Pseudo measurement data generation method, device and equipment

Abstract

The invention relates to a pseudo measurement data generation method, device and equipment. The method comprises the following steps: acquiring historical measurement data of each node in the power distribution network and a topological structure of the power distribution network; according to the historical measurement data of each node, correcting a transformer user relationship in the topological structure of the power distribution network to obtain an updated topological structure; based on node measurement data in the updated topological structure, training the initial data generation model to obtain a target data generation model; the target data generation model is used for generating node pseudo measurement data and branch pseudo measurement data in the power distribution network. By adopting the method, the accuracy of pseudo measurement data can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device and apparatus for generating pseudo measurement data. Background Art

[0002] In the field of power systems, pseudo-measurement data refers to electrical parameters generated through mathematical modeling, simulation or prediction algorithms. Pseudo-measurement can fill the gaps in actual measurement data to fully understand the operating status of the power system.

[0003] In related technologies, when generating pseudo measurement data, the measurement data of known nodes in the power system are usually used as a basis to build a data generation model, and the measurement data of missing nodes in the power system are predicted through the data prediction model to obtain pseudo measurement data of the missing nodes.

[0004] However, the accuracy of pseudo measurement data generated by the data generation model in the related art is low. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, device and equipment for generating pseudo measurement data that can improve the accuracy of pseudo measurement data in order to address the above technical problems.

[0006] In a first aspect, the present application provides a method for generating pseudo measurement data, comprising:

[0007] Obtain historical measurement data of each node in the distribution network and the distribution network topology;

[0008] According to the historical measurement data of each node, the change relationship in the distribution network topology is modified to obtain an updated topology;

[0009] Based on the node measurement data in the updated topology structure, the initial data generation model is trained to obtain the target data generation model; the target data generation model is used to generate node pseudo measurement data and branch pseudo measurement data in the distribution network.

[0010] In one embodiment, the nodes include user nodes and device nodes; based on historical measurement data of each node, the user relationship in the distribution network topology is modified to obtain an updated topology, including:

[0011] According to the historical measurement data of each node, the correlation quantization value between each user node and each device node is determined to obtain the user correlation sequence of each user node;

[0012] Determine the target device node corresponding to each user node and the target user change relationship of each user node according to the user change correlation sequence of each user node;

[0013] According to the target change relationship of each user node, the change relationship of the distribution network topology is modified to obtain an updated topology.

[0014] In one embodiment, the historical measurement data includes voltage amplitude data; determining a correlation quantization value between each user node and each device node based on the historical measurement data of each node, and obtaining a user-to-user correlation sequence of each user node includes:

[0015] Normalize the voltage amplitude data of each node;

[0016] For any user node, the correlation quantization value between the normalized voltage amplitude sequence of the user node and the normalized voltage amplitude sequence of each device node is calculated respectively;

[0017] The quantitative values ​​of each correlation are summarized to obtain the user node's change correlation sequence.

[0018] In one embodiment, determining the target device node corresponding to each user node and the target user change relationship of each user node based on the user change correlation sequence of each user node includes:

[0019] For any user node, determining a maximum user-change correlation quantized value in the user-change correlation sequence of the user node;

[0020] Determine the device node corresponding to the maximum variable user correlation quantization value as the target device node of the user node;

[0021] The association relationship between the user node and the target device node is determined as the target user change relationship of the user node.

[0022] In one embodiment, based on the target user change relationship of each user node, the user change relationship of the distribution network topology is modified to obtain an updated topology structure, including:

[0023] Filter out the target user change relationships of each user node, the user change relationships to be processed that do not match the current user change relationships in the distribution network topology, and the user nodes to be processed corresponding to each of the target user change relationships;

[0024] Based on the voltage amplitude data of each user node to be processed, each user node to be processed is divided into a first user node and a second user node;

[0025] Different correction strategies are adopted to correct the first pending household change relationship corresponding to the first user node and the second pending household change relationship corresponding to the second user node, respectively, to obtain an updated topological structure.

[0026] In one embodiment, different correction strategies are adopted to correct the first pending account change relationship corresponding to the first user node and the second pending account change relationship corresponding to the second user node, respectively, to obtain an updated topology structure, including:

[0027] For any first user node, replacing the current user change relationship of the first user node in the distribution network topology with the target user change relationship of the first user node;

[0028] For any second user node, the user correlation sequence between the second user node and other user nodes in the distribution network topology is calculated, and the current user change relationship of the second user node in the distribution network topology is corrected according to the user correlation sequence of the second user node.

[0029] In one embodiment, modifying a current user relationship of a second user node in a distribution network topology structure according to a user correlation sequence of the second user node includes:

[0030] Obtaining significant feature values ​​of user correlation sequences;

[0031] When the significant feature value is greater than the standard significant feature value, determining a maximum user relevance quantization value in the user relevance sequence;

[0032] Determine the device node corresponding to the maximum user correlation quantization value as the target device node of the second user node;

[0033] According to the user change relationship between the second user node and the target device node, the current user change relationship of the second user node in the power distribution network topology is updated.

[0034] In one embodiment, based on the node measurement data in the updated topology structure, the initial data generation model is trained to obtain the target data generation model, including:

[0035] Construct training data based on the node measurement data in the updated topology structure;

[0036] Input the training data into the initial data generation model to obtain the predicted data of the training data;

[0037] Obtain the power balance residual loss value and voltage stability constraint loss value between the labeled data of the training data and the predicted data, and obtain the composite loss function value of the initial data generation model;

[0038] The initial data generation model is trained based on the composite loss function value to obtain the target data generation model.

[0039] In a second aspect, the present application further provides a pseudo measurement data generating device, comprising:

[0040] An acquisition module is used to obtain historical measurement data of each node in the distribution network and the distribution network topology;

[0041] The correction module is used to correct the change relationship in the distribution network topology structure based on the historical measurement data of each node to obtain an updated topology structure;

[0042] The training module is used to train the initial data generation model based on the node measurement data in the updated topology structure to obtain the target data generation model; the target data generation model is used to generate node pseudo measurement data and branch pseudo measurement data in the distribution network.

[0043] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in any one of the embodiments of the first aspect are implemented.

[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0045] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0046] The above-mentioned pseudo-measurement data generation method, device, and apparatus obtain historical measurement data and the distribution network topology of each node in the distribution network; based on the historical measurement data of each node, modify the change relationships in the distribution network topology to obtain an updated topology; and based on the node measurement data in the updated topology, train an initial data generation model to obtain a target data generation model; the target data generation model is used to generate pseudo-measurement data for nodes and branches in the distribution network. In this method, the change relationships between nodes are obtained based on the node's historical measurement data, thereby modifying the change relationships in the distribution network topology and improving the accuracy and rationality of the updated topology. Next, the data generation model is trained using the node measurement data of the updated topology as training data. Because the training data is reliable, the target data generation model obtained based on this training is also accurate and can accurately represent the node and line relationships in the distribution network topology. Therefore, when the target data generation model is actually applied, the pseudo-measurement data generated by the target data generation model is also accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A diagram showing an application environment of a method for generating pseudo measurement data in one embodiment;

[0049] Figure 2 1 is a flow chart of a method for generating pseudo measurement data according to an embodiment;

[0050] Figure 3 Schematic diagram of a topology structure updating process in one embodiment;

[0051] Figure 4 Schematic diagram of a flow chart of a step of obtaining a change correlation sequence in one embodiment;

[0052] Figure 5 Schematic diagram of a flow chart of a topology structure updating step in another embodiment;

[0053] Figure 6 Schematic diagram of a flow chart of the steps for determining a household relationship change in one embodiment;

[0054] Figure 7 A schematic diagram of a flow chart of a model training step in one embodiment;

[0055] Figure 8 is a flow chart of a method for generating pseudo measurement data in another embodiment;

[0056] Figure 9 FIG. 1 is a schematic diagram of the architecture of a pseudo-measurement generation model in one embodiment;

[0057] Figure 10 FIG. 4 is a structural block diagram of a pseudo measurement data generating device in one embodiment. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0059] With the widespread access of massive new decentralized elements and new business formats such as distributed new energy, new energy storage, diversified loads, microgrids, and virtual power plants to the distribution network, the requirements for them have also increased, and it is necessary to achieve comprehensive observability and measurability and digital transparency in all links.

[0060] However, the relevant technologies have a low effective coverage rate for distribution network measurements. Some distribution networks lack state perception capabilities and have poor real-time performance, which cannot effectively meet existing control needs. Short-term load forecasting based on statistical analysis only explores the potential connection between system measurement information and state quantities in terms of pseudo-measurement generation, and does not consider the topological connection relationship between distribution network nodes, resulting in insufficient data generation frequency. In addition, the existing calculation method fails to achieve accurate perception of measurement-missing nodes, resulting in missing data for some branches and nodes, which reduces the accuracy of distribution network data measurement and affects the reliability of distribution network data monitoring results, which needs to be urgently addressed.

[0061] Pseudo-measurements refer to power system state parameters or operating data generated through mathematical modeling, simulation, or predictive algorithms. These are used to assist in power system monitoring, control, and optimization. Pseudo-measurements can fill gaps in actual measurement data and provide more comprehensive and accurate system status information.

[0062] In related technologies, pseudo-measurement data is typically generated by building a data generation model based on time-series data from distribution network node loads. This model then uses the data prediction model to predict the measurement data for missing nodes in the power system, resulting in pseudo-measurement data for the missing nodes. This results in insufficient data generation frequency and accuracy. Furthermore, current calculation methods fail to accurately detect nodes with missing measurements, effectively addressing the issue of missing data for some branches and nodes.

[0063] Based on this, the present application provides a method, device and equipment for generating pseudo-measurement data, which can further improve the accuracy of pseudo-measurement of the distribution network by topology identification and correction of the distribution network topology structure.

[0064] The pseudo measurement data generation method provided in the embodiment of the present application can be applied to a computer device. The computer device can be a server, and its internal structure diagram can be as follows: Figure 1As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store a pseudo-measurement data generation model for generating pseudo-measurement data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a pseudo-measurement data generation method is implemented.

[0065] It should be noted that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0066] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0067] In an exemplary embodiment, Figure 2 As shown, a method for generating pseudo measurement data is provided, comprising the following steps:

[0068] S201, obtaining historical measurement data of each node in the distribution network and the distribution network topology.

[0069] A distribution network is a power network system that transmits electricity from distribution substations to end users. Distribution network topology describes the connections between various network components, such as power sources, lines, transformers, and loads.

[0070] In actual application scenarios, the distribution network topology can be obtained by analyzing the drawings or design documents of the distribution network, or by on-site surveys and measurements. The distribution network topology can also be drawn through a geographic information system.

[0071] Nodes in the distribution network are used to collect and distribute electrical energy in the power system, including user nodes, transformer nodes, power supply nodes, etc. The measurement data of the nodes refers to the electrical parameters monitored by the measurement equipment (such as sensors) in the distribution network, such as voltage, current, impedance, power, switch telesignaling signals, etc.

[0072] In actual application scenarios, it is possible to communicate with the measurement devices corresponding to each node in the distribution network to obtain the measurement data sent by each measurement device, or to pull the historical measurement data of each node from the cloud server in the distribution network according to the historical time period.

[0073] S202: According to the historical measurement data of each node, the user change relationship in the distribution network topology is modified to obtain an updated topology.

[0074] The transformer-user relationship in the distribution network refers to the power supply relationship between the transformer and the user, which can be the power supply and use relationship between the lines and substations in the power supply enterprise and the users who use the transformers in the substations to supply power.

[0075] According to the historical measurement data of each node, the correlation coefficient between each user node and each transformer node is calculated to obtain the transformer correlation sequence corresponding to each user node. Then, for any user node, based on the transformer correlation sequence corresponding to the user node, the transformer node corresponding to the user node is determined as the actual transformer relationship of the user node, so as to obtain the actual transformer relationship of each user node. Then, the actual transformer relationship of each user node is replaced with the corresponding transformer relationship in the distribution network topology to obtain an updated topology.

[0076] In another implementation scenario, mathematical models can be directly used to analyze historical measurement data from each node in the distribution network to automatically identify the distribution network topology and obtain an updated topology diagram. For example, a graph-theory-based algorithm can abstract the distribution network into a graph and determine the updated topology by analyzing the graph.

[0077] S203 , based on the node measurement data in the updated topology structure, the initial data generation model is trained to obtain a target data generation model; the target data generation model is used to generate node pseudo measurement data and branch pseudo measurement data in the distribution network.

[0078] The data generation model can be a pseudo-measurement generation model built based on a spatiotemporal graph neural network. In the embodiment of the present application, the target data generation model represents the topological structure of the distribution network and the connection relationship between the nodes in the topological structure. In actual applications, the input of the target data generation model is the measurement data of some nodes in the distribution network, and the output is the pseudo-measurement data of the missing nodes in the distribution network and the pseudo-measurement data of the branches in the distribution network.

[0079] In an embodiment of the present application, the node measurement data in the updated topology structure is organized into a graph data set as a model training data set, and the model training data set is divided into a training set, a validation set, and a test set according to a preset ratio, such as 8:1:1; the training set is used to optimize the model parameters, the hyperparameters of the model are adjusted through the validation set, and finally the generalization, robustness, and accuracy of the model are verified on the test set. If the initial data generation model meets the requirements, the training is terminated and the optimized data generation model, that is, the target data generation model, is saved.

[0080] This method obtains historical measurement data and the distribution network topology of each node in the distribution network; based on the historical measurement data of each node, the change relationships in the distribution network topology are modified to obtain an updated topology; based on the node measurement data in the updated topology, an initial data generation model is trained to obtain a target data generation model; and the target data generation model is used to generate pseudo-measurement data for nodes and branches in the distribution network. In this method, the change relationships between nodes are obtained based on the node's historical measurement data, which enables the change relationships in the distribution network topology to be modified, improving the accuracy and rationality of the updated topology. Next, the data generation model is trained using the node measurement data of the updated topology as training data. Because the training data is reliable, the target data generation model obtained based on this training is also accurate and can accurately represent the node and line relationships in the distribution network topology. Therefore, when the target data generation model is actually applied, the pseudo-measurement data generated by the target data generation model is also accurate.

[0081] Next, the steps for modifying the household relationship in the above embodiment will be further described. In an exemplary embodiment, Figure 3 As shown, the nodes include user nodes and device nodes; based on the historical measurement data of each node, the user relationship in the distribution network topology is corrected to obtain an updated topology, which includes the following steps:

[0082] S301, determining the correlation quantization value between each user node and each device node based on the historical measurement data of each node, and obtaining the user correlation sequence of each user node.

[0083] In the embodiment of the present application, the device node refers to a transformer node. The historical measurement data of the user node is sorted according to the time sequence of the measurement data to obtain a historical measurement data sequence of the user node, and the historical measurement data of the device node is sorted to obtain a historical measurement data sequence of the device node.

[0084] For any user node, the correlation quantification value between the historical measurement data sequence of the user node and the historical measurement data sequence of each device node is calculated separately, such as the Pearson correlation coefficient, and then the correlation quantification values ​​between the user node and different device nodes are summarized to obtain the user-related correlation sequence of the user node.

[0085] S302: Determine the target device node corresponding to each user node and the target user change relationship of each user node according to the user change correlation sequence of each user node.

[0086] For any user node, a target quantization value is determined from multiple correlation quantization values ​​in the user node's change correlation sequence, the device node corresponding to the target quantization value is determined as the target device node corresponding to the user node, and the connection relationship between the user node and the target device node is determined as the target change relationship of the user node.

[0087] Exemplarily, for any user node, the maximum user change correlation quantization value in the user node's user change correlation sequence is determined; the device node corresponding to the maximum user change correlation quantization value is determined as the target device node of the user node; and the association relationship between the user node and the target device node is determined as the target user change relationship of the user node.

[0088] It should be noted that a larger quantized value of the user-to-device correlation between a user node and a device node indicates a more tightly coupled connection between the user node and the device node. Conversely, a smaller quantized value of the user-to-device correlation between the user node and the device node indicates a more independent connection between the user node and the device node. Based on this, the device node corresponding to the maximum quantized value of the user-to-device correlation in the user node's user-to-device correlation sequence is determined as the target device node for the user node, i.e., the device node with the closest connection to the user node. The association between the user node and the target device node is then determined as the target user-to-device relationship for the user node. This entire determination process is based on reliable principles, has clear logic, and is easy to implement.

[0089] S303: According to the target user change relationship of each user node, the user change relationship of the distribution network topology structure is modified to obtain an updated topology structure.

[0090] The current user change relationship of each user node is determined from the distribution network topology structure, and the current user change relationship of each user node is compared with the target user change relationship of each user node. According to the comparison result, the user change relationship correction strategy of the distribution network topology structure is determined to obtain the updated topology structure.

[0091] Exemplarily, for any user node, if the current account change relationship of the user node does not match the corresponding target account change relationship, the current account change relationship is replaced with the target account change relationship.

[0092] In an embodiment of the present application, based on the historical measurement data of each node, the quantitative value of the correlation between each user node and each device node is determined, and the target household change relationship corresponding to each user node is further determined, which provides a reliable correction basis for the household change relationship correction of the distribution network topology structure, so that the corrected updated topology structure can more clearly and accurately represent the actual topology architecture of the distribution network, thereby improving the authenticity and accuracy of the updated topology structure.

[0093] In an exemplary embodiment, the historical measurement data includes voltage amplitude data; Figure 4 As shown, the aforementioned step S301 "determining the correlation quantization value between each user node and each device node based on the historical measurement data of each node to obtain the user-to-user correlation sequence of each user node" includes the following steps:

[0094] S401 , normalizing the voltage amplitude data of each node.

[0095] For any node, the normalization expression is as follows:

[0096]

[0097] in, represents the original value of the i-th voltage amplitude data at the k-th moment, The normalized result is, and are the minimum and maximum values ​​of the voltage amplitude data in the historical k moments respectively.

[0098] S402 : For any user node, respectively calculate the correlation quantization value between the normalized voltage amplitude sequence of the user node and the normalized voltage amplitude sequence of each device node.

[0099] For user node i and device node j, the correlation quantization value Defined as:

[0100]

[0101] in, express and The covariance between and They are and The variance of . and Represents the normalized sequence of voltage amplitude of user node i , the voltage amplitude normalized sequence of the device node .

[0102] S403: Aggregate the quantized correlation values ​​to obtain a user-change correlation sequence of the user node.

[0103] The correlation quantization values ​​between the same user node and each device node are sorted into a sequence to obtain the user correlation sequence of the user node.

[0104] In an embodiment of the present application, taking into account the differences between the voltage amplitude data of different types of nodes, the voltage amplitude data of each node is normalized to obtain a voltage amplitude normalized sequence of the user node and a voltage amplitude normalized sequence of the device node, so as to eliminate the differences in the dimensions and value ranges between the indicators, facilitate data comparison and analysis, and then calculate the quantitative value of the correlation between the two to obtain the user node's variable correlation sequence, so as to clearly reflect the degree of correlation between the user node and each device node.

[0105] In an exemplary embodiment, Figure 5 As shown, the aforementioned step S303 "modifying the user change relationship of the distribution network topology structure according to the target user change relationship of each user node to obtain an updated topology structure" includes the following steps:

[0106] S501 , filtering out, from the target user change relationships of each user node, user change relationships to be processed that do not match the current user change relationships in the distribution network topology, and user nodes to be processed corresponding to each user change relationship to be processed.

[0107] The target user change relationship of each user node is compared with the corresponding current user change relationship in the distribution network topology structure, the user change relationship to be processed that does not match the current user change relationship is determined, and the user node corresponding to the user change relationship to be processed is determined as the user node to be processed.

[0108] S502 : Based on the voltage amplitude data of each user node to be processed, divide each user node to be processed into a first user node and a second user node.

[0109] Calculate the average voltage value of the voltage amplitude data of each user node to be processed, sort each user node to be processed according to the average voltage value of each user node to be processed, and obtain a sequence of user nodes to be processed; determine the user node in the sequence of user nodes to be processed that matches a preset number of nodes as the first user node, and determine the user node in the sequence of user nodes to be processed other than the first user node as the second user node. The preset number of nodes can be calculated based on the total number of user nodes in the distribution network and a threshold coefficient, such as 1 / 10, 1 / 20, etc. of the total number of user nodes in the distribution network, or can be calculated based on the total number of user nodes to be processed and a threshold coefficient, and this application does not impose any restrictions on this.

[0110] S503: adopt different correction strategies to correct the first pending account change relationship corresponding to the first user node and the second pending account change relationship corresponding to the second user node, respectively, to obtain an updated topology structure.

[0111] In actual application scenarios, the first pending change relationship corresponding to the first user node can replace the current change relationship corresponding to the first user node in the distribution network topology structure, while keeping the second pending change relationship corresponding to the second user node unchanged.

[0112] In an exemplary embodiment, for any first user node, the current user change relationship of the first user node in the distribution network topology is replaced with the target user change relationship of the first user node; for any second user node, the user correlation sequence between the second user node and other user nodes in the distribution network topology is calculated, and according to the user correlation sequence of the second user node, the current user change relationship of the second user node in the distribution network topology is corrected.

[0113] For any second user node, the quantitative value of the correlation between the second user node and other user nodes in the distribution network topology structure except the second user node is calculated to obtain a user correlation sequence of the second user node, and then the target user node that is most relevant to the second user node is determined according to the user correlation sequence, and the device node corresponding to the target user node is obtained, and the device node corresponding to the target user node is determined as the device node corresponding to the second user node, to obtain a revised user change relationship of the second user node, and the revised user change relationship of the second user node replaces the corresponding current user change relationship in the distribution network topology structure.

[0114] In this way, the current user change relationship of the first user node in the distribution network topology structure is replaced with the target user change relationship of the first user node, and the current user change relationship of the second user node in the distribution network topology structure is corrected according to the user correlation sequence between the second user node and other user nodes in the distribution network topology structure. This is equivalent to using different correction bases for different types of user nodes, providing user change relationship correction strategies that match different types of user nodes, so as to improve the diversity and flexibility of the correction methods, and thus ensure the rationality of the correction results.

[0115] In an embodiment of the present application, the target change relationship of each user node is compared with the current change relationship corresponding to each user node in the distribution network topology structure to determine the user node to be processed. Based on the voltage amplitude data of each user node to be processed, each user node to be processed is divided into a first user node and a second user node. Different correction strategies are adopted to correct the first change relationship to be processed corresponding to the first user node and the second change relationship to be processed corresponding to the second user node, respectively, to obtain an updated topology structure. This is equivalent to adopting different correction strategies for user nodes to be processed with different voltage amplitude data, so as to realize hierarchical correction of the user nodes to be processed, so as to ensure the rationality of the updated topology structure.

[0116] Next, the correction method of the second user node is further described through an embodiment. Figure 6 As shown, according to the user correlation sequence of the second user node, the current user change relationship of the second user node in the distribution network topology is modified, including:

[0117] S601: Obtain significant feature values ​​of user correlation sequences.

[0118] It is emphasized again that the user correlation sequence includes the user correlation quantization values ​​between the second user node and other user nodes.

[0119] The significant characteristic value of the user correlation sequence refers to a statistic that measures the degree of dispersion of the user correlation sequence, which can be the standard deviation of the user correlation sequence. The larger the significant characteristic value, the greater the fluctuation of the data representing the user correlation sequence and the higher the degree of dispersion. The smaller the significant characteristic value, the smaller the fluctuation of the data representing the user correlation sequence and the lower the degree of dispersion. In the embodiment of the present application, the standard deviation calculation formula is used to calculate the standard deviation of the user correlation sequence as the significant characteristic value of the user correlation sequence.

[0120] S602: When the significant feature value is greater than the standard significant feature value, determine the maximum user relevance quantization value in the user relevance sequence.

[0121] If the significant eigenvalue of the user correlation sequence is greater than the standard significant eigenvalue, it indicates that the user correlation sequence is relatively discrete. In this case, it is necessary to further correct the user change relationship of the second user node corresponding to the user correlation sequence. Furthermore, in this embodiment of the present application, the user correlation quantized values ​​in the user correlation sequence are sorted, and the maximum user correlation quantized value is obtained as the basis for correcting the user change relationship of the second user node.

[0122] In another case, the significant characteristic value of the user correlation sequence is smaller than the standard significant characteristic value, indicating that the user correlation sequence is relatively concentrated. In this case, the user relationship of the second user node in the distribution network topology can be maintained.

[0123] S603: Determine the device node corresponding to the maximum user correlation quantization value as the target device node of the second user node.

[0124] It should be noted that the user correlation quantization value refers to the correlation index between the second user node and other user nodes in the distribution network topology other than the second user node. For any user node other than the second user node, a larger user correlation quantization value indicates a closer connection between the second user node and the user node. Conversely, a smaller user correlation quantization value indicates a more distant connection between the second user node and the user node. Based on this, the device node corresponding to the maximum user correlation quantization value is determined as the target device node for that user node.

[0125] S604: Update the current user change relationship of the second user node in the distribution network topology structure according to the user change relationship between the second user node and the target device node.

[0126] The user change relationship between the second user node and the target device node replaces the current user change relationship of the second user node in the distribution network topology structure, thereby updating the user change relationship of the second user node.

[0127] In an embodiment of the present application, when the significance characteristic value of the user correlation sequence of the second user node is greater than the standard significance characteristic value, the maximum user correlation quantization value in the user correlation sequence is determined, and the device node corresponding to the maximum user correlation quantization value is determined as the target device node of the second user node. According to the household change relationship between the second user node and the target device node, the current household change relationship of the second user node in the distribution network topology structure is updated, which is equivalent to using the household change relationship of the adjacent nodes of the second user node as the household change relationship of the second user node, thereby realizing the update of the current household change of the second user node and improving the rationality of the household change relationship of the second user node.

[0128] In an exemplary embodiment, Figure 7 As shown, based on the node measurement data in the updated topology structure, the initial data generation model is trained to obtain the target data generation model, including:

[0129] S701: Construct training data based on the node measurement data in the updated topology structure.

[0130] Based on the node measurement data in the updated topology structure, multi-time step historical graph data describing node and edge features is constructed to construct a dynamic spatiotemporal graph The characteristics of the graph include: node characteristics , edge features and timestamp Among them, the node features , including node injection power and voltage; edge characteristics Including line impedance, branch power, current, power transmission direction The value is or , The timestamp representing the power from node i to j, or from node j to i, is used as the same time section data to measure time synchronization alignment.

[0131] According to the time arrangement order of the spatiotemporal graph data, the node measurement data are divided, 80% of the node measurement data are used as the training data set, and 20% of the node measurement data are used as the pure labeled test set.

[0132] S702: Input the training data into the initial data generation model to obtain predicted data of the training data.

[0133] The training data is input into the initial data generation model, and the node features of the training data are extracted through the initial data generation model. Combined with the updated distribution network topology, the missing nodes in the distribution network other than the nodes of the training data and the data of the missing nodes are predicted as the predicted data of the training data.

[0134] S703 , obtaining a power balance residual loss value and a voltage stability constraint loss value between the label data of the training data and the predicted data, and obtaining a composite loss function value of the initial data generation model.

[0135] Prediction data includes power state and voltage amplitude. The power balance residual loss value is a quantitative measure of the difference between the predicted power state and the power state label. The voltage stability constraint loss value is used to evaluate whether the model's predicted voltage state meets the voltage stability constraints. The power balance residual loss value and the voltage stability constraint loss value corresponding to the training data are weighted and fused to obtain the composite loss function value.

[0136] S704: Train the initial data generation model based on the composite loss function value to obtain a target data generation model.

[0137] The model parameters in the initial data generation model are updated using a staged gradient descent method until the composite loss function value of the updated model is less than the preset threshold. The training is completed and the target data generation model is obtained.

[0138] In an embodiment of the present application, a composite loss function value of the initial data generation model is obtained based on the power balance residual loss value and the voltage stability constraint loss value between the label data and the predicted data of the training data to comprehensively evaluate the training effect of the model, and the initial data generation model is trained based on the composite loss function value to improve the training speed and training accuracy of the model so as to quickly obtain the target data generation model.

[0139] In an exemplary embodiment, a pseudo measurement generation method for a low-voltage distribution network is provided, wherein the data generation model is a spatiotemporal graph neural network model, such as Figure 8 As shown, the following steps are included:

[0140] S801, pre-processing the existing topology, device parameters and measurement data collected by known sensing devices.

[0141] Obtain relevant data of the distribution network and perform preprocessing; wherein, the distribution network related data includes: historical power consumption of smart meters on the user side of the low-voltage distribution network; , voltage amplitude as user node characteristics; edge physical characteristics, line impedance parameters , branch power measurement; switch telesignaling signal, the closing and opening status of each switchgear in the distribution network, real-time tracking of distribution network topology changes; physical derivative characteristics, generating node power balance residuals based on Kirchhoff's law , It represents the sum of all input powers of node i (inflow is positive and outflow is negative), and serves as the semi-supervisory signal feature of node i. The node power balance residual represents the difference between the power input and output at each node. It serves as a regularization term to correct and optimize the power distribution in the network when training the model.

[0142] S802: Based on the voltage amplitude provided by the advanced measurement equipment and the voltage and current historical measurement data provided by the line switch measurement equipment, the phase-to-household relationship and the transformer-to-household relationship of the distribution network are obtained by calculating the voltage Pearson correlation coefficient, and based on this, the obtained phase-to-household relationship and line-to-household relationship are corrected.

[0143] The specific correction steps are as follows:

[0144] (1) Based on the voltage data set of each node , calculate the correlation coefficient value between each node, and get the dimension The node voltage correlation coefficient matrix ,matrix Divided into four block matrices: The low-voltage distribution network to be identified (the voltage correlation coefficient matrix between the low-voltage distribution network busbar and the adjacent low-voltage distribution network busbars; The voltage correlation coefficient matrix between the low-voltage distribution network busbar to be identified and the users in the adjacent low-voltage distribution network; is a matrix The transpose of Represents the voltage correlation coefficient matrix between the low-voltage distribution network to be identified and the users in the adjacent low-voltage distribution networks.

[0145] (2) In the matrix In the column vector, the column vector represents the correlation coefficient between the user and each low-voltage distribution network node. The low-voltage busbar where the maximum value in each column vector is located determines the corresponding low-voltage distribution network and serves as the user-change relationship of the corresponding user, thereby obtaining a preliminary user-change relationship.

[0146] (3) Compare the existing low-voltage distribution network ledger information table with the household change relationship result obtained in step (2). In a low-voltage distribution network, if the connectivity between two different users is inconsistent, they will be included in the first user set. middle, ,in is the number of low voltage distribution networks, and the total The first user.

[0147] (4) For each user in the low-voltage distribution network ledger information, based on the original voltage data, calculate the average voltage of each user during the measurement period and sort them from large to small according to the voltage value. The user with higher voltage is closer to The users at the head end of the low-voltage distribution network are obtained as the head end user set ,in is the threshold coefficient, , For the The number of users to be identified in the low-voltage distribution network is A set of head-end users.

[0148] (5) Each set belongs to The users identified as users with incorrect household change relationships shall be based on the household change relationships in step (2) and shall be selected from the first user set. The updated first user set is recorded as , .

[0149] (6) Divide the users in each low-voltage distribution network into two categories, one of which is the updated first user set , and the other is the second user set that does not need to be processed. Through the matrix , extract the correlation coefficient values ​​between the users in the second user set in all low-voltage distribution networks.

[0150] (7) Calculation set Middle users The mean of the correlation coefficient series and standard deviation By comparing the mean and standard deviation of each user, the relationship between users is corrected.

[0151] If the collection A user node in exist The corresponding mean value under the candidate areas Significantly below the preset threshold, or standard deviation If it is significantly higher than the standard distribution interval, it is judged that the current change relationship of the user is abnormal. Select the correlation value between the user and all low-voltage distribution network nodes, take the substation where the busbar corresponding to the maximum voltage amplitude is located as the new substation, and update the user node The connection relationship between the bus node and the reconstruction graph structure The adjacency matrix and edge set in .

[0152] S803: After the topology is modified, the active power and current amplitude of each branch of the low-voltage distribution network and the voltage amplitude measurement of the branch terminal are generated based on the historical measurement data of the user's smart meter and branch terminal. The data is organized into a graph dataset as a model training dataset and divided into a training set and a test set according to a preset ratio.

[0153] In the training set, the first 80% of the data served as a mixed labeled and unlabeled training set, with 30% of the nodes randomly masked as unlabeled data. The remaining 20% ​​served as a purely labeled test set. In both the training and test sets, node power and voltage amplitudes were used as node features, and a high-frequency data mask was applied. In the training set, spatiotemporal graph data for each load branch and node was constructed as the training set input data; the masked data served as the training set label data.

[0154] The masking process is done by using the mask vector Multiplication is achieved, where is a vector containing 0 and 1, where 0 indicates that the feature is masked and 1 indicates that the feature remains unchanged. The data processed in this way is used as the training target of the model and participates in the loss calculation of each iteration. In the embodiment of the present application, the masking step includes: (1) determining the input data sample, including the low-voltage distribution network load node and its surrounding adjacent nodes; (2) generating a mask vector for each adjacent load node , specify the features to be masked; (3) For each adjacent load node, multiply the input features by the corresponding mask vector to obtain the masked data, which is used as the target for model training and the loss of each iteration is calculated. For unlabeled nodes, a power balance residual constraint is added to the loss function to ensure that the model complies with the physical constraints of the power grid during training.

[0155] S804: Build a pseudo-measurement generation model based on a spatiotemporal graph neural network and use the training set to optimize the model parameters. Adjust the model's hyperparameters through the validation set, and finally verify the model's generalization, robustness, and accuracy on the test set. If the model meets the requirements, save the optimized model parameters and terminate training.

[0156] The architecture of the low-voltage distribution network pseudo-measurement generation model is shown in the figure below: Figure 9 As shown, it includes a physical enhanced multi-head self-attention mechanism module and a physical constraint space processing module. The overall architecture is encoder-decoder. Through multi-module collaboration, it realizes the deep integration of data-driven and physical laws. The specific features are as follows: The physical enhanced multi-head self-attention mechanism module embeds the distribution network topology perception position coding in the traditional self-attention mechanism, based on the node impedance matrix Dynamically adjust the attention weight distribution to make the model focus on the node associations with close electrical distances. At the same time, the sparsity of attention scores is constrained by Kirchhoff's current law to suppress the noise interference of non-physical associated edges. The physical constraint space processing module adopts a bidirectional spectrum graph convolution method to the distribution network topology (edge ​​features). ) The generated graph Laplacian matrix As the spectrum basis, the line power transmission equation is integrated into the frequency domain filter design The linear constraints explicitly model the propagation characteristics of node voltage-current fluctuations in the spatial dimension; the physical regularization time processing module, based on the improved GRU unit, injects the power balance residual term into the hidden state update ; Semi-supervised coding layer design, by integrating historical measurement data and topological features Extract common spatiotemporal patterns and decode them using joint supervision data (labeled nodes) ) and the pseudo-measurements generated by physical equations to achieve the pseudo-measurement generation driven by the collaborative efforts of “a small amount of labeled data + physical laws”.

[0157] Backpropagation and the Adam optimization algorithm are used to adjust the model's weights and parameters while meeting physical constraints (power balance). Hyperparameter optimization in the test set includes learning rate, batch size, network depth, etc. to improve the model's generalization ability and accuracy.

[0158] The overall architecture of the spatiotemporal graph neural network model adopts an encoder-decoder architecture, achieving physical information fusion through the following: ① Inductive bias design. The bidirectional spectral convolution of the spatial processing module is based on the distribution network topology (edge ​​features E), enforcing the physical connectivity constraints between nodes (Kirchhoff's law). The improved GRU of the temporal processing module introduces a priori information on the smoothness of the power load time series (load change rate constraint). ② Learning bias mechanism: The residual term of the physical equation is embedded in the loss function as a soft constraint for semi-supervised learning.

[0159] Next, the training steps of the spatiotemporal graph neural network pseudo-measurement generation model are explained as follows:

[0160] (1) Physical perception parameter initialization.

[0161] The Xavier initialization method is adopted, but pre-trained weights are used for the physical constraint layer (spectral convolution kernel). The distribution network steady-state equation is used to generate synthetic data to pre-train the basic module. The learning rate is set to 0.0008 (supervision term) and 0.0005 (physical constraint term). The optimizer uses Adam multi-task weight decay.

[0162] (2) Calculation of hybrid loss function.

[0163] Randomly extract labeled samples from historical data (number 128) and unlabeled samples (Quantity 128), total batch , define the composite loss function:

[0164]

[0165] in, represents the model output function, represents pseudo measurement data generated by spatiotemporal graph neural network; No. arrive Input features at each moment (time series data); Represents the edge characteristics of the distribution network topology, usually including information such as admittance, impedance, and current; For the Real measurement data at all times is used as a supervisory signal; , are the weight coefficients in the loss function, which control the proportion of supervision loss and physical constraint loss in the total loss respectively.

[0166] The purpose of defining the composite loss function is to balance the accuracy and physical consistency of the model. The supervision loss part guides the training of the model by labeling the error of the data, and the physical constraint loss part ensures that the model follows the physical laws of the power system. It is the weighted sum of the power balance residual and the voltage stability constraint, and is used to introduce physical laws into model training. The specific calculation formula is:

[0167]

[0168] in, and is the weight coefficient in the loss function, which controls the importance of power balance residual and voltage stability constraint in the total loss respectively; power balance residual and voltage stability constraints According to the preset weight coefficient and Weighted search.

[0169] Among them, the power balance residual According to Kirchhoff's current law, the power injected into each node should be equal to the power flowing out of the node. The power balance residual ensures the power balance constraint of the node. The specific calculation method is:

[0170]

[0171] in, represents a collection of unlabeled data; For nodes To neighboring nodes The branch power; For nodes Pseudo-measurement value of the injected power; Representation and Node A set of adjacent nodes; represents the two-norm, i.e., the Euclidean distance. represents that for each unlabeled node, the difference between the sum of all input powers of the node and its predicted value of the injected power is calculated.

[0172] Voltage stability constraints , refers to constraining the stability of the power grid by calculating the relationship between line voltage drop, current and resistance for the voltage pseudo-measurement that needs to be calculated. The specific calculation method is:

[0173]

[0174] in, Indicates that the line is the line voltage drop; For the line The current on For the line Impedance; It represents the difference between voltage drop and current multiplied by impedance.

[0175] (3) Gradient update and physical regularization methods.

[0176] Update model parameters using staged gradients:

[0177]

[0178] in, and are the weight and bias of the model respectively; represents the supervision loss (loss of labeled data); is the physical constraint loss.

[0179] (4) Alternating training strategy.

[0180] During the pre-training phase (the first 2000 iterations), we use Constrain unlabeled data and initialize physical consistency representation; in the joint training phase, the joint training phase alternates sampling and , gradually increase Weights; During the fine-tuning phase (the last 1000 iterations), freeze the spectral convolutional layers and focus on optimizing the time series prediction accuracy.

[0181] (4) Physical compliance assessment.

[0182] Test set performance indicators include: prediction accuracy: MSE, MAE (measurement node), physical consistency: power imbalance .

[0183] S805, deploy the target spatiotemporal graph neural network model in the computing unit of the distribution network to receive data transmitted in real time by the data center and generate pseudo-measurement values ​​for each branch and each node in the distribution network.

[0184] In an embodiment of the present application, a pseudo-measurement generation model for a low-voltage distribution network is constructed by taking into account the voltage correlation analysis method to correct and utilize the load connection relationships such as line-to-user relationships and phase-to-user relationships, equipment parameters, and known sensing devices to collect measurement data, and calculate the time series flow data of each branch and node. Then, a pseudo-measurement generation model is constructed based on a spatiotemporal graph neural network. Through the physical parameterization of topological awareness attention and spectral graph convolution, physical laws such as Kirchhoff's law and power transmission equation of the distribution network are encoded as model inductive bias; combined with the Encoder-Decoder architecture for multimodal fusion of labeled data and physical priors, a semi-supervised learning paradigm is constructed to reduce dependence on full-node labeled data while ensuring the physical rationality of the pseudo-measurement.

[0185] The low-voltage distribution network pseudo-measurement generation model based on spatiotemporal graph neural network makes up for the shortcomings of measurement data missing due to the upload bandwidth limitation of the low-voltage distribution network, and realizes comprehensive perception of the basic data of the distribution network. By calculating time-series flow data, the flow of each distribution network branch and node with different data frequencies can be unified, which helps to construct a pseudo-measurement generation dataset; by considering the correlation analysis method of user voltage measurement data, the phase-to-household relationship and the transformer-to-household relationship are modified as the predefined topological structure of the graph neural network. The designed graph neural network model can deeply explore the spatial correlation of time-series data of different lines and nodes; by constructing a low-voltage distribution network pseudo-measurement generation model based on a spatiotemporal graph neural network, the accuracy of pseudo-measurement generation is improved; in addition, for the real-time generation of distribution network pseudo-measurements, the overall model architecture in this example adopts the Seq2Seq Encoder-Decoder architecture. By integrating Kirchhoff's laws and the physical constraints of the power transmission equation, the topological connection relationship (phase-to-household relationship, transformer-to-household relationship) is encoded as a predefined spectral convolution kernel of the graph neural network. The labeled data and the physical equation residuals are jointly trained through a semi-supervised learning framework to make up for the shortcomings of traditional data-driven methods that have a strong dependence on pseudo-measurement labeled data due to the lack of physical laws.

[0186] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0187] Based on the same inventive concept, embodiments of the present application also provide a pseudo measurement data generation device for implementing the pseudo measurement data generation method described above. The solution provided by this device is similar to the solution described in the method described above. Therefore, the specific limitations of one or more pseudo measurement data generation device embodiments provided below can be found in the above-described limitations of the pseudo measurement data generation method and will not be further elaborated here.

[0188] In an exemplary embodiment, Figure 10 As shown, a pseudo measurement data generating device is provided, comprising: an acquisition module 1001, a correction module 1002 and a training module 1003, wherein:

[0189] An acquisition module 1001 is used to acquire historical measurement data of each node in the distribution network and the distribution network topology;

[0190] The correction module 1002 is used to correct the user relationship in the distribution network topology structure based on the historical measurement data of each node to obtain an updated topology structure;

[0191] The training module 1003 is used to train the initial data generation model based on the node measurement data in the updated topology structure to obtain a target data generation model; the target data generation model is used to generate node pseudo measurement data and branch pseudo measurement data in the distribution network.

[0192] In an exemplary embodiment, the nodes include user nodes and device nodes; the correction module 1002 includes: a sequence determination unit, a relationship determination unit, and a relationship correction unit, wherein:

[0193] A sequence determination unit is used to determine the correlation quantization value between each user node and each device node based on the historical measurement data of each node, and obtain the user correlation sequence of each user node;

[0194] A relationship determination unit, configured to determine the target device node corresponding to each user node and the target user change relationship of each user node according to the user change correlation sequence of each user node;

[0195] The relationship correction unit is used to correct the user change relationship of the distribution network topology structure according to the target user change relationship of each user node to obtain an updated topology structure.

[0196] In an exemplary embodiment, the sequence determination unit is also used to normalize the voltage amplitude data of each node; for any user node, the correlation quantization value between the voltage amplitude normalization sequence of the user node and the voltage amplitude normalization sequence of each device node is calculated respectively; and the correlation quantization values ​​are summarized to obtain the user node's variable correlation sequence.

[0197] In an exemplary embodiment, the relationship determination unit is further used to determine, for any user node, the maximum user change correlation quantization value in the user change correlation sequence of the user node; determine the device node corresponding to the maximum user change correlation quantization value as the target device node of the user node; and determine the association relationship between the user node and the target device node as the target user change relationship of the user node.

[0198] In an exemplary embodiment, the relationship modification unit includes: a node determination subunit, a node division subunit, and a modification execution subunit, wherein:

[0199] The node determination subunit is used to screen out the pending user change relationships that do not match the current user change relationships in the distribution network topology from the target user change relationships of each user node, as well as the pending user nodes corresponding to each pending user change relationship;

[0200] A node division subunit, configured to divide each to-be-processed user node into a first user node and a second user node based on the voltage amplitude data of each to-be-processed user node;

[0201] The correction execution subunit is used to adopt different correction strategies to correct the first pending account change relationship corresponding to the first user node and the second pending account change relationship corresponding to the second user node to obtain an updated topology structure.

[0202] In an exemplary embodiment, the correction execution sub-unit is also used to replace the current user change relationship of the first user node in the distribution network topology structure with the target user change relationship of the first user node for any first user node; for any second user node, calculate the user correlation sequence between the second user node and other user nodes in the distribution network topology structure, and correct the current user change relationship of the second user node in the distribution network topology structure according to the user correlation sequence of the second user node.

[0203] In an exemplary embodiment, the correction execution sub-unit is also used to obtain the significance feature value of the user correlation sequence; when the significance feature value is greater than the standard significance feature value, determine the maximum user correlation quantization value in the user correlation sequence; determine the device node corresponding to the maximum user correlation quantization value as the target device node of the second user node; and update the current user change relationship of the second user node in the distribution network topology structure based on the user change relationship between the second user node and the target device node.

[0204] In an exemplary embodiment, the training module 1003 includes: a data acquisition unit, a model training unit, a function calculation unit, and a model generation unit, wherein:

[0205] A data acquisition unit, configured to construct training data based on node measurement data in the updated topology structure;

[0206] A model training unit, used to input training data into the initial data generation model to obtain predicted data of the training data;

[0207] A function calculation unit is used to obtain a power balance residual loss value and a voltage stability constraint loss value between the label data of the training data and the predicted data, and obtain a composite loss function value of the initial data generation model;

[0208] The model generation unit is used to train the initial data generation model based on the composite loss function value to obtain the target data generation model.

[0209] Each module in the pseudo-measurement data generation device described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0210] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0211] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0212] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0213] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0214] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0215] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0216] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating pseudo measurement data, characterized in that: The method comprises: Obtain historical measurement data of each node in the distribution network and the distribution network topology; According to the historical measurement data of each node, the user change relationship in the distribution network topology structure is modified to obtain an updated topology structure; Based on the node measurement data in the updated topology structure, the initial data generation model is trained to obtain a target data generation model; the target data generation model is used to generate node pseudo measurement data and branch pseudo measurement data in the distribution network.

2. The method according to claim 1, characterized in that The nodes include user nodes and device nodes; and the user change relationship in the distribution network topology structure is modified based on the historical measurement data of each node to obtain an updated topology structure, including: Determine the correlation quantization value between each user node and each device node based on the historical measurement data of each node, and obtain the user correlation sequence of each user node; Determining the target device node corresponding to each user node and the target user change relationship of each user node according to the user change correlation sequence of each user node; According to the target user change relationship of each user node, the user change relationship of the distribution network topology structure is modified to obtain the updated topology structure.

3. The method according to claim 2, characterized in that The historical measurement data includes voltage amplitude data; determining the correlation quantization value between each user node and each device node based on the historical measurement data of each node to obtain a user-related correlation sequence of each user node includes: Normalizing the voltage amplitude data of each node; For any user node, respectively calculating the correlation quantization value between the voltage amplitude normalization sequence of the user node and the voltage amplitude normalization sequence of each of the device nodes; The correlation quantization values ​​are aggregated to obtain a user-change correlation sequence of the user node.

4. The method according to claim 2, characterized in that The determining, based on the user change correlation sequence of each user node, the target device node corresponding to each user node and the target user change relationship of each user node includes: For any user node, determining a maximum user change correlation quantized value in a user change correlation sequence of the user node; Determine the device node corresponding to the maximum user-to-user correlation quantization value as the target device node of the user node; The association relationship between the user node and the target device node is determined as the target user change relationship of the user node.

5. The method according to claim 2, characterized in that The modifying of the user change relationship of the distribution network topology structure according to the target user change relationship of each user node to obtain the updated topology structure includes: Filtering out, from the target user change relationships of each user node, user change relationships to be processed that do not match the current user change relationship in the distribution network topology structure, and user nodes to be processed corresponding to each of the user change relationships to be processed; Based on the voltage amplitude data of each of the user nodes to be processed, dividing each of the user nodes to be processed into a first user node and a second user node; Different correction strategies are adopted to correct the first pending account change relationship corresponding to the first user node and the second pending account change relationship corresponding to the second user node respectively, to obtain the updated topology structure.

6. The method according to claim 5, characterized in that The adopting different correction strategies to respectively correct the first pending account change relationship corresponding to the first user node and the second pending account change relationship corresponding to the second user node to obtain the updated topology structure includes: For any first user node, replacing the current user change relationship of the first user node in the power distribution network topology with the target user change relationship of the first user node; For any second user node, a user correlation sequence between the second user node and other user nodes in the distribution network topology is calculated, and according to the user correlation sequence of the second user node, a current user change relationship of the second user node in the distribution network topology is corrected.

7. The method according to claim 6, characterized in that The modifying, according to the user correlation sequence of the second user node, the current user change relationship of the second user node in the distribution network topology structure includes: Obtaining a significant feature value of the user correlation sequence; In a case where the significant feature value is greater than a standard significant feature value, determining a maximum user relevance quantization value in the user relevance sequence; Determining the device node corresponding to the maximum user correlation quantization value as the target device node of the second user node; According to the user change relationship between the second user node and the target device node, the current user change relationship of the second user node in the power distribution network topology is updated.

8. The method according to any one of claims 1 to 7, characterized in that The training of the initial data generation model based on the node measurement data in the updated topology structure to obtain the target data generation model includes: constructing training data based on the node measurement data in the updated topology structure; Inputting the training data into the initial data generation model to obtain predicted data of the training data; Obtaining a power balance residual loss value and a voltage stability constraint loss value between the label data of the training data and the predicted data, and obtaining a composite loss function value of the initial data generation model; The initial data generation model is trained based on the composite loss function value to obtain the target data generation model.

9. A pseudo measurement data generating device, characterized in that: The device comprises: An acquisition module is used to obtain historical measurement data of each node in the distribution network and the distribution network topology; A correction module, configured to correct the change relationship in the distribution network topology structure according to the historical measurement data of each node, to obtain an updated topology structure; A training module is used to train an initial data generation model based on the node measurement data in the updated topology structure to obtain a target data generation model; the target data generation model is used to generate node pseudo-measurement data and branch pseudo-measurement data in the distribution network.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images