A method and related apparatus for optimizing the configuration of distribution network measurements considering state estimation

Abstract

This invention belongs to the field of distribution network technology and discloses a method and related apparatus for optimizing the configuration of distribution network measurements considering state estimation. The method includes the following steps: selecting the beginning of the distribution network to be optimized as a measurement point; based on existing measurement points, setting each remaining non-measurement point node of the distribution network as a candidate measurement point, and obtaining the distribution network evaluation error corresponding to each candidate measurement point; setting the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point, and determining whether a preset termination condition has been met; if so, outputting the optimized configuration result of the distribution network measurement. The technical solution provided by this invention can improve the accuracy of the evaluation results, reduce the number of measurement devices, and demonstrates superior performance in terms of speed, convergence, and accuracy.

Description

Technical Field

[0001] This invention belongs to the field of power distribution network technology, and specifically relates to a method and related apparatus for optimizing the configuration of power distribution network measurements considering state estimation. Background Technology

[0002] In recent years, with the continuous expansion of urban scale, the power grid has become larger and larger, and the power grid structure has become more and more complex. The amount of nonlinear equation data for power flow calculation has also become larger and larger. Among them, due to various factors such as investment cost and actual conditions, the distribution network cannot build a full-coverage measurement system according to the main grid standard. Therefore, it is necessary to study the optimal configuration scheme of distribution network measurement.

[0003] Optimized configuration of distribution network measurements is a crucial step in ensuring the efficient and reliable operation of the distribution network. This optimization process requires consideration of multiple factors to achieve the optimal configuration in terms of economy while maintaining reliability. Existing optimization schemes mostly aim to improve the observability of the distribution network, identifying key nodes based on standards or experience and deploying measurement devices at these nodes. However, these existing schemes neglect the accuracy of distribution network state estimation, leading to a series of serious consequences that may directly impact the safe, reliable, and economical operation of the distribution network (specific examples include decision-making errors, safety hazards, reduced power supply reliability, decreased energy efficiency, and economic losses). Therefore, a new optimized configuration scheme for distribution network measurements that considers state estimation is urgently needed. Summary of the Invention

[0004] The purpose of this invention is to provide a method and related apparatus for optimizing the configuration of distribution network measurements considering state estimation, in order to solve one or more of the aforementioned technical problems. The technical solution provided by this invention combines state estimation with optimized configuration, establishing an optimized configuration model (i.e., a graph neural network model) with the state estimation results as the objective. This improves the accuracy of the evaluation results, reduces the number of measurement devices, and demonstrates superior performance in terms of speed, convergence, and accuracy.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for optimizing the configuration of distribution network measurements considering state estimation, comprising the following steps:

[0007] The first end of the distribution network to be measured and optimized is selected as the measurement point;

[0008] Based on existing measurement points, the remaining non-measurement point nodes of the distribution network are set as candidate measurement points one by one, and the distribution network evaluation error corresponding to each candidate measurement point is obtained; wherein, the distribution network evaluation error is obtained based on the historical measurement data of the distribution network and the trained graph neural network model.

[0009] Set the candidate measurement point corresponding to the smallest distribution network assessment error as the measurement point, and determine whether the preset termination condition is met; if so, end the measurement optimization configuration process and output the distribution network measurement optimization configuration result.

[0010] A further improvement of the distribution network measurement optimization configuration method of the present invention is that the step of obtaining the distribution network evaluation error corresponding to each candidate measurement point includes:

[0011] For each candidate measurement point, the trained graph neural network model is used to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point, and the node evaluation results are obtained; wherein, the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node;

[0012] Based on historical measurement data of the distribution network, the error of the evaluation results of each node is calculated and summed to obtain the distribution network evaluation error.

[0013] A further improvement of the distribution network measurement optimization configuration method of the present invention is that the step of determining whether the preset termination condition has been met is as follows:

[0014] Determine whether the number of measurement points is greater than the first preset value; or, determine whether the maximum distribution network assessment error is less than the second preset value.

[0015] A further improvement of the distribution network measurement optimization configuration method of the present invention is that the step of obtaining the trained graph neural network model includes:

[0016] Historical measurement data of the distribution network are acquired and normalized to obtain a training dataset; wherein, each training sample in the training dataset includes the active power, reactive power, voltage and phase angle of the node;

[0017] Obtain the distribution network topology data and convert it into an adjacency matrix;

[0018] Before training, the network structure and model parameters of the graph neural network model are set. During training, for the selected training samples, the active power and reactive power of the nodes in the adjacent matrix and the training samples are used as inputs. The estimated values ​​of the voltage and phase angle of the nodes are output by the graph neural network model. The loss function is calculated by comparing the estimated values ​​of the voltage and phase angle of the nodes with the actual values ​​of the voltage and phase angle of the nodes in the training samples. The backpropagation algorithm is used to update the model weights. When the preset convergence condition is reached, the trained graph neural network model is obtained.

[0019] A further improvement of the distribution network measurement optimization configuration method of the present invention lies in the step of acquiring historical measurement data of the distribution network and performing normalization processing to obtain a training dataset.

[0020] Historical measurement data are normalized using standard deviation.

[0021] A further improvement of the distribution network measurement optimization configuration method of the present invention is that noise data with a preset proportion is randomly added to the training dataset.

[0022] A further improvement of the distribution network measurement optimization configuration method of the present invention is that the graph neural network model adopts a two-stage mathematical model including a message passing stage and a reading stage.

[0023] A further improvement of the distribution network measurement optimization configuration method of the present invention is that the step of setting the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point and determining whether the preset termination condition is met; if so, ending the measurement optimization configuration process and outputting the distribution network measurement optimization configuration result, further includes:

[0024] If the preset termination condition is not met, the process of setting the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points and obtaining the distribution network evaluation error corresponding to each candidate measurement point is repeated.

[0025] A second aspect of the present invention provides a distribution network measurement optimization configuration system considering state estimation, comprising:

[0026] The measurement point selection module is used to select the beginning of the distribution network to be measured and optimized as the measurement point;

[0027] The distribution network assessment error acquisition module is used to set the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points, and to acquire the distribution network assessment error corresponding to each candidate measurement point; wherein, the distribution network assessment error is acquired based on the historical measurement data of the distribution network and the trained graph neural network model.

[0028] The measurement point addition and condition judgment module is used to set the candidate measurement point corresponding to the smallest distribution network assessment error as the measurement point and determine whether the preset termination condition is met; if so, the measurement optimization configuration process ends and the distribution network measurement optimization configuration result is output.

[0029] A further improvement of the distribution network measurement optimization configuration system of the present invention is that, in the distribution network evaluation error acquisition module, the step of acquiring the distribution network evaluation error corresponding to each candidate measurement point includes:

[0030] For each candidate measurement point, the trained graph neural network model is used to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point, and the node evaluation results are obtained; wherein, the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node;

[0031] Based on historical measurement data of the distribution network, the error of the evaluation results of each node is calculated and summed to obtain the distribution network evaluation error.

[0032] A further improvement of the distribution network measurement optimization configuration system of the present invention is that, in the measurement point addition and condition judgment module, the step of judging whether the preset termination condition has been met is as follows:

[0033] Determine whether the number of measurement points is greater than the first preset value; or, determine whether the maximum distribution network assessment error is less than the second preset value.

[0034] A further improvement of the distribution network measurement optimization configuration system of the present invention is that the step of obtaining the trained graph neural network model includes:

[0035] Historical measurement data of the distribution network are acquired and normalized to obtain a training dataset; wherein, each training sample in the training dataset includes the active power, reactive power, voltage and phase angle of the node;

[0036] Obtain the distribution network topology data and convert it into an adjacency matrix;

[0037] Before training, the network structure and model parameters of the graph neural network model are set. During training, for the selected training samples, the active power and reactive power of the nodes in the adjacent matrix and the training samples are used as inputs. The estimated values ​​of the voltage and phase angle of the nodes are output by the graph neural network model. The loss function is calculated by comparing the estimated values ​​of the voltage and phase angle of the nodes with the actual values ​​of the voltage and phase angle of the nodes in the training samples. The backpropagation algorithm is used to update the model weights. When the preset convergence condition is reached, the trained graph neural network model is obtained.

[0038] A further improvement of the distribution network measurement optimization configuration system of the present invention lies in the step of acquiring historical measurement data of the distribution network and performing normalization processing to obtain a training dataset.

[0039] Historical measurement data are normalized using standard deviation.

[0040] A further improvement of the power distribution network measurement optimization configuration system of the present invention is that noise data with a preset proportion is randomly added to the training dataset.

[0041] A further improvement of the power distribution network measurement optimization configuration system of the present invention is that the graph neural network model adopts a two-stage mathematical model including a message passing stage and a reading stage.

[0042] A further improvement of the distribution network measurement optimization configuration system of the present invention is that, in the measurement point addition and condition judgment module, the step of setting the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point and judging whether the preset termination condition has been met; if so, ending the measurement optimization configuration process and outputting the distribution network measurement optimization configuration result, further includes:

[0043] If the preset termination condition is not met, the steps of the power distribution network assessment error acquisition module are repeated.

[0044] In a third aspect, 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, when executing the program, implements a distribution network measurement optimization configuration method considering state estimation as described in any one of the first aspects of the present invention.

[0045] In a fourth aspect, 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 measurement optimization configuration method considering state estimation as described in any one of the first aspects of the present invention.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] This invention discloses a distribution network measurement optimization configuration method considering state estimation. Based on a graph neural network, a distribution network measurement optimization configuration model is constructed. The deep computation characteristics of graph neural networks solve the "curse of dimensionality" problem in state estimation, providing accurate state estimation results even under noisy conditions. Simultaneously, the state estimation using the graph neural network model achieves optimized configuration of measurement equipment. This invention utilizes a graph neural network model and incorporates a strategy of progressively optimizing node placement, improving the accuracy of evaluation results, reducing the number of measurement devices, and solving the problems of noise in power grid measurement data and excessive computational load in overall deployment optimization. More specifically, compared to traditional state estimation methods, the graph neural network state estimation method of this invention avoids the "curse of dimensionality" problem and demonstrates superior performance in terms of speed, convergence, and accuracy. Under noisy data conditions, compared to data-driven distribution network state estimation methods, the method of this invention provides more accurate evaluation results. Optimizing measurement configuration by considering the distribution network state estimation results improves the distribution network state estimation effect, reduces the number of measurement devices, and saves costs. Attached Figure Description

[0048] 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.

[0049] Figure 1 This is a flowchart illustrating a distribution network measurement optimization configuration method that considers state estimation, as described in an embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of the training process of the graph neural network model in an embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of a graph neural network model in an embodiment of the present invention.

[0052] Figure 4 This is a schematic diagram of a specific process for a distribution network measurement optimization configuration method that considers state estimation, as described in an embodiment of the present invention.

[0053] Figure 5 This is a schematic diagram of the IEEE 33-node distribution network in an embodiment of the present invention.

[0054] Figure 6 In this embodiment of the invention, the graph convolution process of the graph neural network is shown in the figure.

[0055] Figure 7 This is a schematic diagram comparing the results of IEEE 33 nodes in an embodiment of the present invention; wherein, Figure 7 (a) is a schematic diagram comparing iteration errors. Figure 7 (b) is a schematic diagram comparing node voltages. Figure 7 (c) is a schematic diagram comparing the phase angles of the nodes.

[0056] Figure 8 This is a schematic diagram of the optimized node layout under the IEEE 33-node measurement configuration in an embodiment of the present invention.

[0057] Figure 9 This is a schematic diagram comparing results under the IEEE 33-node measurement configuration in an embodiment of the present invention; wherein, Figure 9 (a) is a schematic diagram comparing node voltages. Figure 9 (b) is a schematic diagram comparing the phase angles of the nodes.

[0058] Figure 10 This is a flowchart illustrating a distribution network measurement optimization configuration method that considers state estimation, as described in an embodiment of the present invention.

[0059] Figure 11This is a schematic diagram of a distribution network measurement optimization configuration system that takes into account state estimation in an embodiment of the present invention.

[0060] Figure 12 This is a schematic diagram of a distribution network measurement optimization configuration system that takes into account state estimation, according to an embodiment of the present invention. Detailed Implementation

[0061] 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 inventive effort are within the scope of protection of this invention. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0062] Please see Figure 1 In this embodiment of the invention, a method for optimizing the configuration of distribution network measurements considering state estimation is provided, specifically including the following steps:

[0063] Step 1: Select the beginning of the distribution network to be measured and optimized as the measurement point;

[0064] Step 2: Based on the existing measurement points, the remaining non-measurement point nodes of the distribution network are set as candidate measurement points one by one, and the distribution network evaluation error corresponding to each candidate measurement point is obtained; wherein, the distribution network evaluation error is obtained based on the historical measurement data of the distribution network and the trained graph neural network model.

[0065] Step 3: Set the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point, and determine whether the preset termination condition has been met; if so, end the measurement optimization configuration process and output the distribution network measurement optimization configuration result. The technical solution provided by this embodiment aims to solve the problems of noise in power grid measurement data and excessive computational load in overall deployment optimization. Specifically, the solution of this embodiment utilizes a Graph Neural Network (GNN) algorithm, combined with a strategy of progressively optimizing node placement, which can improve the accuracy of evaluation results and reduce the number of measurement devices.

[0066] To further explain, the technical solutions of the embodiments of the present invention have the following significant advantages:

[0067] 1) Mining Power Grid Topology and Node Relationships: Graph neural networks are particularly well-suited for processing graph data with complex topologies and node relationships. In a power grid, there are complex electrical connections between various nodes (e.g., transformers, circuit breakers, load points, etc.), which can be represented by a graph. Graph neural networks can learn the relationships between these nodes and edges, thereby more accurately assessing the state of the power grid. Traditional power grid state assessment methods often struggle to obtain accurate results when faced with noisy measurement data; however, graph neural networks can reduce the impact of noisy data on the assessment results by learning the dependencies and patterns between nodes, thus enabling accurate distribution network state assessment even under noisy conditions.

[0068] 2) Avoiding excessive computational load for overall optimization: In power grids, the number of nodes is typically large. Directly performing overall optimization would lead to a sharp increase in computational load, and might even render the solution unsolvable. To address this issue, the technical solution of this invention adopts a strategy of gradually optimizing the node layout, thus avoiding excessive computational load for overall optimization. Furthermore, it first optimizes a subset of nodes, and then, based on the optimization results of these nodes, gradually expands to the optimization of the entire power grid.

[0069] 3) Ensuring the accuracy of state assessment: Although a stepwise optimization strategy is adopted, the technical solution of this embodiment can still ensure the accuracy of state assessment under the condition of minimal deployment of nodes; this is because graph neural networks can learn the key nodes and edges in the power grid, thereby giving priority to these key parts in the optimization process and ensuring the accuracy of the state assessment of the entire power grid.

[0070] In summary, by combining graph neural networks and a strategy of progressively optimizing node placement, this invention successfully solves the problems of noise in power grid measurement data and excessive computational load for overall deployment optimization, providing a new and effective solution for power grid condition assessment.

[0071] In one embodiment of the present invention, step 3, determining whether a preset termination condition has been met, may specifically be as follows:

[0072] Determine whether the number of measurement points exceeds the first preset value;

[0073] Alternatively, determine whether the error of the evaluation result is less than the second preset value.

[0074] Please see Figure 2 and Figure 3 In this embodiment of the invention, the training steps of the graph neural network model include:

[0075] Step S1, construct the topology model data, including:

[0076] Historical measurement data of the distribution network is acquired, and training and validation datasets are formed accordingly. The historical measurement data may include node voltage, current, active power, reactive power, and phase angle, etc.

[0077] A distribution network system model and parameters are constructed, including system topology information, node information, load information, and line parameters, forming distribution network topology data. Further, based on the distribution network structure matrix, it is transformed into an adjacency matrix. Explained, the rows and columns of the adjacency matrix correspond to nodes in the graph, and each element in the matrix represents the connection relationship between two nodes; for example, if there is an edge between nodes, the corresponding matrix element is 1, otherwise it is 0.

[0078] Step S2, initial parameter setting and model construction of the graph neural network, including:

[0079] 1) Data normalization processing:

[0080] To eliminate the influence of units in historical measurement data, or in other words, to ensure that all features of the sample have equal weight, this embodiment of the invention relies on standard deviation to normalize the data. The formula for standard deviation is shown below.

[0081]

[0082] In the formula, k sd The result is the result of normalization; k is the original data; mean(k) is the standard number; standarddeviation(k) is the standard deviation.

[0083] 2) The constructed graph neural network model mainly includes a two-stage mathematical model: a message passing stage and a message reading stage; among which,

[0084] Message passing phase: The node state wave and the line state vector are combined through messages and update functions. The specific process is as follows:

[0085] Within a graph convolutional layer, the neighborhood information of a node s is obtained using the following formula:

[0086] N(s)={p|e sp ∈E, p≠s};

[0087] In the formula, N(s) represents the information of node s; p represents a node different from s; e sp E represents the edge between node s and node p; E represents the connection between node s and other nodes.

[0088] That is, through edge e sp ∈E determines whether they are adjacent nodes;

[0089] information The message is passed along the edge to the node via the message function Mn, and is obtained using the following formula:

[0090]

[0091] In the formula, Represents the information of node s in layer l; f E (e sp ) represents the edge computation function between nodes s and p; M n Represents a message function; This represents the hidden state vector of the (n-1)th layer.

[0092] The 0th hidden state vector is initialized using the node's input feature vector, i.e. Subsequently, the hidden state vector of node s By updating function U n renew Update function U n Combining the hidden state from the previous layer with information received from the domain, it is represented as: The hidden state of a node after n convolutions. It contains local environment information with a radius of n nodes.

[0093] During the reading phase, the hidden node states of the last convolutional layer are combined into a graph representation vector h using a pooling function. G Molecular fingerprint,

[0094] In the formula, the function p is chosen as the average value.

[0095] Finally, the molecular fingerprint vector h G It was used for molecular property regression, employing a multilayer perceptron (MLP):

[0096]

[0097] Step S3, determine the graph neural network structure and parameters, including:

[0098] Before training the model, set the network structure and model parameters of the graph neural network model, including: the total number of layers, the number of pooling layers, the optimizer, the number of iterations, etc.

[0099] Step S4, offline training, includes:

[0100] Input multi-section historical map data, output estimated values ​​of voltage amplitude and phase angle of nodes, calculate loss function by using estimated values ​​and actual values ​​(historical measurement data), and update model weights using backpropagation algorithm.

[0101] In the graph convolution process of a graph neural network, node data is generated by combining its surrounding nodes and node topological relationships.

[0102] In this embodiment of the invention, measurement devices are installed sequentially on the remaining candidate nodes in the network. The active power and reactive power of the selected nodes are used as input values ​​of the graph neural network to calculate the node voltage and phase angle. The node with the smallest error in node voltage and phase angle is selected as the next measurement arrangement point.

[0103] Specifically, by example, measurement nodes are continuously added according to the above steps until the prediction results of each node in the graph neural network are within 5%, and the measurement nodes and the number of measurements are output; if the number of measurements cannot meet the requirements, the second candidate node causing the error is used for calculation (and so on to select nodes).

[0104] Please see Figures 4 to 9 Specifically, this invention provides a method for optimizing the configuration of distribution network measurements considering state estimation, comprising the following steps:

[0105] S101, Time-domain simulation generation of training and validation data, specifically includes:

[0106] In one embodiment of the present invention, the training set and test set are selected from IEEE 33-node computational data, with node load fluctuations around 20%, and their topology is as follows. Figure 5 As shown, the active and reactive power of the circuit nodes are used as the node features of the graph neural network, the circuit lines are used as the edges, the node correlation matrix is ​​used as the edge feature vector, and the voltage and phase angle of the circuit nodes are used as the node output features. At the same time, 5% noise data is randomly added to the training data.

[0107] S102, Construct topology model data, including:

[0108] Construct a power distribution network system model and parameters, including system topology information, node information, load information, and line parameters, to form a topological network structure.

[0109] S103, based on the distribution network structure matrix, transform adjacent matrices, including:

[0110] The rows and columns of the matrix correspond to the nodes in the graph, and each element in the matrix represents the connection between two nodes; if there is an edge between nodes, the corresponding matrix element is 1, otherwise it is 0.

[0111] S104, Data normalization and selection of graph neural network parameters, specifically including:

[0112] The data is normalized using standard deviation.

[0113] The graph neural network model of this invention takes the active power, reactive power and adjacency matrix of the nodes as inputs and the node voltage and phase angle as outputs. For example, according to the IEEE 33-node model, the number of input nodes and output layer nodes is 66 nodes, and the input matrix is ​​the network topology adjacency matrix.

[0114] S105, completing the construction of the message passing and reading stages. To extract features from different dimensions of the data, deep convolutional layers need to be stacked in the message passing stage. In this embodiment, five convolutional layers are selected, such as... Figure 6 As shown, convolution and pooling operations are performed on the feature values ​​X and the new relation matrix A of each layer to obtain the data characteristics Z of each layer. Z is then used as the input of the next layer, and the process is continuously optimized and iterated.

[0115] S106, offline training, includes: calculating the loss function from the output system node voltage amplitude and branch phase angle difference using estimated and actual values, and updating the model weights using the backpropagation algorithm.

[0116] S107, Online model optimization and output, including: the objective function is to minimize the system node deviation; when the node voltage error is greater than the set error, the system returns to step S106 for another iteration; when the node voltage is less than the set error, the model training is complete.

[0117] S108, select accuracy as the optimization target, including: setting the minimum error of the graph neural network state estimation result (voltage amplitude, phase angle value) as the calculation target.

[0118] S109, Node selection, includes: installing measuring devices sequentially on the remaining nodes in the distribution network; using the active and reactive power of the selected nodes as input values ​​for the graph neural network; calculating the node voltage and phase angle; and selecting the node with the smallest voltage and phase angle error as the next measuring point. Measuring nodes are continuously added until the prediction results of each node in the graph neural network have a deviation within 5%. The number of measuring nodes and the quantity of measurements are then output.

[0119] Please see Figure 7 , Figure 7 To introduce noise into the input data and to obtain noise-free input data, a graph neural network was used to calculate the voltage values ​​and phase angles of 33 nodes, comparing these results with the actual values. (See results...) Figure 7 As shown in (a), the difference continuously decreases after multiple iterations; as the results show... Figure 7 As shown in (b) and (c), under noisy conditions, the results obtained by the graph neural network are not significantly different from the true values, thus demonstrating strong anti-interference capabilities.

[0120] Please see Figure 8 , Figure 8To optimize the distribution network measurement configuration method and obtain the optimal deployment of 33 nodes through the technical solution of this invention, the state values ​​(node ​​voltage and phase angle) of the 33 nodes with an error within 5% can be calculated from the five yellow node values ​​in the figure.

[0121] Please see Figure 9 , Figure 9 for Figure 8 The measured values ​​of the five nodes are obtained by using a graph neural network to obtain the voltage and phase angle values ​​of 33 nodes, and the results are compared with the actual values. As can be seen from the comparison, the results obtained by the method of this embodiment have a very small deviation from the actual values.

[0122] Please see Figure 10 In this embodiment of the invention, a method for optimizing the configuration of distribution network measurements considering state estimation is provided, specifically including the following steps:

[0123] Step 1: Select the beginning of the distribution network to be measured and optimized as the measurement point, and proceed to step 2.

[0124] Step 2: Based on the existing measurement points, set the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one, obtain the distribution network evaluation error corresponding to each candidate measurement point, and then proceed to step 3.

[0125] Step 3: Set the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point, and determine whether the preset termination condition is met; if yes, end the measurement optimization configuration process and output the distribution network measurement optimization configuration result; if no, jump to step 2.

[0126] The step of obtaining the distribution network evaluation error corresponding to each candidate measurement point includes:

[0127] For each candidate measurement point, the trained graph neural network model is used to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point, and the node evaluation results are obtained; wherein, the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node;

[0128] Based on historical measurement data of the distribution network, the error of the evaluation results of each node is calculated and summed to obtain the distribution network evaluation error.

[0129] The technical solution provided in this invention uses graph neural networks for distribution network state estimation. Compared with purely data-driven methods, the model has stronger interpretability and higher transparency. In addition, by utilizing the structural information contained in the correlation matrix of the power grid topology, the structural features between nodes are efficiently utilized. Graph structure data contains rich information, and nodes are connected by lines, effectively and fully expressing information such as the relationships between different nodes, thereby maximizing the utilization of the structural features of the distribution network topology. Furthermore, by optimizing the configuration of measurement based on the distribution network state estimation results, the number of measurement devices is reduced.

[0130] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0131] Please see Figure 11 In this embodiment of the invention, a distribution network measurement optimization configuration system considering state estimation is provided, comprising:

[0132] The measurement point selection module is used to select the beginning of the distribution network to be measured and optimized as the measurement point;

[0133] The distribution network assessment error acquisition module is used to set the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points, and to acquire the distribution network assessment error corresponding to each candidate measurement point; wherein, the distribution network assessment error is acquired based on the historical measurement data of the distribution network and the trained graph neural network model.

[0134] The measurement point addition and condition judgment module is used to set the candidate measurement point corresponding to the smallest distribution network assessment error as the measurement point and determine whether the preset termination condition is met; if so, the measurement optimization configuration process ends and the distribution network measurement optimization configuration result is output.

[0135] Please see Figure 12 In this embodiment of the invention, a distribution network measurement optimization configuration system considering state estimation is provided, comprising: a measurement point selection module, a distribution network evaluation error acquisition module, and a measurement point addition and condition judgment module; wherein,

[0136] The measurement point selection module is used to select the beginning of the distribution network to be measured and optimized as the measurement point, and then jump to execute the steps of the distribution network evaluation error acquisition module.

[0137] The distribution network evaluation error acquisition module is used to set the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points, acquire the distribution network evaluation error corresponding to each candidate measurement point, and jump to execute the steps of the measurement point addition and condition judgment module.

[0138] The measurement point addition and condition judgment module is used to set the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point, and to determine whether the preset termination condition is met; if yes, the measurement optimization configuration process ends and the distribution network measurement optimization configuration result is output; if no, the process jumps to the steps of the distribution network evaluation error acquisition module.

[0139] The step of obtaining the distribution network evaluation error corresponding to each candidate measurement point includes:

[0140] For each candidate measurement point, the trained graph neural network model is used to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point, and the node evaluation results are obtained; wherein, the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node;

[0141] Based on historical measurement data of the distribution network, the error of the evaluation results of each node is calculated and summed to obtain the distribution network evaluation error.

[0142] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used to execute the operation of a distribution network measurement optimization configuration method considering state estimation.

[0143] In one embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the distribution network measurement optimization configuration method considering state estimation in the above embodiments.

[0144] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0145] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0146] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0147] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0148] 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 it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing the configuration of distribution network measurements considering state estimation, characterized in that, Includes the following steps: The first end of the distribution network to be measured and optimized is selected as the measurement point; Based on existing measurement points, the remaining non-measurement point nodes of the distribution network are set as candidate measurement points one by one, and the distribution network evaluation error corresponding to each candidate measurement point is obtained; wherein, the distribution network evaluation error is obtained based on the historical measurement data of the distribution network and the trained graph neural network model. Set the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point, and determine whether the preset termination condition is met; if so, end the measurement optimization configuration process and output the distribution network measurement optimization configuration result. The step of obtaining the distribution network evaluation error corresponding to each candidate measurement point includes: for each candidate measurement point, using a trained graph neural network model to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point, and obtaining the node evaluation result; wherein the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node; based on the historical measurement data of the distribution network, the error of the evaluation result of each node is calculated and summed to obtain the distribution network evaluation error.

2. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 1, characterized in that, The step of determining whether the preset termination condition has been met is as follows: Determine whether the number of measurement points is greater than the first preset value; or, determine whether the maximum distribution network assessment error is less than the second preset value.

3. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 1, characterized in that, The steps for obtaining the trained graph neural network model include: Historical measurement data of the distribution network are acquired and normalized to obtain a training dataset; wherein, each training sample in the training dataset includes the active power, reactive power, voltage and phase angle of the node; Obtain the distribution network topology data and convert it into an adjacency matrix; Before training, the network structure and model parameters of the graph neural network model are set. During training, for the selected training samples, the active power and reactive power of the nodes in the adjacent matrix and the training samples are used as inputs. The estimated values ​​of the voltage and phase angle of the nodes are output by the graph neural network model. The loss function is calculated by comparing the estimated values ​​of the voltage and phase angle of the nodes with the actual values ​​of the voltage and phase angle of the nodes in the training samples. The backpropagation algorithm is used to update the model weights. When the preset convergence condition is reached, the trained graph neural network model is obtained.

4. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 3, characterized in that, In the step of acquiring historical measurement data of the distribution network and performing normalization processing to obtain a training dataset, Historical measurement data are normalized using standard deviation.

5. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 3, characterized in that, The training dataset contains randomly added noise data with a preset proportion.

6. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 1, characterized in that, The graph neural network model adopts a two-stage mathematical model that includes a message passing stage and a reading stage.

7. The method for optimizing the configuration of distribution network measurements considering state estimation according to claim 1, characterized in that, The candidate measurement point corresponding to the smallest distribution network evaluation error is set as the measurement point, and it is determined whether the preset termination condition is met. If so, the step of ending the measurement optimization configuration process and outputting the distribution network measurement optimization configuration results also includes: If the preset termination condition is not met, the process of setting the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points and obtaining the distribution network evaluation error corresponding to each candidate measurement point is repeated.

8. A distribution network measurement optimization configuration system considering state estimation, characterized in that, include: The measurement point selection module is used to select the beginning of the distribution network to be measured and optimized as the measurement point; The distribution network assessment error acquisition module is used to set the remaining non-measurement point nodes of the distribution network as candidate measurement points one by one based on the existing measurement points, and to acquire the distribution network assessment error corresponding to each candidate measurement point; wherein, the distribution network assessment error is acquired based on the historical measurement data of the distribution network and the trained graph neural network model. The measurement point addition and condition judgment module is used to set the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point and determine whether the preset termination condition is met; if so, the measurement optimization configuration process ends and the distribution network measurement optimization configuration result is output. The distribution network assessment error acquisition module includes the following steps for obtaining the distribution network assessment error corresponding to each candidate measurement point: for each candidate measurement point, a trained graph neural network model is used to evaluate each of the remaining non-measurement point nodes excluding the candidate measurement point to obtain the node assessment result; wherein the input of the graph neural network model is the active power and reactive power of the node, and the output is the voltage and phase angle of the node; based on the historical measurement data of the distribution network, the error of the assessment result of each node is calculated and summed to obtain the distribution network assessment error.

9. A distribution network measurement optimization configuration system considering state estimation according to claim 8, characterized in that, In the measurement point addition and condition judgment module, the step of judging whether the preset termination condition has been met is as follows: Determine whether the number of measurement points is greater than the first preset value; or, determine whether the maximum distribution network assessment error is less than the second preset value.

10. A distribution network measurement optimization configuration system considering state estimation according to claim 8, characterized in that, The steps for obtaining the trained graph neural network model include: Historical measurement data of the distribution network are acquired and normalized to obtain a training dataset; wherein, each training sample in the training dataset includes the active power, reactive power, voltage and phase angle of the node; Obtain the distribution network topology data and convert it into an adjacency matrix; Before training, the network structure and model parameters of the graph neural network model are set. During training, for the selected training samples, the active power and reactive power of the nodes in the adjacent matrix and the training samples are used as inputs. The estimated values ​​of the voltage and phase angle of the nodes are output by the graph neural network model. The loss function is calculated by comparing the estimated values ​​of the voltage and phase angle of the nodes with the actual values ​​of the voltage and phase angle of the nodes in the training samples. The backpropagation algorithm is used to update the model weights. When the preset convergence condition is reached, the trained graph neural network model is obtained.

11. A distribution network measurement optimization configuration system considering state estimation according to claim 10, characterized in that, In the step of acquiring historical measurement data of the distribution network and performing normalization processing to obtain a training dataset, Historical measurement data are normalized using standard deviation.

12. A distribution network measurement optimization configuration system considering state estimation according to claim 10, characterized in that, The training dataset contains randomly added noise data with a preset proportion.

13. A distribution network measurement optimization configuration system considering state estimation according to claim 8, characterized in that, The graph neural network model adopts a two-stage mathematical model that includes a message passing stage and a reading stage.

14. A distribution network measurement optimization configuration system considering state estimation according to claim 8, characterized in that, In the measurement point addition and condition judgment module, the process of setting the candidate measurement point corresponding to the smallest distribution network evaluation error as the measurement point is executed, and it is judged whether the preset termination condition is met. If so, the step of ending the measurement optimization configuration process and outputting the distribution network measurement optimization configuration results also includes: If the preset termination condition is not met, the steps of the power distribution network assessment error acquisition module are repeated.

15. 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 measurement optimization configuration method considering state estimation as described in any one of claims 1 to 7.

16. 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 measurement optimization configuration method considering state estimation as described in any one of claims 1 to 7.

Images