A power distribution network multi-source heterogeneous operation data association method and device

Abstract

The application relates to a power distribution network multi-source heterogeneous operation data association method and device, and belongs to the technical field of power system operation data processing. According to multi-source heterogeneous operation data of a target device node, a candidate device node set is determined, an association relationship between device nodes is determined through joint similarity of the target device node and each candidate device node in the candidate device node set, finally, multi-source heterogeneous operation data of the target device node and associated candidate device nodes are fused based on the association relationship between the target device node and the candidate device nodes, a reliable asynchronous, multi-source heterogeneous data unified fusion result can be provided, the accuracy and engineering usability of power distribution network unified semantic object construction are improved, and the problem that current multi-source heterogeneous data objects are difficult to unify, the same object has different names and different places have the same name, and association errors occur frequently is solved.

Description

Technical Field

[0001] This invention relates to a method and apparatus for correlating multi-source heterogeneous operation data of a power distribution network, belonging to the field of power system operation data processing technology. Background Technology

[0002] The operation of power distribution networks generates massive amounts of heterogeneous data from multiple sources, which are scattered across different business systems such as measurement, mapping, and maintenance. Due to the different construction standards of each system, the data varies significantly in terms of temporal and spatial granularity and lacks a naturally unified association identifier. This leads to the phenomenon of "same object, different name" for the same physical entity in different systems, and "same name, different location" for equipment in different regions, making it extremely challenging to build a unified semantic object view across systems.

[0003] Existing technologies typically employ field rule matching or plain text semantic matching to address object association problems. However, rule matching demands extremely high data quality and struggles to handle inconsistent legacy data formats and discrepancies in manual data entry. Plain text semantic matching, on the other hand, focuses solely on the similarity of names, severely lacking constraints related to the topological context and electrical connections of the power distribution network diagram. This makes it highly susceptible to misassociations and fails to fundamentally resolve the challenges of intertwined "same object, different names" and "same name, different locations," leading to duplicate equipment ledgers and chaotic graph nodes. Furthermore, in terms of multi-source attribute fusion and conflict resolution, existing technologies often employ time-based latest priority, static reliability weighting, or simple voting mechanisms. However, in the context of strong spatiotemporal coupling in power distribution networks, multi-source data often suffers from asynchronous sampling, sensor anomalies, or delayed manual data entry.

[0004] Existing simple fusion methods not only ignore the time freshness decay characteristics of data, but also neglect the local physical consistency under the topological neighborhood. This leads to the inability to effectively identify abnormal conflict data, bad data and pseudo data and to be incorrectly entered into the database, which seriously affects subsequent fault assessment and unified data services. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and device for associating multi-source heterogeneous operation data of distribution networks. This provides a reliable unified fusion result of asynchronous, multi-source heterogeneous data, improves the accuracy and engineering usability of constructing unified semantic objects for distribution networks, and solves the problems of difficulty in unifying multi-source heterogeneous data objects and frequent association errors caused by different names for the same object and different locations for the same name.

[0006] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution:

[0007] This invention provides a method for correlating multi-source heterogeneous operation data of a power distribution network, including:

[0008] A set of candidate device nodes is determined based on the multi-source heterogeneous operation data of the target device nodes in the power distribution network system; wherein, the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data and text data;

[0009] Calculate the joint similarity between the target device node and each candidate device node in the candidate device node set;

[0010] Based on joint similarity, the association between the target device node and each candidate device node is determined;

[0011] If the target device node is associated with a candidate device node, the multi-source heterogeneous operating data of the target device node and the associated candidate device node will be fused to generate operating data that conforms to the actual operating conditions of the target device node.

[0012] Furthermore, the set of candidate device nodes for the target device node is represented as follows:

[0013] ;

[0014] in, Indicates the target device node The set of candidate device nodes; Indicates candidate device node The type of equipment, Indicates the target device node The type of equipment; Indicates candidate device node voltage level, Indicates the target device node Voltage level; Indicates candidate device node The power supply area, transformer substation, or feeder zone identifier. Indicates the target device node The corresponding set of candidate search regions; Indicates the target device node With candidate device nodes Geographical distance between them Indicates the geographic distance threshold; Indicates the target device node With candidate device nodes Topological hop count between This represents the topology hop count threshold.

[0015] Further, the calculation of the joint similarity between the target device node and each node in the candidate device node set includes:

[0016] Acquire multi-source heterogeneous operational data of target device nodes and candidate device nodes;

[0017] Based on the multi-source heterogeneous operational data of the target device nodes and candidate device nodes, obtain the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target device nodes and candidate device nodes.

[0018] Based on the name semantics, attributes, physical operation behavior, meteorological context, and text description of the target device node and the candidate device node, respectively construct the corresponding multimodal semantic vectors;

[0019] Construct an electrical topology neighborhood set for the target device node and candidate device nodes;

[0020] Calculate the electrical distance based on the electrical topology neighborhood set of the target device node and the candidate device nodes;

[0021] The joint similarity is calculated based on the multimodal semantic vectors and electrical distance between the target device node and the candidate device node.

[0022] Furthermore, the step of constructing corresponding multimodal semantic vectors based on the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target device node and candidate device nodes includes:

[0023] Modal encoding is performed on name semantics, attributes, physical operational behavior, meteorological context, and text description to generate corresponding modal feature vectors;

[0024] The modal feature vectors are concatenated to obtain the joint feature vector;

[0025] The joint feature vector is mapped and normalized to obtain the multimodal semantic vector, which is represented as follows: ;

[0026] in, Represents device node Multimodal semantic vectors Represents positive numbers. Represents the L2 norm;

[0027] Represents device node The unnormalized mapping results ; , Represents the mapping weight matrix. , Indicates the bias term. Represents a nonlinear activation function; Represents device node The joint eigenvectors;

[0028] ;

[0029] Represents device node Name semantic feature vector, Represents device node The attribute feature vector, Represents device node The physical operation behavior feature vector, Represents device node Meteorological context feature vector, Represents device node The text description feature vector.

[0030] Furthermore, the calculation of electrical distance based on the electrical topology neighborhood set of the target device node and the candidate device node includes:

[0031] Based on the electrical topology neighborhood set of the target device node and the candidate device node, determine the shortest effective path and the corresponding line impedance;

[0032] Based on the shortest effective path and the corresponding line impedance, the electrical distance between the target device node and the candidate device node is calculated using the following formula:

[0033] ;

[0034] in, Indicates the target device node With candidate device nodes Electrical distance between; Indicates the weight of the line length item. Indicates the target device node With candidate device nodes Normalized cumulative path length on the shortest effective path; Indicates the weight of the line impedance term. Indicates the target device node With candidate device nodes Normalized cumulative line impedance on the shortest effective path.

[0035] Further, calculating the joint similarity between the target device node and each candidate device node in the candidate device node set includes: calculating the joint similarity between the target device node and each node in the candidate device node set using the following formula:

[0036] ;

[0037] in, Indicates the target device node With candidate device nodes Joint similarity between; Indicates the semantic item weight, Indicates normalized semantic similarity; Indicates the target device node Multimodal semantic vectors Indicates candidate device nodes Multimodal semantic vectors; Indicates the weight of the topology term. Indicates the similarity of electrical topology neighborhoods; Indicates the target device node The set of electrical topological neighborhoods, Indicates candidate device node The set of electrical topological neighborhoods; Indicates the electrical distance penalty factor. Indicates the target device node With candidate device nodes Electrical distance between; This represents the natural exponential function.

[0038] Furthermore, the calculation of the joint similarity between the target device node and each candidate device node in the candidate device node set also includes: the semantic term weight, topological term weight, and electrical distance penalty coefficient satisfying the following update rule:

[0039] ;

[0040] ;

[0041] ;

[0042] in, This represents the initial value of the semantic item weight. This indicates the lower bound of the semantic item weight. This indicates the upper limit of the semantic item weight. This represents the initial value of the electrical distance penalty coefficient. This indicates the lower limit of the electrical distance penalty coefficient. This indicates the upper limit of the electrical distance penalty coefficient. , Indicates the update step size; Indicates semantic completeness. Indicates topological completeness. Indicates the distance dispersion; This represents the interval truncation function.

[0043] Further, determining the association between the target device node and each candidate device node based on joint similarity includes:

[0044] Sort the candidate device nodes in the candidate device node set in descending order according to the magnitude of joint similarity;

[0045] Select the first and second candidate device nodes with the largest and second largest joint similarity from the candidate device node set;

[0046] If the joint similarity of the first candidate device node is not less than the preset similarity threshold, and the difference between the joint similarity of the first candidate device node and the joint similarity of the second candidate device node is greater than the preset difference threshold, then the target device node is associated with the first candidate device node.

[0047] Furthermore, if the target device node is associated with a candidate device node, the multi-source heterogeneous operating data of the target device node and the associated candidate device node are fused to generate operating data that conforms to the actual operating conditions of the target device node, including:

[0048] The multi-source heterogeneous operating data is divided into continuous data and discrete data;

[0049] The following formula is used to fuse continuous data:

[0050] ;

[0051] in, This represents the operational data obtained after fusing continuous data; This indicates the total number of continuous data entries. Indicates the first Continuous data; Indicates the first The fusion weights for continuous data;

[0052] ;

[0053] Indicates the first The data source credibility or scenario-based weight corresponding to each continuous data point. This represents the time decay coefficient, used to control the degree to which the time difference decays the fusion weights. Indicates the first The time difference between continuous data points and the target time. This represents the average or reference value of measurements taken at device nodes in the electrical topology neighborhood. Indicates the first The squared deviation between the continuous data and the reference value in the electrical topology neighborhood. Indicates the spatial consistency scale parameter. Represents the natural exponential function;

[0054] The following formula is used to fuse discrete data:

[0055] ;

[0056] in, This represents the operational data obtained after fusing discrete data. Indicates the first Category labels corresponding to discrete data , This represents the set of candidate categories corresponding to discrete data.

[0057] Another aspect of the present invention provides a device for correlating multi-source heterogeneous operation data of a power distribution network, comprising:

[0058] The candidate device segmentation module is used to determine a set of candidate device nodes based on the multi-source heterogeneous operation data of the target device nodes in the power distribution network system; wherein, the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data and text data;

[0059] The joint similarity calculation module is used to calculate the joint similarity between the target device node and each candidate device node in the candidate device node set;

[0060] The association relationship determination module is used to determine the association relationship between the target device node and each candidate device node based on the joint similarity.

[0061] The fusion module is used to fuse the multi-source heterogeneous operating data of the target device node and the associated candidate device node if the target device node is associated with the candidate device node, and generate operating data that conforms to the actual operating conditions of the target device node.

[0062] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0063] 1. This invention determines a set of candidate device nodes based on the multi-source heterogeneous operation data of the target device node. By determining the joint similarity between the target device node and each candidate device node in the set, the association relationship between the device nodes is determined. Finally, based on the established association relationship between the target device node and the candidate device nodes, the multi-source heterogeneous operation data of the target device node and the associated candidate device nodes are fused. This can provide a reliable unified fusion result of asynchronous, multi-source heterogeneous data, improve the accuracy and engineering usability of the unified semantic object construction of the distribution network, and solve the current problems of difficulty in unifying multi-source heterogeneous data objects and frequent association errors caused by the same object having different names and the same name in different locations.

[0064] 2. In the process of determining the association relationship between device nodes, the present invention combines the multimodal semantic vectors, topological neighborhoods and electrical distance penalties of device nodes, which can simultaneously utilize semantic information and power distribution network topology information in the process of device node association, thereby improving the accuracy of object association.

[0065] 3. This invention provides update rules based on semantic completeness, topological completeness, and distance dispersion for the semantic item weights, topological item weights, and electrical distance penalty coefficients in the joint similarity calculation process. This enables adaptive adjustment under different working conditions and can effectively identify and eliminate abnormal conflict data, bad data, and pseudo data, thereby improving the credibility of the joint similarity calculation results.

[0066] 4. This invention employs a weighted mechanism based on time decay, data source credibility, and topological neighborhood local spatial consistency to perform weighted center fusion on continuous data and weighted majority voting on discrete data, which can improve the reliability of data fusion results in asynchronous sampling and multi-source conflict scenarios. Attached Figure Description

[0067] Figure 1 This is a flowchart of a method for associating multi-source heterogeneous operation data of a power distribution network provided in an embodiment of the present invention;

[0068] Figure 2 This is a flowchart of the joint similarity calculation provided in an embodiment of the present invention;

[0069] Figure 3 This is a schematic diagram of the structure of a device for associating multi-source heterogeneous operation data of a power distribution network, provided in an embodiment of the present invention. Detailed Implementation

[0070] 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.

[0071] Example 1

[0072] like Figure 1 As shown, this embodiment 1 provides a method for associating multi-source heterogeneous operation data of a distribution network, including:

[0073] Step 1: Determine the candidate device node set based on the multi-source heterogeneous operation data of the target device nodes in the distribution network system; the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data, and text data; specifically:

[0074] In this embodiment, the physical operation data comes from the dispatch automation system, distribution automation system and station-end measurement device in the distribution network system, and is used to characterize voltage, current, power and switch status quantities.

[0075] Physical operation data is represented as , ; Indicates the number of devices. Indicates the number of time sampling points. The attribute dimensions representing physical operation data include voltage, current, active power, reactive power, and numerically encoded switching state quantities. Represents the set of real numbers;

[0076] Meteorological data is used to characterize the temperature, humidity, and extreme weather conditions in the area where the equipment nodes are located;

[0077] Meteorological data is expressed as ; The horizontal coordinates of the device node's spatial location. The vertical coordinate represents the spatial location of the device node. Indicates time;

[0078] Topology chart data is represented as ; Represents a set of device nodes. Represents the set of topological edges. This represents the set of attributes associated with device nodes and topology edges; where device nodes correspond to substation busbars, feeder branch points, switches, distribution transformers, transformer substation entrance nodes, and user access points, and topology edges correspond to electrical connection relationships.

[0079] Text data is represented as ; Indicates the first A text string, This represents the total number of entries containing text strings; the text data includes work orders, inspection logs, maintenance records, defect records, alarm descriptions, and maintenance notes.

[0080] It should be noted that before performing calculations, the multi-source heterogeneous operating data needs to be standardized.

[0081] The device name is mapped to a unified field, the time field is converted to a unified timestamp, the spatial location is converted to a unified coordinate, missing values ​​are imputed using a preset method, and duplicate records are merged using the "device identifier + time window + attribute category" key to provide consistent input for subsequent calculations.

[0082] The set of candidate device nodes for the target device node is represented as follows:

[0083] ;

[0084] in, Indicates the target device node The set of candidate device nodes; Indicates candidate device node The type of equipment, Indicates the target device node The type of equipment; Indicates candidate device nodes voltage level, Indicates the target device node Voltage level; Indicates candidate device nodes The power supply area, transformer substation, or feeder zone identifier. Indicates the target device node The corresponding candidate search area set is used to limit the boundary range of the same station area, the same feeder or adjacent feeders; Indicates the target device node With candidate device nodes Geographical distance between them Indicates the geographic distance threshold; Indicates the target device node With candidate device nodes Topological hop count between This represents the topology hop count threshold.

[0085] By first excluding objects with inconsistent equipment types, then excluding objects with inconsistent voltage levels, and then retaining the limited forms of objects in the same substation area and adjacent feeder boundary objects to perform a first-stage range narrowing of candidate device nodes, and then performing a second-stage range narrowing based on geographical distance and topological range, objects that are obviously unlikely to be the same physical device can be eliminated in advance, avoiding objects with similar names but completely different regions from entering the joint similarity calculation stage, thereby reducing the probability of "same name, different location" mismatch.

[0086] Step 2, as follows Figure 2 As shown, the joint similarity between the target device node and each candidate device node in the candidate device node set is calculated; specifically:

[0087] Acquire multi-source heterogeneous operational data of target device nodes and candidate device nodes;

[0088] Based on the multi-source heterogeneous operational data of the target device nodes and candidate device nodes, obtain the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target device nodes and candidate device nodes.

[0089] Based on the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target and candidate device nodes, corresponding multimodal semantic vectors are constructed, including:

[0090] Modal encoding is performed on name semantics, attributes, physical operational behavior, meteorological context, and text description to generate corresponding modal feature vectors;

[0091] The modal feature vectors are concatenated to obtain the joint feature vector;

[0092] The joint feature vector is mapped and normalized to obtain the multimodal semantic vector, which is represented as follows: ;

[0093] in, Represents device node Multimodal semantic vectors Represents positive numbers. Represents the L2 norm;

[0094] Represents device node The unnormalized mapping results ; , Represents the mapping weight matrix. , Indicates the bias term. Represents a nonlinear activation function; Represents device node The joint eigenvectors;

[0095] ;

[0096] Represents device node Name semantic feature vector, Represents device node The attribute feature vector, Represents device node The physical operation behavior feature vector, Represents device node Meteorological context feature vector, Represents device node The textual description feature vector;

[0097] The electrical topology neighborhood set of a device node is constructed using the following formula:

[0098] ;

[0099] in, Represents device node The set of electrical topological neighborhoods;

[0100] Represents device node The first-order electrical topology neighborhood set is selected according to the following rule: it is selected based on the proximity of the device nodes. Connect all device nodes directly through the currently connected edge, and filter out edges that are in a disconnected state;

[0101] Represents device node The selection rule for the second-order electrical topology neighborhood set is as follows: when the number of device nodes in the first-order electrical topology neighborhood set is less than a preset threshold, the device nodes in the first-order electrical topology neighborhood set are extended outward along the conducting edge to obtain the second-order electrical topology neighborhood nodes, and the visited nodes and nodes outside the isolation boundary are excluded.

[0102] when hour, Indicates the target device node The set of electrical topological neighborhoods, Indicates the target device node The first-order electrical topological neighborhood set, Indicates the target device node The set of second-order electrical topological neighborhoods;

[0103] when hour, Indicates candidate device nodes The set of electrical topological neighborhoods, Indicates candidate device nodes The first-order electrical topological neighborhood set, Indicates candidate device nodes The set of second-order electrical topological neighborhoods;

[0104] By using the above rules, we can ensure that neighborhood information is local and supplement contextual information when the neighborhood is too sparse.

[0105] Based on the electrical topology neighborhood set of the target device node and candidate device nodes, the electrical distance is calculated, including:

[0106] Based on the electrical topology neighborhood set of the target device node and the candidate device nodes, determine the shortest effective path and the corresponding line impedance;

[0107] Based on the shortest effective path and the corresponding line impedance, the electrical distance between the target device node and the candidate device node is calculated using the following formula:

[0108] ;

[0109] in, Indicates the target device node With candidate device nodes Electrical distance between; Indicates the weight of the line length item. Indicates the target device node With candidate device nodes Normalized cumulative path length on the shortest effective path; Indicates the weight of the line impedance term. Indicates the target device node With candidate device nodes Normalized cumulative line impedance along the shortest effective path;

[0110] When impedance is missing , When the line length is missing, , ;

[0111] The joint similarity is calculated based on the multimodal semantic vectors and electrical distance between the target device node and the candidate device nodes. The joint similarity between the target device node and each node in the candidate device node set is calculated using the following formula:

[0112] ;

[0113] in, Indicates the target device node With candidate device nodes Joint similarity between; Indicates the semantic item weight, Indicates normalized semantic similarity; Indicates the target device node Multimodal semantic vectors Indicates candidate device nodes Multimodal semantic vectors; Indicates the weight of the topology term. Indicates the similarity of electrical topology neighborhoods; Indicates the electrical distance penalty factor. Indicates the target device node With candidate device nodes Electrical distance between; Represents the natural exponential function;

[0114] By using the electrical distance term, penalties can be imposed on candidate objects with the same or similar names but which are electrically far apart, thereby effectively suppressing false matches of "same name, different location";

[0115] In addition, in this embodiment, the semantic term weight, topology term weight, and electrical distance penalty coefficient satisfy the following update rules:

[0116] ;

[0117] ;

[0118] ;

[0119] in, This represents the initial value of the semantic item weight. This indicates the lower bound of the semantic item weight. This indicates the upper limit of the semantic item weight. This represents the initial value of the electrical distance penalty coefficient. This indicates the lower limit of the electrical distance penalty coefficient. This indicates the upper limit of the electrical distance penalty coefficient. , Indicates the update step size;

[0120] The semantic completeness is represented by a weighted average of the name field's non-nullability rate, the text attachment rate, and the attribute field's completeness rate.

[0121] The topology completeness is represented by a weighted average of the adjacent edge completeness rate, the switch state availability rate, and the adjacent node reachability rate.

[0122] The distance dispersion is represented by the normalized standard deviation of the electrical distance between candidate device nodes;

[0123] Represents the interval cutoff function;

[0124] For example, when Higher than At that time, improve and reduce ,when Higher than At that time, improve and reduce When the electrical distance dispersion between the target device node and the candidate device node is large, improve This is to increase the penalties for distant candidate device nodes.

[0125] Step 3: Determine the association between the target device node and each candidate device node based on the joint similarity; specifically:

[0126] Sort the candidate device nodes in the candidate device node set in descending order according to the magnitude of joint similarity;

[0127] When the candidate device node set is empty, it is determined that there is no candidate device node that meets the preset candidate constraints for the target device node, and the unassociated status is output, or the target device node is written into the manual supplement queue.

[0128] If the set of candidate device nodes contains only one candidate device node, then that candidate device node is designated as the first candidate device node. If the first candidate device node Corresponding joint similarity Not less than the preset similarity threshold Then determine the target device node. With the first candidate device node Associate and link the target device node With the first candidate device node The association relationship is written into the formal unified object library for storage; if the first candidate device node Corresponding joint similarity Less than the preset similarity threshold Then determine the first candidate device node. With the target device node Not associated, or the first candidate device node Write it to the review queue for subsequent review;

[0129] When the candidate device node set contains two or more candidate device nodes, the first candidate device node with the highest joint similarity and the second highest joint similarity is selected from the candidate device node set. Second candidate device node ;

[0130] If the first candidate device node Corresponding joint similarity Not less than the preset similarity threshold And the first candidate device node Corresponding joint similarity With the second candidate device node Corresponding joint similarity The difference is greater than the preset threshold for difference. Then the target device node With the first candidate device node Related;

[0131] target device node With the first candidate device node The relationships between objects are written into the formal unified object library for storage;

[0132] In this embodiment, when and Then the first candidate device node For target device node The objects to be reviewed;

[0133] when At that time, the first candidate device node With the target device node Not related;

[0134] Write the objects to be reviewed into the review queue and record the top two objects to be reviewed and their corresponding joint similarity; then perform a second review, including expanding the time window to re-extract the running features, supplementing the most recent operation and maintenance text, checking the manual maintenance alias table, and checking the most recent diagram change record;

[0135] If the object to be reviewed is associated with the target device node after the second verification, the object to be reviewed will be added to the formal unified object library for storage; otherwise, the object to be reviewed will be kept in the review queue, and only the review status will be output without entering the formal unified object library.

[0136] Step 4: If the target device node is associated with a candidate device node, then the multi-source heterogeneous operational data of the target device node and the associated candidate device nodes are fused to generate operational data that conforms to the actual operating conditions of the target device node; specifically:

[0137] Multi-source heterogeneous operational data is divided into continuous data and discrete data;

[0138] It should be noted that among physical operation data, meteorological data, topology chart data and text data, continuously recorded data such as voltage, current, active power, reactive power, and load measurement are continuous data, while data such as switch status, equipment operation status, equipment category labels, alarm labels, and maintenance status labels are discrete data.

[0139] The following formula is used to fuse continuous data:

[0140] ;

[0141] in, This represents the operational data obtained after fusing continuous data; This indicates the total number of continuous data entries. Indicates the first Continuous data; Indicates the first The fusion weights for continuous data;

[0142] ;

[0143] Indicates the first The data source credibility or scenario-based weight corresponding to each continuous data point. This represents the time decay coefficient, used to control the degree to which the time difference decays the fusion weights. Indicates the first The time difference between continuous data points and the target time. This represents the average or reference value of measurements taken at device nodes in the electrical topology neighborhood. The average value is calculated by prioritizing the measurement values ​​of device nodes that are of the same type and feeder as the target device node and are located in the neighborhood of the first-order electrical topology. Indicates the first The squared deviation between continuous data points and reference values ​​in the electrical topology neighborhood is used to measure local spatial consistency deviation. Indicates the spatial consistency scale parameter. This represents the natural exponential function.

[0144] The voltage fusion scenario of a distribution transformer on a 10kV feeder line is used as an example for illustration:

[0145] Within the same target time window, the object obtained three continuous voltage records from the main station measurement system, the acquisition terminal system, and the manual recording system, respectively; the main station measurement record showed a voltage of 231V, the acquisition terminal record showed a voltage of 229V, and the manual recording record showed a voltage of 258V;

[0146] Based on the measurement values ​​of nodes of the same type in the first-order and second-order electrical topology neighborhood sets of the transformer, the following calculations are obtained: The value is 230V. Since the deviation between the main station measurement and the acquisition terminal measurement and the neighboring reference value is small, and the sources of both are stable and the timestamps are close to the target time, the fusion weight is high.

[0147] Although the manually entered records were not significantly delayed in time, their values ​​deviated significantly from the neighborhood reference values, and the source credibility was low. Therefore, under the combined effect of the neighborhood consistency item and the source credibility item, their weight was significantly reduced.

[0148] When the fusion weight corresponding to this record If the data is less than the preset threshold, it is judged as abnormal conflict data and removed from the fusion sample, and will not be included in the formal fusion result;

[0149] Ultimately, the formal voltage attribute of the unified semantic object is generated by merging the two records, 231V and 229V, thus avoiding the erroneous writing of false out-of-limit data.

[0150] The following formula is used to fuse discrete data:

[0151] ;

[0152] in, This represents the operational data obtained after fusing discrete data. Indicates the first Category labels corresponding to discrete data , This represents the set of candidate categories corresponding to discrete data.

[0153] Let's take discrete switch state conflict as an example for explanation:

[0154] A certain switch is displayed as "closed" in the dispatch automation system, recorded as "open" in the inspection work order, and still retains the "open" status from the previous time period in the maintenance record;

[0155] Through time decay calculation, the weight of maintenance records decreases significantly because the entry time is earlier than the target time;

[0156] Based on the state consistency rate calculation, the "closed" label, which is consistent with the energized state of adjacent switches and adjacent lines, receives a higher weight, while the "open" label, which is inconsistent with the dominant state of the neighborhood, receives a lower weight.

[0157] Ultimately, the weighted majority vote determined that the current unified status of the switch is "closed". The inspection work order and maintenance record will still be retained as historical textual evidence, but will not be written into the current official status.

[0158] Example 2

[0159] like Figure 3 As shown, this embodiment 2 provides a device for associating multi-source heterogeneous operation data of a distribution network, including:

[0160] The candidate device segmentation module is used to determine the set of candidate device nodes based on the multi-source heterogeneous operation data of the target device nodes in the distribution network system; wherein, the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data and text data;

[0161] The joint similarity calculation module is used to calculate the joint similarity between the target device node and each candidate device node in the candidate device node set;

[0162] The association relationship determination module is used to determine the association relationship between the target device node and each candidate device node based on the joint similarity.

[0163] The fusion module is used to fuse the multi-source heterogeneous operating data of the target device node and the associated candidate device node if the target device node is associated with the candidate device node, and generate operating data that conforms to the actual operating conditions of the target device node.

[0164] 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.

[0165] 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.

[0166] 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.

[0167] 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.

[0168] 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.

[0169] 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 method for correlating multi-source heterogeneous operation data of a power distribution network, characterized in that, include: A set of candidate device nodes is determined based on the multi-source heterogeneous operation data of the target device nodes in the power distribution network system; wherein, the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data and text data; Calculate the joint similarity between the target device node and each candidate device node in the candidate device node set, including: calculating the joint similarity between the target device node and each node in the candidate device node set using the following formula: ； in, Indicates the target device node With candidate device nodes Joint similarity between; Indicates the semantic item weight, Indicates normalized semantic similarity; Indicates the target device node Multimodal semantic vectors Indicates candidate device node Multimodal semantic vectors; Indicates the weight of the topology term. Indicates the similarity of electrical topology neighborhoods; Indicates the target device node The set of electrical topological neighborhoods, Indicates candidate device node The set of electrical topological neighborhoods; Indicates the electrical distance penalty factor. Indicates the target device node With candidate device nodes Electrical distance between; Represents the natural exponential function; Based on joint similarity, the association between the target device node and each candidate device node is determined; If the target device node is associated with a candidate device node, the multi-source heterogeneous operational data of the target device node and the associated candidate device nodes are fused to generate operational data that conforms to the actual operating conditions of the target device node, including: The multi-source heterogeneous operating data is divided into continuous data and discrete data; The following formula is used to fuse continuous data: ； in, This represents the operational data obtained after fusing continuous data; This indicates the total number of continuous data entries. Indicates the first Continuous data; Indicates the first The fusion weights for continuous data; ； Indicates the first The data source credibility or scenario-based weight corresponding to each continuous data point. This represents the time decay coefficient, used to control the degree to which the time difference decays the fusion weights. Indicates the first The time difference between continuous data points and the target time. This represents the average or reference value of measurements taken at device nodes in the electrical topology neighborhood. Indicates the first The squared deviation between the continuous data and the reference value in the electrical topology neighborhood. Indicates the spatial consistency scale parameter. Represents the natural exponential function; The following formula is used to fuse discrete data: ； in, This represents the operational data obtained after fusing discrete data. Indicates the first Category labels corresponding to discrete data , This represents the set of candidate categories corresponding to discrete data.

2. The method for associating multi-source heterogeneous operation data of a distribution network according to claim 1, characterized in that, The set of candidate device nodes for the target device node is represented as follows: ； in, Indicates the target device node The set of candidate device nodes; Indicates candidate device node The type of equipment, Indicates the target device node The type of equipment; Indicates candidate device node voltage level, Indicates the target device node Voltage level; Indicates candidate device node The power supply area, transformer substation, or feeder zone identifier. Indicates the target device node The corresponding set of candidate search regions; Indicates the target device node With candidate device nodes Geographical distance between them Indicates the geographic distance threshold; Indicates the target device node With candidate device nodes Topological hop count between This represents the topology hop count threshold.

3. The method for associating multi-source heterogeneous operation data of a distribution network according to claim 1, characterized in that, The calculation of the joint similarity between the target device node and each node in the candidate device node set includes: Acquire multi-source heterogeneous operational data of target device nodes and candidate device nodes; Based on the multi-source heterogeneous operational data of the target device nodes and candidate device nodes, obtain the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target device nodes and candidate device nodes. Based on the name semantics, attributes, physical operation behavior, meteorological context, and text description of the target device node and the candidate device node, respectively construct the corresponding multimodal semantic vectors; Construct an electrical topology neighborhood set for the target device node and candidate device nodes; Calculate the electrical distance based on the electrical topology neighborhood set of the target device node and the candidate device nodes; The joint similarity is calculated based on the multimodal semantic vectors and electrical distance between the target device node and the candidate device node.

4. The method for correlating multi-source heterogeneous operation data of a distribution network according to claim 3, characterized in that, The step involves constructing corresponding multimodal semantic vectors based on the name semantics, attributes, physical operational behavior, meteorological context, and text description of the target and candidate device nodes, including: Modal encoding is performed on name semantics, attributes, physical operational behavior, meteorological context, and text description to generate corresponding modal feature vectors; The modal feature vectors are concatenated to obtain the joint feature vector; The joint feature vector is mapped and normalized to obtain the multimodal semantic vector, which is represented as follows: ; in, Represents device node Multimodal semantic vectors Represents positive numbers. Represents the L2 norm; Represents device node The unnormalized mapping results ; , Represents the mapping weight matrix. , Indicates the bias term. Represents a nonlinear activation function; Represents device node The joint eigenvectors; ； Represents device node Name semantic feature vector, Represents device node The attribute feature vector, Represents device node The physical operation behavior feature vector, Represents device node Meteorological context feature vector, Represents device node The text description feature vector.

5. The method for correlating multi-source heterogeneous operation data of a distribution network according to claim 3, characterized in that, The calculation of electrical distance based on the electrical topology neighborhood set of the target device node and candidate device nodes includes: Based on the electrical topology neighborhood set of the target device node and the candidate device nodes, determine the shortest effective path and the corresponding line impedance; Based on the shortest effective path and the corresponding line impedance, the electrical distance between the target device node and the candidate device node is calculated using the following formula: ； in, Indicates the target device node With candidate device nodes Electrical distance between; Indicates the weight of the line length item. Indicates the target device node With candidate device nodes Normalized cumulative path length on the shortest effective path; Indicates the weight of the line impedance term. Indicates the target device node With candidate device nodes Normalized cumulative line impedance on the shortest effective path.

6. The method for correlating multi-source heterogeneous operation data of a distribution network according to claim 1, characterized in that, The calculation of the joint similarity between the target device node and each candidate device node in the candidate device node set further includes: the semantic term weight, topological term weight, and electrical distance penalty coefficient satisfying the following update rules: ； ； ； in, This represents the initial value of the semantic item weight. This indicates the lower bound of the semantic item weight. This indicates the upper limit of the semantic item weight. This represents the initial value of the electrical distance penalty coefficient. This indicates the lower limit of the electrical distance penalty coefficient. This indicates the upper limit of the electrical distance penalty coefficient. , Indicates the update step size; Indicates semantic completeness. Indicates topological completeness. Indicates the distance dispersion; This represents the interval truncation function.

7. The method for correlating multi-source heterogeneous operation data of a distribution network according to claim 1, characterized in that, The step of determining the association between the target device node and each candidate device node based on joint similarity includes: Sort the candidate device nodes in the candidate device node set in descending order according to the magnitude of their joint similarity. Select the first and second candidate device nodes with the largest and second largest joint similarity from the candidate device node set; If the joint similarity of the first candidate device node is not less than the preset similarity threshold, and the difference between the joint similarity of the first candidate device node and the joint similarity of the second candidate device node is greater than the preset difference threshold, then the target device node is associated with the first candidate device node.

8. A device for correlating multi-source heterogeneous operation data of a power distribution network, characterized in that, include: The candidate device segmentation module is used to determine a set of candidate device nodes based on the multi-source heterogeneous operation data of the target device nodes in the distribution network system; wherein, the multi-source heterogeneous operation data includes physical operation data, meteorological data, topology chart data and text data; The joint similarity calculation module is used to calculate the joint similarity between the target device node and each candidate device node in the candidate device node set, including: calculating the joint similarity between the target device node and each node in the candidate device node set using the following formula: ； in, Indicates the target device node and candidate device nodes Joint similarity between; Indicates the semantic item weight, Indicates normalized semantic similarity; Indicates the target device node Multimodal semantic vectors Indicates candidate device node Multimodal semantic vectors; Indicates the weight of the topology term. Indicates the similarity of electrical topology neighborhoods; Indicates the target device node The set of electrical topological neighborhoods, Indicates candidate device node The set of electrical topological neighborhoods; Indicates the electrical distance penalty factor. Indicates the target device node With candidate device nodes Electrical distance between; Represents the natural exponential function; The association relationship determination module is used to determine the association relationship between the target device node and each candidate device node based on the joint similarity. The fusion module is used to fuse the multi-source heterogeneous operational data of the target device node and the associated candidate device nodes if the target device node is associated with a candidate device node, generating operational data that conforms to the actual operating conditions of the target device node, including: The multi-source heterogeneous operating data is divided into continuous data and discrete data; The following formula is used to fuse continuous data: ； in, This represents the operational data obtained after fusing continuous data; This indicates the total number of continuous data entries. Indicates the first Continuous data; Indicates the first The fusion weights for continuous data; ； Indicates the first The data source credibility or scenario-based weight corresponding to each continuous data point. This represents the time decay coefficient, used to control the degree to which the time difference decays the fusion weights. Indicates the first The time difference between continuous data points and the target time. This represents the average or reference value of measurements taken at device nodes in the electrical topology neighborhood. Indicates the first The squared deviation between the continuous data and the reference value in the electrical topology neighborhood. Indicates the spatial consistency scale parameter. Represents the natural exponential function; The following formula is used to fuse discrete data: ； in, This represents the operational data obtained after fusing discrete data. Indicates the first Category labels corresponding to discrete data , This represents the set of candidate categories corresponding to discrete data.

Images