A method and related device for evaluating transformer short-circuit current based on adaptive multi-view graph neural network

Abstract

This invention provides a method and related apparatus for assessing transformer short-circuit current based on an adaptive multi-view graph neural network. The method includes: obtaining raw data as an input feature matrix through power flow calculation by simulating different operating modes and state fluctuations; performing Z-score normalization on the feature matrix; dividing the normalized data into training and testing sets; constructing a topology graph and feature graph; constructing a power grid operation model based on an adaptive multi-view graph convolutional neural network; training the power grid operation model using the training set; and after model training, running the power grid operation model on the testing set to predict transformer short-circuit current, and calculating evaluation metrics to measure model performance and assess the accuracy of the model's short-circuit current prediction. This invention can effectively improve the accuracy and response speed of short-circuit current prediction, enhance the safety and stability of the power grid, and reduce economic losses and maintenance costs caused by short-circuit faults.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence algorithms, specifically a method and related apparatus for evaluating transformer short-circuit current based on an adaptive multi-view graph neural network. Background Technology

[0002] In power systems, short-circuit faults are among the most common types of faults. They not only damage equipment but can also cause power outages, posing a serious threat to the safe and stable operation of the power grid. Transformers, as key equipment in power systems, bear enormous electrodynamic and thermal stresses during short circuits. Therefore, accurately assessing transformer short-circuit currents is crucial for preventing transformer failures and protecting equipment safety. By precisely calculating short-circuit currents, circuit breaker capacity can be rationally selected and relay protection configurations optimized, thereby effectively improving the reliability and economy of the power system.

[0003] Traditional assessment methods primarily rely on node characteristics, such as impedance and voltage levels, to calculate short-circuit currents. However, these methods often neglect the topological characteristics of the power grid, namely the interconnections between components and their impact on short-circuit current distribution. In complex power grid structures, this simplification can lead to significant discrepancies between the assessment results and actual conditions. Furthermore, traditional methods typically assume a static power grid model, which is ill-suited to the dynamic nature of modern power grids. This limitation becomes even more pronounced in the context of large-scale renewable energy integration and the development of smart grids.

[0004] To address the aforementioned issues, some researchers have begun exploring the application of Graph Convolutional Networks (GCNs) to short-circuit current assessment in power systems. GCNs are a deep learning method based on graph data structures, capable of effectively handling data with complex connections. In a power system, the power grid can be modeled as a graph, where nodes represent equipment (such as generators, transformers, and loads), and edges represent connections between equipment. Compared to traditional methods, GCNs can simultaneously consider node features and topological features, thus more accurately simulating the operating characteristics of the power system. The basic principle of GCNs is to update the feature representation of nodes by aggregating local neighborhood information, thereby capturing the relationships and patterns in the graph structure. In GCNs, each node's features depend not only on itself but also on the features of its direct neighbors. Specifically, GCNs use neighborhood aggregation operations to perform weighted summation or mean calculation of the node's features with those of its neighbors, and then transform this sum using a learnable weight matrix to generate new node feature representations. This process is typically iterated repeatedly in a multi-layered network structure, with each layer gradually expanding the receptive field of the nodes, allowing them to acquire a wider range of graph structure information. In this way, GCN can effectively capture local and global structural features in graph data and is widely used in tasks such as node classification, link prediction, and graph embedding. However, while GCN has performed well in processing graph-structured data in previous short-circuit current assessments, it has the following limitations:

[0005] 1. GCN's limited ability to integrate node characteristics and topology restricts its performance in certain classification tasks. When assessing short-circuit currents, this may result in insufficient consideration of topology information within the power grid, thus affecting the accuracy of the assessment.

[0006] 2. GCN may not be able to adaptively learn the deep-level information between node features and topology, which means it may not be able to adjust weights according to specific circumstances in order to better extract the most relevant information.

[0007] 3. GCNs typically only consider traditional topologies, neglecting the potential value of feature graphs. In real-world power grids, feature graphs (e.g., k-nearest neighbor graphs generated from node features) may reveal more about the grid's intrinsic connections than traditional topologies, which is particularly important for accurately assessing short-circuit currents. Summary of the Invention

[0008] To address the above issues, this invention proposes a transformer short-circuit current assessment method and related device based on an adaptive multi-view graph neural network. This method can effectively improve the accuracy and response speed of short-circuit current prediction, enhance the safety and stability of the power grid, and reduce economic losses and maintenance costs caused by short-circuit faults.

[0009] A transformer short-circuit current assessment method based on adaptive multi-view graph neural network includes the following steps:

[0010] Step 1: Data Acquisition: Obtain raw data by simulating different operating modes and state fluctuations to perform power flow calculations, which will serve as the input feature matrix;

[0011] Step 2, Data Standardization Processing: Perform Z-score standardization on the feature matrix to transform the dimensional raw data into a dimensionless representation X;

[0012] Step 3, Dataset Splitting: Divide the standardized data from Step 2 into a training set for learning model parameters and a test set for finally evaluating the model's generalization ability;

[0013] Step 4: Constructing the topology graph and feature graph: Construct the topology graph based on the power grid topology, and construct the feature graph based on the feature matrix;

[0014] Step 5: Construct a power grid operation model based on an adaptive multi-view graph convolutional neural network: This includes defining graph convolutional layers, obtaining multi-view embeddings, implementing the attention mechanism, and obtaining the target loss value. Specifically, specific graph convolutional modules and common graph convolutional modules are constructed by combining the topology graph and feature graph from Step 4.

[0015] Step 6: Model Training: Determine the optimizer parameters and model parameters, and train the power grid operation model using the training set;

[0016] Step 7, Results Analysis: After completing the model training, the power grid operation model is run on the test set to predict the transformer short-circuit current, and evaluation indicators are calculated to measure the model performance and evaluate the accuracy of the model in predicting the short-circuit current.

[0017] Furthermore, the feature matrix records the characteristics of the transformer bus before a short-circuit fault, including bus voltage, phase angle, active power, and reactive power injection.

[0018] Furthermore, in step 4, the topology diagram is used to describe the degree of connectivity between nodes and the line impedance.

[0019] Furthermore, in step 4, when constructing the feature map based on the feature matrix, the similarity between each pair of nodes in the graph is first calculated. The similarity is achieved through cosine similarity or Gaussian kernel function. Then, for each node, the nodes most similar to it are selected as neighbors according to the similarity level.

[0020] Furthermore, step 5 constructs a power grid operation model based on an adaptive multi-view graph convolutional neural network, specifically including:

[0021] (1) First, define the graph convolutional layer: specify the input feature dimension and the output feature dimension; initialize a weight matrix to map from the input feature space to the output feature space; in the forward propagation, use the adjacency matrix and the feature matrix to multiply to collect local neighborhood information of each node; use the weight matrix to perform a linear transformation on the aggregated features, and then use a non-linear activation function to enhance the non-linear expressive power of the model.

[0022] (2) Obtaining three different perspectives of embedding: First, based on the topology graph and feature graph, three types of embedding are defined as input: topology-based embedding, feature space-based embedding, and common embedding; among them, graph convolution operations are performed on the topology graph and feature graph respectively to obtain topology-based embedding and feature space-based embedding; a common convolution module is introduced to process the embedding information based on the topology graph and feature graph through a parameter sharing strategy, and extract the common information shared by the two spaces as common embedding to enhance the ability of the power grid operation model to capture common features;

[0023] (3) Attention mechanism implementation: First, a shared attention vector q is initialized to calculate the original attention score of each node on different embeddings, which is accomplished by the dot product of the node embedding and q; then, the original attention score is normalized using the softmax function to ensure that the attention weight of each node is equal to 1 when added on the three embeddings, thus obtaining the final weight of each embedding; the weighted embeddings are then merged to form the final embedding representation of each node.

[0024] (4) Obtaining the target loss value: The power grid operation model introduces consistency constraints and difference constraints to improve the expressive and generalization capabilities of the embedding. The consistency constraint ensures that the embedding vectors obtained from different perspectives are consistent through L2 normalization and Frobenius norm differences of the similarity matrix. The difference constraint encourages the model to capture the diversity in the feature space and topology space, and improves the model's ability to distinguish different types of nodes. The consistency constraint and the difference constraint are integrated in the loss function to obtain the final loss value.

[0025] A transformer short-circuit current assessment device based on an adaptive multi-view graph neural network includes:

[0026] The data acquisition module is used to obtain raw data by simulating different operating modes and state fluctuations to perform power flow calculations, which serves as the input feature matrix;

[0027] The data standardization processing module is used to perform Z-score standardization on the feature matrix, transforming the dimensional raw data into a dimensionless expression form X.

[0028] The dataset partitioning module is used to divide the data after it has been standardized by the data standardization module into a training set for learning model parameters and a test set for finally evaluating the model's generalization ability.

[0029] The topology graph and feature graph construction module is used to construct a topology graph based on the topology of the power grid and a feature graph based on the feature matrix.

[0030] The power grid operation model construction module is used to construct a power grid operation model based on an adaptive multi-view graph convolutional neural network. It includes defining graph convolutional layers, obtaining multi-view embeddings, implementing attention mechanisms, and obtaining target loss values. Specifically, it combines topological graphs and feature graphs to construct specific graph convolutional modules and common graph convolutional modules.

[0031] The model training module is used to determine the optimizer parameters and the model parameters, and to train the power grid operation model using the training set.

[0032] The results analysis module is used to run the power grid operation model on the test set after model training is completed to predict transformer short-circuit current and calculate evaluation indicators to measure model performance and evaluate the accuracy of the model in predicting short-circuit current.

[0033] Furthermore, the feature matrix records the characteristics of the transformer bus before a short-circuit fault, including bus voltage, phase angle, active power, and reactive power injection.

[0034] Furthermore, the power grid operation model construction module constructs a power grid operation model based on an adaptive multi-view graph convolutional neural network, specifically including:

[0035] (1) First, define the graph convolutional layer: specify the input feature dimension and the output feature dimension; initialize a weight matrix to map from the input feature space to the output feature space; in the forward propagation, use the adjacency matrix and the feature matrix to multiply to collect local neighborhood information of each node; use the weight matrix to perform a linear transformation on the aggregated features, and then use a non-linear activation function to enhance the non-linear expressive power of the model.

[0036] (2) Obtaining three different perspectives of embedding: First, based on the topology graph and feature graph, three types of embedding are defined as input: topology-based embedding, feature space-based embedding, and common embedding; among them, graph convolution operations are performed on the topology graph and feature graph respectively to obtain topology-based embedding and feature space-based embedding; a common convolution module is introduced to process the embedding information based on the topology graph and feature graph through a parameter sharing strategy, and extract the common information shared by the two spaces as common embedding to enhance the ability of the power grid operation model to capture common features;

[0037] (3) Attention mechanism implementation: First, a shared attention vector q is initialized to calculate the original attention score of each node on different embeddings, which is accomplished by the dot product of the node embedding and q; then, the original attention score is normalized using the softmax function to ensure that the attention weight of each node is equal to 1 when added on the three embeddings, thus obtaining the final weight of each embedding; the weighted embeddings are then merged to form the final embedding representation of each node.

[0038] (4) Obtaining the target loss value: The power grid operation model introduces consistency constraints and difference constraints to improve the expressive and generalization capabilities of the embedding. The consistency constraint ensures that the embedding vectors obtained from different perspectives are consistent through L2 normalization and Frobenius norm differences of the similarity matrix. The difference constraint encourages the model to capture the diversity in the feature space and topology space, and improves the model's ability to distinguish different types of nodes. The consistency constraint and the difference constraint are integrated in the loss function to obtain the final loss value.

[0039] A transformer short-circuit current assessment system based on an adaptive multi-view graph neural network includes: a computer-readable storage medium and a processor;

[0040] The computer-readable storage medium is used to store executable instructions;

[0041] The processor is used to read executable instructions stored in the computer-readable storage medium and execute the transformer short-circuit current assessment method based on adaptive multi-view graph neural network.

[0042] A non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the transformer short-circuit current assessment method based on an adaptive multi-view graph neural network as described above.

[0043] The present invention has the following beneficial effects:

[0044] 1. Effective Integration of Node Features and Topology Information: This invention significantly improves the accuracy of short-circuit current assessment by introducing specific graph convolution modules and common graph convolution modules. The specific graph convolution modules operate independently on the topology graph and feature graph, extracting their respective unique information, thus ensuring the model fully understands the complex structure of the power grid. The common graph convolution module employs a parameter-sharing strategy, allowing the extraction of shared features from both graphs, enhancing the grasp of the commonalities in the information space. This dual processing mechanism ensures the efficient integration of node features and topology information, thereby improving the model's predictive capability.

[0045] 2. An adaptive mechanism is designed to enable the model to dynamically adjust weights among different feature representations to optimize short-circuit current prediction. Specifically, the model utilizes an attention mechanism to dynamically fuse node topological features and feature information. By generating specific and shared embeddings, the model calculates attention weights and applies them to embedding fusion, achieving adaptive integration of information. This innovation allows the model to automatically adjust the importance of features based on different input conditions, thereby improving the flexibility and accuracy of prediction.

[0046] 3. Constructing a Multi-View Graph Neural Network: This invention constructs a multi-view graph neural network to capture the complex relationships between nodes in a power grid and learn richer feature representations. This model not only considers the traditional topology graph but also introduces a feature graph based on the feature matrix. By performing graph convolution operations simultaneously in the topology space and feature space, features are learned from multiple perspectives. This multi-view design enables the model to comprehensively understand the dynamic behavior of the power grid, enhances its ability to distinguish between different types of nodes, and thus improves the accuracy and robustness of short-circuit current assessment. Attached Figure Description

[0047] Figure 1 This is a flowchart of a transformer short-circuit current evaluation method based on an adaptive multi-view graph neural network according to an embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of the power grid operation model according to an embodiment of the present invention;

[0049] Figure 3 This is a trend graph showing how the loss value changes with the number of iterations in an embodiment of the present invention;

[0050] Figure 4 This refers to the error between the predicted short-circuit current and the actual short-circuit current at the fault node in this embodiment of the invention.

[0051] Figure 5 This is a clustering trend diagram according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] The inventors of this application discovered during the development of this invention that, in assessing transformer short-circuit current, traditional GCNs primarily focus on the system's topology, i.e., the adjacency relationships between nodes. However, relying solely on the topology for graph convolution operations leads to the loss of a significant amount of valuable information. This is because this single perspective ignores the feature graphs of nodes. Node features not only include the nodes' intrinsic attributes but also the implicit relationships between nodes—the feature graphs themselves. These feature graphs are crucial for understanding the dynamic behavior of the power grid and predicting short-circuit current. Based on a multi-perspective model, the model can better understand the geometric relationships and directional similarities between nodes. This makes the model more flexible and comprehensive in capturing complex relationships and feature patterns, and the multi-perspective structure enhances the model's robustness.

[0054] With a single adjacency matrix, the performance of graph network models is easily affected by the graph structure, especially when noisy or incomplete graph data is present. By combining different adjacency matrices, the model can perform weighted averaging of multiple relationships, thereby reducing the overall performance degradation caused by inaccuracies in a particular channel. Furthermore, multiple perspectives can enhance the model's expressive power. Different adjacency matrices represent different graph structures and relationship types; by performing convolutional operations on these structures in parallel, the model can learn richer feature representations.

[0055] Furthermore, traditional multi-view feature fusion methods fail to fully utilize the differences in information captured by each channel. Fixed-weight averaging or simple combination cannot dynamically adjust the importance of different feature representations, potentially leading to information loss or confusion. For example, in some tasks, node topology may be more important than structural topology, and vice versa. However, fixed weighting methods cannot identify and utilize this difference, potentially diluting the contribution of a particular important feature and affecting the overall model performance. Adaptive multi-view feature fusion, on the other hand, allows the model to dynamically adjust the weights of each feature representation during the learning process, enabling it to adaptively focus on the most relevant feature representations. This capability enhances the model's flexibility and generalization ability when processing different types of graph data. Secondly, adaptive learning can capture the interrelationships and relative importance between different feature representations, thereby improving the model's ability to recognize complex patterns. For example, under certain conditions, an adjacency matrix based on node features may be more representative of the target information than other adjacency matrices; attention mechanisms can assign it higher weights, thus optimizing the final prediction results.

[0056] Please see Figure 1 This invention provides a method for evaluating transformer short-circuit current based on an adaptive multi-view graph neural network, comprising the following steps:

[0057] Step 1. Data Acquisition: Raw data is obtained by simulating different operating modes and state fluctuations to perform power flow calculations. This raw data serves as the input feature matrix, which records the characteristics of the transformer bus before a short-circuit fault, including bus voltage, phase angle, active power, and reactive power injection.

[0058] Step 2. Data Standardization: The feature matrix is ​​Z-score standardized to transform the dimensional raw data into a dimensionless representation X. Data standardization eliminates the influence of dimensions by adjusting features to the same scale, thereby accelerating model convergence, improving the stability and performance of the algorithm, and ensuring fair model performance across different features.

[0059] Step 3. Dataset partitioning: After standardizing the data, the dataset is divided into two parts according to a certain ratio: a training set, used to learn the model parameters; and a test set, used to finally evaluate the model's generalization ability.

[0060] Step 4. Construction of Topology and Feature Graph: A topology graph is constructed based on the power grid's topology. This graph describes the connectivity between nodes and line impedance. A feature graph is then constructed based on the feature matrix. For the feature graph construction, the similarity between each pair of nodes is first calculated, typically using cosine similarity or a Gaussian kernel function. Next, for each node, the k most similar nodes are selected as its neighbors based on their similarity scores.

[0061] Step 5. Construct a power grid operation model based on an adaptive multi-view graph convolutional neural network: This includes defining graph convolutional layers, obtaining multi-view embedding values, implementing the attention mechanism, and obtaining the target loss. Specifically, it involves constructing specific graph convolutional modules and common graph convolutional modules by combining the topology graph and feature graph from Step 4. For example... Figure 2 As shown, the specific implementation steps of step 5 are as follows:

[0062] (1) First, define the graph convolutional layer: specify the input feature dimension and the output feature dimension; initialize a weight matrix to map from the input feature space to the output feature space. In the forward propagation, the adjacency matrix and the feature matrix are multiplied to collect local neighborhood information of each node. The aggregated features are linearly transformed using the weight matrix, and then a nonlinear activation function is used to enhance the nonlinear expressive power of the model.

[0063] (2) Obtaining three different perspectives of embedding: First, based on the topology graph and feature graph, three types of embedding are defined as inputs: topology-based embedding, feature space-based embedding, and co-embedding. Specifically, graph convolution operations are performed on the topology graph and feature graph to obtain topology-based embedding and feature space-based embedding, respectively. A common convolution module is introduced, and a parameter-sharing strategy is used to process the embedding information based on the topology graph and feature graph, extracting the common information shared by the two spaces as the co-embedding to enhance the power grid operation model's ability to capture common features.

[0064] (3) Attention Mechanism Implementation: First, a shared attention vector q is initialized to compute the original attention score of each node on different embeddings. This is typically done by the dot product of the node embedding and q. Then, these original scores are normalized using the softmax function to ensure that the attention weight of each node adds up to 1 across the three embeddings, thus obtaining the final weight for each embedding. These weighted embeddings are then merged to form the final embedding representation for each node.

[0065] (4) Finally, obtaining the target loss value: To improve the expressive and generalization capabilities of the embeddings, the power grid operation model introduces consistency constraints and difference constraints. Consistency constraints ensure consistency in embedding vectors obtained from different perspectives through L2 normalization and Frobenius norm differences in the similarity matrix. Difference constraints encourage the model to capture diversity in the feature space and topological space, improving the model's ability to distinguish between different types of nodes. Since the assessment of short-circuit current is a regression task, this embodiment selects MSE to measure the loss between predicted and actual values. The consistency and difference constraints are then integrated into the loss function to obtain the final loss value.

[0066] Step 6. Model Training: After constructing the power grid operation model based on an adaptive multi-view graph convolutional neural network, the model parameters are adjusted using the training set and an optimization algorithm based on the calculated gradients to minimize the loss function. Several important hyperparameters are designed in the optimizer, such as learning rate, momentum, and weight decay. The learning rate is a control factor for the magnitude of model parameter updates; it determines the step size of parameter updates in each iteration. A higher learning rate means a larger parameter update step, which may lead to faster convergence but may also skip the optimal solution or cause instability during training. A lower learning rate means a smaller parameter update step, which, while allowing the model to more stably approach the optimal solution, may result in a very slow training process. Weight decay is a regularization technique that suppresses model complexity and prevents overfitting by adding a penalty term of the sum of squared weights to the loss function. Weight decay is essentially equivalent to L2 regularization.

[0067] Step 7. Results Analysis: After completing model training, make predictions on the test set and calculate evaluation metrics such as mean squared error to measure model performance. Figure 3 The trend of the loss value as the number of iterations increases is depicted. It can be observed that as the number of iterations increases, the loss value shows a continuous decrease until it reaches a convergent state. Figure 4 To measure the error between the predicted short-circuit current and the actual short-circuit current at the fault node, the figure shows that all errors are below 0.06, which is within an acceptable range. Specifically, at... Figure 5 In the diagram, the features learned by the model were dimensionality reduced and visualized, revealing a clear clustering trend. This indicates that graph neural network models can effectively capture the inherent structure of data and group similar instances together, thereby improving the performance of classification or other downstream tasks.

[0068] This invention has the following characteristics:

[0069] 1. Adaptive Multi-View Graph Neural Network Structure: This invention proposes a transformer short-circuit current assessment method based on an adaptive multi-view graph neural network. The core of this method lies in its ability to simultaneously consider node characteristics and the topology of the power grid. By constructing different adjacency matrices, it captures the complex relationships between nodes in the power grid, thereby improving the accuracy of short-circuit current prediction.

[0070] 2. Feature Fusion and Adaptive Weight Adjustment Mechanism: This invention employs an adaptive feature fusion method, allowing the model to dynamically adjust the weights of different feature representations during the learning process to optimize short-circuit current prediction. This adaptive mechanism can capture the interrelationships and relative importance between different feature representations, improving the model's ability to recognize complex patterns.

[0071] 3. Optimization Objectives of Consistency and Difference Constraints: During model optimization, this invention introduces consistency and difference constraints to ensure that embedding vectors obtained from different perspectives are consistent and can capture the different characteristics of nodes in the feature space and topological space. This design helps improve the model's generalization ability and its ability to distinguish between different types of nodes.

[0072] This invention also provides a transformer short-circuit current assessment device based on an adaptive multi-view graph neural network, comprising:

[0073] The data acquisition module is used to obtain raw data by simulating different operating modes and state fluctuations to perform power flow calculations, which serves as the input feature matrix;

[0074] The data standardization processing module is used to perform Z-score standardization on the feature matrix, transforming the dimensional raw data into a dimensionless expression form X.

[0075] The dataset partitioning module is used to divide the data after it has been standardized by the data standardization module into a training set for learning model parameters and a test set for finally evaluating the model's generalization ability.

[0076] The topology graph and feature graph construction module is used to construct a topology graph based on the topology of the power grid and a feature graph based on the feature matrix.

[0077] The power grid operation model construction module is used to construct a power grid operation model based on an adaptive multi-view graph convolutional neural network. It includes defining graph convolutional layers, obtaining multi-view embeddings, implementing attention mechanisms, and obtaining target loss values. Specifically, it combines topological graphs and feature graphs to construct specific graph convolutional modules and common graph convolutional modules.

[0078] The model training module is used to determine the optimizer parameters and the model parameters, and to train the power grid operation model using the training set.

[0079] The results analysis module is used to run the power grid operation model on the test set after model training is completed to predict transformer short-circuit current and calculate evaluation indicators to measure model performance and evaluate the accuracy of the model in predicting short-circuit current.

[0080] Another embodiment of the present invention provides a transformer short-circuit current assessment system based on an adaptive multi-view graph neural network, comprising: a computer-readable storage medium and a processor;

[0081] The computer-readable storage medium is used to store executable instructions;

[0082] The processor is used to read executable instructions stored in the computer-readable storage medium and execute the transformer short-circuit current assessment method based on adaptive multi-view graph neural network.

[0083] Another embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned transformer short-circuit current assessment method based on an adaptive multi-view graph neural network.

[0084] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for evaluating transformer short-circuit current based on an adaptive multi-view graph neural network, characterized in that, Includes the following steps: Step 1: Data Acquisition: Obtain raw data by simulating different operating modes and state fluctuations to perform power flow calculations, which will serve as the input feature matrix; Step 2, Data Standardization Processing: Perform Z-score standardization on the feature matrix to transform the dimensional raw data into a dimensionless representation X; Step 3, Dataset Splitting: Divide the standardized data from Step 2 into a training set for learning model parameters and a test set for finally evaluating the model's generalization ability; Step 4: Constructing the topology graph and feature graph: Construct the topology graph based on the power grid topology, and construct the feature graph based on the feature matrix; Step 5: Construct a power grid operation model based on an adaptive multi-view graph convolutional neural network: This includes defining graph convolutional layers, obtaining multi-view embeddings, implementing the attention mechanism, and obtaining the target loss value. Specifically, specific graph convolutional modules and common graph convolutional modules are constructed by combining the topology graph and feature graph from Step 4. Step 6: Model Training: Determine the optimizer parameters and model parameters, and train the power grid operation model using the training set; Step 7, Results Analysis: After completing the model training, the power grid operation model is run on the test set to predict the transformer short-circuit current, and evaluation indicators are calculated to measure the model performance and evaluate the accuracy of the model in predicting the short-circuit current.

2. The transformer short-circuit current assessment method based on adaptive multi-view graph neural network as described in claim 1, characterized in that, The feature matrix records the characteristics of the transformer bus before a short-circuit fault, including bus voltage, phase angle, active power, and reactive power injection.

3. The transformer short-circuit current assessment method based on adaptive multi-view graph neural network as described in claim 1, characterized in that, In step 4, the topology diagram is used to describe the degree of connectivity between nodes and the line impedance.

4. The transformer short-circuit current assessment method based on adaptive multi-view graph neural network as described in claim 1, characterized in that, In step 4, when constructing the feature map based on the feature matrix, the similarity between each pair of nodes in the graph is first calculated. The similarity is achieved by using cosine similarity or Gaussian kernel function. Then, for each node, the nodes that are most similar to it are selected as neighbors according to the similarity.

5. The transformer short-circuit current assessment method based on adaptive multi-view graph neural network as described in claim 1, characterized in that, Step 5 involves constructing a power grid operation model based on an adaptive multi-view graph convolutional neural network, specifically including: (1) First, define the graph convolutional layer: specify the input feature dimension and the output feature dimension; initialize a weight matrix to map from the input feature space to the output feature space; in the forward propagation, use the adjacency matrix and the feature matrix to multiply to collect local neighborhood information of each node; use the weight matrix to perform a linear transformation on the aggregated features, and then use a non-linear activation function to enhance the non-linear expressive power of the model. (2) Obtaining three different perspectives of embedding: First, based on the topology graph and feature graph, three types of embedding are defined as input: topology-based embedding, feature space-based embedding, and common embedding; among them, graph convolution operations are performed on the topology graph and feature graph respectively to obtain topology-based embedding and feature space-based embedding; a common convolution module is introduced to process the embedding information based on the topology graph and feature graph through a parameter sharing strategy, and extract the common information shared by the two spaces as common embedding to enhance the ability of the power grid operation model to capture common features; (3) Attention mechanism implementation: First, a shared attention vector q is initialized to calculate the original attention score of each node on different embeddings, which is accomplished by the dot product of the node embedding and q; then, the original attention score is normalized using the softmax function to ensure that the attention weight of each node is equal to 1 when added on the three embeddings, thus obtaining the final weight of each embedding; the weighted embeddings are then merged to form the final embedding representation of each node. (4) Obtaining the target loss value: The power grid operation model introduces consistency constraints and difference constraints to improve the expressive and generalization capabilities of the embedding. The consistency constraint ensures that the embedding vectors obtained from different perspectives are consistent through L2 normalization and Frobenius norm differences of the similarity matrix. The difference constraint encourages the model to capture the diversity in the feature space and topology space, and improves the model's ability to distinguish different types of nodes. The consistency constraint and the difference constraint are integrated in the loss function to obtain the final loss value.

6. A transformer short-circuit current assessment device based on an adaptive multi-view graph neural network, characterized in that, include: The data acquisition module is used to obtain raw data by simulating different operating modes and state fluctuations to perform power flow calculations, which serves as the input feature matrix; The data standardization processing module is used to perform Z-score standardization on the feature matrix, transforming the dimensional raw data into a dimensionless expression form X. The dataset partitioning module is used to divide the data after it has been standardized by the data standardization module into a training set for learning model parameters and a test set for finally evaluating the model's generalization ability. The topology graph and feature graph construction module is used to construct a topology graph based on the topology of the power grid and a feature graph based on the feature matrix. The power grid operation model construction module is used to construct a power grid operation model based on an adaptive multi-view graph convolutional neural network. It includes defining graph convolutional layers, obtaining multi-view embeddings, implementing attention mechanisms, and obtaining target loss values. Specifically, it combines topological graphs and feature graphs to construct specific graph convolutional modules and common graph convolutional modules. The model training module is used to determine the optimizer parameters and the model parameters, and to train the power grid operation model using the training set. The results analysis module is used to run the power grid operation model on the test set after model training is completed to predict transformer short-circuit current and calculate evaluation indicators to measure model performance and evaluate the accuracy of the model in predicting short-circuit current.

7. The transformer short-circuit current assessment device based on adaptive multi-view graph neural network as described in claim 6, characterized in that, The feature matrix records the characteristics of the transformer bus before a short-circuit fault, including bus voltage, phase angle, active power, and reactive power injection.

8. The transformer short-circuit current assessment device based on adaptive multi-view graph neural network as described in claim 6, characterized in that, The power grid operation model construction module constructs a power grid operation model based on an adaptive multi-view graph convolutional neural network, specifically including: (1) First, define the graph convolutional layer: specify the input feature dimension and the output feature dimension; initialize a weight matrix to map from the input feature space to the output feature space; in the forward propagation, use the adjacency matrix and the feature matrix to multiply to collect local neighborhood information of each node; use the weight matrix to perform a linear transformation on the aggregated features, and then use a non-linear activation function to enhance the non-linear expressive power of the model. (2) Obtaining three different perspectives of embedding: First, based on the topology graph and feature graph, three types of embedding are defined as input: topology-based embedding, feature space-based embedding, and common embedding; among them, graph convolution operations are performed on the topology graph and feature graph respectively to obtain topology-based embedding and feature space-based embedding; a common convolution module is introduced to process the embedding information based on the topology graph and feature graph through a parameter sharing strategy, and extract the common information shared by the two spaces as common embedding to enhance the ability of the power grid operation model to capture common features; (3) Attention mechanism implementation: First, a shared attention vector q is initialized to calculate the original attention score of each node on different embeddings, which is accomplished by the dot product of the node embedding and q; then, the original attention score is normalized using the softmax function to ensure that the attention weight of each node is equal to 1 when added on the three embeddings, thus obtaining the final weight of each embedding; the weighted embeddings are then merged to form the final embedding representation of each node. (4) Obtaining the target loss value: The power grid operation model introduces consistency constraints and difference constraints to improve the expressive and generalization capabilities of the embedding. The consistency constraint ensures that the embedding vectors obtained from different perspectives are consistent through L2 normalization and Frobenius norm differences of the similarity matrix. The difference constraint encourages the model to capture the diversity in the feature space and topology space, and improves the model's ability to distinguish different types of nodes. The consistency constraint and the difference constraint are integrated in the loss function to obtain the final loss value.

9. A transformer short-circuit current assessment system based on an adaptive multi-view graph neural network, comprising: Computer-readable storage media and processors; The computer-readable storage medium is used to store executable instructions; The processor is used to read executable instructions stored in the computer-readable storage medium and execute the transformer short-circuit current assessment method based on adaptive multi-view graph neural network as described in any one of claims 1-5.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the transformer short-circuit current assessment method based on an adaptive multi-view graph neural network as described in any one of claims 1-5.

Images