Power distribution network super-resolution measurement generation method and device based on topology decomposition training

Abstract

The application provides a power distribution network super-resolution measurement generation method and device based on topology decomposition training, comprising: adopting a branch cutting method to divide a power distribution network into a plurality of sub-zones; determining a coordination zone corresponding to each of the sub-zones; based on a multi-source sparse measurement matrix of the sub-zones, adopting a super-resolution measurement coordination consistent manner of the sub-zones and the coordination zones corresponding to each boundary node in the sub-zones to generate a super-resolution node state measurement matrix of the sub-zones; and splicing the super-resolution node state measurement matrices of the plurality of sub-zones to obtain a super-resolution node state measurement matrix of the power distribution network. The application divides the power distribution network into a plurality of sub-zones, assigns super-resolution measurement tasks to different sub-zones for completion, and then makes the super-resolution measurements of the sub-zones and the super-resolution measurements of the coincident nodes of the coordination zones consistent according to a decomposition coordination algorithm, thereby completing the super-resolution measurement generation of the large-scale power distribution network.

Description

Technical Field

[0001] This invention relates to the field of distribution network measurement technology, and in particular to a method and apparatus for generating super-resolution distribution network measurements based on topology decomposition training. Background Technology

[0002] The current distribution network measurement system includes various devices such as PMUs, SCADA systems, AMIs, and smart meters. These devices vary in measurement accuracy and frequency. Furthermore, high-precision, high-frequency measurement equipment is prohibitively expensive for large-scale deployment in the distribution network. This results in uneven spatial distribution, insufficient synchronization, and poor data quality in distribution network measurements, leading to high spatiotemporal sparsity. This makes it difficult to obtain comprehensive and accurate data analysis for situational awareness, causing unnecessary difficulties and risks to the management and operation of power facilities. To address this, researchers have proposed a distribution network super-resolution measurement generation method. This method learns the mapping relationship between sparse distribution network measurements and super-resolution measurements under each distribution network topology using a neural network to obtain a distribution network super-resolution measurement generation model. By processing sparse distribution network measurement data using this model, the corresponding super-resolution measurements can be obtained, improving the temporal and spatial resolution of distribution network measurements.

[0003] However, existing methods for generating super-resolution measurements for distribution networks employ a centralized measurement approach, which involves aggregating sparse measurement data from the entire distribution network into a central control center, which then generates the super-resolution measurements. This approach has the following drawbacks:

[0004] (1): As the scale of the distribution network continues to expand, the sparse measurement data of the distribution network will continue to increase, which will increase the communication burden of the distribution network and affect the communication stability of the distribution network.

[0005] (2): The sparse measurement data of the entire distribution network is collected into a central control center. The central control center centrally stores and processes this data, which will result in a heavy storage and computing burden on the central control center and affect the real-time performance of the super-resolution measurement of the distribution network.

[0006] (3): As the scale of the distribution network topology increases, problems such as insufficient hardware resources (mainly in computing resources), difficulty in model convergence, and increased model training time will occur in the learning of the super-resolution measurement generation model of the distribution network.

[0007] Therefore, there is an urgent need to provide a new method for generating super-resolution measurements for power distribution networks. Summary of the Invention

[0008] To address the aforementioned problems, this invention provides a method and apparatus for generating super-resolution measurements of distribution networks based on topology decomposition training. The distribution network is decomposed into several sub-regions, and the super-resolution measurement task is assigned to different sub-regions. A decomposition coordination algorithm ensures that the super-resolution measurements of overlapping nodes in the sub-regions are consistent with those in the coordination region, thereby completing the generation of super-resolution measurements for large-scale distribution networks.

[0009] In a first aspect, the present invention provides a method for generating super-resolution measurements of a distribution network based on topology decomposition training, the method comprising:

[0010] The distribution network is divided into multiple sub-zones using the branch cutting method;

[0011] Determine the coordination region corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination region, and the coordination region corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition;

[0012] Based on the multi-source sparse measurement matrix of the sub-partition, the super-resolution node state measurement matrix of the sub-partition is generated by adopting the super-resolution measurement coordination consistency method of the sub-partition and the coordination area corresponding to each boundary node in the sub-partition.

[0013] By splicing the super-resolution node state measurement matrices of multiple sub-regions, the super-resolution node state measurement matrix of the distribution network is obtained.

[0014] According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the expression of the multi-source sparse measurement matrix of the sub-partition is as follows:

[0015]

[0016] In the above formula, z N (t n ) represents the Nth node in the sub-partition at time t n Measurement data at time; φ represents no measurement, and N represents the total number of nodes in the sub-partition.

[0017] According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the super-resolution node state measurement matrix of the sub-partition is generated by using a method that coordinates the super-resolution measurements of the sub-partition and the coordination area corresponding to each boundary node in the sub-partition, based on the multi-source sparse measurement matrix of the sub-partition, including:

[0018] For each boundary node in the sub-partition, construct a super-resolution measurement model for the sub-partition, the target sub-partition, and the coordination region respectively;

[0019] Using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination region, an optimization model for the node state measurement data of the boundary node is constructed.

[0020] Using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination region, an optimization model for the node state measurement data of the boundary node is constructed.

[0021] Substitute the multi-source sparse measurement matrix of the sub-partition into the super-resolution measurement model of the sub-partition to obtain the initial super-resolution node state measurement matrix of the sub-partition.

[0022] Substitute the multi-source sparse measurement matrix of the target sub-partition into the super-resolution measurement model of the target sub-partition to obtain the initial super-resolution node state measurement matrix of the target sub-partition.

[0023] The super-resolution node state measurement data of the adjacent nodes of the boundary node in the initial super-resolution node state measurement matrix of the sub-partition and the super-resolution node state measurement data of the adjacent nodes of the target node in the initial super-resolution node state measurement matrix of the target sub-partition are input into the node state measurement data optimization model of the boundary node to obtain the optimized value of the super-resolution node state measurement data of the boundary node.

[0024] Replace the super-resolution node state measurement data of each boundary node in the initial super-resolution node state measurement matrix of the sub-partition with its optimized value to obtain the super-resolution node state measurement matrix of the sub-partition.

[0025] According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the node state measurement data includes: node voltage measurement values ​​and node phase angle measurement values.

[0026] According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the step of constructing an optimized model for the node state measurement data of the boundary node using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination area respectively includes:

[0027] Step 1: Construct the first dataset; wherein each sample in the first dataset consists of a multi-source sparse measurement matrix of the sub-partition under the historical period and a multi-source sparse measurement matrix of the target sub-partition;

[0028] Step 2: For each sample, generate a target super-resolution node state measurement matrix based on the sample, the super-resolution measurement model of the sub-partition, and the super-resolution measurement model of the target sub-partition; the super-resolution node state measurement matrix; wherein, the target super-resolution node state measurement matrix is ​​the super-resolution node state measurement matrix of the coordination region when the super-resolution node state measurement data of the boundary node and the target node are missing.

[0029] Step 3: Input the target super-resolution node state measurement matrix into the super-resolution measurement model of the coordination area, solve for the optimized values ​​of the super-resolution node state measurement data of the boundary node and the target node when the preset optimization objective function is satisfied, and use the optimized objective value at this time as the optimization loss of the sample;

[0030] Step 4: Optimize the parameters of the super-resolution node state measurement model of the coordination region using the optimization loss of each sample in the first sample set;

[0031] Step 5: Repeat steps 3 to 4 above until the super-resolution node state measurement model of the coordination area converges, and use the converged super-resolution node state measurement model of the coordination area as the optimization model for the node state measurement data of the boundary nodes.

[0032] According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the optimization objective function includes:

[0033] The optimization objective is to minimize the deviation between the optimized and sparse measurements of the super-resolution node state measurement data of the boundary node and the sum of the deviations between the optimized and sparse measurements of the super-resolution node state measurement data of the target node.

[0034] And the power flow equations of the coordination zone used to constrain the optimization objective. According to the distribution network super-resolution measurement generation method based on topology decomposition training provided by the present invention, the super-resolution measurement model of the sub-partition / target sub-partition / coordination zone is obtained by learning the mapping relationship between the multi-source sparse measurement matrix and the super-resolution node state measurement matrix of the sub-partition / target sub-partition / coordination zone on the topology of the sub-partition / target sub-partition / coordination zone using a graph attention learning mechanism.

[0035] In a second aspect, the present invention provides a distribution network super-resolution measurement generation device based on topology decomposition training, the device comprising:

[0036] The partitioning module is used to divide the power distribution network into multiple sub-zones using the branch cutting method;

[0037] The first determining module is used to determine the coordination area corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination area, and the coordination area corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition;

[0038] The sub-partition super-resolution measurement generation module is used to generate the super-resolution node state measurement matrix of the sub-partition based on the multi-source sparse measurement matrix of the sub-partition and by adopting a method of super-resolution measurement coordination consistency between the sub-partition and the coordination area corresponding to each boundary node in the sub-partition.

[0039] The distribution network super-resolution measurement generation module is used to stitch together the super-resolution node status measurement matrices of multiple sub-regions to obtain the distribution network super-resolution node status measurement matrix.

[0040] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the distribution network super-resolution measurement generation method based on topology decomposition training as described in the first aspect.

[0041] Fourthly, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the distribution network super-resolution measurement generation method based on topology decomposition training as described in the first aspect.

[0042] This invention provides a method and apparatus for generating super-resolution measurements of a distribution network based on topology decomposition training, comprising: dividing the distribution network into multiple sub-regions using a branch cutting method; determining a coordination zone corresponding to each sub-region; wherein, a boundary node in each sub-region corresponds to a coordination zone, and the coordination zone corresponding to each boundary node in the sub-region is composed of the boundary node, the adjacent nodes of the boundary node in the sub-region, a target node, and the adjacent nodes of the target node in the target sub-region; the boundary node and the target node are the two endpoints of the cut branches between the sub-region and the target sub-region; generating a super-resolution node state measurement matrix of the sub-region based on the multi-source sparse measurement matrix of the sub-region, using a method of super-resolution measurement coordination consistency between the sub-region and the coordination zone corresponding to each boundary node in the sub-region; and concatenating the super-resolution node state measurement matrices of multiple sub-regions to obtain the super-resolution node state measurement matrix of the distribution network. This invention decomposes the distribution network into several sub-regions, assigning super-resolution measurement tasks to different sub-regions. A decomposition and coordination algorithm then ensures that the super-resolution measurements of nodes overlapping with the coordination region in each sub-region are consistent, thereby generating super-resolution measurements for large-scale distribution networks. This avoids the shortcomings of centralized measurement in distribution networks and improves the accuracy of super-resolution measurement results for large-scale distribution networks. Furthermore, this invention offers greater flexibility in handling changes to the distribution network topology nodes; only the super-resolution measurement model of the sub-region containing the changed node and its related models need to be adjusted individually, reducing the impact of frequent topology changes on distribution network super-resolution measurements. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the method for generating super-resolution measurements of distribution networks based on topology decomposition training provided by the present invention.

[0045] Figure 2 This is a schematic diagram of the topology of the two-zone power distribution network provided by the present invention;

[0046] Figure 3 This is the topology diagram of the IEEE 33-node distribution network system after partitioning provided by this invention;

[0047] Figure 4 This is the topology diagram of the IEEE 123 node distribution network system after partitioning provided by this invention;

[0048] Figure 5 This is a schematic diagram illustrating the accuracy of node voltages obtained by the decomposition and coordination super-resolution algorithm in the IEEE 33-node system provided by this invention.

[0049] Figure 6 This is a schematic diagram of the node voltage phase angle accuracy obtained by the decomposition coordination super-resolution algorithm in the IEEE 33-node system provided by the present invention.

[0050] Figure 7 This is a schematic diagram illustrating the accuracy of node voltages obtained by the decomposition and coordination super-resolution algorithm in the IEEE 123-node system provided by this invention.

[0051] Figure 8 This is a schematic diagram illustrating the accuracy of the node voltage phase angle obtained by the decomposition coordination super-resolution algorithm in the IEEE 123-node system provided by this invention.

[0052] Figure 9 This is the topology diagram of the IEEE 123-node system after adding a new node, provided by this invention.

[0053] Figure 10 This is a schematic diagram of the structure of the distribution network super-resolution measurement generation device based on topology decomposition training provided by the present invention;

[0054] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention;

[0055] Figure label:

[0056] 1110: Processor; 1120: Communication interface; 1130: Memory; 1140: Communication bus. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] The following is combined with Figures 1-11 The present invention describes a method and apparatus for generating super-resolution measurements of distribution networks based on topology decomposition training.

[0059] Existing distribution network super-resolution measurement generation technology adopts a centralized measurement method, and the corresponding distribution network super-resolution measurement generation model can be expressed by the following formula:

[0060] SR S(V,θ)=f(G(n N ,E),LR(z))

[0061] In the above formula, G(n) N E) represents node n of the complete distribution network topology S. N The connection relationship between edge E; LR(z) is the multi-source sparse measurement matrix of the distribution network; SR S (V,θ) represents the distribution network super-resolution node state measurement matrix, where V is the node voltage, θ is the voltage phase angle of the node voltage, and f is the distribution network super-resolution measurement generation model, which is defined in G(n N Learning SR on the topology of E) S The mapping relationship between (V,θ) and LR(z).

[0062] Because the sparse measurement data of the entire distribution network needs to be collected in a central control center, and the distribution network super-resolution measurement is generated by the control center, it has the following drawbacks:

[0063] From a network perspective, if high-frequency sensors and additional measurement equipment are used to improve measurement resolution, or if the distribution network scales up, the amount of measurement data in the distribution network will increase accordingly. This may overburden network communication and affect communication stability. Furthermore, if sparse measurement data from the entire distribution network is aggregated into a central control center, and this centralized storage and processing of data results in a heavy storage and computational burden on the central control center, impacting the real-time performance of the distribution network's super-resolution measurements.

[0064] From the perspective of the super-resolution measurement generation model for distribution networks, the super-resolution measurement generation model for distribution networks is designed based on the graph attention mechanism. It solves for a set of neural network parameters WS that meet the numerical accuracy requirements through gradient descent of the loss function, as shown in the following equation:

[0065]

[0066] X S z is the ground truth of the super-resolution node state measurement matrix of the distribution network for the training samples. S For the sparse measurement matrix of the distribution network used in the training samples, N S This represents the number of training samples.

[0067] Based on the structural design of the distribution network super-resolution measurement generation model, with G(n NAs the graph size increases, more nodes and edges need to be processed, and the number of graph attention parameters (WS) and network layers will also increase significantly. These parameters and network layers need to be calculated during training and inference. Increasing the graph size and network complexity will lead to an increase in hardware resource requirements. For example, (1) As the graph size increases, more memory is needed to store model weights and computation graphs, and more GPU memory is needed to store model parameters and intermediate computation results. (2) As the number of network layers increases, the input and output of each layer need to be calculated by convolution, normalization and activation operations, which will also increase the computational burden of the GPU and consume more memory. (3) More data needs to be collected and stored during training, which will also increase the space requirements of the hard disk. The increase in hardware resource requirements may lead to insufficient hardware resources, resulting in training delays or failures. In addition, when the topology becomes larger, the training of graph attention networks may not converge for many reasons. For example, (1) As the graph size becomes more complex, the computational complexity of calculating attention weights will also increase. When there are a large number of nodes and edges in the graph, the computation time and memory consumption will increase sharply, causing the training process to be slow or stagnant. (2) Backpropagation of gradients between nodes may encounter vanishing or exploding gradient problems, which can slow down the convergence speed of the model or stop training altogether. (3) The number of parameters in the model will also increase accordingly. If there is not enough training data to support these parameters, the model will have difficulty converging. In addition, when the graph is large, the coverage of training data will become lower, which will cause the model to encounter more unseen data during the learning process, resulting in training difficulties and slow convergence.

[0068] The aforementioned problems are all caused by the excessive size of the distribution network topology. Therefore, this invention proposes a method for generating super-resolution measurements of distribution networks based on topology decomposition training. The distribution network is decomposed into several sub-regions, and the super-resolution measurement task is assigned to different sub-regions. Then, a decomposition coordination algorithm is used to ensure that the super-resolution measurements of overlapping nodes in the sub-regions are consistent with those in the coordination region, thereby completing the super-resolution measurement of large-scale distribution networks. Figure 1 As shown, the method includes:

[0069] S11. The distribution network is divided into multiple sub-zones using the branch cutting method;

[0070] A power distribution network is a vast power system, typically composed of several regional power grids, each responsible for managing the power supply and distribution within its designated area. Distribution networks are characterized by hierarchical management and layered control, meaning that each regional power grid within the network has its own management and control hierarchy, thus ensuring the stability and reliability of the entire distribution network. Furthermore, distribution networks feature distributed processing, allowing each regional power grid to achieve more efficient operation and management through distributed computing and processing. Therefore, distribution networks possess high scalability and flexibility.

[0071] Regional power grids are interconnected through a small number of transmission lines, meaning that the coupling between these regional power grids is relatively weak. Therefore, by using the branch cutting method to cut off these few transmission lines between regional power grids, the distribution network can be divided into multiple sub-regions.

[0072] S12. Determine the coordination region corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination region, and the coordination region corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition;

[0073] S13. Based on the multi-source sparse measurement matrix of the sub-partition, the super-resolution node state measurement matrix of the sub-partition is generated by adopting the super-resolution measurement coordination consistency method of the sub-partition and the coordination area corresponding to each boundary node in the sub-partition.

[0074] It is understandable that the expression for the multi-source sparse measurement matrix z of the sub-partition is as follows:

[0075]

[0076] In the above formula, z N (t n ) represents the Nth node in the sub-partition at time t n Measurement data at time; φ represents no measurement, and N represents the total number of nodes in the sub-partition.

[0077] The multi-source sparse measurement matrix is ​​obtained using multi-source measurement devices deployed in the sub-partition. These devices include PMUs (Phasor Measurement Units), RTUs (Remote Terminal Units), smart meters, SCADA systems, etc. Measurement data includes, but is not limited to, measured voltage (amplitude, phase angle), current (amplitude, phase angle), active power, reactive power, power factor, and frequency. The node state measurement data includes node voltage measurements and node phase angle measurements. The sub-partition's super-resolution measurement task essentially uses low-resolution node measurement data to predict high-resolution node state data.

[0078] S14. By splicing the super-resolution node state measurement matrices of multiple sub-regions, the super-resolution node state measurement matrix of the distribution network is obtained.

[0079] Figure 2 A topology diagram of a two-zone distribution network is provided. Taking a two-zone distribution network as an example, the distribution network S is divided into two sub-zones S' using the branch cutting method. i and S j S i and S j The two endpoints of the cut branch are S i and S j Boundary node, S i Boundary node A, and adjacent nodes of A (i.e., S) i (Nodes directly connected to A in the middle) S j The boundary node B and its adjacent nodes (i.e., S) j (Nodes directly connected to B in the middle) jointly generate S i and S j Coordination zone S between B Coordination Zone S B Known as S i The corresponding coordination area is also known as S. j The corresponding coordination area.

[0080] Global super-resolution measurement of distribution network is equivalent to sub-region S i and sub-partition S j Super-resolution collaborative measurement, also known as distribution network super-resolution measurement model SR S (V,θ), sub-partition S i Super-resolution measurement model Subpartition S j Super-resolution measurement model and Coordination Area S B Super-resolution node state measurement model satisfy:

[0081]

[0082] in, This indicates that, according to the decomposition and coordination algorithm, the super-resolution node state measurement matrices of sub-partitions Si and Sj are consistent with the super-resolution node state measurement matrices of the coordination region SB, resulting in the same super-resolution generation of nodes.

[0083] Here, sub-partition S i S j and Coordination Area S B The super-resolution node state measurement model is obtained by learning the mapping relationship between the multi-source sparse measurement matrix and the super-resolution node state measurement matrix based on the topology of the corresponding region. It is consistent with the learning method of the distribution network super-resolution node state measurement model, and will not be elaborated here.

[0084] Similarly, a real power distribution network is divided into M sub-regions. Each boundary node in a sub-region corresponds to a coordination area. Each sub-region generates its own super-resolution node state measurement matrix through decomposition and coordination. The concatenation of the super-resolution node state measurement matrices of M sub-regions is the super-resolution node state measurement matrix of the distribution network.

[0085] This invention provides a method for generating super-resolution measurements of distribution networks based on topology decomposition training. The distribution network is decomposed into several sub-regions, and the super-resolution measurement task is assigned to different sub-regions. Then, a decomposition coordination algorithm is used to ensure that the super-resolution measurements of nodes overlapping with the coordination area in each sub-region are consistent, thereby completing the generation of super-resolution measurements for large-scale distribution networks. This avoids the shortcomings of centralized measurement in distribution networks and improves the accuracy of super-resolution measurement generation results for large-scale distribution networks. Furthermore, this invention offers greater flexibility in dealing with changes in distribution network topology nodes. Only the super-resolution measurement model of the sub-region containing the changed node and the corresponding related models of that sub-region need to be adjusted individually, reducing the impact of frequent topology changes on distribution network super-resolution measurements.

[0086] Specifically, the S13 process is as follows:

[0087] First, for each boundary node in the sub-partition, construct the super-resolution measurement model for the sub-partition, the target sub-partition, and the coordination region respectively;

[0088] Secondly, using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination area, an optimization model for the node state measurement data of the boundary node is constructed.

[0089] This step specifically includes:

[0090] Step 1: Construct the first dataset; wherein each sample in the first dataset consists of a multi-source sparse measurement matrix of the sub-partition under the historical period and a multi-source sparse measurement matrix of the target sub-partition;

[0091] Step 2: For each sample, generate a target super-resolution node state measurement matrix based on the sample, the super-resolution measurement model of the sub-partition, and the super-resolution measurement model of the target sub-partition; wherein, the target super-resolution node state measurement matrix is ​​the super-resolution node state measurement matrix of the coordination region when the super-resolution node state measurement data of the boundary node and the target node are missing.

[0092] Step 3: Input the target super-resolution node state measurement matrix into the super-resolution measurement model of the coordination area, solve for the optimized values ​​of the super-resolution node state measurement data of the boundary node and the target node when the preset optimization objective function is satisfied, and use the optimized objective value at this time as the optimization loss of the sample;

[0093] Step 4: Optimize the parameters of the super-resolution node state measurement model of the coordination region using the optimization loss of each sample in the training sample set;

[0094] Step 5: Repeat steps 3 to 4 above until the super-resolution node state measurement model of the coordination area converges, and use the converged super-resolution node state measurement model of the coordination area as the optimization model for the node state measurement data of the boundary nodes.

[0095] Here, the optimization objective function includes:

[0096] The optimization objective is to minimize the deviation between the optimized and sparse measurements of the super-resolution node state measurement data of the boundary node and the sum of the deviations between the optimized and sparse measurements of the super-resolution node state measurement data of the target node.

[0097] And the power flow equations of the coordination region used to constrain the optimization objective. Figure 2 Taking the topology diagram of the two-zone distribution network shown as an example, let's look at sub-zone S. i S j and Coordination Area S B First, train the super-resolution node state measurement model to obtain the sub-partition neural network parameters W. Si W Sj and coordination region neural network parameters W SB At this point, each sub-partition and coordination region can generate its own topology's super-resolution measurement results (SR). Si (V,θ), SR Sj(V,θ) and SR B (V,θ):

[0098]

[0099] Next, the parameters of the fixed sub-partition neural network are... and The super-resolution node state measurement model of coordination region B is optimized using the first sample set. The optimization approach is as follows:

[0100] For each sample, SR is generated using sub-partition LR;

[0101] Next, the sub-partitions SR are concatenated to obtain the coordination region SR, and the measurements of nodes A and B in the coordination region SR are set to 0 to obtain the coordination region SR*.

[0102] Then, the coordination region SR* is input into the coordination region super-resolution node state measurement model to solve for the measurements at nodes A and B when the preset optimization objective function is satisfied. Here, the coordination region optimization objective function consists of the optimization objective and the power flow constraints of the coordination region; the optimization objective is min J(x) B )=ω i (x Bi -x′ Bi ) 2 +ω j (x Bj -x′ Bj ) 2 , where ω i ω represents the weight coefficient of the boundary node A of the sub-partition Si. j x represents the weight coefficient of the boundary node A of sub-partition Sj. Bi and x′ Bi The measurement and true value at node A are respectively, x Bj and x′ Bj These are the measurement and the true value at node B, respectively.

[0103] Then, J(x) B The training loss of all samples is used to optimize the super-resolution node state measurement model of the coordination region. This concludes one iteration of training. It can achieve power flow balancing using nodes A and B in coordination area SB, sub-partition Si, and sub-partition Sj;

[0104] Repeat the iteration until convergence, that is, when J(x) B When )≤ε (ε is the convergence numerical precision), the results of each sub-partition converge and the results of the coordination area converge, and the distribution network super-resolution task SR S Subtasks SR of (V,θ) Si (V,θ), SRSj (V,θ) and SR B (V,θ) both meet the convergence accuracy requirements.

[0105] Because the super-resolution node state measurement model of the coordination area after convergence in this invention can ensure that the difference between the electrical state quantities generated by the measurement of the coordination area and the sub-area is small enough, and can satisfy the power flow balance of the whole network, the node state measurement data of node A / node B generated by it is more accurate, and can be regarded as the node state measurement data optimization model of node A / node B.

[0106] By analogy, an optimization model for node state measurement data of any boundary node in any sub-region of the distribution network can be generated.

[0107] Step 2: Substitute the multi-source sparse measurement matrix of the sub-partition into the super-resolution measurement model of the sub-partition to obtain the initial super-resolution node state measurement matrix of the sub-partition.

[0108] Step 3: Substitute the multi-source sparse measurement matrix of the target sub-partition into the super-resolution measurement model of the target sub-partition to obtain the initial super-resolution node state measurement matrix of the target sub-partition.

[0109] Step 4: Input the super-resolution node state measurement data of the adjacent nodes of the boundary node in the initial super-resolution node state measurement matrix of the sub-partition and the super-resolution node state measurement data of the adjacent nodes of the target node in the initial super-resolution node state measurement matrix of the target sub-partition into the node state measurement data optimization model of the boundary node to obtain the optimized value of the super-resolution node state measurement data of the boundary node.

[0110] Step 5: Replace the super-resolution node state measurement data of each boundary node in the initial super-resolution node state measurement matrix of the sub-partition with its optimized value to obtain the super-resolution node state measurement matrix of the sub-partition.

[0111] It is understandable that this invention achieves power flow balance between sub-regions through a decomposition and coordination algorithm, thus completing the super-resolution node state measurement of the entire distribution network.

[0112] Example verification:

[0113] The proposed method was tested using both an IEEE 33-node distribution network system and an IEEE 123-node distribution network system. The test system parameters are shown in Table 1.

[0114] Table 1

[0115]

[0116] Figure 3 The topology diagram of the IEEE 33-node distribution network system after partitioning is provided, which includes 4 partitions and 3 coordination sides. Nodes 1-18 belong to partition 1, nodes 19-22 to partition 2, nodes 23-25 ​​to partition 3, and nodes 26-33 to partition 4. Partition 1 and partition 2 are connected through nodes 2 and 19, and the coordination area topology includes nodes 1, 2, 3, 19, and 20; partition 1 and partition 3 are connected through nodes 3 and 23, and the coordination area topology includes nodes 2, 3, 4, 23, and 24; partition 1 and partition 4 are connected through nodes 6 and 26, and the coordination area topology includes nodes 5, 6, 7, 26, and 27.

[0117] Experiments showed that decomposing the large topology into several smaller topologies simplified the connections between nodes and edges, thus reducing the training difficulty of each model. The convergence speed of each partitioned model was faster than that of the full topology model. However, since the IEEE 33-node system is still relatively small, the super-resolution model can still converge within 500 rounds when dealing with the full topology of the IEEE 33-node system.

[0118] Figure 4 The topology diagram of the IEEE 123 node distribution network system after partitioning is provided. Nodes 1-20 and 57-71 are in partition 1, and the coordination side topology includes nodes 10, 16, 17, 21, 22, 24, 38, and 57; nodes 21-56 are in partition 2; nodes 72-108 are in partition 3, and the coordination area topology includes nodes 63, 66, 67, 72, and 73; nodes 109-123 are in partition 4, and the coordination area topology includes nodes 103, 108, 109, 110, and 113.

[0119] Experiments showed that as the topology size increased, the IEEE 123 full-topology loss function decreased much more slowly than that of IEEE 33, and could not reach the previous data accuracy after 1000 rounds. However, for partitioned topologies 1-4, since the partitioned topology size remained similar to that of the IEEE 33 node system, convergence was still possible within 500 rounds.

[0120] Therefore, it can be seen that when the decomposition and coordination super-resolution algorithm (i.e. the method of this invention) is applied to large-scale distribution network topology, the large-scale distribution network topology is decomposed and sub-partition models are trained separately, and the sub-partition models are coordinated to generate super-resolution results. This has a faster convergence speed and a more obvious effect in systems with larger topology scale.

[0121] Figure 5 A schematic diagram illustrating the accuracy of node voltages obtained through the decomposition-coordinated super-resolution algorithm in the IEEE 33-bus system is provided. Figure 6A schematic diagram illustrating the accuracy of node voltage phase angles obtained using the decomposition-coordinated super-resolution algorithm for the IEEE 33-node system is provided. Both accuracy curves are above 99.9%, similar to the accuracy of the super-resolution model generated across the entire network. This demonstrates that for the IEEE 33-node system, the decomposition-coordinated super-resolution algorithm can achieve faster model convergence while still ensuring the generation of high-precision data.

[0122] Figure 7 A schematic diagram illustrating the accuracy of node voltages obtained using the decomposition-coordinated super-resolution algorithm in the IEEE 123-bus system is presented. Figure 8 A schematic diagram illustrating the accuracy of node voltage phase angles obtained using the decomposition-coordinated super-resolution algorithm for the IEEE 123-node system is presented. Both accuracy curves maintain an accuracy above 99.6%, further improving the accuracy of the generated data compared to the 98% accuracy of the overall network super-resolution results. This demonstrates that for the IEEE 123-node system, using the decomposition-coordinated super-resolution algorithm not only accelerates model convergence but also improves the accuracy of data generation.

[0123] Table 2 compares the computational resource consumption, training time, and accuracy of full-network training and decomposition-coordinated super-resolution in the IEEE 123-node system. As can be seen from the table, compared to the full-network training method that trains the entire topology together, the decomposition-coordinated super-resolution algorithm consumes fewer computational resources, has a shorter training time, and achieves higher accuracy.

[0124] Table 2

[0125]

[0126] Meanwhile, considering that power distribution networks often connect new electrical nodes to the existing topology during upgrades and renovations, thus altering local topological connections, this change differs from power distribution network reconfiguration in that the adjacency matrix also changes in its dimension. Since the model's training input parameters include the topological adjacency matrix, the model learns the electrical relationships between nodes in the topology. When the dimension of the topological adjacency matrix changes, the original model attention parameter matrix also needs to be modified to handle the new topological connections, requiring retraining and updating the model's neural network parameters. However, newly added load nodes often only affect locally adjacent nodes, so only the parameters of the corresponding partition's model neural network need to be updated, without requiring a full topology retraining.

[0127] like Figure 9The example illustrates the topology of an IEEE 123-node system after the addition of a new node. With the addition of node 124 after node 123, the original adjoint matrix changes from a 123x123 symmetric matrix to a 124x124 symmetric matrix. At this point, the original model input is no longer applicable. If a full-network training approach is used, the super-resolution model needs to be retrained to incorporate the data from the new node. Retraining the entire model is not only time-consuming and labor-intensive but also disrupts online power grid applications. The impact of the new node only affects its neighboring nodes. Using a decomposition-coordinated super-resolution algorithm, node 124 can be assigned to partition 4. The model for partition 4 can then be retrained, while models for other partitions do not require adjustment. Once partition 4 is trained, it can be incorporated into the overall model, significantly reducing the training cost associated with the new node.

[0128] Table 3 compares the training time required for super-resolution generation of the topology after adding new nodes in the IEEE 123-node system, as well as the accuracy of node voltage and voltage phase angle. Table 3 shows that the decomposition-coordinated super-resolution algorithm can significantly reduce the training time required after topology changes, enabling the model to be deployed online as quickly as possible. At the same time, it ensures that the changed topology can still achieve high-precision super-resolution data generation, guaranteeing reliable situational awareness of the distribution network.

[0129] Table 3

[0130]

[0131] As can be seen, this invention addresses the difficulties in training convergence, long training times, and high hardware resource consumption caused by the increased topology size in large-scale distribution network systems. It studies a coordinated super-resolution algorithm based on distribution network topology decomposition, decomposing the large-scale topology into multiple sub-regional systems for separate model training. The super-resolution results of the entire network are then generated by coordinating the super-resolution results of the sub-regions. The models used in the decomposition and coordination algorithm are trained on local sub-topologies, significantly reducing the topology size of a single training topology and avoiding the training difficulties and high hardware requirements caused by large-scale topologies. Furthermore, since the large-scale topology is split into multiple sub-regions for separate training, when nodes are added or deleted, only the parameters of the sub-regional topology super-resolution model need to be adjusted. After the model is retrained, it can be integrated into the original coordinated decomposition model.

[0132] Secondly, the distribution network super-resolution measurement generation device based on topology decomposition training provided by the present invention will be described. The distribution network super-resolution measurement generation device based on topology decomposition training described below can be referred to in correspondence with the distribution network super-resolution measurement generation method based on topology decomposition training described above. Figure 10 A schematic diagram of a distribution network super-resolution measurement generation device based on topology decomposition training is shown in the example, such as... Figure 10 As shown, the device includes:

[0133] The partitioning module 21 is used to divide the power distribution network into multiple sub-zones using the branch cutting method;

[0134] The first determining module 22 is used to determine the coordination area corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination area, and the coordination area corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition;

[0135] Sub-partition super-resolution measurement generation module 23, used for

[0136] Based on the multi-source sparse measurement matrix of the sub-partition, the super-resolution node state measurement matrix of the sub-partition is generated by adopting the super-resolution measurement coordination consistency method of the sub-partition and the coordination area corresponding to each boundary node in the sub-partition.

[0137] The distribution network super-resolution measurement generation module 24 is used to splice the super-resolution node status measurement matrices of multiple sub-regions to obtain the distribution network super-resolution node status measurement matrix.

[0138] This invention provides a distribution network super-resolution measurement generation device based on topology decomposition training. The distribution network is decomposed into several sub-regions, and super-resolution measurement tasks are assigned to different sub-regions. A decomposition coordination algorithm is then used to ensure that the super-resolution measurements of nodes overlapping with the coordination region in each sub-region are consistent, thereby generating super-resolution measurements for large-scale distribution networks. This avoids the shortcomings of centralized measurement in distribution networks and improves the accuracy of super-resolution measurement generation results for large-scale distribution networks. Furthermore, this invention offers greater flexibility in handling changes in distribution network topology nodes. Only the super-resolution measurement model of the sub-region containing the changed node and its related models need to be adjusted individually, reducing the impact of frequent topology changes on distribution network super-resolution measurements.

[0139] Thirdly, Figure 11 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 11As shown, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communication bus 1140, wherein the processor 1110, the communications interface 1120, and the memory 1130 communicate with each other through the communication bus 1140. Processor 1110 can call logic instructions in memory 1130 to execute a distribution network super-resolution measurement generation method based on topology decomposition training. This method includes: dividing the distribution network into multiple sub-regions using a branch cutting method; determining a coordination zone corresponding to each sub-region; wherein a boundary node in each sub-region corresponds to a coordination zone, and the coordination zone corresponding to each boundary node in the sub-region is composed of the boundary node, the adjacent nodes of the boundary node in the sub-region, a target node, and the adjacent nodes of the target node in the target sub-region; the boundary node and the target node are the two endpoints of the cut branches between the sub-region and the target sub-region; generating a super-resolution node state measurement matrix for the sub-region based on the multi-source sparse measurement matrix of the sub-region, using a method that ensures the super-resolution measurement coordination between the sub-region and the coordination zone corresponding to each boundary node in the sub-region; and concatenating the super-resolution node state measurement matrices of multiple sub-regions to obtain the super-resolution node state measurement matrix of the distribution network.

[0140] Furthermore, the logical instructions in the aforementioned memory 1130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] Fourthly, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the above-described methods to perform a distribution network super-resolution measurement generation method based on topology decomposition training. The method includes: dividing the distribution network into multiple sub-regions using a branch cutting method; determining a coordination area corresponding to each sub-region; wherein, a boundary node in the sub-region corresponds to a coordination area, and the coordination area corresponding to each boundary node in the sub-region is composed of the boundary node, the adjacent nodes of the boundary node in the sub-region, a target node, and the adjacent nodes of the target node in the target sub-region; the boundary node and the target node are the two endpoints of the cut branches between the sub-region and the target sub-region; generating a super-resolution node state measurement matrix of the sub-region based on the multi-source sparse measurement matrix of the sub-region and using a method of super-resolution measurement coordination consistency between the sub-region and the coordination area corresponding to each boundary node in the sub-region; and concatenating the super-resolution node state measurement matrices of multiple sub-regions to obtain a super-resolution node state measurement matrix of the distribution network. Fifthly, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the above-described methods to execute a method for generating super-resolution measurements of a distribution network based on topology decomposition training. This method includes: dividing the distribution network into multiple sub-regions using a branch cutting method; determining a coordination zone corresponding to each sub-region; wherein a boundary node in one sub-region corresponds to a coordination zone, and the coordination zone corresponding to each boundary node in the sub-region is composed of the boundary node, adjacent nodes of the boundary node in the sub-region, a target node, and adjacent nodes of the target node in the target sub-region; the boundary node and the target node are the two endpoints of the cut branches between the sub-region and the target sub-region; generating a super-resolution node state measurement matrix of the sub-region based on the multi-source sparse measurement matrix of the sub-region, using a method that ensures the super-resolution measurements of the sub-region and the coordination zone corresponding to each boundary node in the sub-region are coordinated; and concatenating the super-resolution node state measurement matrices of multiple sub-regions to obtain a super-resolution node state measurement matrix of the distribution network.

[0142] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating super-resolution measurements for distribution networks based on topology decomposition training, characterized in that, The method includes: The distribution network is divided into multiple sub-zones using the branch cutting method; Determine the coordination region corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination region, and the coordination region corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition; Based on the multi-source sparse measurement matrix of the sub-partition, the super-resolution node state measurement matrix of the sub-partition is generated by adopting the super-resolution measurement coordination consistency method of the sub-partition and the coordination area corresponding to each boundary node in the sub-partition. By splicing the super-resolution node state measurement matrices of multiple sub-regions, the super-resolution node state measurement matrix of the distribution network is obtained.

2. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 1, characterized in that, The expression for the multi-source sparse measurement matrix of the sub-partition is as follows: ； In the above formula, For the sub-partition Each node Measurement data at any given time; For no measurement, N is the total number of nodes in the sub-partition.

3. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 1 or 2, characterized in that, The multi-source sparse measurement matrix based on the sub-partition, using a method that ensures consistency between the super-resolution measurements of the sub-partition and the coordination region corresponding to each boundary node in the sub-partition, generates the super-resolution node state measurement matrix of the sub-partition, including: For each boundary node in the sub-partition, construct a super-resolution measurement model for the sub-partition, the target sub-partition, and the coordination region respectively; Using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination region, an optimization model for the node state measurement data of the boundary node is constructed. Using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination region, an optimization model for the node state measurement data of the boundary node is constructed. Substitute the multi-source sparse measurement matrix of the sub-partition into the super-resolution measurement model of the sub-partition to obtain the initial super-resolution node state measurement matrix of the sub-partition. Substitute the multi-source sparse measurement matrix of the target sub-partition into the super-resolution measurement model of the target sub-partition to obtain the initial super-resolution node state measurement matrix of the target sub-partition. The super-resolution node state measurement data of the adjacent nodes of the boundary node in the initial super-resolution node state measurement matrix of the sub-partition and the super-resolution node state measurement data of the adjacent nodes of the target node in the initial super-resolution node state measurement matrix of the target sub-partition are input into the node state measurement data optimization model of the boundary node to obtain the optimized value of the super-resolution node state measurement data of the boundary node. Replace the super-resolution node state measurement data of each boundary node in the initial super-resolution node state measurement matrix of the sub-partition with its optimized value to obtain the super-resolution node state measurement matrix of the sub-partition.

4. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 3, characterized in that, The node status measurement data includes: node voltage measurement value and node phase angle measurement value.

5. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 3, characterized in that, The step of constructing an optimized model for the node state measurement data of the boundary nodes using the super-resolution measurement models of the sub-partition, the target sub-partition, and the coordination region includes: Step 1: Construct the first dataset; wherein each sample in the first dataset consists of a multi-source sparse measurement matrix of the sub-partition under the historical period and a multi-source sparse measurement matrix of the target sub-partition; Step 2: For each sample, generate a target super-resolution node state measurement matrix based on the sample, the super-resolution measurement model of the sub-partition, and the super-resolution measurement model of the target sub-partition; the super-resolution node state measurement matrix; wherein, the target super-resolution node state measurement matrix is ​​the super-resolution node state measurement matrix of the coordination region when the super-resolution node state measurement data of the boundary node and the target node are missing. Step 3: Input the target super-resolution node state measurement matrix into the super-resolution measurement model of the coordination area, solve for the optimized values ​​of the super-resolution node state measurement data of the boundary node and the target node when the preset optimization objective function is satisfied, and use the optimized objective value at this time as the optimization loss of the sample; Step 4: Optimize the parameters of the super-resolution node state measurement model of the coordination region using the optimization loss of each sample in the training sample set; Step 5: Repeat steps 3 to 4 above until the super-resolution node state measurement model of the coordination region converges, and use the converged super-resolution node state measurement model of the coordination region as the node state measurement data optimization model of the boundary node.

6. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 5, characterized in that, The optimization objective function includes: The optimization objective is to minimize the deviation between the optimized and sparse measurements of the super-resolution node state measurement data of the boundary node and the sum of the deviations between the optimized and sparse measurements of the super-resolution node state measurement data of the target node. And the power flow equations of the coordination zone used to constrain the optimization objective.

7. The method for generating super-resolution measurements of distribution networks based on topology decomposition training according to claim 3, characterized in that, The super-resolution measurement model of the sub-partition / target sub-partition / coordination region is obtained by learning the mapping relationship between the multi-source sparse measurement matrix and the super-resolution node state measurement matrix of the sub-partition / target sub-partition / coordination region on the topology of the sub-partition / target sub-partition / coordination region using a graph attention learning mechanism.

8. A distribution network super-resolution measurement generation device based on topology decomposition training, characterized in that, The device includes: The partitioning module is used to divide the power distribution network into multiple sub-zones using the branch cutting method; The first determining module is used to determine the coordination area corresponding to each of the sub-partitions; wherein, a boundary node in the sub-partition corresponds to a coordination area, and the coordination area corresponding to each boundary node in the sub-partition is composed of the boundary node, the adjacent nodes of the boundary node in the sub-partition, the target node, and the adjacent nodes of the target node in the target sub-partition; the boundary node and the target node are the two endpoints of the cut branch between the sub-partition and the target sub-partition; The sub-partition super-resolution measurement generation module is used to generate the super-resolution node state measurement matrix of the sub-partition based on the multi-source sparse measurement matrix of the sub-partition and by adopting a method of super-resolution measurement coordination between the sub-partition and the coordination area corresponding to each boundary node in the sub-partition. The distribution network super-resolution measurement generation module is used to stitch together the super-resolution node status measurement matrices of multiple sub-regions to obtain the distribution network super-resolution node status measurement matrix.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the distribution network super-resolution measurement generation method based on topology decomposition training as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the distribution network super-resolution measurement generation method based on topology decomposition training as described in any one of claims 1 to 7.

Images