multimodal data to obtain a first target feature vector, and multimodal feature fusion is performed on the second-type multimodal data to obtain a second target feature vector, realizing the fusion of multimodal data and improving data quality while ensuring data multidimensionality. Then, based on the first and second entity data, a knowledge graph is constructed, and logical rules are obtained based on the knowledge graph, providing a tool to ensure the accuracy of signal matching. Based on the first and second target feature vectors, candidate matching signals are determined, realizing signal matching. Then, based on the logical rules, the candidate matching signals are verified to obtain matching verification results, ensuring the accuracy of signal matching. Finally, confidence analysis is performed on the matching verification results to obtain a target point table. This solution introduces multimodal data, providing data for entity recognition from multiple dimensions and laying the foundation for obtaining an accurate point table. Furthermore, by introducing knowledge graphs, it provides insights into discovering potential relationships between data. Finally, by introducing confidence analysis, it ensures the accuracy of signal matching and realizes the intelligent generation of the point table.

[0058] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0059] Figure 1 This is an application environment diagram of a point table generation method provided in the embodiments of this application;

[0060] Figure 2 This is a flowchart illustrating a point table generation method provided in an embodiment of this application;

[0061] Figure 3 This is a schematic diagram of a process for determining candidate matching signals provided in an embodiment of this application;

[0062] Figure 4 This is a schematic diagram of a process for obtaining logical rules provided in an embodiment of this application;

[0063] Figure 5 This is a schematic diagram of a process for obtaining a matching verification result provided in an embodiment of this application;

[0064] Figure 6 This is a structural block diagram of a dot table generation device provided in the embodiments of this application;

[0065] Figure 7 This is an internal structural diagram of a computer device provided in the embodiments of the application. Detailed Implementation

[0066] The terms "first," "second," and "third," etc., used in this application specification and accompanying drawings are used to distinguish different objects, not to limit a specific order.

[0067] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0068] To make the technical solution of this application clearer and easier to understand, the application scenarios of the technical solution of this application are described below with reference to the accompanying drawings. Figure 1 As shown, this figure is an application environment diagram provided by an embodiment of this application.

[0069] In this application scenario, server 104 analyzes and calculates based on the first type of multimodal data and the second type of multimodal data to obtain a target point table, and transmits the target point table to terminal 102 through a communication network so that terminal 102 can display the target point table.

[0070] To make the technical solution of this application clearer and easier to understand, the following describes how this method is applied in the above-mentioned application environment. Figure 1 Taking server 104 as an example, this application provides a method for generating a point table. Figure 2 As shown in the figure, this figure is a flowchart illustrating a point table generation method provided in an embodiment of this application.

[0071] S201. Obtain the first type of multimodal data of the target power station on the dispatch side, and the second type of multimodal data on the power station side.

[0072] The target power station refers to the specific power facility being implemented, specifically a substation, which is a node in the power system that transforms voltage, receives and distributes electrical energy. It includes a large number of devices that need to be monitored, such as transformers, circuit breakers, disconnect switches, and protection devices.

[0073] The dispatching side refers to the power grid's command and control center, which is responsible for monitoring, controlling, and making decisions about the entire regional power grid. It needs to obtain data from various substations to understand the overall network operation status.

[0074] The plant-side refers to the monitoring system of the target power station itself, which is responsible for collecting real-time operating data and status signals of all power equipment (such as circuit breakers, transformers, protection devices, etc.) within the station.

[0075] Modality refers to the type or form of information; different modalities mean that the data source, structure, and the characteristics of the information contained are completely different. For example, text and images are two completely different modalities.

[0076] Multimodal data refers to a collection of raw data of various types and sources from a target power station, including text data, image data, substation configuration description (SCD) data, and structured data. These data describe the equipment, status, and connection relationships of the target power station from different perspectives. By integrating these different modalities of data, they can complement and verify each other, forming a more comprehensive and accurate digital description of the target power station.

[0077] The first type of multimodal data refers to a collection of raw data in various formats and sources related to the target power station, originating from the dispatching side. Examples include text data (signal description text in the dispatching master station information point table), structured data (signal types, telemetry, etc.), data formats, communication planning tables, etc.), and power grid topology data (model data describing the station's connectivity within the entire power grid).

[0078] The second type of multimodal data refers to a collection of raw data in various formats and sources from the substation side, describing the conditions within the substation. For example, SCD files are substation configuration description files conforming to the IEC 61850 standard, serving as an authoritative source for equipment modeling and signal definition; text data is the signal description text of the monitoring system or device itself; structured data is the equipment and signal association information contained in drawings and documents such as primary main wiring diagrams, secondary schematic diagrams, and equipment lists; and power grid topology data is the hierarchical structure information of "substation-voltage level-bay-equipment" within the substation.

[0079] For example, by calling the standard data interfaces (such as database queries and APIs) of a dispatch automation system (such as D5000), structured point table information (signal names, types, point numbers, etc.) and power grid topology model files in CIM / E format can be automatically exported to obtain dispatch-side data. Substation-side data can be obtained by parsing the entire substation's SCD configuration file to extract the IEC 61850 standard signal model, combining it with the signal list exported by the monitoring system, and using graphic analysis or OCR (Optical Character Recognition) technology to recognize design documents such as primary main wiring diagrams and secondary schematic diagrams to construct equipment topology relationships and associated attributes. Finally, these two types of multimodal data can be integrated into structured input to provide a complete data foundation for subsequent feature fusion and intelligent analysis.

[0080] S202. Perform entity recognition on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data.

[0081] Entity recognition refers to the process of automatically identifying, extracting, and normalizing discrete information units with specific meanings from multimodal data using technologies such as natural language processing and computer vision; these information units represent objectively existing objects or their attributes in the power system.

[0082] Entity data is the structured data result formed after entity identification. It is the direct raw material for subsequent knowledge graph construction. It not only includes the identified entities themselves, but also records the semantic and logical relationships between entities, including multiple entities and entity relationship data between them.

[0083] An entity is a core object or concept extracted from data; it is a clearly identifiable physical device, logical function, semaphore, or spatial location. In the context of power systems, it primarily refers to clearly identifiable physical devices, logical functions, semaphores, or spatial locations.

[0084] First entity data refers to the structured set of entities extracted from the first type of multimodal data after entity recognition.

[0085] Second entity data refers to the structured set of entities extracted from the second type of multimodal data after entity recognition.

[0086] For example, named entity recognition can be performed on text data using the Transformer-based BERT (Bidirectional Encoder Representations from Transformers) model to accurately extract key information such as device names and signal types. Then, object detection and OCR technologies such as YOLOv8 are used to parse the image data and identify device entities. Simultaneously, SCD files and structured databases are parsed to extract standard signal paths and device attributes, collectively forming structured first and second entity data. Subsequently, cross-modal representation alignment technology is used to align the original representations of the same entity in different modalities. A dynamic weight allocation mechanism is introduced to adaptively adjust the fusion weights based on the confidence and quality of each modality's data. Finally, robust and information-rich target feature vectors are generated for each entity, completing the dual transformation from raw data to structured knowledge and standardized features, laying the foundation for subsequent knowledge graph construction and intelligent matching.

[0087] S203. Perform multimodal feature fusion on the first type of multimodal data to obtain the first target feature vector, and perform multimodal feature fusion on the second type of multimodal data to obtain the second target feature vector.

[0088] Multimodal feature fusion is a computational process that integrates heterogeneous raw features from different modal data into a unified, complementary, and more representative joint feature vector through alignment, transformation, and combination. The aim is to overcome the limitations of single-modal information and improve the comprehensiveness and robustness of data understanding.

[0089] The target feature vector is the final output after multimodal feature fusion. It refers to a fixed-dimensional numerical array generated for each entity after multimodal feature fusion. This vector is a comprehensive mathematical representation of the entity under all modal information and is the digital fingerprint of the entity in the whole system.

[0090] The first target feature vector is a fixed-dimensional numerical array generated after multimodal feature fusion of the first type of multimodal data.

[0091] The second target feature vector is a fixed-dimensional numerical array generated after multimodal feature fusion of the second type of multimodal data.

[0092] For example, semantic feature vectors can be extracted from text descriptions using the BERT model, structured feature codes parsed from SCD files and database records can be obtained, and visual feature vectors can be extracted from images using a convolutional neural network. Subsequently, specific fusion strategies, such as feature concatenation, weighted summation, or attention-based fusion networks, can be employed to align and integrate these features at different scales and semantic levels. During this process, an innovative dynamic weight allocation mechanism is introduced, which adaptively adjusts the contribution weight of each modality of the current entity in the final fusion result based on its quality and confidence level. Ultimately, this fusion process outputs a highly condensed first target feature vector for each entity on the scheduling side, and a similarly high-quality second target feature vector for each entity on the plant side. This transforms complex multi-source information into a standardized numerical form suitable for large-scale, efficient computation, laying a solid foundation for subsequent accurate semantic matching.

[0093] S204. Based on the first entity data and the second entity data, construct a knowledge graph and obtain logical rules based on the knowledge graph.

[0094] One possible approach is to extract equipment entities, signal entities, and attribute information based on first entity data and second entity data; determine the hierarchical relationships between equipment entities based on the power grid topology; determine the mapping relationships between equipment entities and signal entities based on equipment function types; determine the logical relationships between signal entities based on historical operating data; and construct a knowledge graph based on equipment entities, signal entities, attribute information, hierarchical relationships, mapping relationships, and logical relationships.

[0095] Among them, a knowledge graph is a large semantic network that organizes and represents knowledge in the form of a graph. It is a structured database that stores knowledge in the field of power systems and reveals the complex relationships between entities such as equipment, signals and locations. For example, a complex network including nodes (such as transformer 1, circuit breaker, high-voltage side, oil temperature signal) and edges (such as location, control, measurement) formally encodes the topology and functional logic of the substation.

[0096] Logical rules are reusable reasoning patterns and hard constraints abstracted from knowledge graphs, usually existing in the form of "if...then..."; they represent widely recognized expert knowledge and engineering specifications in power systems; for example, rules describing the necessity of equipment functions such as "if a device is a circuit breaker, then it must have a position signal", or rules describing the logical association of signals such as "if a bus protection action signal is generated, then the associated circuit breaker position signal should change within a very short time".

[0097] For example, based on the identified first and second entity data, equipment entities (such as transformers and circuit breakers) and signal entities (such as position signals and alarm signals) and their key attributes are extracted. Further, based on the power grid topology, a clear hierarchical relationship is established between equipment entities, such as a circuit breaker belonging to a certain 110kV line bay. According to the functional type of the equipment, the inherent mapping relationship between equipment entities and signal entities is determined, such as all circuit breakers should have position signals. Based on historical operating data and expert experience, the logical relationships between signal entities are mined and defined, such as protection action signals should be strongly correlated with circuit breaker trip signals in time. Finally, all entities and relationships are integrated to construct a structured, machine-readable knowledge graph.

[0098] Based on this knowledge graph, logical rules for verifying the rationality of signal matching are automatically extracted through graph traversal, rule engine, or knowledge representation learning techniques. These rules include device hierarchy rules, signal type consistency rules, and signal linkage logic rules.

[0099] S205. Based on the first target feature vector and the second target feature vector, determine the candidate matching signal.

[0100] Among them, candidate matching signals refer to one or more potential matches that are most similar to a specific dispatching signal in terms of numerical characteristics, selected from a massive amount of plant-side signals through preliminary automated calculations. It is a preliminary, unverified pairing list, which serves as the basis for subsequent fine-grained verification and confirmation, rather than the final result.

[0101] For example, a large-scale similarity calculation can be performed between a first target feature vector (representing a certain scheduling demand signal) on the scheduling side and all second target feature vectors (representing all available signals in the entire station) on the plant side. This calculation is usually performed using an efficient vector space model algorithm (such as cosine similarity), and its output quantifies the degree of matching of each pair of scheduling-plant signals in deep semantics and comprehensive features. Furthermore, for each scheduling-side signal, the top-N results (e.g., top-3) with the highest similarity scores are selected from the plant side, or all plant-side signals with scores exceeding a certain preset threshold are selected. Finally, these initially paired "scheduling-side signal-plant-side signal" combinations, together with their calculated similarity scores, are jointly marked as candidate matching signals, forming a detailed list to be verified. This transforms the complex problem of "full search" into a controllable problem of "fine verification of a limited candidate set," providing accurate input for subsequent knowledge graph-based logical verification and confidence analysis.

[0102] S206. Based on logical rules, the candidate matching signals are verified to obtain the matching verification results.

[0103] The matching verification result refers to the state judgment obtained after the candidate matching signal has been verified by a series of logical rules. It is not a simple score, but a structured conclusion that clearly identifies whether the candidate match is consistent, suspected, inconsistent, missing, or redundant, and may be accompanied by specific verification details and explanations of reasons.

[0104] For example, each candidate matching signal (i.e., a pairing of a dispatch-side signal and a substation-side signal) is used as a verification object, and automated review is performed based on a pre-set logical rule base. The review process includes: using equipment hierarchy affiliation rules to check whether the equipment associated with the paired signal belongs to the same topology bay or has a direct electrical connection to verify the correctness of the physical association; using the correspondence rules between equipment type and standard signal type to determine whether the signal type (such as remote signaling, telemetry) conforms to the functional specifications of the equipment to prevent logical errors such as "temperature monitoring instrument issuing circuit breaker opening / closing signal"; using the logical linkage rules between signals to analyze whether the match contradicts the known signal action sequence (e.g., protection actions must correspond to circuit breaker opening, not closing); and using terminology standardization verification rules to standardize and review the paths, parameters, etc. in the signal description. Finally, based on the comprehensive verification results of all these rules, a clear matching verification result status label can be assigned to each candidate match, and a detailed report including the specific reasons for inconsistency can be generated, thereby elevating the pure numerical similarity matching to a reliable verification that integrates the deep business logic of the power system.

[0105] S207. Perform confidence analysis on the matching verification results to obtain the target point table.

[0106] One possible approach is to calculate a comprehensive confidence score for each signal pair in the matching verification results, based on the verification status, semantic compliance, historical frequency of occurrence, and device correlation of the signal pair; determine the confidence level of each signal pair based on the comprehensive confidence score; and generate a target point table based on the confidence level of each signal pair.

[0107] Optionally, add signal pairs with high confidence levels to the target point table; mark signal pairs with medium confidence levels as items to be reviewed; mark signal pairs with low confidence levels as suspected invalid items; review the items to be reviewed and the suspected invalid items, and update the target point table based on the review results.

[0108] The point table is a data index in the power monitoring system, which is a list that defines information about all monitoring points in detail.

[0109] The target point table is a data mapping table that has undergone fully automated processing and incorporated human-machine collaborative decision-making. It is a standardized and accurate master-slave station interaction information point table that can be directly applied to engineering applications. It not only includes the basic attributes of monitoring points, but also deeply integrates the equipment associations, spatial locations and logical relationships of monitoring points in the physical world. It provides a solid and reliable data foundation for the digital twin, intelligent operation and maintenance and advanced applications of power systems. It is a data list that guides the correct data interaction between substations and dispatch master stations.

[0110] Confidence analysis is a comprehensive quantitative evaluation process that is not limited to a single indicator. Instead, it integrates evidence from multiple dimensions, such as logical verification results, semantic consistency, historical statistical patterns, and device context, to calculate a quantitative score representing the final reliability of each signal matching pair.

[0111] For example, for each signal pair in the matching verification results, a multi-factor comprehensive evaluation mechanism is implemented. This mechanism not only considers the verification status of the signal pair (e.g., consistent or suspected), but also simultaneously calculates its semantic conformity (the degree of matching between the descriptive text and standard terminology), historical frequency of occurrence (the prevalence of the pattern in past correct point tables), and device relevance (the degree of fit between the signal content and the function of the source device). These factors can be aggregated into a comprehensive confidence score using a pre-defined weighted scoring model. Furthermore, a confidence level (e.g., high, medium, low) can be determined based on this score. Finally, a decision combining automated and manual review is executed according to the level. For example, high-confidence signal pairs are automatically adopted and directly added to the target point table; medium-confidence and low-confidence signals are marked as pending review items and suspected invalid items, respectively, and pushed to the manual review interface. Experts make the final decision based on the various scores and reasons provided by the system, and the review results are used to update and complete the final target point table. This process ensures the combination of automated efficiency and artificial intelligence, significantly improving work efficiency while guaranteeing the accuracy and reliability of the final results.

[0112] The aforementioned point table generation method acquires first-type multimodal data from the target power station on the dispatch side and second-type multimodal data from the substation side, providing sufficient data for thorough and accurate analysis. Furthermore, entity recognition is performed on the first and second-type multimodal data to obtain first entity data corresponding to the first-type multimodal data and second entity data corresponding to the second-type multimodal data, laying the foundation for signal matching. Multimodal feature fusion is performed on the first-type multimodal data to obtain a first target feature vector, and multimodal feature fusion is performed on the second-type multimodal data to obtain a second target feature vector, achieving multimodal data fusion and improving data quality while ensuring data multidimensionality. Then, a knowledge graph is constructed based on the first and second entity data, and logical rules are derived from the knowledge graph, providing a tool to ensure the accuracy of signal matching. Candidate matching signals are determined based on the first and second target feature vectors, achieving signal matching. Then, the candidate matching signals are verified based on the logical rules to obtain matching verification results, ensuring the accuracy of signal matching. Finally, confidence analysis is performed on the matching verification results to obtain the target point table. This solution introduces multimodal data, providing data for entity recognition from multiple dimensions and laying the foundation for obtaining an accurate point table. Furthermore, by introducing knowledge graphs, it provides insights into discovering potential relationships between data. Finally, by introducing confidence analysis, it ensures the accuracy of signal matching and realizes the intelligent generation of the point table.

[0113] Based on the above embodiments, this application provides a detailed explanation of S205. Specifically, this application involves a process for determining candidate matching signals, such as... Figure 3 As shown, the specific steps include:

[0114] S301. Combine the first target feature vector and the second target feature vector into a feature pair to be matched.

[0115] Among them, the feature pair to be matched refers to a data unit temporarily constructed for similarity comparison. This unit is composed of a first target feature vector of a scheduling side signal and a second target feature vector of a plant side signal.

[0116] For example, a first target feature vector (representing a specific dispatch demand signal, such as "active power of line XX") is extracted from the feature vector set on the dispatch side, and a second target feature vector (representing a specific signal that can be provided within the station, such as "high voltage side P of main transformer No. 1") is extracted from the feature vector set on the substation side. These two vectors are logically associated to form a feature pair to be matched. This process is performed on a large scale and traversally: that is, the feature vector of each dispatch side signal is paired with the feature vector of each substation side signal, thereby generating all possible signal combination pairs. Finally, a set containing a massive number of feature pairs to be matched is output, laying a solid foundation for the batch, high-efficiency vector similarity calculation and preliminary screening in the next step.

[0117] S302. Use at least two independent semantic mapping methods to calculate the matching degree of the feature pairs to be matched, and obtain at least two matching degree data.

[0118] Semantic mapping refers to a specific algorithm or model that can understand the deep semantics represented by feature vectors and calculate the degree of correlation between two feature vectors (i.e., a pair of features to be matched) in the semantic space.

[0119] Matching data refers to a quantitative result calculated through a certain semantic mapping method, which is used to represent the degree of similarity or matching between two signals in the feature pair to be matched; it is usually a numerical value (such as similarity score, probability value) or a sorted list.

[0120] For example, for each feature pair to be matched, at least two different semantic mapping methods are computed in parallel. These methods are independent, meaning they may be based on different mathematical models—for example, one a vector space model based on cosine similarity, and another a distance metric model based on a neural network, with different internal structures or parameters—ensuring they evaluate the same matching pair from slightly different perspectives. Each method processes the feature pair independently and outputs a matching score; for example, model A outputs a similarity score of 0.92, and model B outputs a matching probability of 95%. Ultimately, at least two matching scores are collected for each original feature pair. These scores form the basis for subsequent consistency decisions and final matching score fusion. This parallel heterogeneous computation strategy significantly improves the accuracy and reliability of the initial matching stage.

[0121] S303. Calculate the similarity between at least two matching data.

[0122] For example, at least two matching data points obtained through parallel computation (e.g., a similarity score S_A from model A and a matching probability P_B from model B) are used as input. Further, a suitable similarity metric algorithm is employed to quantify the proximity between these two (or more) matching data points, such as calculating cosine similarity, absolute difference, or correlation coefficient. This calculated similarity value (e.g., a number between 0 and 1) does not directly represent the degree of matching of the signals themselves, but rather reflects the level of consistency among different models regarding the question of whether these two signals match. A high similarity indicates that different models have given highly consistent judgments, resulting in high reliability; a low similarity indicates significant disagreement between models, signifying uncertainty in the matching pair, requiring special attention.

[0123] It should be noted that although matching data may originate from different calculation paradigms (such as scores and probabilities), they can be standardized to a comparable scale. Determining the suitability of a similarity measurement algorithm requires comprehensive consideration of data characteristics, algorithm objectives, and business needs. First, the basic algorithm should be selected based on the type of matching data (such as continuous scores, probability values, or sorted lists) and its distribution scale. For example, continuous scores can use cosine similarity to measure trend consistency, or absolute difference to measure absolute numerical consistency. Second, the choice of algorithm must closely align with business objectives. If the focus is on the relative order consistency of judgment results between models, cosine similarity should be prioritized. If it is necessary to avoid automated decision-making discrepancies due to subtle numerical differences, absolute difference, which is sensitive to decision boundaries, should be used. Finally, testing on a labeled validation set is necessary to observe the separation effect of different measurement algorithms on high-consistency and low-consistency samples, thereby selecting the algorithm that most effectively supports the subsequent decision-making process as the appropriate solution.

[0124] S304. Based on similarity, determine the final matching data from at least two matching data.

[0125] For example, similarity (representing the degree of consistency between the judgments of different models) is used as an important basis for decision-making, and the following logic is executed based on preset decision rules: When the similarity is high (e.g., greater than or equal to 80%), it indicates that there is a strong consensus among different models. At this time, the judgment of either model can be trusted. The decision strategy can be to randomly select the matching score data output by one model, or to select the one with the higher score as the final matching score data. When the similarity is low (e.g., less than 80%), it indicates that there is a significant divergence between the models. At this time, it is risky to adopt either result alone. Therefore, a more robust strategy is to retain all the matching score data with divergence (i.e., the original outputs of the two models) and let them flow together into the subsequent verification and analysis process, allowing more complex mechanisms (such as knowledge base reasoning and comprehensive confidence analysis) to make the final decision. Through this consistency-based dynamic decision-making mechanism, efficiency can be improved when the model consensus is high, and stricter review can be automatically activated when the consensus is low, thereby optimizing the accuracy and reliability of the matching process overall.

[0126] S305. Based on the final matching degree data, determine the candidate matching signal corresponding to the first type of multimodal data from the second type of multimodal data.

[0127] For example, each dispatch-side signal (originating from the first type of multimodal data) is processed traversally. For each dispatch signal, one or more associated final matching degree data are already available (these data represent the calculated matching results between the dispatch signal and certain plant / station signals). Further, these matching degree data are sorted and filtered. For example, the top N plant / station signals with the highest matching degree are retained for each dispatch signal, or all plant / station signals with matching degrees exceeding a certain preset threshold are selected. Next, based on the identifiers of these filtered, high-matching plant / station signals, the complete, structured descriptive information, attributes, and their specific paths within the plant / station system are located and extracted from the original second type of multimodal data (i.e., the plant / station SCD files, signal lists, and other original materials). Finally, a structured list is generated for each dispatch-side signal, where each item includes a candidate plant / station signal and its matching degree. This list is defined as the candidate matching signals. This completes the initial screening to find the most relevant supply for each demand from massive amounts of data, providing a precise target range for subsequent in-depth logical verification and manual review.

[0128] In this embodiment, feature pairs are computed in parallel using at least two independent semantic mapping methods, and cross-validation effectively avoids bias from a single model. Secondly, by calculating the similarity between different mapping results, the quality assessment and confidence perception of the matching process itself are achieved. Then, based on this similarity, the final matching data is dynamically selected, forming an intelligent decision-making process with high consensus and efficient adoption, and low consensus and prudent review. Finally, the quantitative matching results are accurately mapped back to actual engineering signal entities, ensuring that the output results possess both algorithmic intelligence and engineering operability. This systematically solves the pain points of poor fault tolerance and uncontrollable results in traditional automated matching, providing crucial technical support for the fully automated generation of power meter readings.

[0129] Based on the above embodiments, this application provides a detailed explanation of S204. Specifically, this application involves a process for obtaining logical rules, such as... Figure 4 As shown, the specific steps include:

[0130] S401. Extract the relationships and attribute constraints between entities from the knowledge graph.

[0131] In this context, association refers to a directed or undirected connection in a knowledge graph that links two or more entities and has a specific semantic meaning, describing the structural, functional, or logical interactions between entities. For example, "belonging to" is an association that connects the entity "circuit breaker" and the entity "110kV line bay".

[0132] Attribute constraints are normative conditions or characteristic limitations attached to entities or relationships, defining the rules or ranges that an entity's values ​​must follow. For example, for the attribute "voltage level", the constraint might be "the value must be one of {110kV, 220kV, 500kV}"; for the relationship "belongs to", the constraint might be "a signal entity can only belong to one device entity".

[0133] For example, the constructed power knowledge graph is traversed and parsed to extract the relationships between entities. This includes extracting relationship types with clear power business semantics from the graph's edges, such as "equipment - belongs to -> bay", "signal - originates from -> equipment", and "equipment - connected to -> busbar". Simultaneously, attribute constraints can be extracted, that is, from the attribute-value pairs of entities and relationships, those normative and restrictive conditions can be summarized, such as "protection device entities must have protection action signals", "the unit attribute of telemetry signals cannot be empty", and "the state attribute value range of circuit breakers is {open, closed}". These extracted relationships and attribute constraints together constitute a formalized rule material library, providing direct and structured input for the next step of "generating logical rules for verifying the rationality of signal matching", making knowledge-based automated reasoning possible.

[0134] S402. Based on the relationship and attribute constraints, generate logical rules for verifying the rationality of signal matching.

[0135] The logical rules include at least one of the following: device hierarchy classification rules, device type and standard signal type correspondence rules, signal inter-logic linkage rules, and terminology standardization verification rules.

[0136] For example, based on the extracted relationships and attribute constraints, a rule engine or logic programming technique is used to formalize them into explicit logical rules, specifically including:

[0137] Generate device hierarchy affiliation rules: Based on the relationship of "device - belongs to -> interval", generate rules, such as: "Matching dispatch side and plant side signals, the associated devices must belong to the same interval or have a direct topological connection relationship". This can be used to verify whether the devices associated with the signals are logically consistent in physical structure.

[0138] The rules for generating the correspondence between device types and standard signal types are based on attribute constraints such as "protection device - must have -> protection action signal". For example, "a circuit breaker device must include position in its signal type; if an oil temperature signal appears, it is considered unreasonable". This can be used to verify whether the signal content is consistent with the functional nature of the device.

[0139] Generate logical linkage rules between signals: Based on the causal or temporal relationship described by "protection action - leading to -> circuit breaker tripping", generate rules, such as: "If the line differential protection action signal is successfully matched, then there must be a corresponding circuit breaker tripping signal that is also successfully matched within a very short time, otherwise an alarm will be triggered". This can be used to verify whether the matching result conforms to the dynamic operation logic of the power grid.

[0140] Generate terminology standardization verification rules: Based on attribute constraints such as signal description and path naming (e.g., must follow the standard terminology library), generate rules, such as: "The 'closed' in the signal description must use the standard term 'closed', not 'closed' or 'connected'"; "The 61850 path must contain the corresponding logical node code CSWI", which can be used to ensure the standardization and consistency of point table data.

[0141] Ultimately, these generated logical rules constitute an automated verification knowledge base that can perform in-depth, business semantic-based logical consistency checks on preliminary candidate matching signals, much like an experienced expert.

[0142] In this embodiment, by transforming the relationships and attribute constraints in the knowledge graph into executable logical rules, automated verification is achieved, moving from shallow text matching to deep logical reasoning. It can not only verify the topological and functional rationality of signal associations based on rules such as device hierarchy and type correspondence, but also discover logical conflicts that violate the power grid operation mechanism through signal linkage rules. Furthermore, it precipitates industry standards and expert experience into reusable digital assets such as terminology verification, thus systematically solving the pain point of traditional methods failing to guarantee semantic consistency and logical completeness. This significantly improves the accuracy, interpretability, and adaptability to complex scenarios of the verification process, laying a technical foundation for building a highly reliable intelligent power meter generation system.

[0143] Based on the above embodiments, this application provides a detailed explanation of S206. Specifically, this application involves the process of obtaining the matching verification result, as follows: Figure 5 As shown, the specific steps include:

[0144] S501. Based on the equipment hierarchy affiliation rules, verify whether the equipment associated with the dispatch side and the plant side signals in the candidate matching signals have the same topology hierarchy affiliation.

[0145] Among them, the equipment hierarchy rule refers to the normative logic that defines the subordinate relationship of equipment in the power grid topology, which clarifies that equipment must belong to a specific bay, and the bay belongs to a specific voltage level and substation, forming a strict "substation-voltage level-bay-equipment" hierarchical tree.

[0146] Topology hierarchy refers to the precise location of a power device in the power grid topology hierarchy tree. For example, the topology hierarchy of "220kV XX line circuit breaker" is "XX station -> 220kV voltage level -> 220kV XX line bay".

[0147] For example, after obtaining a candidate matching signal (i.e., a pairing of a dispatch-side signal and a substation-side signal), the system queries the knowledge graph to find the device entities associated with these two signals. Then, using the device hierarchy affiliation rules, the system compares the topology hierarchy affiliation paths of these two devices level by level to check whether they are located in the same interval under the same voltage level of the same substation. For example, if the dispatch signal is associated with "#1 main transformer 501 switch" and the substation signal is associated with "501 switch", and both belong to "XX station-220kV-#1 main transformer interval" in the knowledge graph, then the verification passes. If the devices do not belong to the same interval, the system further determines whether they are logically strongly associated adjacent intervals, such as incoming and outgoing line intervals on the same busbar. This verification can effectively intercept serious errors such as mismatching "110kV line protection signal" with "220kV main transformer switch signal", ensuring the consistency between the point table data and the actual physical structure of the power grid from the source.

[0148] S502. Based on the correspondence rules between equipment type and standard signal type, verify whether the types of signals from the dispatch side and the plant side in the candidate matching signals match the corresponding standardized signal template.

[0149] Among them, the correspondence rules between equipment types and standard signal types refer to the set of specifications that clearly define the signal types that equipment should possess in the power system field, based on the function and principle of the equipment. For example, the rules stipulate that a "circuit breaker" must have a "position" (open / close) signal, while a "transformer" must have non-electrical telemetry signals such as "oil temperature".

[0150] Standardized signal templates refer to a predefined, standardized, and complete list of signal types and their attribute descriptions for each type of equipment, constituting the authoritative standard for which signals that type of equipment "should" have in the point table.

[0151] For example, after obtaining a candidate matching signal, the device types associated with the two ends of the signal pair are first identified from the knowledge graph, such as line protection devices and transformer oil thermometers. Further, the pre-set correspondence rules between device types and standard signal types can be invoked to query the standardized signal templates corresponding to that type of device. For example, the rule stipulates that the standard signal template for "line protection device" should include alarm signals such as "protection action" and "communication interruption", but should not include analog signals such as "A-phase current". Then, the signal types in the candidate matching signals, such as "remote signaling" and "remote measurement" and their specific contents, can be compared with the template to verify whether they are within the reasonable range defined by the template.

[0152] This verification can effectively identify and intercept fundamental functional mismatches such as "collecting circuit breaker position signals from temperature monitoring instruments" or "missing essential oil temperature monitoring signals for main transformers," thereby ensuring the rationality and completeness of the meter readings at the logical level.

[0153] S503. Based on the logical linkage rules between signals, verify whether there is a preset logical conflict in the combination of dispatch side and plant side signals in the candidate matching signals.

[0154] Among them, the logical linkage rules between signals refer to the time sequence, causal or mutual exclusion relationships that should be satisfied between signal states, formally defined based on the working principles and operating experience of power systems. For example, "after any protection action signal is issued, its corresponding circuit breaker trip position signal must change to the open position within a very short time window" is a typical causal and time sequence linkage rule.

[0155] A pre-defined logical conflict refers to a situation where the state presented by a combination of signals violates the established rules for logical linkage between signals. For example, if a combination of signals simultaneously includes "protection action" and "circuit breaker closed," this constitutes a direct logical conflict with the aforementioned rules.

[0156] For example, pre-defined logical linkage rules between signals can be invoked from the knowledge base. These rules define the constraints (such as causality, timing, and mutual exclusion) that a group of signals must satisfy under specific events or states. Furthermore, candidate matching signals (not just the currently verified pair, but also other logically strongly related matched signals) can be grouped together and substituted into these rules for verification. For instance, a rule might require that "the bus protection action signal and the current disappearance signal of the related lines should appear in conjunction." If it is found that the bus protection has acted, but the current signals of all related lines still show normal values, then the combination is determined to have a pre-defined logical conflict.

[0157] This verification can capture systematic logical errors distributed among signals from different devices that traditional point-to-point matching cannot detect, greatly improving the depth of point-to-point verification and the reliability of power grid operation.

[0158] S504. Based on the terminology standardization verification rules, verify whether the path features contained in the signal description in the candidate matching signal conform to the standard terminology specification.

[0159] Among them, the terminology standardization verification rules refer to the mandatory provisions on the format, vocabulary and structure that must be followed by text information such as signal description, equipment naming and logical node paths, which are clearly defined by power industry standards or enterprise internal specifications, ensuring the uniformity and unambiguity of the point table content.

[0160] In the context of point tables, a path feature specifically refers to a hierarchical string that uniquely identifies a signal and contains information such as logical devices, logical nodes, and data attributes in systems conforming to standards such as IEC 61850. For example, XXQKGGIO1$Ind$stVal is a path feature, where XXQK is the logical device, GGIO1 is the logical node, and Ind and stVal are the data object and attribute, respectively.

[0161] For example, the path characteristics (i.e., the complete 61850 signal path string) of the plant-side signals (usually from the SCD file) in the candidate matching signals can be obtained, and the description text of the dispatch-side signals will also be parsed. Further, the preset terminology standardization verification rules are called. This rule set usually includes a standard terminology library, naming abbreviation standards, and a mapping table of logical node types and functions. The key fields in the path characteristics (such as logical node type CSWI and data object Pos) are compared with the standard terminology library to verify whether they conform to the standards. For example, it verifies whether Pos is used to represent a position signal and whether CSWI is used for switch control. At the same time, it also checks whether the composition structure of the path conforms to the established pattern.

[0162] This verification can effectively identify and block non-standard path naming issues caused by configuration errors, non-standard abbreviations, or personal habits, ensuring the accuracy and reliability of dot tables during cross-system transmission and parsing from the data source.

[0163] In this embodiment, a multi-layered verification system is constructed to improve quality and efficiency: Device hierarchical attribution rules ensure the physical location consistency of signal-associated devices from a topological perspective, reducing the risk of object misalignment; device type and standard signal type correspondence rules verify the compatibility of signals and device functions from a functional principle perspective, eliminating logical contradictions; signal inter-logic linkage rules detect causal timing conflicts in signal combinations through simulated expert reasoning, identifying systemic vulnerabilities; and terminology standardization verification rules unify path features and descriptive text from a data standard dimension, ensuring cross-system interconnectivity. This method transforms domain knowledge into executable verification logic, enabling deep semantic understanding and logical reasoning capabilities while completing numerical matching, significantly improving the accuracy, security, and standardization of point table data.

[0164] The above text combined Figures 1 to 5 The point table generation method provided in the embodiments of this application has been described in detail. The apparatus and equipment provided in the embodiments of this application will be described below with reference to the accompanying drawings.

[0165] like Figure 6As shown, this figure is a structural block diagram of a dot table generation device 600 provided in an embodiment of this application. The dot table generation device 600 includes: an acquisition module 601, an identification module 602, a fusion module 603, a construction module 604, a matching module 605, a verification module 606, and a generation module 607, wherein:

[0166] The acquisition module 601 is used to acquire the first type of multimodal data of the target power station on the dispatch side and the second type of multimodal data on the power station side.

[0167] The recognition module 602 is used to perform entity recognition on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data.

[0168] The fusion module 603 is used to perform multimodal feature fusion on the first type of multimodal data to obtain a first target feature vector, and to perform multimodal feature fusion on the second type of multimodal data to obtain a second target feature vector;

[0169] Module 604 is used to construct a knowledge graph based on the first entity data and the second entity data, and to obtain logical rules based on the knowledge graph.

[0170] The matching module 605 is used to determine candidate matching signals based on the first target feature vector and the second target feature vector;

[0171] The verification module 606 is used to verify the candidate matching signals based on logical rules and obtain the matching verification results.

[0172] The generation module 607 is used to perform confidence analysis on the matching verification results to obtain the target point table.

[0173] In one embodiment, the matching module 605 is specifically used for:

[0174] The first target feature vector and the second target feature vector are combined into a feature pair to be matched;

[0175] At least two independent semantic mapping methods are used to calculate the matching degree of the feature pairs to be matched, so as to obtain at least two matching degree data.

[0176] Calculate the similarity between at least two matching data points;

[0177] Based on similarity, the final matching score is determined from at least two matching score data.

[0178] Based on the final matching degree data, candidate matching signals corresponding to the first type of multimodal data are determined from the second type of multimodal data.

[0179] In one embodiment, the construction module 604 is specifically used for:

[0180] Extracting relationships and attribute constraints between entities from knowledge graphs;

[0181] Based on the relationship and attribute constraints, logical rules are generated to verify the rationality of signal matching; among them, the logical rules include at least one of the following: device level attribution rules, device type and standard signal type correspondence rules, signal logical linkage rules, and terminology standardization verification rules.

[0182] In one embodiment, the verification module 606 is specifically used for:

[0183] Based on the equipment hierarchy affiliation rules, verify whether the equipment associated with the dispatch side and the plant side signals in the candidate matching signals has the same topology hierarchy affiliation;

[0184] Based on the correspondence rules between equipment type and standard signal type, verify whether the types of signals from the dispatch side and the plant side in the candidate matching signals match the corresponding standardized signal templates;

[0185] Based on the logical linkage rules between signals, verify whether there is a preset logical conflict in the combination of dispatch side and plant side signals in the candidate matching signals;

[0186] Based on the terminology standardization verification rules, we verify whether the path features contained in the signal description in the candidate matching signal conform to the standard terminology specification.

[0187] In one embodiment, the generation module 607 is specifically used for:

[0188] For each signal pair in the matching verification results, the comprehensive confidence score of the signal pair is calculated based on the verification status, semantic compliance, historical frequency of occurrence and device correlation of the signal pair.

[0189] The confidence level of each signal pair is determined based on the overall confidence score.

[0190] A target point table is generated based on the confidence level of each signal pair.

[0191] In one embodiment, the generation module 607 is specifically used for:

[0192] Add the signal pairs with high confidence levels to the target point table;

[0193] Mark the signal pairs with a confidence level of 0 as items to be reviewed;

[0194] Signal pairs with low confidence levels are marked as suspected invalid items;

[0195] The items pending review and those suspected of being invalid will be reviewed, and the target point table will be updated based on the review results.

[0196] In one embodiment, the construction module 604 is specifically used for:

[0197] Based on the first entity data and the second entity data, extract the device entity, signal entity, and attribute information;

[0198] Based on the power grid topology, determine the hierarchical relationships between equipment entities;

[0199] Based on the device function type, determine the mapping relationship between device entities and signal entities;

[0200] Based on historical operational data, determine the logical relationships between signal entities;

[0201] A knowledge graph is constructed based on device entities, signal entities, attribute information, hierarchical relationships, mapping relationships, and logical relationships.

[0202] The dot table generation apparatus 600 according to the embodiments of this application can correspond to the execution of the method described in the embodiments of this application, and the other operations and / or functions of each module / unit of the dot table generation apparatus 600 are respectively for implementing Figures 2-5 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.

[0203] This application also provides a computing device. This computing device can be a local computing device or an application server.

[0204] like Figure 7 As shown in the figure, this is an internal structural diagram of a computing device provided in an embodiment of this application. The computing device 700 includes a bus 701, a processor 702, a communication interface 703, and a memory 704. The processor 702, the memory 704, and the communication interface 703 communicate with each other via the bus 701.

[0205] The 701 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0206] The processor 702 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0207] The communication interface 703 is used for communication with external devices.

[0208] Memory 704 may include volatile memory, such as random access memory (RAM). Memory 704 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0209] The memory 704 stores executable code, and the processor 702 executes the executable code to perform the aforementioned point table generation method.

[0210] Specifically, in achieving Figure 6 In the case of the illustrated embodiment, and Figure 6 When the modules or units of the dot table generation device described in the embodiment are implemented by software, the execution... Figure 6 The software or program code required for the functions of each module / unit can be partially or wholly stored in memory 704. Processor 702 executes the program code corresponding to each unit stored in memory 704 and performs the aforementioned point table generation method.

[0211] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the above-described dot table generation method.

[0212] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.

[0213] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0214] When the computer program product is executed by a computer, the computer performs any of the aforementioned point table generation methods. The computer program product can be a software installation package; when any of the aforementioned point table generation methods is required, the computer program product can be downloaded and executed on the computer.

[0215] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0216] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.

Claims

1. A method for generating a point table, characterized in that, The method includes: Acquire the first type of multimodal data of the target power station on the dispatch side, and the second type of multimodal data on the power plant side; Entity recognition is performed on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data; Multimodal feature fusion is performed on the first type of multimodal data to obtain a first target feature vector, and multimodal feature fusion is performed on the second type of multimodal data to obtain a second target feature vector; wherein, the multimodal feature fusion adopts a fusion strategy of feature concatenation, weighted summation or attention-based fusion network; A knowledge graph is constructed based on the first entity data and the second entity data; Extracting relationships and attribute constraints between entities from knowledge graphs; Based on the aforementioned relationships and attribute constraints, logical rules are generated to verify the rationality of signal matching; wherein, the logical rules include at least one of the following: device hierarchy affiliation rules, device type and standard signal type correspondence rules, signal inter-logic linkage rules, and terminology standardization verification rules; Based on the first target feature vector and the second target feature vector, candidate matching signals are determined; Based on the aforementioned logical rules, the candidate matching signals are verified to obtain the matching verification results; For each signal pair in the matching verification results, a comprehensive confidence score is calculated based on the verification status, semantic compliance, historical frequency of occurrence, and device association of the signal pair. The confidence level of each signal pair is determined based on the comprehensive confidence score. The target point table is generated based on the confidence level of each signal pair.

2. The method according to claim 1, characterized in that, The step of determining candidate matching signals based on the first target feature vector and the second target feature vector includes: The first target feature vector and the second target feature vector are combined into a feature pair to be matched; At least two independent semantic mapping methods are used to calculate the matching degree of the feature pairs to be matched, respectively, to obtain at least two matching degree data. Calculate the similarity between at least two matching data points; Based on the similarity, the final matching score data is determined from at least two matching score data. Based on the final matching degree data, candidate matching signals corresponding to the first type of multimodal data are determined from the second type of multimodal data.

3. The method according to claim 1, characterized in that, The process of verifying candidate matching signals based on the logical rules to obtain matching verification results includes: Based on the device hierarchy affiliation rules, verify whether the devices associated with the scheduling side and the plant side signals in the candidate matching signals have the same topology hierarchy affiliation. Based on the correspondence rules between the equipment type and the standard signal type, verify whether the types of the dispatch side and the plant side signals in the candidate matching signals match the corresponding standardized signal templates; Based on the logical linkage rules between the signals, verify whether there is a preset logical conflict in the combination of the scheduling side and the power plant side signals in the candidate matching signals; Based on the terminology standardization verification rules, verify whether the path features contained in the signal description of the candidate matching signal conform to the standard terminology specification.

4. The method according to claim 1, characterized in that, The confidence levels include high, medium, and low. Generating the target point table based on the confidence levels of each signal pair includes: Add the signal pairs with high confidence levels to the target point table; Mark the signal pairs with a confidence level of 0 as items to be reviewed; Signal pairs with low confidence levels are marked as suspected invalid items; The items pending review and those suspected of being invalid will be reviewed, and the target point table will be updated based on the review results.

5. The method according to claim 1, characterized in that, The construction of the knowledge graph based on the first entity data and the second entity data includes: Based on the first entity data and the second entity data, extract the device entity, signal entity, and attribute information; Based on the power grid topology, the hierarchical relationships between the device entities are determined; Based on the device function type, determine the mapping relationship between the device entity and the signal entity; Based on historical operational data, determine the logical relationships between signal entities; A knowledge graph is constructed based on the device entities, signal entities, attribute information, hierarchical relationships, mapping relationships, and logical relationships.

6. A dot table generation device, characterized in that, The device includes: The acquisition module is used to acquire the first type of multimodal data of the target power station on the dispatch side, and the second type of multimodal data on the power station side; The recognition module is used to perform entity recognition on the first type of multimodal data and the second type of multimodal data to obtain the first entity data corresponding to the first type of multimodal data and the second entity data corresponding to the second type of multimodal data. The fusion module is used to perform multimodal feature fusion on the first type of multimodal data to obtain a first target feature vector, and to perform multimodal feature fusion on the second type of multimodal data to obtain a second target feature vector; wherein the multimodal feature fusion adopts a fusion strategy of feature concatenation, weighted summation or fusion network based on attention mechanism; The building module is used to construct a knowledge graph based on the first entity data and the second entity data, and to obtain logical rules based on the knowledge graph. The matching module is used to determine candidate matching signals based on the first target feature vector and the second target feature vector; The verification module is used to verify the candidate matching signals based on the logical rules and obtain the matching verification results; The generation module is used to perform confidence analysis on the matching verification results to obtain a target point table; The module is used to extract the relationships and attribute constraints between entities from the knowledge graph; Based on the aforementioned relationships and attribute constraints, logical rules are generated to verify the rationality of signal matching; wherein, the logical rules include at least one of the following: device hierarchy affiliation rules, device type and standard signal type correspondence rules, signal inter-logic linkage rules, and terminology standardization verification rules; The generation module is used to calculate the comprehensive confidence score of each signal pair in the matching verification results based on the verification status, semantic compliance, historical frequency of occurrence and device correlation of the signal pair; determine the confidence level of each signal pair based on the comprehensive confidence score; and generate the target point table based on the confidence level of each signal pair.

7. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 5.

Images