Three-phase impedance matrix identification method for distribution station area based on self-supervised graphical attention network

Abstract

The present application provides a method for identifying the three-phase impedance matrix of a distribution substation based on a self-supervised graph attention network, which relates to the technical field of impedance matrix identification of distribution networks. The method can construct a node set of the distribution substation to be identified, including the substation's secondary distribution main and at least one user meter. The method uses the feature vector data of each node collected in real time and, based on a trained three-phase impedance matrix identification model for the distribution substation, identifies the three-phase impedance matrix of the distribution substation to be identified. This avoids the problems of existing methods such as reliance on known network topology and insufficient fitting of physical laws. The method is suitable for distribution substations with unknown or dynamically changing topological structures, and meets the requirements for refined operation and control of distribution networks.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network impedance matrix identification, and in particular to a distribution station area three-phase impedance matrix identification method based on a self-supervised graph attention network. Background Art

[0002] In smart distribution networks, the three-phase impedance matrix of a distribution substation is a core parameter describing its electrical characteristics. Its accurate identification is crucial for fault diagnosis, power flow calculation, and power quality analysis. Existing methods rely on known network topology and component parameters (such as line length, unit resistance, and reactance) to derive the impedance matrix using Kirchhoff's laws. Figure 1 The figure illustrates the topological structure between multiple nodes in a distribution substation. Node 1 is a distribution transformer, and nodes 2 to 33 are user terminals on the low-voltage line. Existing methods derive the impedance matrix using Kirchhoff's law based on the node topology and element parameters.

[0003] However, in most distribution subareas, line parameters are often difficult to obtain, and topology may be unknown or dynamically changing. The distribution subarea here refers to the power supply area in the distribution network, from the low-voltage side of the distribution transformer (usually 0.4kV or 380V) to the user's meter. This is the terminal link of the power network and directly supplies power to users.

[0004] Therefore, it is necessary to provide a new three-phase impedance matrix identification method for distribution station area to solve the above technical problems. Summary of the Invention

[0005] In view of the above problems, this application is proposed to provide a method and device for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network, and an electronic device that overcomes or at least partially solves the above problems. The technical solution is as follows:

[0006] In a first aspect, a method for identifying a three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network is provided, the method comprising:

[0007] Taking the secondary side of the distribution center and the user's electric meter as nodes, a node set including the secondary side of the distribution center and at least one user's electric meter of the distribution center to be identified is constructed;

[0008] Acquire the real-time collected three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution substation to be identified, and acquire the real-time collected three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of each user meter in the distribution substation to be identified;

[0009] According to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution substation to be identified and each user's electricity meter, the characteristic vector of each node in the node set is constructed;

[0010] The feature vector of each node is input into the pre-trained distribution substation three-phase impedance matrix identification model, and the three-phase impedance matrix of the distribution substation to be identified is output and marked as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. n is the total number of nodes; the three-phase impedance matrix identification model of the distribution station area is based on the self-supervised graph attention network, and is trained by combining the fusion loss function of three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; the consistency regularization loss is used to constrain the model to maintain consistency before and after when the input is disturbed by noise.

[0011] In one possible implementation, the three-phase impedance matrix identification model of the distribution station area is trained by the following steps:

[0012] The sample node set of the sample distribution substation is constructed by taking the substation secondary side and at least one user meter of the sample distribution substation as each sample node;

[0013] The historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the secondary side of the distribution station and each user meter of the sample distribution station area are selected to construct the feature vector of each sample node in the sample node set;

[0014] Establish a self-supervised graph attention network model structure, where the graph attention part is composed of graph attention layers. Multiple attention heads are set up in each graph attention layer for parallel calculation. Deep feature extraction is achieved by stacking multiple graph attention layers.

[0015] Construct a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss;

[0016] Using multiple preset neural network parameter initialization strategies, differentiated initialization is performed for different types of network layers. The feature vectors of each sample node in the sample node set are input into the model structure of the self-supervised graph attention network for training. The preset optimizer is used to minimize the fusion loss function on the training node to achieve gradient optimization of the model structure of the self-supervised graph attention network. At the same time, the preset early stopping technology is used on the validation set to automatically trigger the training termination condition, and a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

[0017] In a possible implementation, a sample node set of a sample distribution substation is constructed using the substation secondary side and at least one user meter of the sample distribution substation as sample nodes, including:

[0018] The substation distribution main secondary side and at least one user meter of the sample distribution substation are used as independent sample nodes, and the switch of the line between the substation distribution main secondary side and at least one user meter is not included to construct a sample node set of the sample distribution substation.

[0019] In one possible implementation, the adjacency matrix of the model structure of the established self-supervised graph attention network is fully connected, and the model structure of the self-supervised graph attention network does not need to perform the topological relationship identification step of the sample distribution station area, that is, there is no need to obtain the topological structure of the sample distribution station area, and the initial impedance parameter between any two nodes in the sample node set is defined as the edge feature of the model structure of the self-supervised graph attention network.

[0020] In one possible implementation, a fusion loss function is constructed that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss, including:

[0021] The formula for constructing the fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss is:

[0022] (1)

[0023] In formula (1), is the three-phase voltage consistency loss function, is the error constraint function of three-phase injected active power and three-phase injected reactive power, is the consistency regularization loss function, are the preset weight coefficients respectively.

[0024] In one possible implementation, Represents the constraint on the voltage amplitude error of each phase, and the total loss is the average of the three phases:

[0025] (2)

[0026] In formula (2), n is the total number of nodes, p is the phase, A 、 B 、 C is the phase value, and A 、 B 、 C It forms three phases, The nodes collected i of p Phase voltage amplitude, and Node i The real and imaginary parts of the predicted voltage are calculated as follows:

[0027] According to Kirchhoff's law, For nodes i of p Phase complex voltage satisfies formula (3):

[0028] (3)

[0029] In formula (3), For nodes i, j between pm Phase impedance value, For nodes j of m Phase injected complex current; For nodes j of m The complex power injected into the phase satisfies equations (4) and (5):

[0030] (4)

[0031] (5)

[0032] In formula (4), For nodes j of m Phase complex voltage, represents conjugation;

[0033] In formula (5), The nodes collected j of m Phase injected active power, w is the imaginary unit, The nodes collected j of m Phase injected reactive power;

[0034] According to formula (4) and formula (5), we can get:

[0035] (6)

[0036] In formula (6), The nodes collected j of m Phase voltage amplitude, For nodes j of m Phase angle;

[0037] Assume the secondary side of the substation distribution system is the reference node. i =1; the user's meter is a non-reference node, i =2,3,… n ;

[0038] definition p=A The phase angle of the phase voltage is 0, that is, in formula (6) =0, then:

[0039] , then formula (6) is converted into formula (7):

[0040] (7)

[0041] Substituting equation (7) into equation (3), we can separate The real and imaginary parts of , we get formula (8):

[0042] (8)

[0043] In formula (8), Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase resistance value, Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase reactance value;

[0044] According to formula (8), we can get formula (9):

[0045] (9)

[0046] According to formula (9), we can get formula (10) and formula (11):

[0047] (10)

[0048] (11).

[0049] In one possible implementation, Constrain the three-phase injected active power and three-phase injected reactive power errors, as shown in the following formula (12):

[0050] (12)

[0051] In formula (12), The predicted node i of p Phase injected active power, The nodes collected i of p Phase injected active power, The predicted node i of p Phase injected reactive power, The nodes collected i of p Phase injected reactive power; obtained by presetting the known three-phase power flow calculation formula and .

[0052] In one possible implementation, The constraint model maintains consistency when the input is disturbed by noise, which enhances the model's robustness to noise. The formulas are as follows (13) and (14):

[0053] (13)

[0054] (14)

[0055] In formula (13), Represents the model prediction results, represents the model prediction result after disturbance, represents the node feature perturbation, x The feature vector representing the node;

[0056] In formula (14), It conforms to a Gaussian distribution with mean 0.

[0057] In a second aspect, a three-phase impedance matrix identification device for a distribution station area based on a self-supervised graph attention network is provided, the device comprising:

[0058] The first construction unit is used to construct a node set of the distribution substation to be identified, including the distribution substation secondary side and at least one user meter, with the distribution substation secondary side and the user meter as nodes;

[0059] An acquisition unit is used to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution station area to be identified, and to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of each user meter in the distribution station area to be identified;

[0060] The second construction unit is used to construct a feature vector of each node in the node set according to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution station area to be identified and each user's electricity meter;

[0061] The identification unit is used to input the feature vector of each node into the pre-trained distribution substation three-phase impedance matrix identification model, output the three-phase impedance matrix of the distribution substation to be identified, and mark it as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. n is the total number of nodes; the three-phase impedance matrix identification model of the distribution station area is based on the self-supervised graph attention network, and is trained by combining the fusion loss function of three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; the consistency regularization loss is used to constrain the model to maintain consistency before and after when the input is disturbed by noise.

[0062] In a third aspect, an electronic device is provided, comprising a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute any of the above-mentioned methods for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network.

[0063] By means of the above-mentioned technical solution, the embodiment of the present application provides a method and device for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network, and an electronic device. The method can construct a node set of the distribution station area to be identified, including the secondary side of the distribution station area and at least one user meter. The three-phase impedance matrix of the distribution station area to be identified is identified based on the trained distribution station area three-phase impedance matrix identification model by utilizing the feature vector data of each node collected in real time. This avoids the problems of existing methods relying on known network topology and insufficient fitting of physical laws. It is suitable for distribution stations with unknown or dynamically changing topological structures, and meets the requirements of refined operation and control of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments of the present application.

[0065] Figure 1 It illustrates the topological structure between multiple nodes in a distribution area;

[0066] Figure 2 A flowchart of a method for identifying a three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to an embodiment of the present application is shown;

[0067] Figure 3 An example of a feature vector change process provided by an embodiment of the present application is shown;

[0068] Figure 4 The training set loss function of the model provided in the embodiment of the present application is shown;

[0069] Figure 5 The structure diagram of the three-phase impedance matrix identification device for a distribution station area based on a self-supervised graph attention network provided by an embodiment of the present application is shown;

[0070] Figure 6 A structural diagram of a three-phase impedance matrix identification device for a distribution station area based on a self-supervised graph attention network according to another embodiment of the present application is shown;

[0071] Figure 7 A structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0072] The following describes exemplary embodiments of the present application in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0073] It should be noted that the terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that such usage is interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "including" and its variations are to be interpreted as open-ended terms meaning "including but not limited to."

[0074] In order to solve the above technical problems, the embodiment of the present application provides a method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network, such as Figure 2As shown, the method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network may include the following steps S201 to S204:

[0075] Step S201 : Taking the substation distribution main secondary side and user electric meters as nodes, a node set including the substation distribution main secondary side and at least one user electric meter of the distribution substation to be identified is constructed.

[0076] In this step, the distribution area refers to the area supplied by the distribution transformer, and the area distribution secondary side refers to the low-voltage output end of the distribution transformer. The distribution area to be identified here refers to the distribution area of ​​the three-phase impedance matrix to be identified.

[0077] Here, the substation distribution main secondary side and the user meter are independent nodes, and the switches of the lines between the substation distribution main secondary side and the user meter are not included. A node set including the substation distribution main secondary side and at least one user meter is constructed for the distribution substation to be identified. There is no need to obtain the topological structure between the substation distribution main secondary side and at least one user meter. It is suitable for distribution substations to be identified with unknown or dynamically changing topological structures.

[0078] Step S202, obtain the three-phase voltage amplitude, three-phase injection current amplitude, three-phase injection active power and three-phase injection reactive power of the secondary side of the distribution station to be identified, and obtain the three-phase voltage amplitude, three-phase injection current amplitude, three-phase injection active power and three-phase injection reactive power of each user meter in the distribution station to be identified.

[0079] In this step, the three phases are A Mutually, B Mutually, C Phase, is composed of three sinusoidal alternating currents with the same frequency and a phase difference of 120°.

[0080] Step S203 , constructing a feature vector of each node in the node set according to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the distribution main secondary side of the distribution substation to be identified and each user's electricity meter.

[0081] In this step, the feature vector of each node x Expressed as:

[0082] [ ],in, , , ; The nodes collected i The three phases (i.e. A Mutually, B Mutually, C Phase) voltage amplitude; The nodes collected i The three-phase injection current amplitude; The nodes collected i The three-phase injected active power; The nodes collected i The three-phase injected reactive power.

[0083] Step S204: Input the feature vector of each node into the pre-trained distribution area three-phase impedance matrix identification model, output the three-phase impedance matrix of the distribution area to be identified, and mark it as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. n is the total number of nodes; the three-phase impedance matrix identification model of the distribution station area is based on the self-supervised graph attention network, and is trained by combining the fusion loss function of three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; the consistency regularization loss is used to constrain the model to maintain consistency before and after when the input is disturbed by noise.

[0084] This embodiment can construct a node set of the distribution substation to be identified, including the substation's main secondary side and at least one user meter. By using the feature vector data of each node collected in real time, based on the trained distribution substation three-phase impedance matrix identification model, the three-phase impedance matrix of the distribution substation to be identified is identified. This avoids the problems of existing methods relying on known network topology and insufficient fitting of physical laws. It is suitable for distribution substations with unknown or dynamically changing topological structures, and meets the requirements of refined operation and control of the distribution network.

[0085] The present application provides a possible implementation method. The three-phase impedance matrix identification model of the distribution station area mentioned in step S204 can be trained by the following steps a1 to a5:

[0086] Step a1: construct a sample node set of the sample distribution station area by taking the area distribution main secondary side and at least one user meter of the sample distribution station area as sample nodes.

[0087] In this step, the sample node set for the sample distribution substation can be constructed using the substation distribution main secondary side and at least one user meter in the sample distribution substation as independent sample nodes, excluding switches in the line between the substation distribution main secondary side and the at least one user meter. This eliminates the need to obtain the topology between the substation distribution main secondary side and the at least one user meter. Here, the sample distribution substation refers to the distribution substation serving as the sample.

[0088] Step a2: Select the historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the substation secondary side and each user meter of the sample distribution substation, and construct the feature vector of each sample node in the sample node set.

[0089] In step a3, a model structure of a self-supervised graph attention network is established, in which the graph attention part is composed of graph attention layers. Multiple attention heads are set in each graph attention layer for parallel calculation, and deep feature extraction is achieved by stacking multiple graph attention layers.

[0090] In this step, the adjacency matrix of the model structure of the established self-supervised graph attention network is fully connected. The model structure of the self-supervised graph attention network does not need to perform the topological relationship identification step of the sample distribution station area, that is, there is no need to obtain the topological structure of the sample distribution station area, and the initial impedance parameter between any two nodes in the sample node set is defined as the edge feature of the model structure of the self-supervised graph attention network.

[0091] The model stacks multiple attention layers and retains the original feature information through residual connections to prevent gradient disappearance, extracting local features and global correlation relationships of node neighborhoods layer by layer; the underlying network focuses on feature extraction of original measurement data, while the high-level network realizes the abstract representation of cross-node electrical coupling relationships, forming a hierarchical feature expression system.

[0092] The evolution mechanism from node features to edge features obtains the source node and target node according to the index to perform adjacent edge sampling, concatenates the adjacent node features and inputs them into the multi-layer perceptron. The softplus activation function (a smooth, nonlinear activation function) is used to constrain the real part of the impedance to be non-negative, while the imaginary part is allowed to be freely adjusted.

[0093] During the impedance matrix construction process, edge features are sequentially populated into a complex matrix of a preset dimension, based on the source and target node order. The matrix is ​​then symmetrized to enforce compliance with the reciprocity theorem. The preset dimensions can be, for example, 99×99 or 90×90, but this embodiment does not impose any restrictions on this.

[0094] In step a4, a fusion loss function is constructed that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss.

[0095] In step a5, a plurality of preset neural network parameter initialization strategies are used to perform differentiated initialization for different types of network layers, and the feature vectors of each sample node in the sample node set are input into the model structure of the self-supervised graph attention network for training. The preset optimizer is used to minimize the fusion loss function on the training node to achieve gradient optimization of the model structure of the self-supervised graph attention network. At the same time, the preset early stopping technology is used on the verification set to automatically trigger the training termination condition, and a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

[0096] In this step, the preset neural network parameter initialization strategies can be Xavier or Glorot initialization (an initialization that adaptively adjusts the input and output dimensions of the network layer), which is suitable for the weight matrix of the attention mechanism. The initialization range is adjusted according to the input and output dimensions to avoid gradient disappearance or explosion. It can also be He initialization (an adaptive variance initialization). If ReLU (a linear rectification function) or its variants are used as the activation function, He initialization is used to alleviate the asymmetry problem of neuron output.

[0097] The preset optimizer can be Adam optimizer, an adaptive moment estimator that can adapt the learning rate and is suitable for non-convex optimization problems of graph attention networks.

[0098] The preset early stopping technology may be, for example, early stopping based on verification loss or early stopping based on task indicators, and this embodiment does not impose any limitation on this.

[0099] This embodiment uses the substation distribution main secondary side and at least one user meter of the sample distribution station area as each sample node to construct a sample node set of the sample distribution station area, and uses the historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the substation distribution main secondary side and each user meter of the sample distribution station area as the feature vector of each sample node in the sample node set to construct input samples and use the model structure of the self-supervised graph attention network for training; during the training process, the model combines the fusion loss function of the three-phase voltage consistency, the three-phase injected active power and the three-phase injected reactive power error constraints, and the consistency regularization loss, and continuously adjusts the network parameters to learn the electrical coupling relationship between the nodes, and finally obtains the three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network.

[0100] In the embodiment of the present application, a possible implementation method is provided. In step a4 above, a fusion loss function is constructed by combining three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss. Specifically, it can be:

[0101] The formula for constructing the fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss is:

[0102] (1)

[0103] In formula (1), is the three-phase voltage consistency loss function, is the error constraint function of three-phase injected active power and three-phase injected reactive power, is the consistency regularization loss function, The preset weight coefficients can be set according to actual needs, and this embodiment does not limit this.

[0104] A possible implementation method is provided in the embodiment of the present application. Represents the constraint on the voltage amplitude error of each phase, and the total loss is the average of the three phases:

[0105] (2)

[0106] In formula (2), n is the total number of nodes, p is the phase, A 、 B 、 C is the phase value, and A 、 B 、 C It forms three phases, The nodes collected i of p Phase voltage amplitude, and Node i The real and imaginary parts of the predicted voltage are calculated as follows:

[0107] According to Kirchhoff's law, For nodes i of p Phase complex voltage satisfies formula (3):

[0108] (3)

[0109] In formula (3), For nodes i, j between pm Phase impedance value, For nodes j of m Phase injected complex current; For nodes j of mThe complex power injected into the phase satisfies equations (4) and (5):

[0110] (4)

[0111] (5)

[0112] In formula (4), For nodes j of m Phase complex voltage, represents conjugation;

[0113] In formula (5), The nodes collected j of m Phase injected active power, w is the imaginary unit, The nodes collected j of m Phase injected reactive power;

[0114] According to formula (4) and formula (5), we can get:

[0115] (6)

[0116] In formula (6), The nodes collected j of m Phase voltage amplitude, For nodes j of m Phase angle;

[0117] Assume the secondary side of the substation distribution system is the reference node. i =1; the user's meter is a non-reference node, i =2,3,… n ;

[0118] definition p=A The phase angle of the phase voltage is 0, that is, in formula (6) =0, then:

[0119] , then formula (6) is converted into formula (7):

[0120] (7)

[0121] Substituting equation (7) into equation (3), we can separate The real and imaginary parts of , we get formula (8):

[0122] (8)

[0123] In formula (8), Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase resistance value, Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase reactance value;

[0124] According to formula (8), we can get formula (9):

[0125] (9)

[0126] According to formula (9), we can get formula (10) and formula (11):

[0127] (10)

[0128] (11).

[0129] According to Kirchhoff's law, the node i Predict the real part of the voltage and the imaginary part , and then calculate according to formula (2) , the voltage amplitude error of each phase is constrained, so that the fusion loss function has physical constraints and no real three-phase impedance matrix labels are required to train the model.

[0130] A possible implementation method is provided in the embodiment of the present application. Constrain the three-phase injected active power and three-phase injected reactive power errors, as shown in the following formula (12):

[0131] (12)

[0132] In formula (12), The predicted node i of p Phase injected active power, The nodes collected i of p Phase injected active power, The predicted node i of p Phase injected reactive power, The nodes collected i of p Phase injected reactive power; obtained by presetting the known three-phase power flow calculation formula and .

[0133] A possible implementation method is provided in the embodiment of the present application. The constraint model maintains consistency before and after the input is disturbed by noise, which enhances the model's robustness to noise. The formulas are as follows (13) and (14):

[0134] (13)

[0135] (14)

[0136] In formula (13), Represents the model prediction results, represents the model prediction result after disturbance, represents the node feature perturbation, x The feature vector representing the node;

[0137] In formula (14), It conforms to a Gaussian distribution with mean 0.

[0138] A possible implementation method is provided in an embodiment of the present application. Step a2 selects the historical three-phase voltage amplitude, historical three-phase injection current amplitude, historical three-phase injection active power and historical three-phase injection reactive power of the substation distribution total secondary side and each user meter of the sample distribution substation, and constructs the characteristic vector of each sample node in the sample node set. Here, the historical three-phase voltage amplitude, historical three-phase injection current amplitude, historical three-phase injection active power and historical three-phase injection reactive power of the substation distribution total secondary side and each user meter of the sample distribution substation can be preprocessed to further construct the characteristic vector of each sample node in the sample node set.

[0139] Data preprocessing includes two parts: data cleaning and data normalization, which process the original data to improve the accuracy of model prediction.

[0140] (1) Data cleaning

[0141] When measurement devices in the distribution network collect and transmit measurement data, both the device itself and the external environment may affect the data. Therefore, raw measurement data must be cleaned to restore the true data as much as possible to ensure the quality of the input data for the self-supervised graph attention network model structure. Missing data at a certain moment can be filled using linear interpolation to avoid inconsistent feature vector dimensions.

[0142] In this embodiment, a nearly symmetrical compact support orthogonal wavelet denoising function can be used to perform preliminary processing on the measurement data. The nearly symmetrical compact support orthogonal wavelet has good symmetry and can reduce distortion when performing data denoising.

[0143] (2) Data normalization

[0144] In power system data analysis, using per-unit values ​​to normalize three-phase electrical quantities effectively eliminates dimensional differences, improves model convergence efficiency, and enhances the comparability of data across different voltage levels. Per-unit values ​​are a dimensionless relative value representation that normalizes data by dividing actual physical quantities (such as voltage, current, and power) by a preselected reference value.

[0145] The following describes the training process of the three-phase impedance matrix identification model for a distribution station area through a specific embodiment.

[0146] In this specific embodiment, a 0.4kV distribution system in a certain distribution area was selected as the experimental object. The three-phase measurement data of the secondary side outlet of the distribution area and the user meters (32) in the area were selected from October 10 to October 14, 2024 as the sample data set. Specifically, the data include: three-phase voltage amplitude (unit: V), three-phase injected current amplitude (unit: A), three-phase injected active power (unit: kW), and three-phase injected reactive power (unit: kvar). The sampling interval is 60 minutes, and 24 sample points are collected every day, that is, the number of data samples included is 120. Among them, the first 80% of the data is used as the training set and validation set, and the last 20% of the data is used as the test set to verify the model performance.

[0147] 1) Data preprocessing

[0148] Press each sample node A 、 B 、 C The phase is divided into 3 nodes, each sample has 99 nodes, and the characteristic vector of the sample node (in terms of node i of A phase as an example), , the characteristic matrix is ​​4×99 dimensional, and the three-phase electrical quantities are normalized using per-unit values.

[0149] 2) Edge Index

[0150] Set as a fully connected adjacency matrix (9081 dimensions), the edge index sequence is as follows, the first dimension represents the source node, and the second dimension represents the target node:

[0151]

[0152] 3) Model construction

[0153] The model structure of the self-supervised graph attention network has two attention layers, and the regularization coefficient is set to 0.4 (that is, 40% of the data in the sample is randomly set to zero to prevent the model from overfitting). A dynamic residual mapping layer is set after each attention layer to reduce the loss of original features. The node mask ratio is set to 20%, the initial learning rate is 0.005, and the learning rate scheduler is set to continuously adjust the feature vector of each node. x The change process is as follows Figure 3 As shown. Figure 3 middle," x before SSGAT1" indicates the feature vector x Before the first attention layer of the Self-Supervised Graph Attention Network (SSGAT) model structure; x after SSGAT1" indicates the feature vector x After the first attention layer of the self-supervised graph attention network; x before SSGAT2" indicates the feature vector x Before the second attention layer of the model structure of the self-supervised graph attention network; x after SSGAT2" indicates the feature vector x After the second attention layer of the self-supervised graph attention network model structure, tensor represents the tensor format.

[0154] 4) Result analysis

[0155] The training set loss function of the model is as follows Figure 4 As shown, in Figure 4 In the figure, the black curve represents the fusion loss (TotalLoss) , the blue curve represents the three-phase voltage consistency loss (Vol Loss) , the red curve represents the three-phase injected active power and three-phase injected reactive power error constraint loss (P Loss) , the green curve represents the consistency regularization loss (Cons Loss) The vertical axis represents the loss value, and the horizontal axis represents the training round (Epoch). The horizontal axis scale of the black curve is the same as that of the blue curve, red curve, and green curve. Figure 4 It can be seen that as the number of training rounds increases, the loss value decreases and tends to be stable. The training can be stopped, and finally a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

[0156] It should be noted that the order of execution of the steps in the above embodiments does not necessarily imply a specific order of execution. The order of execution of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. In practical applications, all possible implementation methods described above can be combined in any manner to form possible embodiments of the present application, and will not be described in detail here.

[0157] Based on the distribution station area three-phase impedance matrix identification method based on the self-supervised graph attention network provided in the above embodiments, based on the same inventive concept, the embodiment of the present application also provides a distribution station area three-phase impedance matrix identification device based on the self-supervised graph attention network.

[0158] Figure 5 This is a structural diagram of a three-phase impedance matrix identification device for a distribution station area based on a self-supervised graph attention network provided by an embodiment of the present application. Figure 5 As shown, the distribution station area three-phase impedance matrix identification device based on the self-supervised graph attention network can specifically include a first construction unit 510, an acquisition unit 520, a second construction unit 530 and an identification unit 540.

[0159] The first construction unit 510 is configured to construct a node set of the distribution substation to be identified, including the distribution substation secondary side and at least one user meter, using the distribution substation secondary side and the user meter as nodes;

[0160] An acquisition unit 520 is configured to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power, and three-phase injected reactive power of the secondary side of the distribution substation to be identified, and to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power, and three-phase injected reactive power of each user's electric meter in the distribution substation to be identified, which are acquired in real time.

[0161] The second constructing unit 530 is used to construct a feature vector of each node in the node set according to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution area to be identified and each user's electricity meter;

[0162] The identification unit 540 is used to input the feature vector of each node into the pre-trained distribution station area three-phase impedance matrix identification model, output the three-phase impedance matrix of the distribution station area to be identified, and mark it as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. nis the total number of nodes; the three-phase impedance matrix identification model of the distribution station area is based on the self-supervised graph attention network, and is trained by combining the fusion loss function of three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; the consistency regularization loss is used to constrain the model to maintain consistency before and after when the input is disturbed by noise.

[0163] A possible implementation method is provided in the embodiment of the present application, such as Figure 6 As shown above Figure 5 The device may further include a training unit 610 for obtaining a three-phase impedance matrix identification model of a distribution station area through training by the following steps:

[0164] The sample node set of the sample distribution substation is constructed by taking the substation secondary side and at least one user meter of the sample distribution substation as each sample node;

[0165] The historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the secondary side of the distribution station and each user meter of the sample distribution station area are selected to construct the feature vector of each sample node in the sample node set;

[0166] Establish a self-supervised graph attention network model structure, where the graph attention part is composed of graph attention layers. Multiple attention heads are set up in each graph attention layer for parallel calculation. Deep feature extraction is achieved by stacking multiple graph attention layers.

[0167] Construct a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss;

[0168] Using multiple preset neural network parameter initialization strategies, differentiated initialization is performed for different types of network layers. The feature vectors of each sample node in the sample node set are input into the model structure of the self-supervised graph attention network for training. The preset optimizer is used to minimize the fusion loss function on the training node to achieve gradient optimization of the model structure of the self-supervised graph attention network. At the same time, the preset early stopping technology is used on the validation set to automatically trigger the training termination condition, and a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

[0169] An embodiment of the present application provides a possible implementation method, wherein the training unit 610 is further configured to:

[0170] The substation distribution main secondary side and at least one user meter of the sample distribution substation are used as independent sample nodes, and the switch of the line between the substation distribution main secondary side and at least one user meter is not included to construct a sample node set of the sample distribution substation.

[0171] In an embodiment of the present application, a possible implementation method is provided, in which the adjacency matrix of the model structure of the established self-supervised graph attention network is fully connected, and the model structure of the self-supervised graph attention network does not require a topological relationship identification step of a sample distribution station area, that is, there is no need to obtain the topological structure of the sample distribution station area, and the initial impedance parameter between any two nodes in the sample node set is defined as the edge feature of the model structure of the self-supervised graph attention network.

[0172] An embodiment of the present application provides a possible implementation method, wherein the training unit 610 is further configured to:

[0173] The formula for constructing the fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss is:

[0174] (1)

[0175] In formula (1), is the three-phase voltage consistency loss function, is the error constraint function of three-phase injected active power and three-phase injected reactive power, is the consistency regularization loss function, are the preset weight coefficients respectively.

[0176] A possible implementation method is provided in the embodiment of the present application. Represents the constraint on the voltage amplitude error of each phase, and the total loss is the average of the three phases:

[0177] (2)

[0178] In formula (2), n is the total number of nodes, p is the phase, A 、 B 、 C is the phase value, and A 、 B 、 C It forms three phases, The nodes collected i of p Phase voltage amplitude, and Node i The real and imaginary parts of the predicted voltage are calculated as follows:

[0179] According to Kirchhoff's law, For nodes i of p Phase complex voltage satisfies formula (3):

[0180] (3)

[0181] In formula (3), For nodes i, j between pm Phase impedance value, For nodes j of m Phase injected complex current; For nodes j of m The complex power injected into the phase satisfies equations (4) and (5):

[0182] (4)

[0183] (5)

[0184] In formula (4), For nodes j of m Phase complex voltage, represents conjugation;

[0185] In formula (5), The nodes collected j of m Phase injected active power, w is the imaginary unit, The nodes collected j of m Phase injected reactive power;

[0186] According to formula (4) and formula (5), we can get:

[0187] (6)

[0188] In formula (6), The nodes collected j of m Phase voltage amplitude, For nodes j of m Phase angle;

[0189] Assume the secondary side of the substation distribution system is the reference node. i =1; the user's meter is a non-reference node, i =2,3,… n ;

[0190] definition p=A The phase angle of the phase voltage is 0, that is, in formula (6) =0, then:

[0191] , then formula (6) is converted into formula (7):

[0192] (7)

[0193] Substituting equation (7) into equation (3), we can separate The real and imaginary parts of , we get formula (8):

[0194] (8)

[0195] In formula (8), Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase resistance value, Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase reactance value;

[0196] According to formula (8), we can get formula (9):

[0197] (9)

[0198] According to formula (9), we can get formula (10) and formula (11):

[0199] (10)

[0200] (11).

[0201] A possible implementation method is provided in the embodiment of the present application. Constrain the three-phase injected active power and three-phase injected reactive power errors, as shown in the following formula (12):

[0202] (12)

[0203] In formula (12), The predicted node i of p Phase injected active power, The nodes collected i of p Phase injected active power, The predicted node i of p Phase injected reactive power, The nodes collected i of p Phase injected reactive power; obtained by presetting the known three-phase power flow calculation formula and .

[0204] A possible implementation method is provided in the embodiment of the present application. The constraint model maintains consistency before and after the input is disturbed by noise, which enhances the model's robustness to noise. The formulas are as follows (13) and (14):

[0205] (13)

[0206] (14)

[0207] In formula (13), Represents the model prediction results, represents the model prediction result after disturbance, represents the node feature perturbation, x The feature vector representing the node;

[0208] In formula (14), It conforms to a Gaussian distribution with mean 0.

[0209] Based on the same inventive concept, an embodiment of the present application also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the above embodiments of the three-phase impedance matrix identification method of the distribution station area based on the self-supervised graph attention network.

[0210] In an exemplary embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7 The electronic device 700 shown includes a processor 701 and a memory 703. The processor 701 and the memory 703 are connected, for example, via a bus 702. Optionally, the electronic device 700 may further include a transceiver 704. It should be noted that in actual applications, the number of transceivers 704 is not limited to one, and the structure of the electronic device 700 does not constitute a limitation on the embodiments of the present application.

[0211] Processor 701 may be a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 701 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0212] Bus 702 may include a path for transmitting information between the above components. Bus 702 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 702 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0213] The memory 703 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0214] The memory 703 is used to store computer program codes for executing the solution of the present application, and the execution is controlled by the processor 701. The processor 701 is used to execute the computer program codes stored in the memory 703 to implement the contents shown in the above method embodiments.

[0215] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0216] Those skilled in the art will clearly understand that the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they will not be further described here.

[0217] Those skilled in the art will understand that the technical solution of the present application, in essence, or in whole or in part, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes a number of program instructions that cause an electronic device (such as a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application when the program instructions are executed. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0218] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware related to program instructions (such as electronic devices such as personal computers, servers, or network devices), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of an electronic device, the electronic device executes all or part of the steps of the methods described in the various embodiments of the present application.

[0219] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that, within the spirit and principles of the present application, they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate from the protection scope of the present application.

Claims

1. A method for identifying three-phase impedance matrix of distribution station area based on self-supervised graph attention network, characterized in that: The method comprises: Taking the secondary side of the distribution center and the user's electric meter as nodes, a node set including the secondary side of the distribution center and at least one user's electric meter of the distribution center to be identified is constructed; Acquire the real-time collected three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution substation to be identified, and acquire the real-time collected three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of each user meter in the distribution substation to be identified; According to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution substation to be identified and each user's electricity meter, the characteristic vector of each node in the node set is constructed; The feature vector of each node is input into the pre-trained distribution substation three-phase impedance matrix identification model, and the three-phase impedance matrix of the distribution substation to be identified is output and marked as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. n is the total number of nodes; the three-phase impedance matrix identification model for the distribution station area is trained based on a self-supervised graph attention network and a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss. The consistency regularization loss is used to constrain the model to maintain consistency when noise perturbations occur in the input. The three-phase impedance matrix identification model of the distribution station area is obtained through training through the following steps: The sample node set of the sample distribution substation is constructed by taking the substation secondary side and at least one user meter of the sample distribution substation as each sample node; The historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the secondary side of the distribution station and each user meter of the sample distribution station area are selected to construct the feature vector of each sample node in the sample node set; Establish a self-supervised graph attention network model structure, where the graph attention part is composed of graph attention layers. Multiple attention heads are set up in each graph attention layer for parallel calculation. Deep feature extraction is achieved by stacking multiple graph attention layers. Construct a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; Using multiple preset neural network parameter initialization strategies, differentiated initialization is performed for different types of network layers. The feature vectors of each sample node in the sample node set are input into the model structure of the self-supervised graph attention network for training. The preset optimizer is used to minimize the fusion loss function on the training node to achieve gradient optimization of the model structure of the self-supervised graph attention network. At the same time, the preset early stopping technology is used on the validation set to automatically trigger the training termination condition, and a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

2. The method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to claim 1 is characterized in that: The substation secondary side and at least one user meter of the sample distribution substation are used as sample nodes to construct a sample node set of the sample distribution substation, including: The substation distribution main secondary side and at least one user meter of the sample distribution substation are used as independent sample nodes, and the switch of the line between the substation distribution main secondary side and at least one user meter is not included to construct a sample node set of the sample distribution substation.

3. The method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to claim 1 is characterized in that: The adjacency matrix of the model structure of the established self-supervised graph attention network is fully connected. The model structure of the self-supervised graph attention network does not require the topological relationship identification step of the sample distribution station area, that is, there is no need to obtain the topological structure of the sample distribution station area. The initial impedance parameter between any two nodes in the sample node set is defined as the edge feature of the model structure of the self-supervised graph attention network.

4. The method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to claim 1 is characterized in that: Construct a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss, including: The formula for constructing the fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss is: （1） In formula (1), is the three-phase voltage consistency loss function, is the error constraint function of three-phase injected active power and three-phase injected reactive power, is the consistency regularization loss function, are the preset weight coefficients respectively.

5. The method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to claim 4 is characterized in that: Represents the constraint on the voltage amplitude error of each phase, and the total loss is the average of the three phases: （2） In formula (2), n is the total number of nodes, p is the phase, A 、 B 、 C is the phase value, and A 、 B 、 C It forms three phases, The nodes collected i of p Phase voltage amplitude, and Node i The real and imaginary parts of the predicted voltage are calculated as follows: According to Kirchhoff's law, For nodes i of p Phase complex voltage satisfies formula (3): （3） In formula (3), For nodes i, j between pm Phase impedance value, For nodes j of m Phase injected complex current; For nodes j of m The complex power injected into the phase satisfies equations (4) and (5): （4） （5） In formula (4), For nodes j of m Phase complex voltage, represents conjugation; In formula (5), The nodes collected j of m Phase injected active power, w is the imaginary unit, The nodes collected j of m Phase injected reactive power; According to formula (4) and formula (5), we can get: （6） In formula (6), The nodes collected j of m Phase voltage amplitude, For nodes j of m Phase angle; Assume the secondary side of the substation distribution system is the reference node. i =1; The user's meter is a non-reference node. i =2,3,… n ; definition p=A The phase angle of the phase voltage is 0, that is, in formula (6) =0, then: , then formula (6) is converted into formula (7): （7） Substituting equation (7) into equation (3), we can separate The real and imaginary parts of , we get formula (8): （8） In formula (8), Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase resistance value, Nodes output by the model structure of the self-supervised graph attention network i, j between pm Phase reactance value; According to formula (8), we can get formula (9): （9） According to formula (9), we can get formula (10) and formula (11): （10） （11）。 6. The method for identifying three-phase impedance matrix of distribution station area based on self-supervised graph attention network according to claim 4 is characterized in that: Constrain the three-phase injected active power and three-phase injected reactive power errors, as shown in the following formula (12): （12） In formula (12), The predicted node i of p Phase injected active power, The nodes collected i of p Phase injected active power, The predicted node i of p Phase injected reactive power, The nodes collected i of p Phase injected reactive power; obtained by presetting the known three-phase power flow calculation formula and .

7. The method for identifying the three-phase impedance matrix of a distribution station area based on a self-supervised graph attention network according to claim 4 is characterized in that: The constraint model maintains consistency before and after the input is disturbed by noise, which enhances the model's robustness to noise. The formulas are as follows (13) and (14): （13） （14） In formula (13), Represents the model prediction results, represents the model prediction result after disturbance, represents the node feature perturbation, x The feature vector representing the node; In formula (14), It conforms to a Gaussian distribution with mean 0.

8. A three-phase impedance matrix identification device for distribution station area based on self-supervised graph attention network, characterized in that: The device comprises: The first construction unit is used to construct a node set of the distribution substation to be identified, including the distribution substation secondary side and at least one user meter, with the distribution substation secondary side and the user meter as nodes; An acquisition unit is used to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution station area to be identified, and to acquire the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of each user meter in the distribution station area to be identified; The second construction unit is used to construct a feature vector of each node in the node set according to the three-phase voltage amplitude, three-phase injected current amplitude, three-phase injected active power and three-phase injected reactive power of the secondary side of the distribution station area to be identified and each user's electricity meter; The identification unit is used to input the feature vector of each node into the pre-trained distribution substation three-phase impedance matrix identification model, output the three-phase impedance matrix of the distribution substation to be identified, and mark it as , is a 3 n ×3 n The real part of each element of the complex matrix is ​​the resistance value, and the imaginary part is the reactance value. n is the total number of nodes; the three-phase impedance matrix identification model for the distribution station area is trained based on a self-supervised graph attention network and a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss. The consistency regularization loss is used to constrain the model to maintain consistency when noise perturbations occur in the input. The device further includes a training unit for obtaining a three-phase impedance matrix identification model of a distribution station area through training by the following steps: The sample node set of the sample distribution substation is constructed by taking the substation secondary side and at least one user meter of the sample distribution substation as each sample node; The historical three-phase voltage amplitude, historical three-phase injected current amplitude, historical three-phase injected active power and historical three-phase injected reactive power of the secondary side of the distribution station and each user meter of the sample distribution station area are selected to construct the feature vector of each sample node in the sample node set; Establish a self-supervised graph attention network model structure, where the graph attention part is composed of graph attention layers. Multiple attention heads are set up in each graph attention layer for parallel calculation. Deep feature extraction is achieved by stacking multiple graph attention layers. Construct a fusion loss function that combines three-phase voltage consistency, three-phase injected active power and three-phase injected reactive power error constraints, and consistency regularization loss; Using multiple preset neural network parameter initialization strategies, differentiated initialization is performed for different types of network layers. The feature vectors of each sample node in the sample node set are input into the model structure of the self-supervised graph attention network for training. The preset optimizer is used to minimize the fusion loss function on the training node to achieve gradient optimization of the model structure of the self-supervised graph attention network. At the same time, the preset early stopping technology is used on the validation set to automatically trigger the training termination condition, and a three-phase impedance matrix identification model of the distribution station area based on the self-supervised graph attention network is obtained.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the three-phase impedance matrix identification method of the distribution station area based on the self-supervised graph attention network according to any one of claims 1 to 7.

Images