Generator-level transient stability evaluation model construction method, generator-level transient stability evaluation model prediction method, computer equipment and storage medium

Abstract

The invention relates to a generator-level transient stability evaluation model construction method, a generator-level transient stability evaluation model prediction method, computer equipment and a storage medium. The construction method comprises the steps of obtaining a training data set; inputting the training data set into the transient stability evaluation model to obtain a prediction result; determining a loss function based on the prediction result and the actual label; and updating model parameters of the transient stability evaluation model based on the gradient of the loss function until training is completed. The physical guidance global graph pooling module in the transient stability evaluation module can finely analyze the association relationship between the generator-generator node pair and the generator-non-generator node pair, so that the difference between high-dimensional characteristics of the generator is improved, and the generalization performance of stability evaluation is improved. And through a generator-level stability scanner combining graph pooling and parameter sharing, generator-level transient stability evaluation can be quickly completed based on steady-state information, and the model calculation efficiency is improved, so that the requirement of online analysis is met.

Description

Technical Field

[0001] The present application relates to the technical field of transient stability assessment, and in particular to a method for constructing a generator-level transient stability assessment model, a prediction method, a computer device, and a storage medium. Background Art

[0002] With the development of power technology, transient stability assessment technology has emerged. Transient stability assessment is a key part of power system risk analysis. It usually takes the entire system as the research object to analyze the transient stability state and level of the power system after the expected fault. Generator-level transient stability assessment (GTSA) focuses on the dominant element of transient stability, that is, the generator group. By evaluating the interaction between generators, it can identify the dominant generator and the severity of generator disturbance, etc., which can provide transient stability risk warning and provide guidance for preventive control. In related technologies, short-term dynamic information obtained by multi-dependency time domain simulation is used as input, and transient stability assessment is completed through traditional machine learning models such as MLP, CNN, and LSTM. However, with the increase of node scale and control parameters, the model calculation complexity is large, and it is difficult to meet the needs of online analysis. Summary of the invention

[0003] Based on this, it is necessary to provide a construction method, prediction method, computer equipment and storage medium for a generator-level transient stability assessment model that can meet the needs of online analysis in response to the above technical problems.

[0004] In the first aspect, the present application provides a method for constructing a generator-level transient stability assessment model. The method comprises: obtaining a training data set; wherein the training data set comprises: graph structure data determined by power grid topology information, node steady-state operation information, and node fault information; inputting the training data set into a transient stability assessment model to obtain a prediction result; wherein the transient stability assessment model comprises a graph embedding module, a physical-guided global graph pooling module, and a generator-level stability scanner, which are sequentially arranged, and the prediction result is used to indicate whether the generator is dominantly unstable; determining a loss function based on the prediction result and the actual label; and updating the model parameters of the transient stability assessment model based on the gradient of the loss function until the training is completed.

[0005] In one embodiment, the graph structure data includes: first graph data, second graph data and third graph data; wherein, the first graph data is composed of a first adjacency matrix and a steady-state node characteristic matrix, the second graph data is composed of the first adjacency matrix and a faulty node characteristic matrix, and the third graph data is composed of a second adjacency matrix and the steady-state node characteristic matrix; the first adjacency matrix is ​​determined by the grid topology information in a steady state, the steady-state node characteristic matrix is ​​determined by the node steady-state operation information, the faulty node characteristic matrix is ​​determined by the node fault information, and the second adjacency matrix is ​​determined by the grid topology information in a faulty state.

[0006] In one of the embodiments, the step of inputting the training data set into the transient stability assessment model to obtain a prediction result includes: inputting the training data set into the graph embedding module to perform neighborhood aggregation of node features to obtain an aggregated feature matrix; inputting the aggregated feature matrix into the physical-guided global graph pooling module to perform high-dimensional representation aggregation to obtain a generator high-dimensional representation matrix; inputting the generator high-dimensional representation matrix into the generator-level stability scanner to perform dimensionality reduction to obtain an output matrix; and determining the prediction result based on the output matrix.

[0007] In one of the embodiments, the graph embedding module is used to perform intra-graph attention processing or inter-graph attention processing.

[0008] In one embodiment, the step of inputting the aggregated feature matrix into the physical-guided global graph pooling module for high-dimensional representation aggregation includes: determining a query matrix, a key matrix and a value matrix based on the aggregated feature matrix; wherein the query matrix is ​​composed of a query vector of each generator, the key matrix is ​​composed of a key vector of each node, and the value matrix is ​​composed of a value vector of each node; determining a global attention matrix based on the query matrix and the key matrix; and determining the generator high-dimensional representation matrix based on the global attention matrix and the value matrix.

[0009] In one of the embodiments, the step of determining the global attention matrix based on the query matrix and the key matrix includes: performing dot product operations on the query vector of each generator and the key vectors of all nodes in turn to obtain multiple attention coefficient column vectors; and combining all the attention coefficient column vectors to obtain the global attention matrix.

[0010] In one of the embodiments, the generator level stability scanner is composed of a plurality of fully connected layers.

[0011] In the second aspect, the present application also provides a prediction method for a generator-level transient stability assessment model. The method comprises: obtaining current operating data; wherein the current operating data comprises: graph structure data determined by power grid topology information and node operating information; inputting the current operating data into a trained transient stability assessment model to obtain a stability result; wherein the transient stability assessment model is obtained by the construction method of the generator-level transient stability assessment model described in the embodiment of the first aspect.

[0012] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0013] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0014] The construction method, prediction method, computer equipment and storage medium of the above generator-level transient stability assessment model use graph structure data determined by power grid topology information, node steady-state operation information and node fault information to form a training data set, and train the transient stability assessment model composed of a graph embedding module, a physical-guided global graph pooling module and a generator-level stability scanner. The physical-guided global graph pooling module can finely analyze the association between generator-generator nodes and generator-non-generator node pairs, thereby improving the differences between high-dimensional features of generators and improving the generalization performance of stability assessment. And by combining graph pooling with a parameter-sharing generator-level stability scanner, the generator-level transient stability assessment can be quickly completed based on steady-state information, improving the model calculation efficiency, thereby meeting the needs of online analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a flow chart of a method for constructing a generator-level transient stability assessment model in one embodiment;

[0016] Figure 2 is a schematic diagram of a transient stability assessment model in one embodiment;

[0017] Figure 3 A schematic diagram of a process for determining a prediction result in one embodiment;

[0018] Figure 4 A schematic diagram of a process for determining a high-dimensional characterization matrix of a generator in one embodiment;

[0019] Figure 5 A schematic diagram of a process for determining a global attention matrix in one embodiment;

[0020] Figure 6 A schematic diagram of an IEEE 39 node system in one embodiment;

[0021] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0023] The generator-level transient stability assessment model construction method and prediction method provided in the embodiments of the present application can be applied to terminal devices, wherein the terminal devices can be, but are not limited to, various personal computers, laptops, tablet computers, and the like.

[0024] In one embodiment, Figure 1 As shown, a method for constructing a generator-level transient stability assessment model is provided, and the method is applied to a terminal device as an example for explanation, including the following steps:

[0025] Step S110, obtaining a training data set.

[0026] Specifically, when the terminal device constructs the transient stability assessment model, it first obtains the training data set. The training data set of the present application includes: graph structure data determined by power grid topology information, node steady-state operation information and node fault information. Power grid topology information is used to describe the connection relationship between each node in the power grid (such as generator, substation, load, etc.). Node steady-state operation information is used to describe the parameters of each node in the power grid under normal operation, such as the voltage amplitude, phase angle, active output, reactive output, active load, reactive load, etc. of each node. Node fault information is used to describe the parameters of each node in the power grid under fault conditions, such as the voltage amplitude, phase angle, active output, reactive output, active load, reactive load, etc. of each node. In the graph structure data determined by power grid topology information, node steady-state operation information and node fault information, the nodes represent the various elements in the power grid (including generator and non-generator nodes), and the edges represent the connection relationship between nodes (such as transmission lines). At the same time, considering the different development stages of power grid faults, multiple graphs can be used to represent the operation information of different sections. For example, let the node size be N and the generator node size be N G , considering the different development stages of power grid faults, multiple graphs can be used to represent the operation information of different sections, denoted as a set of M graphs {G m}, for the graph , is an adjacency matrix whose elements Describes the connection weight between nodes i and j in the mth graph; is the node feature matrix, the row vector Represents the feature of the i-th node, and its dimension is C.

[0027] Step S120: input the training data set into the transient stability assessment model to obtain a prediction result.

[0028] Specifically, after obtaining the training data set, the training data set is input into the transient stability assessment model for processing to obtain the prediction result output by the transient stability assessment model. Figure 2 As shown, the transient stability assessment model of the embodiment of the present application includes a graph embedding module, a physics-informed global graph pooling module (Physics-Informed Global Attention Pooling, PIGAP) and a generator-level stability scanner (Generator-level Stability Scanner, GSC) arranged in sequence. The graph embedding module is used for neighborhood aggregation of node features to convert graph structure data into low-dimensional embedding vectors. These vectors can capture the key features of the topological structure and operating status of the power grid, which can adopt algorithms such as intra-graph attention and inter-graph attention. The physics-informed global graph pooling module is used to further process the embedded vectors to extract higher-dimensional features by considering the physical characteristics and global information of the power grid. The generator-level stability scanner is used to further process the high-dimensional features, so as to evaluate the transient stability of each generator based on the processing results to obtain a prediction result.

[0029] Step S130, determining a loss function based on the prediction result and the actual label.

[0030] Specifically, the prediction results include the prediction of transient stability of each generator, and the actual label is the label of generator stability obtained from actual grid operation data or simulation data. By comparing the prediction results and the actual labels, the loss function is calculated to quantify the accuracy of the model prediction. The loss function can be cross entropy loss, mean square error, etc.

[0031] Step S140: updating the model parameters of the transient stability assessment model based on the gradient of the loss function until the training is completed.

[0032] Specifically, after determining the loss function, the back propagation algorithm is used to calculate the gradient of the loss function with respect to the model parameters, and the model parameters of the transient stability assessment model are updated according to the gradient to minimize the loss function. Common optimization algorithms include gradient descent, Adam, etc. The training is completed by repeating the above steps until the performance of the transient stability assessment model on the validation set converges or reaches the preset number of training rounds.

[0033] The method for constructing the above generator-level transient stability assessment model is to form a training data set through graph structure data determined by power grid topology information, node steady-state operation information and node fault information, and to train the transient stability assessment model composed of a graph embedding module, a physical-guided global graph pooling module and a generator-level stability scanner. The physical-guided global graph pooling module can finely analyze the relationship between generator-generator nodes and generator-non-generator node pairs, thereby improving the difference between high-dimensional features of generators and improving the generalization performance of stability assessment. And by combining graph pooling with a parameter-sharing generator-level stability scanner, the generator-level transient stability assessment can be quickly completed based on steady-state information, improving the model calculation efficiency, thereby meeting the needs of online analysis.

[0034] In one embodiment, the graph structure data includes: first graph data G1, second graph data G2 and third graph data G3; wherein the first graph data G1 is composed of a first adjacency matrix A 0- and the steady-state node feature matrix X 0- The second graph data G2 is composed of the first adjacency matrix A 0- and the fault node feature matrix X 0+ The third graph data G3 is composed of the second adjacency matrix A c+ and the steady-state node feature matrix constitutes X 0- ; That is, G1→(A 0- , X 0- )、G2→(A 0- , X 0+ )、G3→(A c+ , X 0- ). The first adjacency matrix A 0- Determined by the grid topology information in the steady state, its elements reflect the branch admittance connecting two nodes. Steady-state node characteristic matrix X 0- Determined by the node steady-state operation information, its elements include the voltage amplitude, phase angle, active output, reactive output, active load, and reactive load of each node. Fault node feature matrix X 0+ Determined by the node fault information, it also includes the voltage amplitude, phase angle, active output, reactive output, active load, reactive load of each node, and the fault node feature matrix X 0+ The steady-state node characteristic matrix X 0- The second adjacency matrix A is obtained by calculating the short circuit of the expected fault. c+ It is determined by the grid topology information under the fault state, and it can be understood that the second adjacency matrix A is the same as the first adjacency matrix A when and only when the fault line connecting nodes i and j is removed. c+ Only distinguished from the first adjacency matrix A 0- , that is, the first adjacency matrix A 0-The elements at the corresponding positions in Set to zero.

[0035] In one embodiment, Figure 3 As shown, in step S120, the step of inputting the training data set into the transient stability assessment model to obtain the prediction result includes:

[0036] Step S121, input the training data set into the graph embedding module to perform neighborhood aggregation of node features to obtain an aggregated feature matrix.

[0037] Specifically, in this embodiment, the graph structure data in the training data set is input into the graph embedding module to perform neighborhood aggregation of node features, and then a corresponding aggregate feature matrix is ​​obtained. In some embodiments, the graph embedding module is used to perform intra-graph attention processing or inter-graph attention processing. For example, after the mth graph is input into the graph embedding module, the output aggregate feature matrix is .

[0038] Step S122, input the aggregated feature matrix into the physical guided global graph pooling module for high-dimensional representation aggregation to obtain a generator high-dimensional representation matrix.

[0039] Specifically, the aggregate feature matrix is ​​used as the input of the physical guided global graph pooling module for high-dimensional representation aggregation. The physical guided global graph pooling module can generate a high-dimensional representation matrix for each generator. For example, the aggregate feature matrix is After high-dimensional representation aggregation, the output generator high-dimensional representation matrix is .

[0040] Step S123, inputting the generator high-dimensional characterization matrix into the generator-level stability scanner for dimensionality reduction to obtain an output matrix.

[0041] Specifically, after obtaining the high-dimensional representation matrix of the generator, the high-dimensional representation matrix of each generator is input into the generator-level stability scanner for dimensionality reduction, thereby obtaining the output matrix. , the generator-level stability scanner is used for dimensionality reduction, which is compressed into a 2-dimensional vector z i , thus obtaining the output matrix ,and In some embodiments, the generator-level stability scanner is composed of multiple fully connected layers, and when there are N G In the case of a generator level stability scanner, its parameters remain consistent.

[0042] Step S124, determining the prediction result based on the output matrix.

[0043] Specifically, after the output matrix is ​​determined, the prediction result is determined according to the output matrix, and the prediction result is used to indicate whether the generator is dominantly unstable. The vector z in i Perform softmax operation, as shown in the following formula:

[0044]

[0045] Among them, r represents the feature number, if The i-th generator is considered to be in a dominant instability state, which is recorded as 0. Otherwise, the generator is considered not to be in a dominant instability state, which is recorded as 1. It can be understood that the dominant instability generator refers to the generator with the largest absolute power angle, which is accelerated relative to the rest of the group and is unstable in the first swing. In model training, this label can be obtained by engineering criteria. Finally, the set { } Logically converted into generator prediction results, that is, the length is N G The binary vector .

[0046] In one embodiment, Figure 4 As shown, in step S122, the step of inputting the aggregated feature matrix into the physical guided global graph pooling module for high-dimensional representation aggregation includes:

[0047] Step S126, determining a query matrix, a key matrix, and a value matrix based on the aggregated feature matrix.

[0048] Specifically, in this embodiment, when determining the high-dimensional characterization matrix of the generator based on the aggregated feature matrix, the query matrix, key matrix and value matrix are first determined based on the aggregated feature matrix. Among them, the query matrix is ​​composed of the query vector of each generator, and the query vector is used to extract information related to the generator from the entire graph. The query vector of each generator is unique, reflecting its specific physical properties and position in the graph. The key matrix is ​​composed of the key vector of each node. The key vector is used as a node feature in the graph to match the query vector to determine the correlation between the node and the generator. The value matrix is ​​composed of the value vector of each node. The value vector contains the actual feature information of the node, which will be used to generate a high-dimensional representation of the generator in subsequent steps. For example, in a given aggregated feature matrix, the feature matrix set is included And the matrix set consisting of the high-dimensional features of the generator When the generator-level stability assessment focuses on the generator, its associated object is the whole network node, so the query matrix Q, key matrix K and value matrix V are determined as follows:

[0049]

[0050] in, , , Both The parameter matrix of .

[0051] Step S127, determine the global attention matrix based on the query matrix and the key matrix.

[0052] Specifically, the global attention matrix determined by the query matrix and the key matrix reflects the attention weight between the generator and each node in the graph. The calculation of the attention weight is based on the similarity or correlation between the query vector and the key vector. Each row of the global attention matrix corresponds to a generator, and each column corresponds to a node in the graph. The value in the matrix represents the attention weight of the generator to the node. In some embodiments, the step of determining the global attention matrix based on the query matrix and the key matrix includes: performing dot product operations on the query vector of each generator and the key vectors of all nodes in turn to obtain multiple attention coefficient column vectors; combining all attention coefficient column vectors to obtain a global attention matrix. For example, the global attention matrix It can be calculated by the following formula:

[0053]

[0054] Among them, d k is the characteristic dimension of the transformed matrix. The physical meaning of the above formula is to make a normalized dot product of the single generator feature to be queried with the key vector of each node to obtain the attention coefficient column vector, which describes the interaction relationship and proportion between each generator and the nodes in the entire network.

[0055] Step S128, determining a high-dimensional representation matrix of the generator based on the global attention matrix and the value matrix.

[0056] Specifically, after determining the global attention matrix, the generator high-dimensional representation matrix can be calculated based on the value matrix. For example, for each graph, directly multiplying the global attention matrix and the value matrix can obtain the high-dimensional representation matrix aggregated for the generator. :

[0057]

[0058] Finally, concatenate the high-dimensional representation matrices corresponding to each graph to obtain the final generator high-dimensional representation matrix .

[0059] In one embodiment, Figure 5 As shown, the present application also proposes a prediction method for a generator-level transient stability assessment model, which is described by taking the application of the method to a terminal device as an example, and includes the following steps:

[0060] Step S210, obtaining current operation data.

[0061] Specifically, in the process of online analysis through the transient stability assessment model, the current operation data of the collected power grid is first obtained. It can be understood that the current operation data has the same data type as that in the training data set. After data preprocessing, the current operation data includes: graph structure data determined by the power grid topology information and node operation information.

[0062] Step S220: input the current operation data into the trained transient stability assessment model to obtain a stability result.

[0063] Specifically, after obtaining the current operating data, the current operating data is input into the trained transient stability assessment model for model calculation, thereby obtaining a stability result output by the model, and the stability result is used to indicate whether the generator is in a dominant instability. The trained transient stability assessment model is obtained by the construction method of the generator-level transient stability assessment model in the above embodiment.

[0064] The prediction method of the above-mentioned generator-level transient stability assessment model obtains the current operating data of the power grid and performs model calculation on the current operating data according to the trained transient stability assessment model to obtain the stability result. When training the transient stability assessment model, the training data set is formed by the graph structure data determined by the power grid topology information, the node steady-state operation information and the node fault information, and the transient stability assessment model composed of the graph embedding module, the physical-guided global graph pooling module and the generator-level stability scanner is trained. The physical-guided global graph pooling module can finely analyze the association between the generator-generator node and the generator-non-generator node pairs, thereby improving the difference between the high-dimensional features of the generator and realizing the improvement of the generalization performance of the stability assessment. And by combining the graph pooling with the parameter-sharing generator-level stability scanner, the generator-level transient stability assessment can be quickly completed based on the steady-state information, and the model calculation efficiency can be improved, thereby meeting the needs of online analysis.

[0065] The following is a detailed description of the construction method and prediction method of the generator-level transient stability assessment model of the present application using a specific embodiment. Figure 6As shown in the figure, the IEEE 10-machine 39-node system is used as the test system. The system includes 39 nodes, 10 generators and 46 transmission lines, among which the generator connected to Bus39 is the reference motor. The node scale N=39, and the generator scale N=10. The transient stability assessment model is trained with 37056 samples, which are divided into training set and validation set at a ratio of 3:1. In terms of model parameter setting, the number of graphs in the graph structure data is M=3, the feature dimension C=5, and F=288 is set. The graph embedding module uses intra-graph attention and inter-graph convolution algorithms. GSC uses 3 layers of fully connected layers, and the input and output pair parameters of each layer are (864,128), (128,16) and (16,2) respectively. At the same time, the node steady-state operation information and node fault information are converted into graph structure data containing 3 graphs, namely the first graph data G1, the second graph data G2 and the third graph data G3, and G1→(A 0- , X 0- )、G2→(A 0- , X 0+ )、G3→(A c+ , X 0- ).

[0066] Since the calculation rules of each figure are the same, the detailed calculation process of the first figure is used as an example for explanation. The example disturbance is: before the fault, the lines Bus15-Bus16 and Bus16-Bus24 are disconnected. The fault occurs at Bus19-Bus20, and the line trip is cleared after 0.1s. The corresponding first adjacency matrix A 0- The non-zero elements of the upper triangular part are shown in the following table, and in the following table, at most four significant digits are retained.

[0067]

[0068] The corresponding steady-state node feature matrix X 0- As shown in the following table:

[0069]

[0070] The first adjacency matrix A 0- and the steady-state node feature matrix X 0- After being used as the input of the graph embedding module, the corresponding aggregate feature matrix H1 is shown in the following table (first 10 dimensions):

[0071]

[0072] After the aggregate feature matrix H1 is input into the physical guided global graph pooling module, the corresponding high-dimensional representation matrix H of the generator is obtained. G ,1As shown in the following table:

[0073]

[0074] After being processed by 10 GSC modules with the same parameters, the output matrix Z is shown in the following table:

[0075]

[0076] After the softmax link and logic processing, the final stability result, that is, the leading instability prediction result, is From the results, it can be seen that the transient stability assessment model identified the 9th generator as belonging to the dominant unstable group, and after verification, it was consistent with the actual label.

[0077] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0078] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a construction method and a prediction method are implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0079] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0080] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0081] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented.

[0082] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0083] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing a generator-level transient stability assessment model, characterized in that: The method comprises: Acquire a training data set; wherein the training data set includes: graph structure data determined by power grid topology information, node steady-state operation information, and node fault information; Inputting the training data set into a transient stability assessment model to obtain a prediction result; wherein the transient stability assessment model includes a graph embedding module, a physical guided global graph pooling module and a generator level stability scanner which are arranged in sequence, and the prediction result is used to indicate whether the generator is dominantly unstable; Determine a loss function based on the prediction result and the actual label; The model parameters of the transient stability assessment model are updated based on the gradient of the loss function until the training is completed.

2. The method for constructing a generator-level transient stability assessment model according to claim 1, characterized in that: The graph structure data includes: first graph data, second graph data and third graph data; wherein, the first graph data is composed of a first adjacency matrix and a steady-state node characteristic matrix, the second graph data is composed of the first adjacency matrix and a faulty node characteristic matrix, and the third graph data is composed of a second adjacency matrix and the steady-state node characteristic matrix; the first adjacency matrix is ​​determined by the grid topology information in a steady state, the steady-state node characteristic matrix is ​​determined by the node steady-state operation information, the faulty node characteristic matrix is ​​determined by the node fault information, and the second adjacency matrix is ​​determined by the grid topology information in a faulty state.

3. The method for constructing a generator-level transient stability assessment model according to claim 1, characterized in that: The step of inputting the training data set into the transient stability assessment model to obtain a prediction result comprises: Inputting the training data set into the graph embedding module to perform neighborhood aggregation of node features to obtain an aggregated feature matrix; Inputting the aggregated feature matrix into the physical guided global graph pooling module for high-dimensional representation aggregation to obtain a generator high-dimensional representation matrix; Inputting the generator high-dimensional characterization matrix into the generator-level stability scanner for dimensionality reduction to obtain an output matrix; The prediction result is determined based on the output matrix.

4. The method for constructing a generator-level transient stability assessment model according to claim 3, characterized in that: The graph embedding module is used to perform intra-graph attention processing or inter-graph attention processing.

5. The method for constructing a generator-level transient stability assessment model according to claim 3, characterized in that: The step of inputting the aggregated feature matrix into the physical guided global graph pooling module for high-dimensional representation aggregation includes: Determine a query matrix, a key matrix and a value matrix based on the aggregate feature matrix; wherein the query matrix is ​​composed of a query vector of each generator, the key matrix is ​​composed of a key vector of each node, and the value matrix is ​​composed of a value vector of each node; determining a global attention matrix based on the query matrix and the key matrix; The generator high-dimensional representation matrix is ​​determined based on the global attention matrix and the value matrix.

6. The method for constructing a generator-level transient stability assessment model according to claim 5, characterized in that: The step of determining a global attention matrix based on the query matrix and the key matrix comprises: Perform dot product operations on the query vector of each generator and the key vectors of all nodes in turn to obtain multiple attention coefficient column vectors; All the attention coefficient column vectors are combined to obtain the global attention matrix.

7. The method for constructing a generator-level transient stability assessment model according to claim 3, characterized in that: The generator-level stability scanner is composed of multiple fully connected layers.

8. A prediction method for a generator-level transient stability assessment model, characterized in that: The method comprises: Acquire current operation data; wherein the current operation data includes: graph structure data determined by power grid topology information and node operation information; The current operating data is input into a trained transient stability assessment model to obtain a stability result; wherein the transient stability assessment model is obtained by the method for constructing a generator-level transient stability assessment model according to any one of claims 1 to 7.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images