Distributed energy system distribution network parameter identification method, device, equipment and medium

Abstract

The invention belongs to the technical field of intelligent power grids, and particularly relates to a distributed energy system distribution network parameter identification method and device, equipment and a medium. The method comprises the following steps: inputting topological structure information and node feature data of a distributed energy system distribution network into a distribution network parameter identification model to obtain a parameter identification result; the distribution network parameter identification model comprises a graph convolutional neural network layer and a long short-term memory network layer; performing feature extraction on the topological structure information and the node feature data by using a graph convolutional neural network layer to obtain a node feature matrix; and arranging the node feature matrixes according to a time sequence to serve as an input sequence of a long short-term memory network layer, and outputting a parameter identification result by the long short-term memory network layer. The method fully utilizes the advantages of the GCN in the aspect of processing graph structure data and the capability of the LSTM in the aspect of time sequence analysis, and achieves the accurate, rapid and adaptive identification of the distribution network parameters of the distributed energy system.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart grids, and specifically relates to a method, device, equipment and medium for identifying distribution network parameters of a distributed energy system. Background Art

[0002] With the continuous increase in global energy demand and the enhancement of environmental protection awareness, distributed energy systems are playing an increasingly important role in modern power networks. Distributed energy systems use renewable energy such as solar energy and wind energy to reduce dependence on traditional centralized power generation through decentralized power generation, improve energy efficiency, and reduce carbon emissions. The large-scale access of distributed energy also brings many challenges to the power system.

[0003] The output of renewable energy is intermittent and random. For example, solar power generation is affected by light intensity, and wind power generation is affected by wind speed changes, which makes the power generation of distributed energy difficult to predict and control. This uncertainty increases the difficulty of dispatching the power system, affecting the balance of supply and demand and the stability of the power grid. The distribution network has changed from a one-way power supply mode to a two-way power supply mode. The traditional distribution network design mainly considers power supply from the upper power grid to the lower load, and the power flow direction is single. After the distributed energy is connected, the power can flow from multiple directions, forming a complex two-way power flow. This change puts forward new requirements for the protection strategy, control method and operation management of the distribution network. In addition, the topology and operation status of the power system have become more complex. The access points and access capacities of distributed energy are different, resulting in dynamic changes in the distribution network structure and fluctuations in the electrical parameters of nodes and lines.

[0004] Traditional distribution network parameter identification methods rely on stable topology and known operating status, which makes it difficult to adapt to this complex and changing environment. Summary of the invention

[0005] The purpose of the present invention is to provide a distributed energy system distribution network parameter identification method, device, equipment and medium to solve the problem in the prior art that traditional distribution network parameter identification methods are difficult to adapt to parameter identification in distributed energy systems.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for identifying parameters of a distributed energy system distribution network, comprising: Obtain the topological structure information and node characteristic data of the distributed energy system distribution network; The topological structure information and node feature data are input into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features of the topological structure information and node feature data to obtain a node feature matrix; the node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

[0007] Furthermore, the topology information and node feature data are input into a pre-trained distribution network parameter identification model, and in the step of the distribution network parameter identification model outputting a parameter identification result, the distribution network parameter identification model is trained in the following manner: Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; Determine the loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; Get the pre-selected training set; Based on the training set and the loss function, the distribution network parameter identification model is trained, and after the preset conditions are met, a trained distribution network parameter identification model is obtained.

[0008] Furthermore, the loss function is expressed as follows:

[0009] in, For mission loss, is the multi-scale collaborative consistency loss, is the trade-off coefficient.

[0010] Furthermore, the multi-scale collaborative consistency loss is expressed as follows:

[0011] Among them, S is the number of scales, and s represents the sth scale; is the weight of the sth scale; They are the spatial domain consistency loss, temporal domain consistency loss and cross-domain consistency loss at the sth scale respectively.

[0012] Furthermore, each loss term in the multi-scale collaborative consistency loss is expressed as follows: Spatial domain consistency loss:

[0013] Temporal consistency loss:

[0014] Cross-domain collaborative consistency loss:

[0015] in, is the length of the time series at the sth scale; is the sth scale, l Spatial feature representation of layers; is the length of the time series at the sth scale; is the hidden state at time step t at the sth scale; It is the spatial feature output at time step t at the sth scale.

[0016] In a second aspect, the present invention provides a distributed energy system distribution network parameter identification device, comprising: A data acquisition module is used to obtain the topological structure information and node characteristic data of the distributed energy system distribution network; A parameter identification module is used to input the topological structure information and node feature data into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features of the topological structure information and node feature data to obtain a node feature matrix; the node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

[0017] Furthermore, in the parameter identification module, the distribution network parameter identification model is trained in the following manner: Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; Determine the loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; Get the pre-selected training set; Based on the training set and the loss function, the distribution network parameter identification model is trained, and after the preset conditions are met, a trained distribution network parameter identification model is obtained.

[0018] Furthermore, the loss function is expressed as follows:

[0019] in, For mission loss, is the multi-scale collaborative consistency loss, is the trade-off coefficient.

[0020] According to a third aspect of the present invention, an electronic device is provided, comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the above-mentioned method for identifying distribution network parameters of a distributed energy system.

[0021] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the distributed energy system distribution network parameter identification method as described above is implemented.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: In response to the problems raised in the prior art, the present invention proposes a distributed energy system distribution network parameter identification method based on a graph convolutional neural network (GCN) and a long short-term memory network (LSTM). This method makes full use of the advantages of GCN in processing graph structure data and the ability of LSTM in time series analysis to achieve accurate, fast and adaptive identification of distributed energy system distribution network parameters. A distributed energy system distribution network parameter identification device, electronic device and computer-readable storage medium provided by the present invention also solve the problems raised in the background technology section.

[0023] In this solution, a distribution network parameter identification model including a GCN layer and an LSTM layer is constructed, and training is performed in combination with a multi-scale collaborative consistency loss function. The method of the present invention not only improves the identification accuracy, but also enhances the generalization ability and adaptability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a flow chart of a method for identifying parameters of a distributed energy system distribution network in an embodiment of the present invention; Figure 2 Schematic diagram of the hidden layer of the GCN network in an embodiment of the present invention; Figure 3 Schematic diagram of the LSTM calculation process in an embodiment of the present invention; Figure 4 It is a structural block diagram of a distributed energy system distribution network parameter identification device according to an embodiment of the present invention; Figure 5 It is a structural block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0026] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0027] The present invention combines a graph convolutional neural network (GCN) with a long short-term memory network (LSTM). First, GCN is used to extract the spatial correlation characteristics between nodes. In the time dimension, the present invention introduces the LSTM model to capture the dynamic change characteristics of the distribution network operation status. The spatial features extracted by GCN are fused with the time features extracted by LSTM to realize the identification of the distribution network parameters of the distributed energy system. The output of the model includes the voltage, current and load changes of each node under disturbance conditions.

[0028] Example 1 like Figure 1 As shown, a distributed energy system distribution network parameter identification method includes: S1. Obtain the topological structure information and node characteristic data of the distributed energy system distribution network.

[0029] In step S1, since the variables generated during the operation and scheduling of the distributed energy system distribution network are complex and numerous, and multiple variables are directly dependent and indirectly dependent, using all variables will make the model coefficients unstable, and the coefficients of the model will become highly sensitive. Slight changes may cause large fluctuations in the results of the distributed energy system distribution network parameter prediction, and will also reduce the generalization ability of the model, resulting in good performance on the training set but completely opposite performance on the test set. Using principal component analysis, fewer new variables can replace the original more new variables, and these fewer new variables retain as much of the original information as possible, enhance the robustness and generalization of the model, and have a better processing effect on numerous and complex data sets.

[0030] Therefore, this scheme uses principal component analysis to reduce feature dimensionality, and the steps are: First calculate the covariance matrix of the standardized sample: (1) Among them, x is the variable indicator, the denominator is the standard deviation of the data sample characteristics, and the numerator is the sample mean difference of the data sample characteristics.

[0031] Calculate the eigenvalues ​​and eigenvectors of R: (2) The eigenvalues ​​λ1≥λ2≥⋯≥λp≥0, and R is a semi-positive definite matrix.

[0032] Calculate the principal component sharing rate and cumulative contribution rate: (3) (4) The principal component contribution rate and the cumulative contribution rate are two important indicators in principal component analysis (PCA), which are used to evaluate the explanatory power of each principal component to the total variance and help determine how many principal components to retain to effectively express the characteristics of the original data. The principal component contribution rate indicates the contribution of each principal component to the total variance of the data, that is, to what extent the principal component explains the variation of the data. The principal component contribution rate indicates to what extent the principal component captures the amount of information in the data. A higher contribution rate means that the principal component can better explain the variation of the data. The cumulative contribution rate refers to the sum of the contribution rates of the first K principal components, indicating how much percentage of the variance in the data these principal components explain in total. The cumulative contribution rate is used to measure the overall explanatory power of the first few principal components. A higher cumulative contribution rate (such as more than 85% or 90%) usually indicates that the first few principal components are sufficient to express the main information in the data.

[0033] (5) S2. Input the topological structure information and node feature data into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; use the graph convolutional neural network layer to extract features of the topological structure information and node feature data to obtain a node feature matrix; arrange the node feature matrix in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

[0034] In the above scheme, the Graph Convolutional Network (GCN) can effectively extract the topological structure information of the distribution network and capture the complex associations between nodes; the Long Short-Term Memory (LSTM) is good at processing time series data and capturing the dynamic changes in the system operation status. Therefore, the spatiotemporal fusion model combining GCN and LSTM can improve the accuracy and efficiency of distribution network parameter identification and disturbance signal identification, providing a new technical approach to solving the problems brought about by distributed energy systems.

[0035] In step S2, the distribution network parameter identification model is trained in the following manner: S21. Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; S22. Determine a loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; S23, obtaining a pre-selected training set; S24. Based on the training set and the loss function, the distribution network parameter identification model is trained to obtain a trained distribution network parameter identification model after a preset condition is met.

[0036] When combining the consistency constraints in the spatial domain and the temporal domain, the existing technology usually adopts a direct linear superposition method, that is, simply adding the spatial domain consistency loss, the temporal domain consistency loss and the cross-domain collaborative consistency loss. However, this method has obvious defects. First, the loss function of direct addition is prone to gradient conflicts during the optimization process, and the gradients of different loss items may interfere with each other, making it difficult for the model to converge or fall into a local optimum, affecting the performance of the model. Secondly, it is very difficult to weigh the weight coefficients of each loss item. Improper parameter setting will cause the model to be overly sensitive or ignored to a certain loss item, increase the complexity and uncertainty of model tuning, and reduce the versatility of the model. In addition, simple superposition may lead to information redundancy, increase computational overhead, and reduce the learning efficiency and prediction speed of the model. Due to the above problems, it is difficult for the existing technology to effectively integrate spatial and temporal features, and it is impossible to fully capture the complex dynamic characteristics of the distribution network, resulting in insufficient accuracy and robustness of parameter identification and interference signal identification, and it is difficult to meet the needs of practical applications of complex distributed energy systems.

[0037] Therefore, the loss function provided in step S22 is optimized in a hierarchical manner under a unified framework, combining the consistency constraints of the spatial domain, the temporal domain and the cross-domain. The loss function is expressed as follows:

[0038] in, For mission loss, is the multi-scale collaborative consistency loss, It is a trade-off coefficient used to balance the impact of task loss and consistency loss.

[0039] The multi-scale collaborative consistency loss is expressed as follows:

[0040] Among them, S is the number of scales, and s represents the sth scale; is the weight of the sth scale, which can be set or adaptively learned according to data characteristics and task requirements; They are the spatial domain consistency loss, temporal domain consistency loss and cross-domain consistency loss at the sth scale respectively.

[0041] The loss terms in the multi-scale collaborative consistency loss are expressed as follows: At the sth scale, the spatial features between different layers of GCN are constrained to be consistent, and the spatial domain consistency loss is: ; At the sth scale, the hidden states of LSTM at different times are smoothed and constrained, and the time domain consistency loss is: ; At the sth scale, the spatial features are fused with the temporal features to calculate the consistency loss and the cross-domain collaborative consistency loss: ; in, is the length of the time series at the sth scale; is the sth scale, l Spatial feature representation of layers; is the length of the time series at the sth scale; is the hidden state at time step t at the sth scale; It is the spatial feature output at time step t at the sth scale.

[0042] It should be noted that the spatial domain refers to the topological structure of the distribution network, that is, the connection relationship and spatial distribution between various nodes in the power grid (such as substations, transmission lines, distributed energy equipment, loads, etc.). In the distribution network, the electrical characteristics and physical connection methods between nodes have an important impact on power transmission and system operation. The concept of spatial domain consistency aims to ensure that the model maintains consistency in the spatial feature representation of the distribution network at different spatial scales and levels. This means: 1) Maintain the consistency of node features. At different layers or different spatial scales of the model, the feature representation of nodes should be kept uniform to prevent the loss of important spatial information during information transmission. 2) The model needs to accurately capture the topological relationship of the distribution network to ensure that the understanding of the connection method and electrical correlation between nodes is consistent at different levels. 3) In deep networks, spatial features may be distorted due to multiple transformations. The spatial domain consistency constraint can prevent this and maintain the integrity of spatial information. For example: In the parameter identification of the distribution network, parameters such as line impedance and node voltage are closely related to the network topology. Spatial domain consistency ensures that the model accurately reflects the relationship between these parameters and the topological structure at different levels.

[0043] The time domain involves the dynamic changes and time series data of the distribution network, including time-related phenomena such as load fluctuations, changes in distributed energy output, and the occurrence and disappearance of faults. The concept of time domain consistency aims to ensure that the model maintains consistency in the representation of time features at different time scales and time steps. Including: 1) Capturing time dynamic characteristics, the model needs to accurately reflect the changes in the operating status of the distribution network over time, and capture dynamic characteristics such as load changes and renewable energy fluctuations. 2) In time series, the data may contain noise or outliers. Time domain consistency constraints help smooth time features and prevent the model from being overly sensitive to short-term anomalies. 3) At different time scales (such as short-term, medium-term, and long-term), the model's understanding of time features should be consistent to avoid contradictory feature representations in different time windows. For example: In disturbance signal recognition, the occurrence of a fault may cause a sudden change in voltage and current. Time domain consistency ensures that the model can distinguish between normal fluctuations and abnormal disturbances and accurately identify fault signals.

[0044] Cross-domain consistency (also called spatiotemporal consistency) refers to the collaborative consistency between spatial features and temporal features. It emphasizes that in the model, the feature representations of the spatial domain and the temporal domain should be coordinated and supported with each other to achieve a comprehensive analysis of the spatiotemporal characteristics of the distribution network. The model needs to combine the topological information of the spatial domain with the dynamic changes of the temporal domain to form a comprehensive feature representation. At different scales, spatial features and temporal features should maintain consistent change trends to ensure that the model's understanding of events is unified in spatial and temporal dimensions. Many distribution network events (faults, disturbances) have both spatial and temporal characteristics, and cross-domain consistency helps the model accurately capture these spatiotemporal correlations. For example: When a fault occurs in a certain area of ​​the distribution network, the nodes near the fault point will show abnormal changes in the same time period. Cross-domain consistency ensures that the model can combine these spatial anomalies (node ​​location) and temporal anomalies (occurrence time) to accurately locate the location and time of the fault.

[0045] In an optional embodiment, Figure 2 As shown, the GCN layer performs feature extraction, including: (1) Model input parameters For the input of graph convolutional neural network, a graph G = (V, E) is required, where V is the number of vertices and E is the number of edges. On this basis, the input X = N×D is defined to represent the features of each node, where N is the number of nodes and D is the number of dimensions; and an adjacency matrix A of the graph is defined as input, hoping to obtain an N× F feature matrix Z (representing the features of each node learned), where F is the dimension that the final output can obtain.

[0046] (2) Graph Convolutional Layer For a network with L layers, it can be expressed as: (6) The traditional convolution algorithm can be described as adding the information of neighboring nodes to each node, and adjusting the corresponding weight on this node according to the importance of the neighboring nodes. Therefore, the operation can be converted into the following formula: (7) Here, σ represents the ReLU activation function. Each node in the above formula combines the information of adjacent nodes. However, since only the A matrix is ​​used as the adjacency matrix, the diagonal is 0 and the node's own information cannot be reflected. Therefore, a ring is added: A=A+I, which is equivalent to combining the node's own information. At the same time, since the A matrix is ​​not regularized, it is hoped that the sum of each row of the A matrix is ​​1, so we use , where D is Degree MatrixTherefore, the layer-wise propagation rule of the network can be written as: (8) This scheme uses a two-layer GCN for node classification. First, the grid node feature representation X is obtained and the adjacency matrix is ​​calculated. (9) Formula (9) is then input into the two-layer hidden layer, using a single weight matrix for each layer, and the different node degrees are handled by properly normalizing the adjacency matrix. Normalization makes the input features on the same scale, avoiding gradient problems caused by large differences in eigenvalues, thereby accelerating the convergence of model training. Normalization can also ensure numerical stability and reduce errors in numerical calculations, especially in neural networks, avoiding training problems caused by eigenvalues ​​that are too large or too small.

[0047] (10) In formula (10), is the weight matrix of the first layer, which is used to map the feature representation of the node to the corresponding hidden layer state. is the weight matrix of the second layer, which is used to map the hidden layer representation of the node to the corresponding output (F corresponds to the number of node labels).

[0048] (11) In an optional embodiment, LSTM time series modeling specifically includes: After completing the spatial feature extraction of the topological structure and node features of the distribution network by using the graph convolutional neural network (GCN), the present invention further introduces the long short-term memory network (LSTM) to model the time series data to capture the temporal dynamic characteristics of the distribution network operation status. The combination of LSTM and GCN enables the model to simultaneously process the spatial and temporal characteristics of the distribution network, and realizes the spatiotemporal fusion modeling of the distribution network parameter identification and disturbance signal identification of the distributed energy system.

[0049] (1) Model input parameters The output of the previous GCN is used as the input of this part of the LSTM. The node feature matrix extracted by GCN is , where N is the number of nodes and F is the feature dimension. For time series data, the tree pool H(t) of GCN at different times t is organized into a sequence ; where T is the number of time steps.

[0050] Construction of time series: For node-level time series, at each node v(i), the sequence of changes in its features over time ,in For the global feature sequence, the features of all nodes can be stacked in the time dimension to form an input tensor .

[0051] (2) LSTM model structure, such as Figure 3 As shown: Input gate calculation: ,in is the activation value of the input gate, and are the weight matrix and bias vector of the input gate respectively, is the hidden state at the previous moment, is the current input vector (i.e., the feature output by GCN , σ is the Sigmoid activation function.

[0052] Forget gate calculation: , is the activation value of the forget gate, and are the weight matrix and bias vector of the forget gate respectively.

[0053] Candidate cell state calculation: , is the candidate cell state, and are the weight matrix and bias vector respectively, and tanh is the hyperbolic tangent activation function.

[0054] Cell status update: , in this formula, is the current cell state, is the cell state at the previous moment, Represents element-wise multiplication.

[0055] Output gate calculation: , in this formula is the activation value of the output gate, and are the weight matrix and bias vector of the output gate respectively.

[0056] Hide status updates: , is the current hidden state, which will be used as the input at the next moment and the final output.

[0057] Through the above calculations, LSTM is able to model the temporal dynamics of node features and capture the temporal dependency of the distribution network operation status. The final prediction results, such as the identification values ​​of distribution network parameters or the identification results of disturbance signals, can be further obtained through the fully connected layer and the appropriate activation function.

[0058] The present invention is aimed at distributed energy systems and proposes a distribution network parameter identification method based on the combination of graph convolutional neural network (GCN) and long short-term memory network (LSTM) with time and space fusion. In view of the complex topology and frequent dynamic changes of the distribution network in the distributed energy system, GCN uses the topological structure and electrical parameters of the distribution network (such as node voltage, current, impedance, etc.) in the spatial dimension to effectively capture the electrical correlation and power flow characteristics between nodes; LSTM models the time series of electrical quantities in the time dimension and captures dynamic characteristics such as load changes and fluctuations in renewable energy output. Through the deep fusion of the time and space characteristics of GCN and LSTM, the model can accurately reflect the operating status and dynamic changes of the distribution network in the distributed energy system, improve the accuracy and reliability of distribution network parameter identification, and meet the needs of distributed energy systems for high precision and high reliability.

[0059] The present invention proposes a new multi-scale collaborative consistency loss function (Multi-Scale Collaborative Consistency Loss, MSCC Loss for short). The traditional method simply adds the consistency losses of the spatial domain, temporal domain and cross-domain, which easily leads to gradient conflicts in the optimization process and makes it difficult for the model to converge. In addition, manually adjusting the weight coefficients of each loss item is complicated, which affects the versatility of the model. The MSCC loss function considers the consistency of spatial, temporal and cross-domain features at different scales through a hierarchical multi-scale design. At each scale, the model calculates the spatial consistency loss, temporal consistency loss and cross-domain consistency loss, and then combines these loss items by product fusion and taking the mean. By introducing a multi-scale architecture (for example, different neighborhood radii or time window lengths) into the model, spatial, temporal and cross-domain consistency are optimized at each scale. The hierarchical design effectively avoids the gradient conflict problem caused by the direct addition of the loss function. Specifically, at each scale, the spatial consistency loss, temporal consistency loss and cross-domain consistency loss are fused in the form of a product, and their geometric mean (i.e., cube root) is taken. The combination of loss functions takes advantage of the unique advantage of geometric mean in balancing multiple positive variables, making the impact of each loss term on the total loss log-linear, thereby maintaining the relative balance of each term during the gradient update process and avoiding excessive dominance or neglect of a certain loss term. It not only promotes the stable convergence of the model, but also enhances the model's ability to capture data features at different scales through multi-scale processing, improves the model's modeling accuracy for complex nonlinear relationships, and enhances the model's robustness and generalization ability in the presence of noise and outliers.

[0060] In an optional embodiment, the parameter input of the distribution network parameter identification model includes: Topological structure information: The adjacency matrix A of the distribution network represents the connection relationship between nodes.

[0061] Node feature data: initial features of the node , such as voltage, current, load and other measurement values.

[0062] Time series data: node features at different time steps t, forming a time series .

[0063] The structure of the distribution network parameter identification model consists of a GCN layer and an LSTM layer. The GCN layer is responsible for extracting the spatial topological features of the distribution network. Through the convolution operation of GCN, the node feature representation of each time step is obtained. The LSTM layer is responsible for modeling time series data. The feature sequence output by GCN In this way, GCN and LSTM are effectively combined. GCN captures the spatial topological characteristics of the distribution network, and LSTM captures the temporal dynamic changes of node characteristics. The fusion of the two enables the model to comprehensively characterize the spatiotemporal characteristics of the distribution network, improving the accuracy and robustness of parameter identification and disturbance signal identification. Ultimately, the output of the model is the estimated values ​​of the distribution network parameters, line impedance, node voltage amplitude and phase, including the identification results of the disturbance signal, fault type, location and time of occurrence. Through training and optimization, the model can be effectively applied in complex distributed energy systems to meet the requirements of high precision and high reliability of the power system.

[0064] In a preferred embodiment, after the distribution network parameter identification model training is completed, a model evaluation is also performed, including: (1) Accuracy assessment Accuracy evaluation focuses on the accuracy of the model's recognition of disturbance signals. This type of evaluation measures the performance of the model by comparing the degree of match between the model's prediction results and the actual labels. The indicators of this solution include accuracy and F1-score.

[0065] Accuracy: (13) F1-score: (16) These indicators can reflect the performance of the model on positive and negative samples in detail. Through confusion matrix analysis, you can intuitively see whether the model accurately recognizes the disturbance signal and what types of signals the model may make mistakes on in actual applications. This evaluation method can comprehensively measure the classification performance of the model and is particularly suitable for scenarios that require high recognition accuracy.

[0066] (2) Robustness evaluation Robustness evaluation focuses on measuring the stability and reliability of the model when facing unknown disturbance signals or noise inputs. There may be various non-standard disturbance signals in the distribution network of distributed energy systems. Therefore, the model must have a certain degree of robustness to cope with the interference of data noise and outliers. By introducing noise into the test data or conducting adversarial sample tests and observing the changes in model performance, the robustness of the model can be evaluated. When a certain proportion of random noise or small disturbances are introduced into the data, whether the output results of the model remain stable. This evaluation method helps ensure that the model can still make robust decisions in the actual deployment environment. This solution uses the cross-validation method for classification problems to evaluate the robustness of the model.

[0067] (16) Where Erri represents the number of classification errors of the i-th model on the i-th test set.

[0068] (3) Timeliness evaluation Timeliness evaluation focuses on the real-time performance and computational efficiency of the model at runtime, which is especially important for distributed energy distribution systems that require fast response. By evaluating the model's performance indicators such as reasoning time, computational complexity, and memory consumption, we can understand the response speed of the model in actual applications. The evaluation method includes testing the model's reasoning time under different hardware environments and testing the model's computational efficiency under different data levels. Timeliness evaluation can ensure that the model can complete the identification of disturbance signals within the specified time and provide timely responses, which is crucial for application scenarios such as fault detection and real-time control.

[0069] Example 2 like Figure 4 As shown, based on the same inventive concept as the above embodiment, the present invention also provides a distributed energy system distribution network parameter identification device, comprising: A data acquisition module is used to obtain the topological structure information and node characteristic data of the distributed energy system distribution network; A parameter identification module is used to input the topological structure information and node feature data into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features of the topological structure information and node feature data to obtain a node feature matrix; the node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

[0070] Furthermore, in the parameter identification module, the distribution network parameter identification model is trained in the following manner: Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; Determine the loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; Get the pre-selected training set; Based on the training set and the loss function, the distribution network parameter identification model is trained, and after the preset conditions are met, a trained distribution network parameter identification model is obtained.

[0071] Furthermore, the loss function is expressed as follows:

[0072] in, For mission loss, is the multi-scale collaborative consistency loss, is the trade-off coefficient.

[0073] Example 3 like Figure 5 As shown, the present invention also provides an electronic device 100 for implementing a distributed energy system distribution network parameter identification method; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0074] The memory 101 can be used to store a computer program 103. The processor 102 implements the steps of a distributed energy system distribution network parameter identification method in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0075] The memory 101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0076] At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0077] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a distributed energy system distribution network parameter identification method, and the processor 102 can execute the plurality of instructions to implement: Obtain the topological structure information and node characteristic data of the distributed energy system distribution network; The topological structure information and node feature data are input into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features of the topological structure information and node feature data to obtain a node feature matrix; the node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

[0078] Example 4 If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0079] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0083] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A distributed energy system distribution network parameter identification method, characterized in that: include: Obtain the topological structure information and node characteristic data of the distributed energy system distribution network; The topological structure information and node feature data are input into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features from the topological structure information and node feature data to obtain a node feature matrix; The node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

2. The distributed energy system distribution network parameter identification method according to claim 1 is characterized in that: In the step of inputting the topology structure information and node feature data into a pre-trained distribution network parameter identification model, wherein the distribution network parameter identification model outputs a parameter identification result, the distribution network parameter identification model is trained in the following manner: Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; Determine the loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; Get the pre-selected training set; Based on the training set and the loss function, the distribution network parameter identification model is trained, and after the preset conditions are met, a trained distribution network parameter identification model is obtained.

3. The distributed energy system distribution network parameter identification method according to claim 2 is characterized in that: The loss function is expressed as follows: in, For mission loss, is the multi-scale collaborative consistency loss, is the trade-off coefficient.

4. The distributed energy system distribution network parameter identification method according to claim 3 is characterized in that: The multi-scale collaborative consistency loss is expressed as follows: Among them, S is the number of scales, and s represents the sth scale; is the weight of the sth scale; They are the spatial domain consistency loss, temporal domain consistency loss and cross-domain consistency loss at the sth scale respectively.

5. The distributed energy system distribution network parameter identification method according to claim 4 is characterized in that: The loss terms in the multi-scale collaborative consistency loss are expressed as follows: Spatial domain consistency loss: Temporal consistency loss: Cross-domain collaborative consistency loss: in, is the length of the time series at the sth scale; is the sth scale, l Spatial feature representation of layers; is the length of the time series at the sth scale; is the hidden state at time step t at the sth scale; It is the spatial feature output at time step t at the sth scale.

6. A distributed energy system distribution network parameter identification device, characterized in that: include: A data acquisition module is used to obtain the topological structure information and node characteristic data of the distributed energy system distribution network; A parameter identification module, used for inputting the topological structure information and node feature data into a pre-trained distribution network parameter identification model, and the distribution network parameter identification model outputs a parameter identification result; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; the graph convolutional neural network layer is used to extract features from the topological structure information and node feature data to obtain a node feature matrix; The node feature matrix is ​​arranged in chronological order as an input sequence of the long short-term memory network layer, and the long short-term memory network layer outputs a parameter identification result.

7. The distributed energy system distribution network parameter identification device according to claim 6, characterized in that: In the parameter identification module, the distribution network parameter identification model is trained in the following manner: Determine the structural composition of the distribution network parameter identification model; wherein the distribution network parameter identification model includes a graph convolutional neural network layer and a long short-term memory network layer; Determine the loss function of the distribution network parameter identification model; wherein the loss function includes task loss and multi-scale collaborative consistency loss; Get the pre-selected training set; Based on the training set and the loss function, the distribution network parameter identification model is trained, and after the preset conditions are met, a trained distribution network parameter identification model is obtained.

8. The distributed energy system distribution network parameter identification device according to claim 7, characterized in that: The loss function is expressed as follows: in, For mission loss, is the multi-scale collaborative consistency loss, is the trade-off coefficient.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the distributed energy system distribution network parameter identification method as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, the distributed energy system distribution network parameter identification method according to any one of claims 1 to 5 is implemented.

Images