Low-voltage distribution network topology identification method and device, electronic equipment and storage medium

Abstract

The invention relates to a low-voltage power distribution network topology identification method and device, electronic equipment and a storage medium, and the method comprises the steps: complementing different measurement features in a preset training data set through employing a pre-constructed measurement complementation network model, so as to obtain the complemented measurement data; simulating the complemented measurement data to obtain a historical data set meeting a preset integrity condition under different topology conditions, and constructing an identification model training set by using the historical data set; and training a pre-constructed multi-topology identification network model by using the identification model training set to obtain a target multi-topology identification network model meeting a preset convergence condition, and performing topology identification of the target low-voltage power distribution network by using the target multi-topology identification network model to obtain a corresponding low-voltage topology identification result. Therefore, the technical problems that in the related technology, the efficiency is low and the real-time performance is poor when a large amount of calculation is carried out in a traditional method, and a topology identification method based on artificial intelligence depends on complete intelligent electric meter measurement data greatly and is poor in economy and practicability are solved.

Description

Technical Field

[0001] The present application relates to the technical field of electrical digital data processing, and particularly relates to a low-voltage distribution network topology identification method, device, electronic device, and storage medium. Background Art

[0002] Due to the large-scale integration of new energy sources such as photovoltaic into the low-voltage distribution network, the uncertainty of the system has increased significantly, and the safe and stable operation of the low-voltage distribution network faces severe challenges. To cope with the uncertainty brought by the access of new energy, it is urgent to dynamically update and accurately identify the topology of the low-voltage distribution network to ensure its safe and stable operation. The topology identification of the low-voltage distribution network provides key structural information for subsequent tasks such as line loss analysis, capacity allocation, and fault location. Therefore, developing a high-precision and high-efficiency topology identification technology for distribution networks with distributed energy sources is of great significance for the intelligent management of distribution networks.

[0003] However, at present, the topology information of most low-voltage distribution areas in China is not updated in a timely manner, the user phase information and topology structure are not clear, which seriously restricts the development of distribution area operation and maintenance work such as distributed energy planning, line loss management, and power supply maintenance. At the same time, the current collection of topology information in distribution areas mostly relies on manual operations, which not only consumes a large amount of manpower and material resources but also has low work efficiency. In addition, there are a variety of intelligent meters installed in the distribution area, with uneven quality and performance, and the collected data often contains noise and missing phenomena, which significantly reduces the accuracy of the topology identification method of related technologies.

[0004] At present, the low-voltage distribution network topology identification methods in related technologies are mainly divided into traditional methods and artificial intelligence-based methods. Traditional methods have obvious limitations in dealing with the two-way power flow caused by the access of distributed power sources and are difficult to meet the requirements of high-precision topology identification of distribution networks. In addition, these methods are inefficient in the face of a large amount of calculations and are difficult to achieve real-time performance. In contrast, the artificial intelligence-based topology identification method has stronger adaptability by mining the correlation between measurement data and topology structure and can achieve online real-time identification. However, the artificial intelligence identification method relies heavily on complete intelligent meter measurement data. When facing the situation of missing measurement data or poor data transmission quality, its accuracy and robustness decrease significantly, which limits its economy and practical application feasibility.

[0005] In summary, in related technologies, traditional methods are inefficient in the face of a large amount of calculations and are difficult to achieve real-time performance. The artificial intelligence-based topology identification method relies heavily on complete intelligent meter measurement data and has poor economy and practicability, and needs to be improved. Summary of the Invention

[0006] The present application provides a method, apparatus, electronic device, and storage medium for identifying the topology of a low-voltage distribution network, so as to solve the technical problems in the related art that the traditional method is inefficient and has poor real-time performance when facing a large amount of calculations, and the topology identification method based on artificial intelligence relies heavily on complete smart meter measurement data and has poor economy and practicability.

[0007] In the first aspect of the embodiments of the present application, a method for identifying the topology of a low-voltage distribution network is provided, which is applied to the model construction stage. The method includes the following steps: using a pre-constructed measurement completion network model to complete different measurement features in a preset training dataset to obtain the completed measurement data; simulating the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies, and using the historical data set to construct an identification model training set; using the identification model training set to train a pre-constructed multi-topology identification network model to obtain a target multi-topology identification network model that meets the preset convergence conditions, so as to use the target multi-topology identification network model to identify the topology of the target low-voltage distribution network and obtain the corresponding low-voltage topology identification result.

[0008] Optionally, in an embodiment of the present application, the step of using a pre-constructed measurement completion network model to complete different measurement features in a preset training dataset to obtain the completed measurement data includes: obtaining measurement data of multiple low-voltage user meters, and generating the preset training set based on the measurement data of the multiple low-voltage user meters; inputting the preset training set into the measurement completion network model, encoding the data in the preset training set, and using multiple autoencoders to fill in the measurement data that meets the preset missing conditions to output the completed measurement data.

[0009] Optionally, in an embodiment of the present application, the step of simulating the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies and using the historical data set to construct an identification model training set includes: representing the topology structure corresponding to the completed measurement data as an undirected graph to obtain graph model data; using a multi-head attention mechanism to quantify the graph model data to obtain a feature association and information aggregation model between nodes; establishing a node state update mechanism based on the feature association and information aggregation model between nodes, so as to use the node state update mechanism to obtain the historical data set.

[0010] Optionally, in an embodiment of the present application, before training a pre-constructed multi-topology identification network model using the identification model training set, it further includes: constructing an initial topology identifier and performing a residual connection between the input features of the initial topology identifier and the output features of the current layer; based on the residual connection, introducing gate control in the initial topology identifier to obtain an actual topology identifier that meets the preset performance conditions; using the actual topology identifier to calculate the final characteristic matrix of the edges of the entire graph and determine the connection state between the nodes, so as to instantiate the autoencoder loss function of different measurement data; based on the topology identification task of the low-voltage distribution network, constructing a corresponding model loss function, and combining the model loss function and the instantiated autoencoder loss function to obtain the multi-topology identification network model.

[0011] An embodiment of the second aspect of the present application provides a low-voltage distribution network topology identification method, which is applied to the model usage stage. The method includes the following steps: obtaining the current measurement data of the target low-voltage distribution network; inputting the current measurement data into a pre-constructed multi-topology identification network model to obtain the complete measurement data and topology identification result of the target low-voltage distribution network, where the multi-topology identification network model is trained from graph model data.

[0012] An embodiment of the third aspect of the present application provides a low-voltage distribution network topology identification device, which is applied to the model construction stage. The device includes: a completion module for using a pre-constructed measurement completion network model to complete different measurement features in a preset training dataset to obtain the completed measurement data; a first construction module for simulating the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies, and using the historical data set to construct an identification model training set; a training module for training a pre-constructed multi-topology identification network model using the identification model training set to obtain a target multi-topology identification network model that meets the preset convergence conditions, and using the target multi-topology identification network model to perform topology identification of the target low-voltage distribution network to obtain the corresponding low-voltage topology identification result.

[0013] Optionally, in an embodiment of the present application, the completion module includes: an acquisition unit for acquiring multiple low-voltage user electricity meter measurement data and generating the preset training set based on the multiple low-voltage user electricity meter measurement data; a completion unit for inputting the preset training set into the measurement completion network model, encoding the data in the preset training set, and using multiple autoencoders to fill in the measurement data that meets the preset missing conditions to output the completed measurement data.

[0014] Optionally, in an embodiment of the present application, the first construction module includes: a processing unit, configured to represent the topological structure corresponding to the complemented measurement data as an undirected graph to obtain graph model data; a quantization unit, configured to use a multi-head attention mechanism to quantize the graph model data to obtain a feature association and information aggregation model between nodes; and an update unit, configured to establish a node state update mechanism based on the feature association and information aggregation model between nodes to obtain the historical data set by using the node state update mechanism.

[0015] Optionally, in an embodiment of the present application, it further includes: a second construction module, configured to construct an initial topology identifier and perform a residual connection between the input features of the initial topology identifier and the output features of the current layer; a processing module, configured to introduce gate control in the initial topology identifier based on the residual connection to obtain an actual topology identifier that meets preset performance conditions; a calculation module, configured to use the actual topology identifier to calculate a final characteristic matrix of the edges of the entire graph and determine the connection state between the nodes to instantiate the autoencoder loss function of the different measurement data; and a third construction module, configured to construct a corresponding model loss function based on the topology identification task of the low-voltage distribution network, and instantiate the multi-topology identification network model by combining the model loss function and the instantiated autoencoder loss function.

[0016] An embodiment of the fourth aspect of the present application provides a low-voltage distribution network topology identification device, which is applied to the model usage stage. The device includes: an acquisition module, configured to acquire current measurement data of a target low-voltage distribution network; and an identification module, configured to input the current measurement data into a pre-constructed multi-topology identification network model to obtain complete measurement data and a topology identification result of the target low-voltage distribution network, where the multi-topology identification network model is trained by graph model data.

[0017] An embodiment of the fifth aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the low-voltage distribution network topology identification method as described in the above embodiments.

[0018] An embodiment of the sixth aspect of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the low-voltage distribution network topology identification method as described in the above embodiments.

[0019] An embodiment of the seventh aspect of the present application provides a computer program product, including a computer program, which is used to implement the above low-voltage distribution network topology identification method when executed.

[0020] The embodiments of the present application can complete the missing data to obtain the completed measurement data, so as to address the problem of missing measurement data in the low-voltage distribution network, and simulate the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies, and use the historical data set to construct an identification model training set to train the pre-constructed multi-topology identification network model, and then obtain the target multi-topology identification network model that meets the preset convergence conditions, realizing the topology identification of the low-voltage distribution network, with improved practicability, and can accurately assist in identifying the actual connection status of the low-voltage distribution network, improving the identification efficiency and accuracy. Thus, it solves the technical problems in the related art that the traditional method is inefficient and has poor real-time performance when facing a large amount of calculations, while the topology identification method based on artificial intelligence has a large dependence on complete smart meter measurement data and poor economy and practicability.

[0021] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, wherein:

[0023] Figure 1 is a flowchart of a method for identifying the topology of a low-voltage distribution network according to an embodiment of the present application;

[0024] Figure 2 is a schematic diagram of the principle of a method for identifying the topology of a low-voltage distribution network according to an embodiment of the present application;

[0025] Figure 3 is a schematic diagram of the structure of a device for identifying the topology of a low-voltage distribution network according to an embodiment of the present application;

[0026] Figure 4 is a flowchart of another method for identifying the topology of a low-voltage distribution network according to an embodiment of the present application;

[0027] Figure 5 is a schematic diagram of the structure of another device for identifying the topology of a low-voltage distribution network according to an embodiment of the present application;

[0028] Figure 6 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, but should not be construed as limiting the present application.

[0030] The following describes a low-voltage distribution network topology identification method, apparatus, electronic device, and storage medium according to embodiments of the present application. In view of the technical problems in the related art mentioned in the above background art, that is, the traditional method is inefficient and has poor real-time performance when facing a large amount of calculations, and the topology identification method based on artificial intelligence relies heavily on complete smart meter measurement data, with poor economy and practicability, the present application provides a low-voltage distribution network topology identification method. In this method, missing data can be completed to obtain the completed measurement data to address the problem of missing measurement data in the low-voltage distribution network, and the completed measurement data is simulated to obtain a historical data set that meets the preset integrity conditions under different topologies, and the historical data set is used to construct an identification model training set to train a pre-constructed multi-topology identification network model, and then a target multi-topology identification network model that meets the preset convergence conditions is obtained to achieve the topology identification of the low-voltage distribution network, with improved practicability, and can accurately assist in identifying the actual connection status of the low-voltage distribution network, improving the identification efficiency and accuracy. Thus, the technical problems in the related art, that is, the traditional method is inefficient and has poor real-time performance when facing a large amount of calculations, and the topology identification method based on artificial intelligence relies heavily on complete smart meter measurement data, with poor economy and practicability, are solved.

[0031] Specifically, Figure 1 is a flowchart of a low-voltage distribution network topology identification method provided by an embodiment of the present application.

[0032] As Figure 1 shown, this low-voltage distribution network topology identification method is applied to the model construction stage, and the method includes the following steps:

[0033] In step S101, a pre-constructed measurement completion network model is used to complete different measurement features in a preset training data set to obtain the completed measurement data.

[0034] In the actual execution process, based on a preset training set, such as low-voltage user smart meter measurement data, embodiments of the present application can use the measurement completion network model to complete the missing of different measurement features in the distribution network respectively. The data completion process of the measurement completion network model will be described below.

[0035] Optionally, in an embodiment of the present application, a pre-constructed measurement completion network model is used to complete different measurement features in a preset training dataset to obtain the completed measurement data, including: obtaining measurement data of multiple low-voltage user meters, and generating a preset training set based on the measurement data of the multiple low-voltage user meters; inputting the preset training set into the measurement completion network model, encoding the data in the preset training set, and using multiple autoencoders to fill in the measurement data that meets the preset missing conditions to output the completed measurement data.

[0036] Herein, the data completion process of the measurement completion network model is elaborated.

[0037] The measurement completion network model of the embodiment of the present application can encode the measurement data of low-voltage user smart meters. The encoder fills all the missing values in the input measurement data with missing values with 0, extracts and captures the key information of the input features through a fully connected neural network layer, and outputs the compressed representation after encoding. The expression is as follows:

[0038] z (k) =σ(W 1 (k) x + b 1 (k) ),

[0039] Further, the embodiment of the present application can decode the encoded compressed high-dimensional data, input the compressed feature representation z (k) into the decoder part, and reconstruct the data through the decoder so that the output is close to the original measurement data (non-missing or completed features). This step generates the completed measurement feature data. The calculation method of this step is as follows:

[0040]

[0041] The activation function used is:

[0042]

[0043] wherein, σ in the network model represents the corresponding activation function, and z (k) is the compressed representation after encoding of the k-th layer; and are respectively the trainable weight parameter and bias of the l-th layer of the encoder; x is the measurement input feature; is the output vector after decoding of the m-th layer (the measurement value after data repair); and are respectively the trainable weight parameter and bias of the m-th layer of the decoder; h (m-1) is the output vector of the m-1 layer of the decoder.

[0044] In step S102, the complemented measurement data is simulated to obtain a historical data set that meets the preset integrity conditions under different topological situations, and the recognition model training set is constructed using the historical data set.

[0045] In the embodiment of the present application, the complemented measurement features can be converted into the representation mode of graph model data, and the recognition model training set is constructed for subsequent graph model training.

[0046] Optionally, in an embodiment of the present application, simulating the complemented measurement data to obtain a historical data set that meets the preset integrity conditions under different topological situations, and constructing a recognition model training set using the historical data set includes: representing the topological structure corresponding to the complemented measurement data as an undirected graph to obtain graph model data; quantifying the graph model data using the multi-head attention mechanism to obtain a feature association and information aggregation model between nodes; establishing a node state update mechanism based on the feature association and information aggregation model between nodes to obtain a historical data set using the node state update mechanism.

[0047] As a possible implementation manner, the embodiment of the present application can represent the low-voltage distribution network topological structure as an undirected graph G=(V, E), and the steps are as follows:

[0048] V represents the node set, and each node v i ∈V represents a user electricity meter or a distribution box, E represents the edge set, and each edge e ij ∈E represents the connection relationship between node v i and node v j (the line connection between the electricity meter boxes), which is defined as the node feature matrix X∈i N×F , constructing the adjacent relationship into an adjacency matrix A, where N is the number of nodes and F is the feature dimension of each node. For each node v i , its feature vector x i ∈i F represents the complemented measurement data, that is, X={x 1 , x 2 ,…x N}. Among them, the adjacency matrix represents the connection relationship between nodes using the adjacency matrix A∈i N×N . If there is an edge e i between node v j and v ij , then A ij = 1, otherwise A ij = 0.

[0049] Furthermore, the embodiment of the present application can establish an attention network mechanism model:

[0050] Embodiments of this application can quantify the intensity of information transmission between adjacent nodes through a multi-head attention mechanism. For each pair of adjacent nodes v i and v j , the attention coefficient α ij is calculated to quantify the information transmission intensity from node v j to node v i . Its calculation involves the generation and combination of query vectors, key vectors, and relation vectors. In the model, the query vector encodes the feature of node i into a query vector through a weight matrix obtained through training and a bias . The key vector encodes the feature of the target node j into a key vector using a weight matrix and a bias . The relation vector encodes the edge feature e ij using a weight matrix W c,e and a bias b c,e to obtain the relation vector of this edge. The model calculation method is as follows:

[0051]

[0052] Finally, combining the above, the calculation of the attention coefficient of the graph model is as follows:

[0053]

[0054] Among them, is the learnable weight vector of layer l, N(i) is the neighbor set of node v i . The combination of query vectors, key vectors, and relation vectors is used to calculate the node attention coefficient is to splice the query vector, key vector, and relation vector into a comprehensive representation to achieve weighted aggregation of information between nodes. The attention score is obtained through dot product operation and activation function. Finally, LeakyReLU is the activation function with a negative slope set to 0.2, where exp(x) = e x .

[0055] Furthermore, embodiments of this application can establish a node state update mechanism:

[0056] Use the attention coefficient to update the feature vector of each node. The calculation formula is as follows:

[0057]

[0058] Among them, x i is the node v iUpdated feature representation; introducing K independent attention heads to capture diverse relationships between nodes, and the multi-head attention aggregation method is as follows:

[0059]

[0060] where ∥ represents the vector concatenation operation, and W (k) are the parameters of the k-th attention head.

[0061] Through the above calculations, the updated feature matrix X′ = {x 1′ , x 2′ , … x N′} of each node is obtained, and this matrix captures the information of the nodes and their neighbor relationships in the topological structure.

[0062] In step S103, the pre-constructed multi-topology identification network model is trained using the recognition model training set to obtain a target multi-topology identification network model that meets the preset convergence conditions, so as to use the target multi-topology identification network model to perform topology identification of the target low-voltage distribution network and obtain the corresponding low-voltage topology identification result.

[0063] Embodiments of the present application can use the recognition model training set to train the model, so as to obtain a target multi-topology identification network model, input real-time limited measurement data into the model, and output complete measurement data and low-voltage topology identification results.

[0064] Optionally, in an embodiment of the present application, before training the pre-constructed multi-topology identification network model using the recognition model training set, it further includes: constructing an initial topology identifier, and performing residual connection on the input feature of the initial topology identifier and the output feature of the current layer; based on the residual connection, introducing gate control in the initial topology identifier to obtain an actual topology identifier that meets the preset performance conditions; using the actual topology identifier to calculate the final characteristic matrix of the edges of the entire graph and determine the connection state between nodes, so as to instantiate the autoencoder loss function of different measurement data; based on the topology identification task of the low-voltage distribution network, constructing a corresponding model loss function, and combining the model loss function and the instantiated autoencoder loss function to obtain a multi-topology identification network model.

[0065] Here, the construction of the multi-topology identification network model is described.

[0066] Embodiments of the present application can add gated residual connections between layers of the topology identifier to improve the effectiveness of layer structure and information transmission. The model calculation process is as follows:

[0067] Perform residual connection on the input feature and the output feature of the current layer. The residual connection formula can be expressed as:

[0068]

[0069] Based on the residual connection, the embodiments of the present application can further control the information flow through a gating mechanism to avoid the propagation of invalid information. A gating weight g is introduced to adjust the ratio of the input feature to the output feature of the current layer:

[0070]

[0071] Among them, the gating weight g is usually controlled by a sigmoid activation function, and its calculation can be:

[0072]

[0073] Among them, W g and b g are trainable parameters.

[0074] The embodiments of the present application can use the output after the gated residual connection as the input of the next layer for subsequent hierarchical calculations. This structure can maintain the effective transmission of information in a deep network, improve the accuracy and stability of topology identification, and obtain better performance in the topology identification task by introducing a controllable information path in each layer of the topology identifier.

[0075] Furthermore, for the calculation of the final feature matrix of the whole graph edge and the determination of the connection state, and the instantiation of the autoencoder loss function for different measurement types, the embodiments of the present application can calculate the final feature matrix of the whole graph edge and determine the connection state. Let h i and h j be the feature vectors of the head node and the tail node respectively. The construction method of the final feature matrix of the whole graph edge is as follows:

[0076] Δh i,j = h i - h j .

[0077] The head and tail node feature vectors formed by concatenating the head and tail node features are expressed as:

[0078] H i,j = concat(h i , h j ).

[0079] Use the original feature of the line and the complemented feature to splice into a complete line feature vector:

[0080]

[0081] Combine these vectors to form the final feature matrix F of the whole graph edges:

[0082] F = concat(Δh i,j , H i,j , E i,j ).

[0083] Let the line connection status be S ij , for each line, the determination method of mapping the features to the line status through the Sigmoid function:

[0084] S ij = σ(F ij ).

[0085] After that, determine the line status, compare the determination value with the set threshold. If S ij > threshold, it is determined as 1 (a certain user is in a connected state with the transformer).

[0086] In the embodiments of this application, a model loss function can be set. The expression of the loss function of the low-voltage distribution network topology identification model is as follows:

[0087] The target model adopts a loss function, which is optimized for different sub-tasks including missing measurement value completion, topology identification, and measurement completion. The calculation method of the loss function is as follows:

[0088]

[0089] Among them, T represents the total number of tasks, L t is the loss function of the t-th task, and λ t is the weight coefficient corresponding to the task, which is used to adjust the importance of each task in the total loss.

[0090] For the measurement completion task, the mean square error (MSE) is used as the loss function, and the calculation method is:

[0091]

[0092] Among them, N is the number of samples, y i is the true value of the i-th sample, is the predicted value after completion.

[0093] For the topology identification task of the low-voltage distribution network, the cross-entropy loss function is used to judge the phase transition relationship of low-voltage users and the corresponding status of the user-transformer relationship; the calculation method is as follows:

[0094]

[0095] Among them, y iis the true label (0 indicates that two nodes are connected, and 1 indicates not connected), is the decision probability predicted by the model.

[0096] The final loss function combination is:

[0097] L = λ topo L CE + λ comp L MSE ,

[0098] where λ topo and λ comp are weight coefficients, which are respectively used to adjust the weights of the topology identification task of the low-voltage distribution network and the task of complementing the data of intelligent measurement devices.

[0099] Combined with Figure 2 shown, the working principle of the low-voltage distribution network topology identification method in the embodiments of the present application is described with an embodiment.

[0100] The embodiments of the present application can use a variety of pre-constructed autoencoder models to complement the missing data of different types of measurement features (voltage, power) in the low-voltage distribution network. The autoencoder fills the missing data with zero values and extracts the key information of the input features through a fully connected layer, and outputs a compressed representation of the features, thereby complementing the phase relationship data of the low-voltage distribution network users; the complemented low-voltage distribution network topology structure is represented as a graph model. The nodes in this graph model represent electricity meters or distribution boxes, the edges represent user connection relationships, and the complemented measurement features are used as node features to form graph structure data. The graph structure data is input into a pre-constructed graph neural network topology identifier, and the message aggregation process from the source node to the target node in the graph is calculated using the multi-head attention mechanism, where the source node represents the currently calculated node and the target node represents other nodes in the graph. The aggregation of information is controlled by the attention coefficient, so that the model can better capture the association characteristics between nodes; in the graph, the features of each edge are calculated according to the node association matrix. First, calculate the difference between the updated head and tail node feature vectors, and splice it with the sum of the head and tail node feature vectors and the two-dimensional features of the line itself to form the final feature matrix of the edge, which characterizes the connection characteristics of the edges of the entire graph; each line feature in the final feature matrix of the edge is input into a fully connected neural network layer, and the features are mapped to the state determination values of each line. By setting the threshold of the determination value, the connection relationship between the user and the substation transformer is judged, and the identification of the low-voltage distribution network topology structure is realized; after the model training is completed, the real-time measurement data is input into the model, and the topology identification result is automatically output to assist in judging the actual connection structure of the current distribution network.

[0101] As Figure 2 shown, the embodiments of the present application may include the following steps in actual execution:

[0102] Step S1: Build a measurement data repair model for low-voltage electricity meters, and determine the hyperparameters of the network model and the model training settings: First, input the constructed training dataset into the measurement completion network model based on the autoencoder, and perform encoding dimensionality reduction and decoding reconstruction operations on different measurement types to obtain the repaired power, voltage, and power data of the intelligent electricity meters in the low-voltage substation area.

[0103] Step S2: Based on the complete actual electricity meter power data, use the pandapower power system simulation analysis software package in python to calculate the power flow results by automatically modifying the phase-household relationship and phase-change relationship in the substation area, generate a complete set of historical measurement data of the intelligent electricity meters in the low-voltage distribution network under different topological conditions, record the relevant topological representations, and construct an identification model training set.

[0104] Step S3: Build a network model for multi-topology identification and determine the hyperparameters of the network model;

[0105] Model the joint low-voltage distribution network topology as graph data for the graph neural network to learn, input it into the pre-built topology identifier, calculate the attention coefficient between the source node and the target node, and weight and aggregate the information of the source node and the target node. Take the mean value of the node features under multi-head attention as the output, and output the model topology through the fully connected neural network layer;

[0106] At the same time, consider the optimization objectives of the two stages of measurement completion and topology identification, construct the total loss function of the low-voltage distribution network topology identification model, use the adaptive moment estimation Adma optimization function for backpropagation to update the network, and evaluate the model to optimize the parameters;

[0107] Repeat the above steps until convergence, save the optimal model parameters, and obtain the target model.

[0108] Step S4: Input real-time limited measurement data into the model, and output complete measurement data and low-voltage topology identification results.

[0109] In summary, the embodiments of the present application can, based on the problem of missing measurement data in a low-voltage distribution network, use an autoencoder model to calculate the high-dimensional feature representation after missing data completion, solving the problem of reduced topology identification accuracy caused by incomplete data in traditional methods; then convert the low-voltage distribution network topology structure into a graph model according to the completed measurement features, effectively overcoming the technical problem in the prior art of difficultly accurately capturing the topological relationship of a complex distribution network; establish a feature association and information aggregation model between nodes through a graph attention network, significantly improving the robustness and accuracy of topology identification, and achieving the technical effect of quickly and accurately identifying the topology structure in a low-voltage distribution network with missing data and complex connections. In the topology identification stage, a multi-head attention mechanism is used to calculate the feature weights of nodes and edges in the graph, where the non-linear node association expression is dynamically calculated through the attention coefficients generated by training, and the edge feature matrix of the user connection line state is mapped and solved.

[0110] In the real-time operation stage, the embodiments of the present application can input real-time measurement data into the model, and quickly calculate the topology structure of the current distribution network based on the pre-trained graph attention network, accurately assisting in identifying the actual connection state of the low-voltage distribution network, and improving the identification efficiency and accuracy.

[0111] According to the low-voltage distribution network topology identification method proposed by the embodiments of the present application, missing data can be completed to obtain the completed measurement data to cope with the problem of missing measurement data in the low-voltage distribution network, and the completed measurement data is simulated to obtain a historical data set that meets the preset integrity conditions under different topologies, and the historical data set is used to construct an identification model training set to train the pre-constructed multi-topology identification network model, and then a target multi-topology identification network model that meets the preset convergence conditions is obtained, realizing the topology identification of the low-voltage distribution network, with improved practicability, and can accurately assist in identifying the actual connection state of the low-voltage distribution network, improving the identification efficiency and accuracy. Thereby, the technical problems in the related art are solved, where traditional methods are inefficient and have poor real-time performance in the face of a large amount of calculations, while the topology identification method based on artificial intelligence has a large dependence on complete smart meter measurement data and poor economy and practicability.

[0112] Next, a low-voltage distribution network topology identification device proposed according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0113] Figure 3 It is a block diagram of a low-voltage distribution network topology identification device according to an embodiment of the present application.

[0114] As Figure 3 shown, the low-voltage distribution network topology identification device 10 is applied to the model construction stage, where the device 10 includes: a completion module 101, a first construction module 102, and a training module 103.

[0115] Specifically, the completion module 101 is used to complete different measurement features in a preset training dataset by using a pre-constructed measurement completion network model to obtain the completed measurement data.

[0116] The first construction module 102 is used to simulate the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies, and construct an identification model training set by using the historical data set.

[0117] The training module 103 is used to train a pre-constructed multi-topology identification network model by using the identification model training set to obtain a target multi-topology identification network model that meets the preset convergence conditions, and use the target multi-topology identification network model to perform topology identification on the target low-voltage distribution network to obtain the corresponding low-voltage topology identification result.

[0118] Optionally, in an embodiment of the present application, the completion module 101 includes: an acquisition unit and a completion unit.

[0119] Among them, the acquisition unit is used to acquire multiple low-voltage user electricity meter measurement data and generate a preset training set based on the multiple low-voltage user electricity meter measurement data.

[0120] The completion unit is used to input the preset training set into the measurement completion network model, encode the data in the preset training set, and use multiple autoencoders to fill the measurement data that meets the preset missing conditions to output the completed measurement data.

[0121] Optionally, in an embodiment of the present application, the first construction module 102 includes: a processing unit, a quantization unit, and an update unit.

[0122] Among them, the processing unit is used to represent the topological structure corresponding to the completed measurement data as an undirected graph to obtain graph model data.

[0123] The quantization unit is used to quantify the graph model data by using the multi-head attention mechanism to obtain a feature association and information aggregation model between nodes.

[0124] The update unit is used to establish a node state update mechanism based on the feature association and information aggregation model between nodes to obtain a historical data set by using the node state update mechanism.

[0125] Optionally, in an embodiment of the present application, the low-voltage distribution network topology identification device 10 further includes: a second construction module, a processing module, a calculation module, and a third construction module.

[0126] Among them, the second construction module is used to construct an initial topology identifier and perform a residual connection between the input features of the initial topology identifier and the output features of the current layer.

[0127] A processing module, configured to introduce gate control in an initial topology identifier based on residual connection to obtain an actual topology identifier that meets preset performance conditions.

[0128] A calculation module, configured to calculate a final characteristic matrix of edges of the whole graph by using the actual topology identifier and determine connection states between nodes, so as to instantiate an autoencoder loss function of different measurement data.

[0129] A third construction module, configured to construct a corresponding model loss function based on a topology identification task of a low-voltage distribution network, and instantiate a multi-topology identification network model by combining the model loss function and the instantiated autoencoder loss function.

[0130] It should be noted that the foregoing explanation of the embodiments of the low-voltage distribution network topology identification method also applies to the low-voltage distribution network topology identification device of this embodiment, and details are not described herein again.

[0131] The low-voltage distribution network topology identification device provided according to an embodiment of the present application can complete missing data to obtain completed measurement data to cope with the problem of missing measurement data in the low-voltage distribution network, and simulate the completed measurement data to obtain a historical data set that meets preset integrity conditions in different topologies, and construct an identification model training set by using the historical data set to train a pre-constructed multi-topology identification network model, and further obtain a target multi-topology identification network model that meets preset convergence conditions, realizing topology identification of the low-voltage distribution network, with better practicability, and can accurately assist in identifying the actual connection state of the low-voltage distribution network, improving the identification efficiency and accuracy. Thereby, the technical problems in the related art that traditional methods are inefficient and have poor real-time performance when facing a large amount of calculations, and topology identification methods based on artificial intelligence rely heavily on complete smart meter measurement data and have poor economy and practicability are solved.

[0132] The above is an elaboration of the embodiment of the present application in the model construction stage. The following explains the working principle of the embodiment of the present application in the model usage stage.

[0133] Specifically, Figure 4 is a schematic flowchart of a low-voltage distribution network topology identification method provided by an embodiment of the present application.

[0134] As Figure 4 shown, this low-voltage distribution network topology identification method is applied to the model usage stage, and the method includes the following steps:

[0135] In step S401, current measurement data of a target low-voltage distribution network is obtained.

[0136] In step S402, the current measurement data is input into a pre-constructed multi-topology identification network model to obtain the complete measurement data and topology identification result of the target low-voltage distribution network, where the multi-topology identification network model is trained from graph model data.

[0137] According to the low-voltage distribution network topology identification method proposed in the embodiments of the present application, missing data can be complemented to obtain the complemented measurement data to address the problem of missing measurement data in the low-voltage distribution network, and the complemented measurement data is simulated to obtain a historical data set that meets the preset integrity conditions under different topologies, and the historical data set is used to construct an identification model training set to train the pre-constructed multi-topology identification network model, and then a target multi-topology identification network model that meets the preset convergence conditions is obtained to realize the topology identification of the low-voltage distribution network, with improved practicability, and it can accurately assist in identifying the actual connection state of the low-voltage distribution network, improving the identification efficiency and accuracy. Thus, the technical problems in the related art are solved, where traditional methods are inefficient and have poor real-time performance when facing a large amount of calculations, while the topology identification method based on artificial intelligence relies heavily on complete smart meter measurement data and has poor economy and practicability.

[0138] Next, a low-voltage distribution network topology identification device according to an embodiment of the present application is described with reference to the accompanying drawings.

[0139] Figure 5 It is a block diagram of the low-voltage distribution network topology identification device according to an embodiment of the present application.

[0140] As Figure 5 shown, the low-voltage distribution network topology identification device 20 is applied in the model usage stage, where the device 20 includes: an acquisition module 201 and an identification module 202.

[0141] Specifically, the acquisition module 201 is configured to acquire the current measurement data of the target low-voltage distribution network.

[0142] The identification module 202 is configured to input the current measurement data into a pre-constructed multi-topology identification network model to obtain the complete measurement data and topology identification result of the target low-voltage distribution network, where the multi-topology identification network model is trained from graph model data.

[0143] It should be noted that the foregoing explanation of the embodiments of the low-voltage distribution network topology identification method also applies to the low-voltage distribution network topology identification device of this embodiment, and will not be elaborated here.

[0144] The low-voltage distribution network topology identification device proposed according to the embodiments of the present application can complete the missing data to obtain the completed measurement data to address the problem of missing measurement data in the low-voltage distribution network, and simulate the completed measurement data to obtain a historical data set that meets the preset integrity conditions under different topologies, and use the historical data set to construct an identification model training set to train the pre-constructed multi-topology identification network model, and then obtain a target multi-topology identification network model that meets the preset convergence conditions, realizing the topology identification of the low-voltage distribution network, with improved practicability, and can accurately assist in identifying the actual connection status of the low-voltage distribution network, improving the identification efficiency and accuracy. Thus, it solves the technical problems in the related art that the traditional method is inefficient and has poor real-time performance in the face of a large amount of calculations, and the topology identification method based on artificial intelligence has a large dependence on complete smart meter measurement data and poor economy and practicability.

[0145] Figure 6 The structural schematic diagram of the electronic device provided by the embodiment of the present application. The electronic device may include:

[0146] A memory 601, a processor 602, and a computer program stored on the memory 601 and executable on the processor 602.

[0147] When the processor 602 executes the program, it implements the low-voltage distribution network topology identification method provided in the above embodiment.

[0148] Further, the electronic device further includes:

[0149] A communication interface 603 for communication between the memory 601 and the processor 602.

[0150] The memory 601 is used to store a computer program executable on the processor 602.

[0151] The memory 601 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0152] If the memory 601, the processor 602, and the communication interface 603 are implemented independently, the communication interface 603, the memory 601, and the processor 602 can be interconnected through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 only a thick line is used to represent it in Figure 6 , but it does not mean that there is only one bus or one type of bus.

[0153] Optionally, in a specific implementation, if the memory 601, the processor 602, and the communication interface 603 are integrated on a single chip, the memory 601, the processor 602, and the communication interface 603 can communicate with each other through an internal interface.

[0154] The processor 602 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0155] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned low-voltage distribution network topology identification method is implemented.

[0156] The embodiments of the present application also provide a computer program product, including a computer program, and when the computer program is executed by a processor, the low-voltage distribution network topology identification method provided by the embodiments of the present invention is implemented.

[0157] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0158] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically and clearly defined.

[0159] Any process or method description shown in a flowchart or described in other ways herein may be understood to represent a module, segment, or portion of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of this application includes additional implementations, where the functions may be executed in a manner that is not in the order shown or discussed, including in a substantially simultaneous manner according to the involved functions or in the reverse order, which should be understood by those skilled in the art to which the embodiments of this application pertain.

[0160] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or N wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0161] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0162] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0163] In addition, each functional unit in various embodiments of the present application may be integrated into one processing module, may exist separately physically for each unit, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0164] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A low voltage distribution network topology identification method, characterized in that: Applied to the model building stage, wherein the method comprises the following steps: Using a pre-built measurement completion network model to complete different measurement features in a preset training data set to obtain completed measurement data; Simulating the completed measurement data to obtain a historical data set that meets a preset integrity condition under different topological conditions, and using the historical data set to construct a recognition model training set; The pre-constructed multi-topology identification network model is trained using the identification model training set to obtain a target multi-topology identification network model that meets preset convergence conditions, and the target multi-topology identification network model is used to perform topology identification of the target low-voltage distribution network to obtain a corresponding low-voltage topology identification result.

2. The method according to claim 1, characterized in that The method of using the pre-built measurement completion network model to complete different measurement features in the preset training data set to obtain completed measurement data includes: Acquire measurement data of multiple low-voltage user electric meters, and generate the preset training set based on the measurement data of multiple low-voltage user electric meters; The preset training set is input into the measurement completion network model, the data in the preset training set is encoded, and the measurement data satisfying the preset missing conditions is filled by using a plurality of autoencoders to output the completed measurement data.

3. The method according to claim 1, characterized in that The simulating of the completed measurement data to obtain a historical data set satisfying a preset integrity condition under different topological conditions, and using the historical data set to construct a recognition model training set, includes: Representing the topological structure corresponding to the completed measurement data as an undirected graph to obtain graph model data; Quantify the graph model data using a multi-head attention mechanism to obtain a feature association and information aggregation model between nodes; A node status update mechanism is established based on the feature association and information aggregation model between the nodes, so as to obtain the historical data set by utilizing the node status update mechanism.

4. The method according to claim 3, characterized in that Before using the recognition model training set to train the pre-built multi-topology recognition network model, the method further includes: Constructing an initial topology identifier, and performing a residual connection between the input features of the initial topology identifier and the output features of the current layer; Based on the residual connection, gate control is introduced into the initial topology identifier to obtain an actual topology identifier that meets preset performance conditions; Utilizing the actual topology identifier to calculate the final characteristic matrix of the entire graph edge and determine the connection status between the nodes to instantiate the autoencoder loss function of the different measurement data; Based on the topology identification task of the low-voltage distribution network, a corresponding model loss function is constructed, and the multi-topology identification network model is instantiated by combining the model loss function and the autoencoder loss function.

5. A method for identifying low voltage distribution network topology, characterized in that: Applied to the model use phase, wherein the method comprises the following steps: Obtain current measurement data of the target low-voltage distribution network; The current measurement data is input into a pre-built multi-topology identification network model to obtain complete measurement data and topology identification results of the target low-voltage distribution network, wherein the multi-topology identification network model is trained by graph model data.

6. A low voltage distribution network topology identification device, characterized in that: Applied to the model building stage, wherein the device comprises: A completion module is used to use a pre-built measurement completion network model to complete different measurement features in a preset training data set to obtain completed measurement data; A construction module, used to simulate the completed measurement data to obtain a historical data set that meets a preset integrity condition under different topological conditions, and use the historical data set to construct a recognition model training set; A training module is used to use the recognition model training set to train a pre-constructed multi-topology identification network model to obtain a target multi-topology identification network model that meets preset convergence conditions, and to use the target multi-topology identification network model to perform topology identification of a target low-voltage distribution network to obtain a corresponding low-voltage topology identification result.

7. The device according to claim 6, characterized in that The completion module includes: An acquisition unit, configured to acquire measurement data of a plurality of low-voltage user electric meters, and generate the preset training set based on the measurement data of the plurality of low-voltage user electric meters; The completion unit is used to input the preset training set into the measurement completion network model, encode the data in the preset training set, and use multiple autoencoders to fill the measurement data that meets the preset missing conditions to output the completed measurement data.

8. A low voltage distribution network topology identification device, characterized in that: Applied to the model use stage, wherein the device comprises: An acquisition module, used to acquire current measurement data of the target low-voltage distribution network; An identification module is used to input the current measurement data into a pre-built multi-topology identification network model to obtain complete measurement data and topology identification results of the target low-voltage distribution network, wherein the multi-topology identification network model is trained by graph model data.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the low-voltage distribution network topology identification method as described in any one of claims 1-4 or 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the low-voltage distribution network topology identification method as described in any one of claims 1-4 or 5.

Images