Transient voltage evaluation method and device based on power grid topological structure

Abstract

The invention provides a transient voltage evaluation method and device based on a power grid topological structure, and the method comprises the steps: obtaining heterogeneous graph data of the power grid topological structure, the heterogeneous graph data comprises initial node features and initial edge features, and nodes of different node types have different node features; performing feature processing on the initial node features and the initial edge features to generate current node features and current edge features; through a pre-constructed heterogeneous graph attention neural network evaluation model, transient voltage evaluation is carried out according to the current node features and the current edge features, a transient voltage evaluation result of each node is generated, accurate transient voltage online evaluation can be carried out for an anticipated accident before a fault occurs, and the method is suitable for an operation scene with frequent topology change; by fully considering the high-dimensional interaction relationship between each node feature and the line feature under different operation conditions, the model evaluation performance for the anticipated fault is effectively improved, the model generalization ability is relatively high, and the evaluation precision is relatively high.

Description

Technical Field

[0001] The present invention relates to the technical field of power system voltage safety, and in particular to a transient voltage assessment method and device based on a power grid topology structure. Background Art

[0002] In recent years, with the increasing penetration rate of renewable energy in the power system, both the transmission network and the distribution network have shown strong volatility and uncertainty, and the daily operation mode of the power grid has changed greatly. Due to the relative scarcity of flexible and adjustable resources in the system, the safety and stability margin of the power system has been continuously compressed, and the operating point is closer to the boundary of the safety domain than before, making the online safety and stability analysis of the power grid more and more important. Different from model-driven analysis methods such as numerical simulation, transient energy function, and bifurcation theory, data-driven machine learning provides the possibility for online analysis and control of transient safety of complex large-scale power grids. For the transient voltage safety problem of the system, it has a strong nonlinear relationship with the operating conditions, equipment control parameters, fault location, and network topology. At present, transient voltage safety assessment is mostly oriented to transient scenarios after faults. Based on the wide-area measurement system, the transient time series data of each node is used as input, while the change trajectory of the state quantity contains rich information such as component dynamics and fault location. It is impossible to perform accurate transient voltage safety assessment before the fault. In the face of scenarios with frequent changes in topology structure, the assessment accuracy is lacking, and the model generalization ability is poor. Summary of the invention

[0003] One object of the present invention is to provide a transient voltage assessment method based on a power grid topology structure, which can perform accurate online transient voltage assessment for anticipated accidents before a fault occurs, and is suitable for operation scenarios with frequent topology changes; by fully considering the high-dimensional interactive relationship between the characteristics of each node and the characteristics of the line under different operating conditions, the model assessment performance for anticipated faults is effectively improved, the model has strong generalization ability, and the assessment accuracy is high. Another object of the present invention is to provide a transient voltage assessment device based on a power grid topology structure. Another object of the present invention is to provide a computer-readable medium. Another object of the present invention is to provide a computer device.

[0004] In order to achieve the above objectives, the present invention discloses a transient voltage assessment method based on a power grid topology structure, comprising:

[0005] Obtain heterogeneous graph data of the power grid topology structure, the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features;

[0006] Perform feature processing on the initial node features and the initial edge features to generate current node features and current edge features;

[0007] Through the pre-built heterogeneous graph attention neural network evaluation model, transient voltage evaluation is performed according to the current node characteristics and current edge characteristics to generate transient voltage evaluation results for each node.

[0008] Preferably, the method further comprises:

[0009] The topology of the power grid is constructed by taking generators, loads, high-voltage direct current and distributed renewable energy as nodes and transmission lines as edges;

[0010] The node characteristics of the generator node include bus voltage amplitude, bus phase angle, generator active power and generator reactive power;

[0011] The node characteristics of the load node include bus voltage amplitude, bus phase angle, load active power, load reactive power and induction motor proportion;

[0012] The node characteristics of the HVDC node include bus voltage amplitude, bus phase angle, HVDC active power and HVDC reactive power;

[0013] The node characteristics of distributed renewable energy include bus voltage amplitude, bus phase angle, distributed renewable energy active power and distributed renewable energy reactive power;

[0014] The edge characteristics of the transmission line include equivalent resistance, reactance, active power, reactive power and line state quantity.

[0015] Preferably, feature processing is performed on the initial node features and the initial edge features to generate current node features and current edge features, including:

[0016] Normalizing the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features;

[0017] The normalized node features and the normalized edge features are projected into the shared space through virtual isomorphism technology to obtain the current node features and the current edge features.

[0018] Preferably, the heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer;

[0019] Through the pre-built heterogeneous graph attention neural network evaluation model, transient voltage evaluation is performed according to the current node characteristics and current edge characteristics, and the transient voltage evaluation results of each node are generated, including:

[0020] Through the feature conversion layer, the current node feature and the current edge feature are dimensionally expanded and converted to generate high-dimensional node features and high-dimensional edge features;

[0021] Through the feature transfer layer, feature fusion and feature transfer are performed on high-dimensional node features and high-dimensional edge features to generate updated node features;

[0022] Through the feature readout layer, the updated node features are pooled and dimension reduced to generate transient voltage assessment results.

[0023] Preferably, the feature transfer layer is used to perform feature fusion and feature transfer on high-dimensional node features and high-dimensional edge features to generate updated node features, including:

[0024] Through the multi-layer perceptron, high-dimensional edge features are mapped to generate an edge feature matrix;

[0025] Performing feature fusion on the edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix to generate fused features;

[0026] Perform message transmission based on fusion features, high-dimensional edge features, and high-dimensional node features to generate a message feature transmission vector;

[0027] Through the gated recurrent unit, feature updates are performed based on high-dimensional node features and message feature transfer vectors to generate updated node features.

[0028] Preferably, message transmission is performed according to the fusion feature, the high-dimensional edge feature and the high-dimensional node feature to generate a message feature transmission vector, including:

[0029] Concatenate the fusion feature with the high-dimensional edge feature to obtain a first concatenated feature;

[0030] Generate an attention coefficient according to the first concatenated feature, the high-dimensional node feature and the preset first weight matrix through a shared attention mechanism;

[0031] Based on the attention coefficient, message aggregation is performed on the first concatenated feature and the preset second weight matrix to generate a message feature transfer vector.

[0032] Preferably, the update node features are pooled and dimensionally reduced through the feature readout layer to generate transient voltage assessment results, including:

[0033] Perform pooling processing on the update node features to generate pooled features;

[0034] Concatenate the pooled feature with the updated node feature to generate a second concatenated feature;

[0035] The second concatenated feature is transformed through a multi-layer perceptron to generate a transient voltage assessment result.

[0036] Preferably, the method further comprises:

[0037] Obtain sample graph data and corresponding real indicators of transient voltage;

[0038] Through the initial graph neural network model, the sample graph data is simulated and calculated to generate transient voltage simulation indicators;

[0039] Through the preset loss function, the transient voltage simulation index and the transient voltage real index are calculated to generate the loss result;

[0040] Through the back-propagation algorithm, the model parameters of the initial graph neural network model are updated according to the loss results, and a heterogeneous graph attention neural network evaluation model is constructed.

[0041] Preferably, the method further comprises:

[0042] Obtain training graph data and predefined transient voltage safety quantitative indicators;

[0043] By training graph data and transient voltage safety quantitative indicators, the graph neural network model is trained to construct the initial graph neural network model.

[0044] The present invention also discloses a transient voltage evaluation device based on a power grid topology structure, comprising:

[0045] A heterogeneous graph data acquisition unit is used to acquire heterogeneous graph data of a power grid topology structure, wherein the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features;

[0046] A feature processing unit, used for performing feature processing on the initial node feature and the initial edge feature to generate the current node feature and the current edge feature;

[0047] The transient voltage evaluation unit is used to perform transient voltage evaluation according to the current node characteristics and the current edge characteristics through a pre-built heterogeneous graph attention neural network evaluation model, and generate transient voltage evaluation results for each node.

[0048] Preferably, the device further comprises:

[0049] Determine units for constructing a power grid topology structure by taking generators, loads, high voltage direct current and distributed renewable energy as nodes and transmission lines as edges;

[0050] The node characteristics of the generator node include bus voltage amplitude, bus phase angle, generator active power and generator reactive power;

[0051] The node characteristics of the load node include bus voltage amplitude, bus phase angle, load active power, load reactive power and induction motor proportion;

[0052] The node characteristics of the HVDC node include bus voltage amplitude, bus phase angle, HVDC active power and HVDC reactive power;

[0053] The node characteristics of distributed renewable energy include bus voltage amplitude, bus phase angle, distributed renewable energy active power and distributed renewable energy reactive power;

[0054] The edge characteristics of the transmission line include equivalent resistance, reactance, active power, reactive power and line state quantity.

[0055] Preferably, the feature processing unit is specifically used to normalize the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features; and project the normalized node features and the normalized edge features to a shared space respectively through virtual isomorphism technology to obtain current node features and current edge features.

[0056] Preferably, the heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer;

[0057] The transient voltage assessment unit is specifically used to perform dimension expansion conversion on the current node features and the current edge features through the feature conversion layer to generate high-dimensional node features and high-dimensional edge features; to perform feature fusion and feature transfer on the high-dimensional node features and high-dimensional edge features through the feature transfer layer to generate updated node features; and to perform pooling and dimensionality reduction processing on the updated node features through the feature readout layer to generate transient voltage assessment results.

[0058] Preferably, the transient voltage evaluation unit is specifically used to map high-dimensional edge features through a multi-layer perceptron to generate an edge feature matrix; perform feature fusion on the edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix to generate a fused feature; perform message transmission based on the fused features, high-dimensional edge features and high-dimensional node features to generate a message feature transmission vector; perform feature update based on the high-dimensional node features and the message feature transmission vector through a gated recurrent unit to generate an updated node feature.

[0059] Preferably, the transient voltage evaluation unit is specifically used to splice the fused feature with the high-dimensional edge feature to obtain a first spliced ​​feature; generate an attention coefficient according to the first spliced ​​feature, the high-dimensional node feature and the preset first weight matrix through a shared attention mechanism; based on the attention coefficient, perform message aggregation on the first spliced ​​feature and the preset second weight matrix to generate a message feature transfer vector.

[0060] Preferably, the transient voltage assessment unit is specifically used to perform pooling processing on the update node features to generate pooled features; splice the pooled features with the update node features to generate second spliced ​​features; and transform the second spliced ​​features through a multi-layer perceptron to generate a transient voltage assessment result.

[0061] Preferably, the device further comprises:

[0062] A sample data acquisition unit, used to acquire sample graph data and corresponding transient voltage real indicators;

[0063] A simulation calculation unit, used to perform simulation calculation on sample graph data through an initial graph neural network model to generate transient voltage simulation indicators;

[0064] A loss calculation unit, used to perform loss calculation on the transient voltage simulation index and the transient voltage real index through a preset loss function to generate a loss result;

[0065] The model optimization unit is used to update the model parameters of the initial graph neural network model according to the loss results through the back-propagation algorithm, and to build a heterogeneous graph attention neural network evaluation model.

[0066] Preferably, the device further comprises:

[0067] A training data acquisition unit, used to acquire training graph data and predefined transient voltage safety quantitative indicators;

[0068] The model training unit is used to train the graph neural network model by training graph data and transient voltage safety quantitative indicators, and to build an initial graph neural network model.

[0069] The present invention also discloses a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the method described above is implemented.

[0070] The present invention also discloses a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, the processor is used to control the execution of program instructions, and the processor implements the above method when executing the program.

[0071] The present invention also discloses a computer program product, including a computer program / instruction, and the method described above is implemented when the computer program / instruction is executed by a processor.

[0072] The present invention obtains heterogeneous graph data of a power grid topology structure, the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; the initial node features and initial edge features are feature processed to generate current node features and current edge features; a transient voltage assessment is performed according to the current node features and the current edge features through a pre-constructed heterogeneous graph attention neural network assessment model, and a transient voltage assessment result of each node is generated. Before a fault occurs, accurate transient voltage online assessment can be performed for anticipated accidents, and the method is suitable for operating scenarios with frequent topology changes; by fully considering the high-dimensional interactive relationship between the features of each node and the line features under different operating conditions, the model assessment performance for anticipated faults is effectively improved, the model has strong generalization ability, and the assessment accuracy is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0074] Figure 1 A flowchart of a transient voltage assessment method based on a power grid topology structure provided by an embodiment of the present invention;

[0075] Figure 2 A flowchart of another method for transient voltage assessment based on a power grid topology structure provided by an embodiment of the present invention;

[0076] Figure 3 A flowchart of a method for constructing a heterogeneous graph attention neural network evaluation model provided by an embodiment of the present invention;

[0077] Figure 4 A schematic diagram of the structure of a transient voltage assessment device based on a power grid topology structure provided by an embodiment of the present invention;

[0078] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] It should be noted that the method and device for transient voltage assessment based on power grid topology structure disclosed in the present application can be used in the field of artificial intelligence technology, and can also be used in any field outside the field of artificial intelligence technology. The application field of the method and device for transient voltage assessment based on power grid topology structure disclosed in the present application is not limited.

[0081] In order to facilitate the understanding of the technical solution provided by the present application, the relevant contents of the technical solution of the present application are first explained below. For the transient voltage safety assessment for anticipated accidents, it depends on an effective evaluation model construction method, and it is necessary to deeply explore the complex high-dimensional relationship between the current steady-state operating conditions and the transient process of the system under the anticipated fault. Due to the locality of reactive voltage, the model must fully consider the impact of network topology and anticipated fault location on the transient interaction of the system, which puts higher requirements on the design of the evaluation model. Online evaluation in transient voltage stability assessment usually refers to a method that can analyze and judge the transient voltage stability of the power system in real time or near real time. This type of evaluation usually requires a rapid response to disturbances in the power system, such as faults or load changes, and gives a conclusion on whether the system can maintain voltage stability. The online evaluation method needs to be efficient, accurate and real-time to meet the needs of safe and stable operation of the power system.

[0082] Based on the graph neural network message passing mechanism, this application constructs a heterogeneous graph attention neural network with edge feature integration to improve evaluation performance and topology generalization capabilities. The edge features are fused with adjacent nodes to reflect the impact of line transmission, edge features are introduced in feature aggregation to reflect the impact of the line's own features, and a multi-head attention mechanism is introduced to reflect the differences in the impact of different node features. Among them, considering the impact of the characteristics of various heterogeneous devices in the power grid, a graph data construction method based on virtual isomorphism is proposed to achieve quantitative evaluation of each node in the online operation scenario.

[0083] The following takes the transient voltage assessment device based on the power grid topology structure as an example of the execution subject to illustrate the implementation process of the transient voltage assessment method based on the power grid topology structure provided by the embodiment of the present invention. It can be understood that the execution subject of the transient voltage assessment method based on the power grid topology structure provided by the embodiment of the present invention includes but is not limited to the transient voltage assessment device based on the power grid topology structure.

[0084] Figure 1 A flowchart of a transient voltage assessment method based on a power grid topology structure provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes:

[0085] Step 101: Obtain heterogeneous graph data of a power grid topology structure, where the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features.

[0086] Step 102: Perform feature processing on the initial node features and the initial edge features to generate current node features and current edge features.

[0087] Step 103: Perform transient voltage evaluation based on current node features and current edge features through a pre-built heterogeneous graph attention neural network evaluation model to generate transient voltage evaluation results for each node.

[0088] In an embodiment of the present invention, the heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer.

[0089] In the technical solution provided by the embodiment of the present invention, heterogeneous graph data of the power grid topology structure is obtained, and the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; the initial node features and initial edge features are feature processed to generate current node features and current edge features; transient voltage evaluation is performed according to the current node features and current edge features through a pre-constructed heterogeneous graph attention neural network evaluation model, and a transient voltage evaluation result of each node is generated, which can perform accurate transient voltage online evaluation for anticipated accidents before a fault occurs, and is suitable for operating scenarios with frequent topology changes; by fully considering the high-dimensional interactive relationship between the characteristics of each node and the line characteristics under different operating conditions, the model evaluation performance for anticipated faults is effectively improved, the model has strong generalization ability, and high evaluation accuracy.

[0090] Figure 2 A flowchart of another method for transient voltage assessment based on a power grid topology structure provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, the method includes:

[0091] Step 201: construct a power grid topology structure by taking generators, loads, high voltage direct current (LCC-HVDC) and distributed renewable energy (DR) as nodes and transmission lines as edges.

[0092] In the embodiment of the present invention, each step is performed by a transient voltage assessment device based on a power grid topology structure.

[0093] In the embodiment of the present invention, the generator, load, high voltage direct current (LCC-HVDC) and distributed renewable energy (DR) are key components of the power grid system. The power grid system may also include other devices, which are not listed here one by one.

[0094] In an embodiment of the present invention, if there is interaction between the generator, the load, the high-voltage direct current (LCC-HVDC) and the distributed renewable energy (DR), there is a transmission line between the two; the generator, the load, the high-voltage direct current (LCC-HVDC) and the distributed renewable energy (DR) are taken as nodes, and the transmission line is taken as an edge to draw a power grid topology structure, and the power grid topology structure belongs to a heterogeneous graph, and the nodes and edges have different types.

[0095] In the embodiment of the present invention, in the power system analysis, the use of heterogeneous graph models can help researchers and engineers better understand and analyze the structural characteristics, dynamic behavior and operating status of the power grid. In addition, heterogeneous graph models also provide the possibility of applying graph theory, machine learning and network science methods to solve problems such as power grid planning, operation and optimization.

[0096] Step 202: Obtain heterogeneous graph data of the power grid topology structure, where the heterogeneous graph data includes initial node features and initial edge features.

[0097] In the embodiment of the present invention, based on the steady-state operating conditions of the system, nodes of different node types have different node characteristics.

[0098] Specifically, the node characteristics of the generator node include the bus voltage amplitude V bus 、Bus phase angle θ bus , Generator active power P G and generator reactive power Q G .

[0099] The node characteristics of the load node include bus voltage amplitude V bus 、Bus phase angle θ bus , Load active power P L , load reactive power Q L and induction motor ratio R IM .

[0100] The node characteristics of the high-voltage DC node include the bus voltage amplitude V bus 、Bus phase angle θ bus , high voltage DC active power P HVDC and high voltage DC reactive power Q HVDC .

[0101] The node characteristics of distributed renewable energy include bus voltage amplitude V bus 、Bus phase angle θ bus , Distributed new energy active power P DR and distributed renewable energy reactive power Q DR .

[0102] The edge characteristics of the transmission line include the equivalent resistance R e Reactance Xe , Active power P e , reactive power Q e and line state quantity fault. That is: 5-dimensional vector [R e , X e , P e , Q e , fault].

[0103] Among them, the bus voltage amplitude V bus and bus phase angle θ bus It is a common feature of all nodes in the heterogeneous graph of the power grid topology. Considering that the transient voltage dynamics are strongly related to the induction motor load, the load is modeled as a combination of the induction motor dynamic model and the ZIP static model in the simulation. IM It describes the amount of power contributed by the induction motor to the total load and is therefore chosen as the key characteristic of the load.

[0104] It is worth noting that the ZIP model refers to a static load model that can describe the static characteristics of the load at different voltage levels. The ZIP model usually consists of three parts: the power consumed by constant impedance (Z), the power consumed by constant current (I), and the power consumed by constant power (P). This model can be used as a tool for physical simulation or mathematical description of load characteristics in power system analysis and calculation.

[0105] In the embodiment of the present invention, in order to describe the expected fault information, one-hot encoding is introduced in the edge feature construction. For the line state quantity fault, "fault=1" indicates that the line is set as an expected fault line, and "fault=0" indicates that the line is a normal line.

[0106] Step 203: normalize the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features.

[0107] In the embodiment of the present invention, the initial node features and the initial edge features are respectively normalized by maximum-minimum normalization to obtain normalized node features and normalized edge features.

[0108] Taking node as an example, through Normalize the initial node features to obtain normalized node features. v,i is the normalized node feature, x i is the eigenvalue of the i-th node feature, is the minimum eigenvalue of the i-th node feature, is the maximum eigenvalue of the feature of the i-th node.

[0109] It is worth noting that the normalization process of edge features is the same as the normalization process of node features, and will not be repeated here.

[0110] In the embodiment of the present invention, after feature normalization processing, all node features and edge features in the heterogeneous graph data are mapped to the [0, 1] interval, thereby improving subsequent calculation efficiency.

[0111] Step 204: project the normalized node features and the normalized edge features into the shared space respectively through the virtual isomorphism technology to obtain the current node features and the current edge features.

[0112] In the actual physical process, the interaction between the buses is completed based on the transmission line, and the interaction mode has nothing to do with the node type. Therefore, virtual isomorphism technology is introduced to deal with the heterogeneous characteristics of nodes.

[0113] Taking node as an example, through The normalized node features and the normalized edge features are projected into the shared space to obtain the current node features and the current edge features. is the projection feature vector of node i, that is, the current node feature; x v,i is the normalized node feature; is the feature transformation matrix for nodes of type ψ.

[0114] It is worth noting that different features of heterogeneous node types are projected into the shared space, and features that do not exist in the node are represented by 0.

[0115] It is worth noting that the virtual isomorphism processing of edge features is the same as that of node features, and will not be repeated here.

[0116] Step 205: Perform dimension expansion conversion on the current node feature and the current edge feature through the feature conversion layer to generate high-dimensional node features and high-dimensional edge features.

[0117] In the embodiment of the present invention, in the feature conversion layer, a multi-layer perceptron (MLP) is used to expand the features of each node and edge in the heterogeneous graph data from low dimensions to high dimensions. The MLP is composed of a fully connected neural network, and the MLP parameters of the nodes and edges in the heterogeneous graph data are shared separately. The rectified linear unit (ReLU) is selected as the activation function of each layer of the fully connected neural network.

[0118] Specifically, through Perform dimension expansion transformation on the current node feature to generate high-dimensional node features. is the current node feature; is the feature conversion function of the node; It is a high-dimensional node feature.

[0119] Specifically, through Perform dimension expansion transformation on the current edge feature to generate a high-dimensional edge feature. e,ij is the current edge feature; is the characteristic conversion function of the edge; It is a high-dimensional edge feature.

[0120] Step 206: Through the feature transfer layer, feature fusion and feature transfer are performed on the high-dimensional node features and the high-dimensional edge features to generate updated node features.

[0121] In the embodiment of the present invention, step 206 specifically includes:

[0122] Step 2061: Map the high-dimensional edge features through a multi-layer perceptron (MLP) to generate an edge feature matrix.

[0123] Specifically, through Map the high-dimensional edge features into a matrix form to generate an edge feature matrix. is the edge feature matrix of edge ij; is the characteristic matrix mapping function of the edge; It is a high-dimensional edge feature.

[0124] Step 2062: perform feature fusion on the edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix to generate a fused feature.

[0125] Specifically, through The edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix are fused to generate fused features. To fusion features, is the edge feature matrix of edge ij, It is the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix.

[0126] In the embodiment of the present invention, the fusion feature indicates that the coupling of the edge is less than , which includes the transmission influence of the transmission line on the hidden characteristics of the voltage of the adjacent node.

[0127] Step 2063: Perform message transmission according to the fusion features, high-dimensional edge features, and high-dimensional node features to generate a message feature transmission vector.

[0128] In the embodiment of the present invention, step 2063 specifically includes:

[0129] Step a1: concatenate the fusion feature with the high-dimensional edge feature to obtain a first concatenated feature.

[0130] In the embodiment of the present invention, the expected fault information in the line edge feature, that is, the line state quantity fault, directly determines the voltage drop center and disturbance area in the system. Therefore, the fusion feature is spliced ​​with the high-dimensional edge feature to consider the common influence, that is: the first splicing feature

[0131] Step a2: Generate an attention coefficient through a shared attention mechanism according to the first concatenated feature, the high-dimensional node feature and the preset first weight matrix.

[0132] In the embodiment of the present invention, in order to reflect the differences in the impact of different neighborhood nodes and edges on the characteristics of the central node, a shared attention mechanism is adopted. The shared attention mechanism is a single-layer feedforward neural network, and the attention coefficient is calculated as follows:

[0133]

[0134] Among them, α ij is the attention coefficient, which indicates the importance of the neighboring node j to the central node i in the heterogeneous graph data, and also involves the characteristics of the edges between nodes The impact of T is the transpose of the preset first weight matrix; is a high-dimensional node feature; It is the first splicing feature.

[0135] It is worth noting that the first weight matrix can be set according to actual needs, and the embodiment of the present invention does not limit this.

[0136] Step a3: Based on the attention coefficient, message aggregation is performed on the first concatenated feature and the preset second weight matrix to generate a message feature transfer vector.

[0137] In the embodiment of the present invention, based on the attention coefficient, the message features of the neighboring nodes and edges are aggregated, wherein multiple independent and parallel attention mechanisms are adopted to increase the model capacity. The message aggregation calculation method is as follows:

[0138]

[0139] Among them, α ij is the attention coefficient; is the message feature transfer vector of node i; σ is the activation function based on sigmoid; k is the kth attention mechanism; K is the number of attention heads; W h is the second weight matrix; C represents the averaging operation of each attention mechanism result; It is the first splicing feature.

[0140] In the embodiment of the present invention, by splicing the neighborhood node features and edge features Fully consider the impact of information such as expected line faults to improve model accuracy.

[0141] It is worth noting that the second weight matrix can be set according to actual needs, and the embodiment of the present invention does not limit this.

[0142] Step 2064: Perform feature update based on the high-dimensional node features and the message feature transfer vector through a gated recurrent unit (GRU) to generate updated node features.

[0143] In the embodiment of the present invention, GRU is a variant of a recurrent neural network, which uses weights to process transfer features and current features.

[0144] Specifically, through The high-dimensional node features and message feature transfer vectors are calculated to generate updated node features. To update node features, is a high-dimensional node feature, is the message feature delivery vector of node i.

[0145] Step 207: Perform pooling and dimensionality reduction processing on the updated node features through the feature readout layer to generate transient voltage assessment results.

[0146] In the embodiment of the present invention, after being processed by the feature transfer layer, the hidden features of each node contain information related to transient voltage safety, and are decoded into final quantitative evaluation indicators at the readout layer.

[0147] In the embodiment of the present invention, step 207 specifically includes:

[0148] Step 2071: perform pooling processing on the update node features to generate pooled features.

[0149] In the embodiment of the present invention, in order to enhance the perception ability of each node in the graph data to the global voltage, a pooling operation based on Set2Set is used to perform feature compression on the nodes in the entire graph.

[0150] Specifically, through Pooling is performed on the update node features to generate pooled features. pool is the global hidden feature after the graph pooling operation, that is, the pooling feature, It is the node feature after the node passes through the feature transfer layer, including local voltage information, that is, updating the node feature.

[0151] Step 2072: Concatenate the pooled feature and the updated node feature to generate a second concatenated feature.

[0152] In the embodiment of the present invention, after the heterogeneous graph nodes are pooled, the pooling result is spliced ​​with the hidden features of each node to generate a second splicing feature

[0153] Step 2073: transform the second splicing feature through MLP to generate a transient voltage assessment result.

[0154] In the embodiment of the present invention, MLP is used to transform the high-dimensional second splicing features of each node into the required transient voltage safety assessment index:

[0155]

[0156] Among them, y i is the transient voltage evaluation result, It is the second splicing feature.

[0157] This application fully considers the high-dimensional interactive relationship between the characteristics of each node and the line characteristics under different operating conditions, effectively improves the model evaluation performance for anticipated faults, and can still have a high evaluation accuracy after the power grid topology changes, and is suitable for operating scenarios with frequent topology changes.

[0158] Furthermore, a heterogeneous graph attention neural network evaluation model is pre-built. Figure 3 A flowchart of a method for constructing a heterogeneous graph attention neural network evaluation model provided by an embodiment of the present invention, such as Figure 3 As shown, the method includes:

[0159] Step 301: Acquire training graph data and predefined transient voltage safety quantitative indicators.

[0160] In an embodiment of the present invention, the training graph data may be historical data or defined data of a power grid topology heterogeneous graph, including but not limited to the bus voltage amplitude, bus phase angle, active power, reactive power, and proportion of induction motors in the load at each node; and also includes: equivalent resistance, reactance, active power, reactive power, and line state quantity of the transmission line, wherein the line state is the line state of an anticipated fault.

[0161] In the embodiment of the present invention, based on the transient voltage trajectory of each node under different expected faults, the transient voltage safety quantitative index is defined as follows:

[0162]

[0163] Among them, STVI i,c It is a quantitative indicator of transient voltage safety; ΔV i,c (t) The transient voltage of node i under the expected fault c deviates from the threshold range at time t Deviation of faultis the starting time of the line state of the expected fault, t end The end time of the line state of the expected fault.

[0164] It is worth noting that transient voltage instability usually develops rapidly, and the time domain simulation duration is set to 3 seconds. According to the requirements of dynamic security analysis of power grid, the indicator can set the voltage range threshold to [0.9, 1.1] pu.

[0165] Step 302: Perform model training on the graph neural network by training graph data and transient voltage safety quantitative indicators to construct an initial graph neural network model.

[0166] In an embodiment of the present invention, the training graph data and the transient voltage safety quantitative index are input into the graph neural network model for model iterative training, and the initial graph neural network model is output.

[0167] Step 303: Obtain sample graph data and corresponding transient voltage real indicators.

[0168] In an embodiment of the present invention, the sample graph data may be historical data or defined data of a power grid topology heterogeneous graph, including but not limited to the bus voltage amplitude, bus phase angle, active power, reactive power, and proportion of induction motors in the load at each node; and also includes: equivalent resistance, reactance, active power, reactive power, and line state quantity of the transmission line, wherein the line state is the line state of an anticipated fault.

[0169] In the embodiment of the present invention, the real indicator of the transient voltage is the real value of the transient voltage corresponding to the sample graph data.

[0170] Step 304: simulate and calculate the sample graph data through the initial graph neural network model to generate a transient voltage simulation index.

[0171] In the embodiment of the present invention, the sample graph data is input into the initial graph neural network model for simulation calculation, and a transient voltage simulation index is output.

[0172] Step 305: Perform loss calculation on the transient voltage simulation index and the transient voltage real index by using a preset loss function to generate a loss result.

[0173] In the embodiment of the present invention, the model performs iterative learning by optimizing the loss function. The loss function LSTVI is calculated based on the evaluation results and simulation results of each node under different scenarios and expected failures:

[0174]

[0175] Among them, L STVI is the loss result; N is the total number of nodes in the graph data, y i is the real indicator of transient voltage, yi ' is the transient voltage simulation index.

[0176] In the embodiment of the present invention, in order to pay more attention to the penalty of nodes with large deviations of the evaluation index, the loss function is calculated in a mean square error manner.

[0177] Step 306: Use the back propagation algorithm to update the model parameters of the initial graph neural network model according to the loss results, and build a heterogeneous graph attention neural network evaluation model.

[0178] Specifically, based on the loss results, the model parameters such as the internal weights and biases of the initial graph neural network model are updated through the back-propagation algorithm, and the model is iteratively optimized until the loss results meet the preset model accuracy, and the heterogeneous graph attention neural network evaluation model is output.

[0179] The present invention can still have high evaluation accuracy after the power grid topology changes, and is suitable for operation scenarios with frequent topology changes.

[0180] It is worth noting that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of laws and regulations. The user information in the embodiments of this application is obtained through legal and compliant channels, and the acquisition, storage, use, and processing of user information are authorized and agreed by the customer.

[0181] It is worth noting that the information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0182] It is worth noting that the technical solution provided in this application provides users with corresponding operation entrances for them to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0183] In the technical solution of the transient voltage assessment method based on the power grid topology structure provided by the embodiment of the present invention, heterogeneous graph data of the power grid topology structure is obtained, and the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; the initial node features and initial edge features are feature processed to generate current node features and current edge features; through a pre-constructed heterogeneous graph attention neural network assessment model, transient voltage assessment is performed according to the current node features and current edge features, and a transient voltage assessment result of each node is generated, which can perform accurate transient voltage online assessment for anticipated accidents before a fault occurs, and is suitable for operating scenarios with frequent topology changes; by fully considering the high-dimensional interactive relationship between the characteristics of each node and the line characteristics under different operating conditions, the model assessment performance for anticipated faults is effectively improved, the model generalization ability is strong, and the assessment accuracy is high.

[0184] Figure 4 A schematic diagram of a transient voltage assessment device based on a power grid topology structure provided by an embodiment of the present invention is provided. The device is used to execute the transient voltage assessment method based on a power grid topology structure, such as Figure 4 As shown, the device includes: a heterogeneous graph data acquisition unit 11, a feature processing unit 12 and a transient voltage evaluation unit 13.

[0185] The heterogeneous graph data acquisition unit 11 is used to acquire heterogeneous graph data of the power grid topology structure. The heterogeneous graph data includes initial node features and initial edge features. Nodes of different node types have different node features.

[0186] The feature processing unit 12 is used to perform feature processing on the initial node features and the initial edge features to generate current node features and current edge features.

[0187] The transient voltage evaluation unit 13 is used to perform transient voltage evaluation according to current node features and current edge features through a pre-built heterogeneous graph attention neural network evaluation model, and generate transient voltage evaluation results for each node.

[0188] In the embodiment of the present invention, the device further includes: a determining unit 14.

[0189] The determination unit 14 is used to construct a power grid topology structure by taking generators, loads, high-voltage direct current and distributed renewable energy as nodes and transmission lines as edges; the node characteristics of generator nodes include bus voltage amplitude, bus phase angle, generator active power and generator reactive power; the node characteristics of load nodes include bus voltage amplitude, bus phase angle, load active power, load reactive power and induction motor proportion; the node characteristics of high-voltage direct current nodes include bus voltage amplitude, bus phase angle, high-voltage direct current active power and high-voltage direct current reactive power; the node characteristics of distributed renewable energy include bus voltage amplitude, bus phase angle, distributed renewable energy active power and distributed renewable energy reactive power; the edge characteristics of transmission lines include equivalent resistance, reactance, active power, reactive power and line state quantity.

[0190] In an embodiment of the present invention, the feature processing unit 12 is specifically used to normalize the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features; and project the normalized node features and the normalized edge features to a shared space respectively through virtual isomorphism technology to obtain current node features and current edge features.

[0191] In an embodiment of the present invention, the heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer; the transient voltage evaluation unit 13 is specifically used to perform dimension expansion conversion on the current node feature and the current edge feature through the feature conversion layer to generate high-dimensional node features and high-dimensional edge features; through the feature transfer layer, perform feature fusion and feature transfer on the high-dimensional node features and high-dimensional edge features to generate updated node features; through the feature readout layer, perform pooling and dimensionality reduction processing on the updated node features to generate a transient voltage evaluation result.

[0192] In the embodiment of the present invention, the transient voltage evaluation unit 13 is specifically used to map high-dimensional edge features through a multi-layer perceptron to generate an edge feature matrix; perform feature fusion on the edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix to generate a fused feature; perform message transmission based on the fused features, high-dimensional edge features and high-dimensional node features to generate a message feature transmission vector; perform feature update based on the high-dimensional node features and the message feature transmission vector through a gated recurrent unit to generate an updated node feature.

[0193] In an embodiment of the present invention, the transient voltage evaluation unit 13 is specifically used to splice the fused feature with the high-dimensional edge feature to obtain a first spliced ​​feature; through a shared attention mechanism, an attention coefficient is generated according to the first spliced ​​feature, the high-dimensional node feature and a preset first weight matrix; based on the attention coefficient, the first spliced ​​feature and the preset second weight matrix are message aggregated to generate a message feature transfer vector.

[0194] In the embodiment of the present invention, the transient voltage assessment unit 13 is specifically used to perform pooling processing on the update node features to generate pooled features; splice the pooled features with the update node features to generate second spliced ​​features; and transform the second spliced ​​features through a multi-layer perceptron to generate a transient voltage assessment result.

[0195] In the embodiment of the present invention, the device further includes: a sample data acquisition unit 15 , a simulation calculation unit 16 , a loss calculation unit 17 and a model optimization unit 18 .

[0196] The sample data acquisition unit 15 is used to acquire sample graph data and corresponding transient voltage real indicators.

[0197] The simulation calculation unit 16 is used to perform simulation calculation on the sample graph data through the initial graph neural network model to generate a transient voltage simulation index.

[0198] The loss calculation unit 17 is used to perform loss calculation on the transient voltage simulation index and the transient voltage real index through a preset loss function to generate a loss result.

[0199] The model optimization unit 18 is used to update the model parameters of the initial graph neural network model according to the loss results through the back propagation algorithm, and construct a heterogeneous graph attention neural network evaluation model.

[0200] In the embodiment of the present invention, the device further includes: a training data acquisition unit 19 and a model training unit 20.

[0201] The training data acquisition unit 19 is used to acquire training graph data and predefined transient voltage safety quantitative indicators.

[0202] The model training unit 20 is used to perform model training on the graph neural network by training graph data and transient voltage safety quantitative indicators to construct an initial graph neural network model.

[0203] In the scheme of the embodiment of the present invention, heterogeneous graph data of the power grid topology structure is obtained, and the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; the initial node features and initial edge features are feature processed to generate current node features and current edge features; transient voltage evaluation is performed according to the current node features and current edge features through a pre-constructed heterogeneous graph attention neural network evaluation model, and a transient voltage evaluation result of each node is generated, which can perform accurate transient voltage online evaluation for anticipated accidents before a fault occurs, and is suitable for operating scenarios with frequent topology changes; by fully considering the high-dimensional interactive relationship between each node feature and the line feature under different operating conditions, the model evaluation performance for anticipated faults is effectively improved, the model has strong generalization ability, and high evaluation accuracy.

[0204] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer device, and specifically, the computer device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0205] An embodiment of the present invention provides a computer device, including a memory and a processor, the memory is used to store information including program instructions, the processor is used to control the execution of the program instructions, and when the program instructions are loaded and executed by the processor, the steps of the embodiment of the above-mentioned transient voltage assessment method based on the power grid topology structure are implemented. For a specific description, please refer to the embodiment of the above-mentioned transient voltage assessment method based on the power grid topology structure.

[0206] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer device 600 suitable for implementing an embodiment of the present application.

[0207] like Figure 5 As shown, the computer device 600 includes a central processing unit (CPU) 601, which can perform various appropriate operations and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer device 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0208] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal feedback device (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed as needed as the storage section 608.

[0209] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 609, and / or installed from the removable medium 611.

[0210] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0211] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0212] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0213] 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.

[0214] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0215] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0216] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0217] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0218] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt 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.) that contain computer-usable program code.

[0219] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0220] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0221] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A transient voltage assessment method based on a power grid topology structure, characterized in that: The method comprises: Acquire heterogeneous graph data of a power grid topology structure, wherein the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; Performing feature processing on the initial node features and the initial edge features to generate current node features and current edge features; Through the pre-built heterogeneous graph attention neural network evaluation model, transient voltage evaluation is performed according to the current node characteristics and the current edge characteristics to generate transient voltage evaluation results for each node.

2. The transient voltage assessment method based on the power grid topology structure according to claim 1 is characterized in that: The method further comprises: The power grid topology is constructed by taking generators, loads, high voltage direct current and distributed renewable energy as nodes and transmission lines as edges; The node characteristics of the generator node include bus voltage amplitude, bus phase angle, generator active power and generator reactive power; The node characteristics of the load node include bus voltage amplitude, bus phase angle, load active power, load reactive power and induction motor proportion; The node characteristics of the HVDC node include bus voltage amplitude, bus phase angle, HVDC active power and HVDC reactive power; The node characteristics of the distributed new energy include bus voltage amplitude, bus phase angle, distributed new energy active power and distributed new energy reactive power; The edge characteristics of the transmission line include equivalent resistance, reactance, active power, reactive power and line state quantity.

3. The transient voltage assessment method based on power grid topology structure according to claim 1 is characterized in that: The performing feature processing on the initial node feature and the initial edge feature to generate the current node feature and the current edge feature includes: Normalizing the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features; The normalized node features and the normalized edge features are projected to a shared space respectively through a virtual isomorphism technology to obtain the current node features and the current edge features.

4. The transient voltage assessment method based on power grid topology structure according to claim 1 is characterized in that: The heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer; The pre-built heterogeneous graph attention neural network evaluation model performs transient voltage evaluation according to the current node features and the current edge features to generate transient voltage evaluation results for each node, including: Through the feature conversion layer, the current node feature and the current edge feature are dimensionally expanded and converted to generate high-dimensional node features and high-dimensional edge features; Through the feature transfer layer, feature fusion and feature transfer are performed on the high-dimensional node features and the high-dimensional edge features to generate updated node features; Through the feature readout layer, the update node features are subjected to pooling and dimensionality reduction processing to generate the transient voltage assessment result.

5. The transient voltage assessment method based on power grid topology structure according to claim 4 is characterized in that: The step of performing feature fusion and feature transfer on the high-dimensional node features and the high-dimensional edge features through the feature transfer layer to generate updated node features includes: Mapping the high-dimensional edge features through a multi-layer perceptron to generate an edge feature matrix; Performing feature fusion on the edge feature matrix and high-dimensional node features of adjacent nodes of the edge indicated by the edge feature matrix to generate fused features; Perform message transmission according to the fusion feature, the high-dimensional edge feature and the high-dimensional node feature to generate a message feature transmission vector; Through the gated recurrent unit, feature update is performed according to the high-dimensional node feature and the message feature transfer vector to generate the updated node feature.

6. The transient voltage assessment method based on power grid topology structure according to claim 5 is characterized in that: The performing message transmission according to the fusion feature, the high-dimensional edge feature and the high-dimensional node feature to generate a message feature transmission vector includes: Splicing the fusion feature with the high-dimensional edge feature to obtain a first splicing feature; Generate an attention coefficient according to the first concatenated feature, the high-dimensional node feature and a preset first weight matrix through a shared attention mechanism; Based on the attention coefficient, message aggregation is performed on the first concatenated feature and a preset second weight matrix to generate a message feature transfer vector.

7. The transient voltage assessment method based on power grid topology structure according to claim 4 is characterized in that: The step of performing pooling and dimensionality reduction processing on the update node features through the feature readout layer to generate the transient voltage assessment result includes: Performing pooling processing on the update node features to generate pooled features; Splicing the pooled feature with the updated node feature to generate a second spliced ​​feature; The second splicing feature is transformed through a multi-layer perceptron to generate the transient voltage evaluation result.

8. The transient voltage assessment method based on power grid topology structure according to claim 1 is characterized in that: The method further comprises: Obtain sample graph data and corresponding real indicators of transient voltage; The sample graph data is simulated and calculated by an initial graph neural network model to generate a transient voltage simulation index; By using a preset loss function, loss calculation is performed on the transient voltage simulation index and the transient voltage real index to generate a loss result; Through the back-propagation algorithm, the model parameters of the initial graph neural network model are updated according to the loss results, and the heterogeneous graph attention neural network evaluation model is constructed.

9. The transient voltage assessment method based on power grid topology structure according to claim 8, characterized in that: The method further comprises: Obtain training graph data and predefined transient voltage safety quantitative indicators; The graph neural network model is trained using the training graph data and the transient voltage safety quantitative index to construct the initial graph neural network model.

10. A transient voltage assessment device based on a power grid topology structure, characterized in that: The device comprises: A heterogeneous graph data acquisition unit, used to acquire heterogeneous graph data of a power grid topology structure, wherein the heterogeneous graph data includes initial node features and initial edge features, and nodes of different node types have different node features; A feature processing unit, used to perform feature processing on the initial node feature and the initial edge feature to generate a current node feature and a current edge feature; The transient voltage evaluation unit is used to perform transient voltage evaluation according to the current node characteristics and the current edge characteristics through a pre-built heterogeneous graph attention neural network evaluation model, and generate transient voltage evaluation results for each node.

11. The transient voltage assessment device based on power grid topology structure according to claim 10, characterized in that: The device also includes: A determination unit is used to construct the power grid topology structure by taking generators, loads, high voltage direct current and distributed renewable energy as nodes and transmission lines as edges; The node characteristics of the generator node include bus voltage amplitude, bus phase angle, generator active power and generator reactive power; The node characteristics of the load node include bus voltage amplitude, bus phase angle, load active power, load reactive power and induction motor proportion; The node characteristics of the HVDC node include bus voltage amplitude, bus phase angle, HVDC active power and HVDC reactive power; The node characteristics of the distributed new energy include bus voltage amplitude, bus phase angle, distributed new energy active power and distributed new energy reactive power; The edge characteristics of the transmission line include equivalent resistance, reactance, active power, reactive power and line state quantity.

12. The transient voltage assessment device based on power grid topology structure according to claim 10, characterized in that: The feature processing unit is specifically used to normalize the initial node features and the initial edge features respectively to obtain normalized node features and normalized edge features; and project the normalized node features and the normalized edge features to a shared space respectively through virtual isomorphism technology to obtain the current node features and the current edge features.

13. The transient voltage assessment device based on power grid topology structure according to claim 10, characterized in that: The heterogeneous graph attention neural network evaluation model includes a feature conversion layer, a feature transfer layer and a feature readout layer; The transient voltage assessment unit is specifically used to perform dimension expansion conversion on the current node features and the current edge features through the feature conversion layer to generate high-dimensional node features and high-dimensional edge features; perform feature fusion and feature transfer on the high-dimensional node features and high-dimensional edge features through the feature transfer layer to generate updated node features; and perform pooling and dimensionality reduction processing on the updated node features through the feature readout layer to generate the transient voltage assessment result.

14. The transient voltage assessment device based on power grid topology structure according to claim 13, characterized in that: The transient voltage evaluation unit is specifically used to map the high-dimensional edge features through a multi-layer perceptron to generate an edge feature matrix; perform feature fusion on the edge feature matrix and the high-dimensional node features of the adjacent nodes of the edge indicated by the edge feature matrix to generate a fusion feature; perform message transmission according to the fusion feature, the high-dimensional edge feature and the high-dimensional node feature to generate a message feature transmission vector; Through the gated recurrent unit, feature update is performed according to the high-dimensional node feature and the message feature transfer vector to generate the updated node feature.

15. The transient voltage assessment device based on power grid topology structure according to claim 14, characterized in that: The transient voltage evaluation unit is specifically used to splice the fusion feature with the high-dimensional edge feature to obtain a first splicing feature; generate an attention coefficient according to the first splicing feature, the high-dimensional node feature and the preset first weight matrix through a shared attention mechanism; based on the attention coefficient, perform message aggregation on the first splicing feature and the preset second weight matrix to generate a message feature transfer vector.

16. The transient voltage assessment device based on power grid topology structure according to claim 13, characterized in that: The transient voltage assessment unit is specifically used to perform pooling processing on the update node features to generate pooled features; splice the pooled features with the update node features to generate second spliced ​​features; and transform the second spliced ​​features through a multi-layer perceptron to generate the transient voltage assessment result.

17. The transient voltage assessment device based on power grid topology structure according to claim 10, characterized in that: The device also includes: A sample data acquisition unit, used to acquire sample graph data and corresponding transient voltage real indicators; A simulation calculation unit, used to perform simulation calculation on the sample graph data through an initial graph neural network model to generate a transient voltage simulation index; A loss calculation unit, used to perform loss calculation on the transient voltage simulation index and the transient voltage real index through a preset loss function to generate a loss result; The model optimization unit is used to update the model parameters of the initial graph neural network model according to the loss result through a back-propagation algorithm, and construct the heterogeneous graph attention neural network evaluation model.

18. The transient voltage assessment device based on power grid topology structure according to claim 17, characterized in that: The device also includes: A training data acquisition unit, used to acquire training graph data and predefined transient voltage safety quantitative indicators; The model training unit is used to perform model training on the graph neural network through the training graph data and the transient voltage safety quantitative index to construct the initial graph neural network model.

19. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the transient voltage assessment method based on the power grid topology structure described in any one of claims 1 to 9 is implemented.

20. A computer device comprising a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, characterized in that: When the program instructions are loaded and executed by the processor, the transient voltage assessment method based on the power grid topology structure described in any one of claims 1 to 9 is implemented.

21. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the transient voltage assessment method based on the power grid topology structure according to any one of claims 1 to 9 is implemented.

Images