Power optical communication network alarm analysis method and system based on time-space diagram neural network

Abstract

The invention discloses an electric power optical communication network alarm analysis method and system based on a time-space diagram neural network. The method comprises the following steps: collecting data of an electric power optical communication network, preprocessing the collected data, and modeling the preprocessed data into a graph structure; extracting and learning the features of the nodes in the graph structure by using a space-time graph neural network, and generating space-time embedded representation of the nodes in combination with space feature extraction and time feature extraction; classifying the nodes based on the spatio-temporal embedding representation of the nodes to obtain prediction labels of devices corresponding to the nodes; and according to the prediction label of the equipment corresponding to the node, carrying out analysis and influence evaluation on the alarm condition in the power optical communication network. According to the method, the electric power optical communication network is represented as the graph structure containing the space-time information, the space-time dependence of the electric power optical communication network can be effectively captured based on the space-time graph neural network, the space relationship in the graph structure and the dynamic change in the time sequence are processed at the same time, and the accuracy and timeliness of alarm analysis are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power optical communication networks, and specifically relates to an electric power optical communication network alarm analysis method and system based on spatiotemporal graph neural network. Background Art

[0002] The power optical communication network refers to a network that connects the nodes of the power system through optical fiber communication technology, which is used to transmit data and control information in real time to ensure the stable operation of the power system. With the expansion of the scale and complexity of the power system, the power optical communication network, as the communication backbone of the power system, faces increasingly complex alarm and fault management challenges. Traditional alarm analysis methods mainly rely on rule-based systems or traditional machine learning models, which have limitations in dealing with complex network structures and spatiotemporal dependencies. For example, there are methods based on the combination of long short-term memory networks (LSTM) and graph convolutional networks (GCN). These methods attempt to process time series data through LSTM and use GCN to process graph structure data to capture the spatiotemporal dependencies in the power optical communication network. However, these methods usually separate the processing of time and space, and it is difficult to fully integrate spatiotemporal characteristics. Therefore, when dealing with dynamically changing power optical communication networks, the prediction effect is limited. For example Figure 1 As shown in the figure, the graph convolutional network is the most classic graph neural network model. The core idea is to update the representation of the current node by aggregating the information of neighboring nodes. layer, whose output feature matrix It can be expressed as:

[0003] in, It is The input feature matrix of the layer, for the input layer, That is the original feature matrix ; It is The weight matrix of the layer is used to perform linear transformation on the features; is a nonlinear activation function, such as ; is the adjacency matrix with self-loops added, is the original adjacency matrix, It is the identity matrix. The purpose of adding self-loops is to allow each node to consider its own features when performing feature aggregation. yes The degree matrix of is a diagonal matrix with the elements on the diagonal Representation Node degree (i.e., the degree of the node The degree matrix is ​​used to normalize the adjacency matrix so that the neighbor features of each node can be weighted according to the degree when aggregated, avoiding nodes with larger degrees from occupying too large a weight when features are aggregated (this normalization method is also called symmetric normalized Laplace moment).

[0004] In the above-mentioned prior art, although the combination of LSTM and GCN can handle some spatiotemporal dependencies, they cannot efficiently integrate the two under the same framework because they process time and space features separately. Therefore, when dealing with complex power optical communication networks, this separate processing method leads to the model's insufficient ability to capture spatiotemporal dependencies. In addition, when faced with the rapidly changing dynamic features of power optical communication networks, the accuracy and real-time performance of these models still need to be improved. Summary of the invention

[0005] The purpose of the present invention is to address the problems in the above-mentioned prior art and provide an electric power optical communication network alarm analysis method and system based on a spatiotemporal graph neural network, which represents the electric power optical communication network as a graph structure containing spatiotemporal information. Based on the spatiotemporal graph neural network (ST-GNN), the spatiotemporal dependencies of the electric power optical communication network can be effectively captured, while processing the spatial relationships in the graph structure and the dynamic changes in the time series, thereby improving the accuracy and timeliness of the alarm analysis.

[0006] In order to achieve the above object, the present invention has the following technical solutions: In a first aspect, a method for analyzing alarms in a power optical communication network based on a spatiotemporal graph neural network is provided, comprising: Collecting data from the electric power optical communication network, and preprocessing the collected data to obtain preprocessed data; Model the preprocessed data into a graph structure; The spatiotemporal graph neural network is used to extract and learn the features of the nodes in the graph structure, and the spatiotemporal embedding representation of the nodes is generated by combining the spatial feature extraction and the temporal feature extraction; Based on the spatiotemporal embedding representation of nodes, nodes are classified to obtain the predicted labels of the devices corresponding to the nodes. According to the predicted labels of the devices corresponding to the nodes, the alarm situations in the electric power optical communication network are analyzed and the impacts are evaluated.

[0007] As a preferred solution, in the step of collecting the power optical communication network data, the power optical communication network data includes equipment information, connection status between equipments and historical alarm information of the power optical communication network.

[0008] As a preferred solution, in the step of preprocessing the collected data, the preprocessing includes one or more of collating and extracting alarm information, denoising, processing missing values, and eliminating duplicate records on the collected data.

[0009] As a preferred solution, in the step of modeling the preprocessed data into a graph structure, the nodes in the graph structure represent devices, the edges represent the connections between devices, and corresponding time dimension information is attached to each time point, thereby generating a graph at each moment.

[0010] As a preferred solution, the step of extracting and learning the features of the nodes in the graph structure by using the spatiotemporal graph neural network, combining spatial feature extraction and temporal feature extraction, and generating a spatiotemporal embedding representation of the nodes includes: The spatial features of nodes and edges in the graph are extracted using a spatiotemporal graph neural network, and the dynamic changes of node states in the spatiotemporal graph neural network are captured through time steps. The spatiotemporal embedding representation of the nodes is generated by fusing the spatial features of nodes and edges and the dynamic changes of node states. The spatiotemporal embedding representation of the nodes is used to predict future alarm events and generate early warning signals.

[0011] As a preferred solution, each time step in the spatiotemporal graph neural network is a graph, and is passed through the graph neural network to obtain a result encoding graph that embeds the spatial interdependencies of the data; if the power optical communication network contains only one type of device, a homogeneous graph neural network model can be used; if the power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used.

[0012] As a preferred solution, the homogeneous graph neural network model includes any one or more combinations of the graph convolution network GCN, the graph sampling and aggregation network GraphSAGE, and the graph attention network GAT; the heterogeneous graph neural network model includes any one or more combinations of the heterogeneous graph neural network HGNN and the heterogeneous graph attention network HAN.

[0013] As a preferred solution, in the step of extracting and learning the features of the nodes in the graph structure by using the spatiotemporal graph neural network, combining the spatial feature extraction and the temporal feature extraction to generate the spatiotemporal embedding representation of the nodes, the spatial features are extracted by neighborhood aggregation through the graph neural network model, and the temporal features are extracted by time series capture through the temporal model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as:

[0014] In the formula, express The elements in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure to predict the current graph structure; The time model includes any one or more combinations of an auto-regressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

[0015] As a preferred scheme, in the step of classifying nodes based on the spatiotemporal embedding representation of nodes to obtain predicted labels for devices corresponding to the nodes, a label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm; by setting multiple labels, multiple types of alarms for devices corresponding to the nodes can be predicted.

[0016] As a preferred solution, in the step of analyzing and impact assessing the alarm situation in the electric power optical communication network based on the prediction label of the device corresponding to the node, potential fault nodes and alarm types in the electric power optical communication network are identified, and the possible impact range is assessed.

[0017] In a second aspect, a power optical communication network alarm analysis system based on a spatiotemporal graph neural network is provided, comprising: The data collection and preprocessing module is used to collect data from the power optical communication network, preprocess the collected data, and obtain preprocessed data; Graph structure modeling module, used to model the preprocessed data into a graph structure; A node spatiotemporal embedding representation generation module is used to extract and learn the features of the nodes in the graph structure using a spatiotemporal graph neural network, and generate a spatiotemporal embedding representation of the nodes by combining spatial feature extraction and temporal feature extraction; The node label prediction module is used to classify nodes based on the spatiotemporal embedding representation of nodes and obtain the predicted label of the device corresponding to the node; The alarm analysis and evaluation module is used to analyze and evaluate the impact of alarm conditions in the power optical communication network based on the prediction labels of the equipment corresponding to the nodes.

[0018] As a preferred solution, when the graph structure modeling module models the graph structure, the nodes in the graph structure represent devices, the edges represent the connections between devices, and the corresponding time dimension information is added to each time point, thereby generating a graph at each moment.

[0019] As a preferred solution, the node spatiotemporal embedding representation generation module uses a spatiotemporal graph neural network to extract the spatial features of nodes and edges in the graph structure, and captures the dynamic changes of node states in the spatiotemporal graph neural network through time steps; By fusing the spatial features of nodes and edges and the dynamic changes of node states, the spatiotemporal embedding representation of nodes is generated. The spatiotemporal embedding representation of nodes is used to predict future alarm events and generate early warning signals. Each time step in the spatiotemporal graph neural network is a graph, and is passed through the graph neural network to obtain a result encoding graph that embeds the spatial interdependencies of the data; if the electric power optical communication network contains only one type of equipment, a homogeneous graph neural network model can be used; if the electric power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used.

[0020] As a preferred solution, the node spatiotemporal embedding representation generation module extracts spatial features through neighborhood aggregation using a graph neural network model, and extracts temporal features through time series capture using a time model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as:

[0021] In the formula, express The elements in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure and predict the graph structure at the current moment.

[0022] As a preferred solution, the homogeneous graph neural network model includes any one or more combinations of graph convolutional network GCN, graph sampling and aggregation network GraphSAGE and graph attention network GAT; The heterogeneous graph neural network model includes any one or more combinations of a heterogeneous graph neural network HGNN and a heterogeneous graph attention network HAN; The time model includes any one or more combinations of an auto-regressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

[0023] As a preferred scheme, the node label prediction module classifies nodes based on the spatiotemporal embedding representation of the nodes to obtain the predicted labels of the devices corresponding to the nodes. A label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm. By setting multiple labels, multiple types of alarms of the devices corresponding to the nodes can be predicted.

[0024] As a preferred solution, the alarm analysis and evaluation module identifies potential faulty nodes and alarm types in the power optical communication network and evaluates the possible impact range when analyzing and evaluating the alarm conditions in the power optical communication network based on the predicted labels of the devices corresponding to the nodes.

[0025] In a third aspect, an electronic device is provided, comprising a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the power optical communication network alarm analysis method based on a spatiotemporal graph neural network as described in the first aspect.

[0026] In a fourth aspect, 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 method for analyzing alarms in an electric power optical communication network based on a spatiotemporal graph neural network as described in the first aspect is implemented.

[0027] Compared with the prior art, the first aspect of the present invention has at least the following beneficial effects: The present invention can effectively analyze and predict alarm events in the power optical communication network through a method based on a spatio-temporal graph neural network (Spatio-Temporal Graph Neural Network, ST-GNN). The power optical communication network is modeled as a graph structure containing spatiotemporal information, and then the graph structure is modeled using a spatio-temporal graph neural network. By combining spatial feature extraction and temporal feature extraction, the spatio-temporal embedding representation of the node is generated, and the spatio-temporal embedding representation of the node is analyzed to predict alarm events that may occur in the future. The present invention uses a graph neural network algorithm to accurately capture the spatio-temporal dependency in the network, improve the accuracy and timeliness of alarm prediction, and thus improve the stability and safety of the power system. Since the spatio-temporal graph neural network can simultaneously process the dependency of time and space under the same framework, the method of the present invention can more effectively capture the spatio-temporal dynamic characteristics of the power optical communication network. Compared with the method of processing time and space separately in the prior art, the present invention adopts a spatio-temporal graph neural network to more naturally integrate the two, thereby performing better in the accuracy and response speed of alarm analysis. Therefore, the alarm analysis method of the present invention can significantly improve the safety and stability of the power optical communication network and reduce the risk of failure.

[0028] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0030] Figure 1 Schematic diagram of the model structure of the prior art graph convolutional network; Figure 2 Flow chart of the method for analyzing alarms in a power optical communication network based on a spatiotemporal graph neural network according to an embodiment of the present invention; Figure 3 The embodiment of the present invention utilizes a spatiotemporal graph neural network to generate a spatiotemporal embedding representation flow chart of nodes. DETAILED DESCRIPTION

[0031] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0032] See also Figure 2 , the power optical communication network alarm analysis method based on the spatiotemporal graph neural network in the embodiment of the present invention includes: S1. Collecting data from the power optical communication network, and preprocessing the collected data to obtain preprocessed data; S2, modeling the preprocessed data into a graph structure; S3, using a spatiotemporal graph neural network to extract and learn the features of the nodes in the graph structure, combining spatial feature extraction and temporal feature extraction to generate a spatiotemporal embedding representation of the nodes; S4. Based on the spatiotemporal embedding representation of the nodes, the nodes are classified to obtain the predicted labels of the devices corresponding to the nodes; S5. Analyze and evaluate the impact of alarm conditions in the electric power optical communication network based on the prediction labels of the devices corresponding to the nodes.

[0033] In a possible implementation manner, the electric power optical communication network data collected in step S1 includes equipment information, connection status between equipments and historical alarm information of the electric power optical communication network.

[0034] In the step of preprocessing the collected data, the preprocessing includes operations such as alarm information collation and extraction, denoising, missing value processing, and elimination of duplicate records to ensure the quality and integrity of the data.

[0035] In a possible implementation, in step S2, the nodes in the graph structure represent devices, the edges represent connections between devices, and corresponding time dimension information is added to each time point, so that a graph is generated at each moment.

[0036] The present invention applies the spatiotemporal graph neural network (ST-GNN) to the alarm analysis of the power optical communication network, represents the power optical communication network as a graph structure containing spatiotemporal information, uses nodes to represent the communication devices in the network, and edges to represent the communication connections between devices. The ST-GNN model models the graph structure in a unified framework, combining temporal and spatial features.

[0037] In a possible implementation, in step S3, first, the spatial features of nodes and edges in the graph are extracted using a spatiotemporal graph neural network, and the dynamic changes of node states in the spatiotemporal graph neural network are captured by time steps; then, the spatiotemporal embedding representation of the nodes is generated by fusing the spatial features of these nodes and edges and the dynamic changes of node states. The spatiotemporal embedding representation of these nodes is used to predict future alarm events and generate early warning signals. Compared with traditional methods, the present invention can more effectively capture the spatiotemporal dependency of the power optical communication network, thereby improving the accuracy and timeliness of alarm analysis.

[0038] See also Figure 3 The spatio-temporal graph neural network (ST-GNN) of the embodiment of the present invention is a type of neural network model specifically used to process graph structure data with both temporal and spatial dependencies. ST-GNN can effectively capture the dynamic changes of nodes and edges in the graph and their mutual relationships by integrating graph neural network (GNN) and time series modeling technology in a unified framework.

[0039] Each time step in the spatiotemporal graph neural network is a graph and is passed through the graph neural network to obtain the resulting encoded graph that embeds the spatial interdependencies of the data. These encoded graphs can be modeled like time series data as long as the graph structure integrity of the data at each time step is preserved. Figure 3 These two steps are demonstrated in .

[0040] If the power optical communication network contains only one type of equipment, a homogeneous graph neural network model can be used; if the power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used. Among them, homogeneous graph neural network models include: graph convolution network GCN, graph sampling and aggregation network GraphSAGE, graph attention network GAT, etc.; heterogeneous graph neural network models include: heterogeneous graph neural network HGNN, heterogeneous graph attention network HAN, etc.

[0041] In a possible implementation, Figure 3 As shown, step S3 extracts spatial features through neighborhood aggregation using a graph neural network model, and extracts temporal features through time series capture using a time model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as:

[0042] In the formula, express The elements in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure and predict the graph structure at the current moment.

[0043] In a possible implementation, the time model of the embodiment of the present invention includes any sequence model such as an autoregressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

[0044] Graph Neural Network (GNN) is a type of deep learning model specifically designed to process graph structured data. Traditional neural networks and convolutional neural networks (CNN) mainly process Euclidean data (such as images and text), while GNN can effectively process non-Euclidean data (such as social networks, molecular structures, etc.) and perform tasks such as node classification, edge prediction, and graph classification on graph structures.

[0045] Graph Convolutional Network (GCN) is a classic model of graph neural network, which is used to process graph structure data. GCN expands the concept of traditional convolutional neural network, enabling it to perform convolution operations on graphs, thereby extracting local structural information in the graph. The core idea of ​​GCN is to generate a new feature representation of a node by aggregating the features of a node and its neighboring nodes. This process is similar to the convolution operation of convolutional neural network in images, but GCN performs convolution on the topological structure of the graph.

[0046] Graph Sample and Aggregate Network (GraphSAGE) is specifically designed to process large-scale graph data. Unlike traditional GCN, GraphSAGE generates node embedding representations by sampling and aggregating the neighbors of nodes, rather than performing convolution calculations on all neighbor nodes like GCN. The purpose of this is to improve computational efficiency, especially on large-scale graphs.

[0047] The Graph Attention Network (GAT) introduces an attention mechanism to dynamically assign different weights when aggregating neighbor node features. GAT can adjust the contribution of neighbor nodes to the representation of the target node according to their importance, allowing the model to learn complex relationships in the graph more flexibly and effectively.

[0048] Heterogeneous Graph Neural Network (HGNN) is a graph neural network model designed specifically for processing heterogeneous graphs. Nodes and edges in heterogeneous graphs can be of different types, which is common in many practical applications, such as social networks, recommendation systems, knowledge graphs, etc. HGNN uses these diverse structural features to capture the complex relationships in heterogeneous graphs, thereby generating more accurate node representations.

[0049] Heterogeneous Graph Attention Network (HAN) is a graph neural network model designed for heterogeneous graphs. It combines the attention mechanism of Graph Attention Network (GAT) and can effectively process complex graph structures containing different types of nodes and edges. Heterogeneous graphs refer to graphs containing multiple types of nodes and edges, which is common in many practical applications, such as social networks, knowledge graphs, etc.

[0050] AutoRegressive Integrated Moving Average (ARIMA) is a statistical model widely used in time series analysis. It combines the characteristics of autoregression (AR) and moving average (MA), and processes non-stationary time series through differential operations, making it suitable for predicting future values.

[0051] Recurrent Neural Network (RNN) is mainly used in natural language processing (NLP) tasks that need to process sequence information. The specific manifestation is that the network will remember the previous information and apply it to the calculation of the current output, that is, the nodes between the hidden layers are no longer disconnected but connected, and the input of the hidden layer includes not only the output of the input layer but also the output of the hidden layer at the previous moment.

[0052] Transformers are a deep learning model architecture for processing sequence data, first proposed by Vaswani et al. in 2017. Transformers have become the mainstream model in the field of natural language processing (NLP), and have also been widely used in other fields such as computer vision and time series analysis. The core idea is to use the self-attention mechanism to model the relationship between elements in the sequence, overcoming the limitations of traditional recurrent neural networks (RNNs) in processing long sequences.

[0053] In one possible implementation, in step S4, a label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm; by setting multiple labels, the embodiment of the present invention can predict multiple types of alarms of the device corresponding to the node.

[0054] In a possible implementation, step S5 analyzes and evaluates the alarm situation in the power optical communication network based on the predicted label of the device corresponding to the node, identifies potential faulty nodes and alarm types in the power optical communication network, and evaluates the possible impact range, thereby providing support and early warning information for decision-making.

[0055] In recent years, the development of deep learning, especially graph neural networks, has provided new possibilities for alarm analysis in power optical communication networks. Spatio-Temporal Graph Neural Network (ST-GNN) can simultaneously process spatial relationships in graph structures and dynamic changes in time series, and is suitable for alarm analysis and prediction tasks in power optical communication networks.

[0056] Another embodiment of the present invention further provides an electric power optical communication network alarm analysis system based on a spatiotemporal graph neural network, comprising: The data collection and preprocessing module is used to collect data from the power optical communication network, preprocess the collected data, and obtain preprocessed data; Graph structure modeling module, used to model the preprocessed data into a graph structure; A node spatiotemporal embedding representation generation module is used to extract and learn the features of the nodes in the graph structure using a spatiotemporal graph neural network, and generate a spatiotemporal embedding representation of the nodes by combining spatial feature extraction and temporal feature extraction; The node label prediction module is used to classify nodes based on the spatiotemporal embedding representation of nodes and obtain the predicted label of the device corresponding to the node; The alarm analysis and evaluation module is used to analyze and evaluate the impact of alarm conditions in the power optical communication network based on the prediction labels of the equipment corresponding to the nodes.

[0057] In a possible implementation, when the graph structure modeling module models the graph structure, the nodes in the graph structure represent devices, the edges represent connections between devices, and corresponding time dimension information is added to each time point, thereby generating a graph at each moment.

[0058] In a possible implementation, the node spatiotemporal embedding representation generation module uses a spatiotemporal graph neural network to extract spatial features of nodes and edges in the graph structure, and captures dynamic changes in node states in the spatiotemporal graph neural network through time steps; By fusing the spatial features of nodes and edges and the dynamic changes of node states, the spatiotemporal embedding representation of nodes is generated. The spatiotemporal embedding representation of nodes is used to predict future alarm events and generate early warning signals. Each time step in the spatiotemporal graph neural network is a graph, and is passed through the graph neural network to obtain a result encoding graph that embeds the spatial interdependencies of the data; if the electric power optical communication network contains only one type of equipment, a homogeneous graph neural network model can be used; if the electric power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used.

[0059] In a possible implementation, the node spatiotemporal embedding representation generation module extracts spatial features through neighborhood aggregation using a graph neural network model, and extracts temporal features through time series capture using a time model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as:

[0060] In the formula, express The elements in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure and predict the graph structure at the current moment.

[0061] In a possible implementation, the homogeneous graph neural network model includes any one or more combinations of a graph convolutional network GCN, a graph sampling and aggregation network GraphSAGE, and a graph attention network GAT; The heterogeneous graph neural network model includes any one or more combinations of a heterogeneous graph neural network HGNN and a heterogeneous graph attention network HAN; The time model includes any one or more combinations of an auto-regressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

[0062] In one possible implementation, the node label prediction module classifies nodes based on the spatiotemporal embedding representation of the nodes to obtain predicted labels for the devices corresponding to the nodes. A label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm. By setting multiple labels, multiple types of alarms for the devices corresponding to the nodes can be predicted.

[0063] In a possible implementation, the alarm analysis and evaluation module identifies potential faulty nodes and alarm types in the power optical communication network and evaluates the possible impact range when analyzing and evaluating the alarm conditions in the power optical communication network according to the predicted labels of the devices corresponding to the nodes.

[0064] The present invention designs a spatiotemporal graph modeling and alarm analysis process for the electric power optical communication network. Through data collection and preprocessing, spatiotemporal graph construction, feature extraction and representation learning, the spatiotemporal graph neural network ST-GNN model is used for training to obtain node prediction labels, and finally alarm analysis is performed based on the labels. By using the spatiotemporal graph neural network ST-GNN model, historical alarm information and equipment information are comprehensively analyzed from both time and space dimensions. This model combines the spatial feature extraction capability of the graph neural network and the sequence modeling capability of the time model. It can track the evolution of alarm events in the time dimension, and capture the association and mutual influence between devices in the spatial dimension, thereby achieving accurate prediction of future alarm conditions.

[0065] Another embodiment of the present invention further proposes an electronic device, including a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the power optical communication network alarm analysis method based on the spatiotemporal graph neural network.

[0066] Another embodiment of the present invention further proposes a computer-readable storage medium, which stores at least one instruction. When the at least one instruction is executed by a processor, the power optical communication network alarm analysis method based on the spatiotemporal graph neural network is implemented.

[0067] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, USB flash drive, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals. For ease of explanation, the above content only shows the part related to the embodiment of the present invention. For specific technical details not disclosed, please refer to the method part of the embodiment of the present invention. The computer-readable storage medium is non-temporary and can be stored in a storage device formed by various electronic devices, which can realize the execution process recorded in the method of the embodiment of the present invention.

[0068] It should be understood by those skilled in the art that the 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 aspects. Moreover, 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.

[0069] 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 process in the flowchart. 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.

[0070] 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.

[0071] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0072] 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 method for analyzing alarms in a power optical communication network based on a spatiotemporal graph neural network, characterized in that: include: Collecting data from the electric power optical communication network, and preprocessing the collected data to obtain preprocessed data; Model the preprocessed data into a graph structure; The spatiotemporal graph neural network is used to extract and learn the features of the nodes in the graph structure, and the spatiotemporal embedding representation of the nodes is generated by combining the spatial feature extraction and the temporal feature extraction; Based on the spatiotemporal embedding representation of nodes, nodes are classified to obtain the predicted labels of the devices corresponding to the nodes. According to the predicted labels of the devices corresponding to the nodes, the alarm situations in the electric power optical communication network are analyzed and the impacts are evaluated.

2. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 1 is characterized in that: In the step of collecting the power optical communication network data, the power optical communication network data includes the equipment information, the connection status between the equipments and the historical alarm information of the power optical communication network.

3. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 1 is characterized in that: In the step of preprocessing the collected data, the preprocessing includes one or more of sorting and extracting alarm information, denoising, processing missing values, and eliminating duplicate records on the collected data.

4. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 1 is characterized in that: In the step of modeling the preprocessed data into a graph structure, the nodes in the graph structure represent devices, the edges represent connections between devices, and corresponding time dimension information is added to each time point, so as to generate a graph at each moment.

5. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 4 is characterized in that: The step of extracting and learning the features of the nodes in the graph structure by using the spatiotemporal graph neural network, combining spatial feature extraction and temporal feature extraction, and generating a spatiotemporal embedding representation of the nodes includes: The spatial features of nodes and edges in the graph are extracted using a spatiotemporal graph neural network, and the dynamic changes of node states in the spatiotemporal graph neural network are captured through time steps. The spatiotemporal embedding representation of the nodes is generated by fusing the spatial features of nodes and edges and the dynamic changes of node states. The spatiotemporal embedding representation of the nodes is used to predict future alarm events and generate early warning signals.

6. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 5 is characterized in that: Each time step in the spatiotemporal graph neural network is a graph, and is passed through the graph neural network to obtain a result encoding graph that embeds the spatial interdependencies of the data; if the electric power optical communication network contains only one type of equipment, a homogeneous graph neural network model can be used; if the electric power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used.

7. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 6 is characterized in that: The homogeneous graph neural network model includes any one or more combinations of the graph convolution network GCN, the graph sampling and aggregation network GraphSAGE, and the graph attention network GAT; the heterogeneous graph neural network model includes any one or more combinations of the heterogeneous graph neural network HGNN and the heterogeneous graph attention network HAN.

8. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 6 is characterized in that: In the step of extracting and learning the features of the nodes in the graph structure by using the spatiotemporal graph neural network, combining the spatial feature extraction and the temporal feature extraction to generate the spatiotemporal embedding representation of the nodes, the spatial features are extracted by neighborhood aggregation through the graph neural network model, and the temporal features are extracted by time series capture through the time model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as: In the formula, express The element in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure to predict the current graph structure; The time model includes any one or more combinations of an auto-regressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

9. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 1, characterized in that: In the step of classifying nodes based on the spatiotemporal embedding representation of nodes to obtain predicted labels for devices corresponding to the nodes, a label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm; by setting multiple labels, multiple types of alarms for devices corresponding to the nodes can be predicted.

10. The method for analyzing power optical communication network alarms based on spatiotemporal graph neural network according to claim 1, characterized in that: In the step of analyzing and impact assessing the alarm situation in the electric power optical communication network according to the prediction label of the device corresponding to the node, potential fault nodes and alarm types of the electric power optical communication network are identified, and the possible impact range is assessed.

11. An electric power optical communication network alarm analysis system based on spatiotemporal graph neural network, characterized in that: include: The data collection and preprocessing module is used to collect data from the power optical communication network, preprocess the collected data, and obtain preprocessed data; Graph structure modeling module, used to model the preprocessed data into a graph structure; A node spatiotemporal embedding representation generation module is used to extract and learn the features of the nodes in the graph structure using a spatiotemporal graph neural network, and generate a spatiotemporal embedding representation of the nodes by combining spatial feature extraction and temporal feature extraction; The node label prediction module is used to classify nodes based on the spatiotemporal embedding representation of nodes and obtain the predicted label of the device corresponding to the node; The alarm analysis and evaluation module is used to analyze and evaluate the impact of alarm conditions in the power optical communication network based on the prediction labels of the equipment corresponding to the nodes.

12. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 11, characterized in that: When the graph structure modeling module models the graph structure, the nodes in the graph structure represent devices, the edges represent the connections between devices, and corresponding time dimension information is added to each time point, so as to generate a graph at each moment.

13. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 12, characterized in that: The node spatiotemporal embedding representation generation module uses a spatiotemporal graph neural network to extract the spatial features of nodes and edges in the graph structure, and captures the dynamic changes of node states in the spatiotemporal graph neural network through time steps; By fusing the spatial features of nodes and edges and the dynamic changes of node states, the spatiotemporal embedding representation of nodes is generated. The spatiotemporal embedding representation of nodes is used to predict future alarm events and generate early warning signals. Each time step in the spatiotemporal graph neural network is a graph, and is passed through the graph neural network to obtain a result encoding graph that embeds the spatial interdependencies of the data; if the electric power optical communication network contains only one type of equipment, a homogeneous graph neural network model can be used; if the electric power optical communication network contains multiple types of equipment, a heterogeneous graph neural network model is used.

14. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 13, characterized in that: The node spatiotemporal embedding representation generation module extracts spatial features through neighborhood aggregation using a graph neural network model, and extracts temporal features through time series capture using a time model. The mathematical expression of the graph neural network model is: , the mathematical expression of the time model is ; Build graph models based on time. express The adjacency matrix of the time graph is formulated as: In the formula, express The element in the adjacency matrix at that moment, if the node and nodes There are edges connecting them, then The value of is 1; Represents the initial feature matrix of the nodes in the graph; Is a learnable weight matrix used to linearly transform the node embedding representation after convolution; use The result after activating the linear transformation is The node embedding matrix at time; Use the time model to analyze the historical graph structure and predict the graph structure at the current moment.

15. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 14, characterized in that: The homogeneous graph neural network model includes any one or more combinations of a graph convolutional network GCN, a graph sampling and aggregation network GraphSAGE, and a graph attention network GAT; The heterogeneous graph neural network model includes any one or more combinations of a heterogeneous graph neural network HGNN and a heterogeneous graph attention network HAN; The time model includes any one or more combinations of an auto-regressive integral moving average model (ARIMA), a recurrent neural network (RNN), and transformers.

16. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 11, characterized in that: The node label prediction module classifies nodes based on the spatiotemporal embedding representation of the nodes to obtain the predicted labels of the devices corresponding to the nodes. A label of 0 indicates that the device corresponding to the node will not generate a specific type of alarm, and a label of 1 indicates that the device corresponding to the node will generate this type of alarm. By setting multiple labels, multiple types of alarms of the devices corresponding to the nodes can be predicted.

17. The power optical communication network alarm analysis system based on spatiotemporal graph neural network according to claim 11, characterized in that: The alarm analysis and evaluation module analyzes and evaluates the alarm situation in the power optical communication network according to the prediction label of the device corresponding to the node, identifies the potential fault nodes and alarm types of the power optical communication network, and evaluates the possible impact range.

18. 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 power optical communication network alarm analysis method based on a spatiotemporal graph neural network as described in any one of claims 1 to 10.

19. 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, it implements the power optical communication network alarm analysis method based on the spatiotemporal graph neural network as described in any one of claims 1 to 10.

Images