Power system transient stability evaluation method based on embedded feature enhancement

Abstract

The invention discloses a power system transient stability evaluation method based on embedded feature enhancement, and relates to the technical field of power systems. Comprising the steps that node voltages of multiple stages are regarded as track clusters, and input features are constructed according to corresponding geometric attributes; based on deep learning, a convolutional neural network model is adopted to establish a transient stability evaluation model feature extractor, and an embedded feature enhancement module is designed and established; and integrating the embedded feature enhancement module into a transient stability evaluation model feature extractor. According to the method, the influence of useless components on feature expression is effectively eliminated by introducing an embedded enhancement module and utilizing orthogonality constraint and cosine similarity constraint.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method for evaluating transient stability of power systems based on embedded feature enhancement. Background Art

[0002] With the widespread deployment of new energy and power electronic equipment, their inherent intermittent and unpredictable nature, as well as their significant impact on the dynamic characteristics of the power grid, have made the inherent mechanism of safe operation of the power grid increasingly complex. In this context, the importance of transient stability issues as a key factor affecting the stable operation of the power grid has become increasingly prominent. Therefore, building a reliable and accurate transient stability assessment model is of urgent practical significance for ensuring the stable operation of the power grid.

[0003] Time domain simulation method and direct method are traditional TSA methods. Due to their limitations in solution speed and calculation accuracy, they have gradually failed to meet the needs of modern power grids for fast and dynamically changing power flow distribution analysis. The rapid development of machine learning and the popularization of synchronized phasor measurement units have laid a solid foundation for data-driven TSA. In the early days, research mainly relied on shallow machine learning algorithms such as decision trees (DT), support vector machines (SVM), random forests (RF) and K nearest neighbors (KNN). However, due to their insufficient feature expression capabilities, when processing high-dimensional power grid data, the generalization performance of the model is often limited, and it is difficult to achieve ideal prediction accuracy. In recent years, breakthroughs in deep learning technology have provided new solutions for TSA. Specifically, algorithms including stacked autoencoders (SAE), convolutional neural networks (CNN), deep belief networks (DBN) and long short-term memory networks (LSTM) focus on the nonlinear feature aggregation requirements of high-dimensional inputs and have been proven to have significant advantages in improving the accuracy and efficiency of TSA.

[0004] However, in TSA based on deep learning, the accuracy of feature extraction has a fundamental impact on the accuracy of prediction results. At present, the methods for improving the feature extraction ability of deep learning models can be divided into two categories: one is to design a suitable backbone network architecture; the other is to introduce an attention mechanism. There are several classic methods for the design of backbone network architecture, such as multi-branch technology, pyramid technology, and multi-scale technology. These methods are usually customized and optimized for specific tasks, but they often lack a universal mathematical theoretical basis, resulting in a certain degree of unexplainability in the design process. On the other hand, attention mechanisms are usually integrated into deep learning models as modular components, including channel attention and multi-head attention. However, these attention mechanisms mainly focus on the relationship within a single sample, and may ignore the implicit global information and relationships in the entire training data set. Therefore, if the model feature extraction ability can be improved while avoiding the defects of the above two solutions, it will be beneficial to further improve the interpretability and generalization of the evaluation model. Summary of the invention

[0005] In view of the problem that the existing feature extraction capability is insufficient, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to use orthogonality constraints to eliminate the influence of useless components on decisive components, and use cosine similarity constraints to gradually remove useless components to enhance the feature expression ability of the model.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a method for transient stability assessment of a power system based on embedded feature enhancement, which includes: considering node voltages in multiple stages as trajectory clusters, and constructing input features according to corresponding geometric properties; based on deep learning, a convolutional neural network model is used to build a transient stability assessment model feature extractor, and an embedded feature enhancement module is designed and established; and the embedded feature enhancement module is integrated into the transient stability assessment model feature extractor.

[0009] As a preferred scheme of the method for transient stability assessment of power system based on embedded feature enhancement described in the present invention, wherein: the multiple stages include three stages: before fault, during fault and after fault; the geometric attributes include basic characteristics, rate of change, curvature and acceleration.

[0010] As a preferred solution of the method for transient stability assessment of power system based on embedded feature enhancement according to the present invention, wherein: the feature extractor for building a transient stability assessment model includes m 1 convolutional pooling layers and m 2 There are m fully connected layers, where 1 and m 2is a constant.

[0011] As a preferred solution of the method for transient stability assessment of power system based on embedded feature enhancement described in the present invention, the design and establishment of embedded feature enhancement module includes: designing orthogonality constraints; designing cosine similarity constraints to push useless components to zero vectors and completely eliminate useless components from feature representation; constructing an embedded feature enhancement module to comprehensively apply the orthogonality constraints and the cosine similarity constraints; connecting the embedded feature enhancement module to the transient stability assessment model to evaluate the transient stability of the power system.

[0012] As a preferred solution of the method for transient stability assessment of power system based on embedded feature enhancement described in the present invention, the orthogonality constraint includes: the feature extractor converts the power system measurement data into a set of feature vectors x, decomposes them into deterministic components x 1 and the useless component x 0 ; If x is linearly changed 1 Converted to x:

[0013] x 1 =A(x 1 +x 0 )=Ax 1 +0

[0014] Where A is a matrix;

[0015] Determining Component x 1 is the eigenvector of A, with useless components x 0 Located in the null space of A, i.e. Ax 0 =0, if x 1 ⊥x 0 , which is most beneficial for feature extraction; the row space and null space of A are used to extract the decision component and useless component respectively; the matrix decomposition technology is used to decompose A into the form of the product of two non-full rank matrices by introducing a relaxation factor, that is:

[0016] R A =D 1 ×E 1

[0017] N A =D 0 ×E 0

[0018] Among them, R A and N A are the row space and null space of A respectively; x 1 ∈R A , x 0 ∈N A ;D 1 and D0 They correspond to the row space and the subspace of the null space of matrix A respectively; when the row space is orthogonal to the null space, x 1 With x 0 Orthogonal.

[0019] As a preferred solution of the method for transient stability assessment of power system based on embedded feature enhancement described in the present invention, the cosine similarity constraint is:

[0020] cos(x,N A x)=1.

[0021] As a preferred solution of the method for transient stability assessment of power system based on embedded feature enhancement described in the present invention, wherein: integrating the embedded feature enhancement module into the assessment model feature extractor includes: connecting the embedded feature enhancement module between the third convolutional pooling layer and the first fully connected layer; the overall loss function L is:

[0022]

[0023] Among them, L TSA To evaluate the classification loss of the model for samples; ||·|| F is the F norm; α and β are weight parameters.

[0024] In the second aspect, an embodiment of the present invention provides an electric power system transient stability assessment system based on embedded feature enhancement, which includes: a trajectory cluster construction module, which is used to regard the node voltages of multiple stages as trajectory clusters, and construct input features according to the corresponding geometric properties; a feature extractor module, which is used to build an assessment model feature extractor based on deep learning and using a convolutional neural network model; an embedded feature enhancement module, which is used to design and establish an embedded feature enhancement module, and integrate it into the assessment model feature extractor.

[0025] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the method for transient stability assessment of a power system based on embedded feature enhancement as described in the first aspect of the present invention are implemented.

[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the method for transient stability assessment of a power system based on embedded feature enhancement as described in the first aspect of the present invention are implemented.

[0027] The beneficial effects of the present invention are as follows: the TSA-EFE method proposed in the present invention is improved on the basis of 1D-CNN, and by introducing an embedded enhancement module and utilizing orthogonality constraints and cosine similarity constraints, the influence of useless components on feature expression is effectively eliminated, thereby enhancing the feature expression capability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0029] Figure 1 Diagram of the power system transient stability assessment model enhanced for embedded features.

[0030] Figure 2 It is a feature decomposition diagram. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0033] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0034] Example 1

[0035] Reference Figure 1-2 , which is the first embodiment of the present invention, and provides a method for evaluating transient stability of a power system based on embedded feature enhancement, comprising:

[0036] S1: The node voltages at the three stages before, during and after the fault are regarded as trajectory clusters, and the input features are constructed according to their geometric properties (basic characteristics, rate of change and curvature, acceleration).

[0037] The model structure of the present invention is as follows Figure 1 The specific contents are as follows:

[0038] Among them, the construction of input features includes:

[0039] As an instantaneous time-domain reflection of a power system fault, the voltage disturbance trajectory directly reflects the disturbance degree and stability characteristics of the system. These trajectories are key indicators for identifying the stability of the system during transient processes. Based on this, the voltage trajectory cluster characteristics are used to evaluate the transient stability of the power system.

[0040] The voltage trajectory cluster features include three categories: basic features, change rate and curvature, and acceleration, as shown in Table 1:

[0041] Table 1 Voltage trace cluster feature definition

[0042] feature Feature Attribute Description feature Feature Attribute Description 1 Centroid 16 Centerline change rate 2 Discreteness 17 Envelope height change rate 3 Upper envelope 18 Curvature of trajectory 4 Lower envelope 19 Centroid Curvature 5 Centerline 20 Discreteness curvature 6 The upper envelope and the centroid 21 Upper envelope curvature 7 The lower envelope and the centroid 22 Lower envelope curvature 8 Upper envelope and center line 23 Centerline curvature 9 Lower envelope and center line 24 Acceleration of the center of mass line 10 Envelope height 25 Discreteness change acceleration 11 The centroid of the trajectory 26 Upper envelope acceleration 12 Instantaneous rate of change of centroid 27 Lower envelope acceleration 13 Instantaneous rate of change of discreteness 28 Centerline acceleration 14 Upper envelope change rate 29 Envelope height change plus 15 Lower envelope change rate

[0043] The calculation formulas for various features in Table 1 are as follows:

[0044] 1) Basic properties of trajectory clusters.

[0045] The cluster matrix of the disturbed voltage trajectory of the power system is expressed as {x ij} m×n , where m is the number of trajectories and n is the number of sampling points. The centroid of the trajectory cluster can be defined as:

[0046]

[0047] The discreteness of each trajectory relative to the centroid line is defined as:

[0048]

[0049] The upper and lower envelopes are defined as:

[0050] {max{x j}},j=1,2,3....,n;

[0051] {min{x j}},j=1,2,3....,n;

[0052] The centerline is defined as:

[0053]

[0054] 2) Change rate and curvature characteristics.

[0055] The rate of change characteristic of the trajectory cluster is defined as:

[0056]

[0057] Where h is the time sampling interval, which is simplified to unit 1.

[0058] The curvature characteristic of the trajectory cluster is defined as:

[0059]

[0060] 3) Acceleration characteristics.

[0061] The acceleration characteristics of the trajectory cluster are calculated using the second-order derivative, defined as:

[0062]

[0063] Among them, r c Indicates the rate of change characteristic.

[0064] S2: Based on deep learning, a convolutional neural network model is used to build an evaluation model feature extractor, and an embedded feature enhancement module is designed and established.

[0065] Preferably, constructing the evaluation model feature extractor is based on a convolutional neural network, and constructing the evaluation model feature extractor includes 3 convolutional pooling layers and 3 fully connected layers.

[0066] Furthermore, an embedded feature enhancement module is designed and established, including:

[0067] Orthogonality constraints are designed to ensure the orthogonality of the decisive component and the useless component in the feature vector, so that any disturbance of the useless component will not affect the feature representation of the decisive component;

[0068] The cosine similarity constraint is proposed to push the useless components to the zero vector as much as possible, so as to completely eliminate the useless components from the feature representation;

[0069] An embedded feature enhancement module is constructed, and orthogonality constraints and cosine similarity constraints are comprehensively applied to eliminate the influence of useless components of features on the decisive components;

[0070] The embedded feature enhancement module is connected to the transient stability assessment model to enhance the model's feature extraction capability, thereby achieving a rapid and accurate assessment of the transient stability of the power system.

[0071] Furthermore, orthogonality constraints are designed, including:

[0072] In the deep learning model, the feature extractor converts the power system measurement data into a set of compressed and information-rich feature representations. These feature vectors x can be decomposed into two categories: the decisive component x 1 and the useless component x 0 , as follows:

[0073] x=x 1 +x 0

[0074] like Figure 2 As shown, if the useless component x 0 With the decisive component x 1 Not orthogonal, that is, x 0 Not perpendicular to x 1 , then x 0 Any small disturbance in x may affect 1 .

[0075] In this case, the transient stability assessment model must take into account x 0 x 1 This inevitably increases the complexity of model training; on the contrary, if x 0 With x 1 Orthogonal, that is, x 0 Perpendicular to x 1 , then x 0 Any changes in x will not affect 1 have an impact.

[0076] In this ideal case, the transient stability assessment model can ignore x 0 Changes, just focus on learning x 1 , which greatly simplifies the learning and generalization tasks of the model. The advantage of this orthogonality can be concisely expressed by the following mathematical expression:

[0077] x=x 1 +x 0 ,st x 1 ⊥x 0

[0078] Based on the above theory, this application proposes an orthogonality constraint, the specific mechanism is as follows:

[0079] By linearly changing x 1 Converted to x:

[0080] x 1 =Ax

[0081] Where A is a matrix, and the above formula can also be rewritten as:

[0082] x 1 =A(x 1 +x 0 )=Ax 1 +0

[0083] In the formula, the decisive component x 1 is the eigenvector of A; the useless component x 0 Located in the null space of A, i.e. Ax 0 = 0. If x 1 ⊥x0 , it is most beneficial for feature extraction. Therefore, the above formula is understood from the perspective of the four basic subspaces of matrix A.

[0084] First, the row space and null space of A are used to extract the decision component and useless component respectively, and the formula is as follows:

[0085] R A =D 1 ×E 1

[0086] N A =D 0 ×E 0

[0087] In the formula, R A and N A are the row space and null space of A respectively; x 1 ∈R A , x 0 ∈N A ;D 1 and D 0 They correspond to the row space and the subspace of the null space of the matrix A respectively;

[0088] It is worth noting that in matrix theory, if A is a non-singular matrix, that is, the null space of A contains only zero vectors, it is impossible to extract useless components with the help of the null space.

[0089] To this end, the matrix decomposition technique is used to decompose A into the product of two non-full rank matrices, namely R A =D 1 ×E 1 , N A =D 0 ×E 0 , to ensure the existence of non-zero vectors in the null space.

[0090] When the row space is orthogonal to the null space, it is guaranteed that x 1 With x 0 Orthogonal.

[0091] Therefore, in order to eliminate x 0 x 1 The influence of , defines the orthogonality constraint:

[0092] ||(D 1 E 1 ) T D 0 E 0 || F =0

[0093] In the formula, ||·|| F is the F-norm.

[0094] Through the orthogonality constraint, it is ensured that the invalid components x in the decomposition of the eigenvector x 0 The null space and the determining component x 1 This constraint strategy effectively ensures that x 0 x 1 The orthogonal property of 0 Any fluctuation or change in will not affect x 1 stability and characteristic expression.

[0095] Furthermore, cosine similarity constraints are designed, including:

[0096] The orthogonality constraint helps the model to clearly separate the decisive components x in the eigenvector 1 and the useless component x 0 , but cannot directly eliminate x 0 Therefore, the cosine similarity constraint is proposed:

[0097] cos(x,N A x)=1

[0098] The cosine similarity constraint helps the model eliminate more useless components by correcting the weights in the null space.

[0099] Build embedded feature enhancement modules, including:

[0100] Combining orthogonality constraints and cosine similarity, the embedded feature enhancement module is defined as follows:

[0101] f(R A ,N A ,x)=R A x+(xN A x)

[0102]

[0103] Where f(·) is the embedded feature enhancement function.

[0104] The embedded feature enhancement module ensures the orthogonality of the decisive and useless components in the feature vector by introducing orthogonality constraints, so that any perturbation of the useless components will not affect the feature representation of the decisive components. In addition, the cosine similarity constraint is used to push the useless components as close to the zero vector as possible, so as to completely eliminate the useless components from the feature representation.

[0105] S3: Integrate the embedded feature enhancement module into the assessment model feature extractor to enhance the prediction performance of the transient stability assessment model.

[0106] Integrate embedded feature enhancement modules into transient stability assessment models, including:

[0107] The design of the embedded feature enhancement module not only enhances the representation of effective features, but also due to its portability, the module can be easily embedded in different parts of the model, enhancing its applicability in the target task. In the transient stability assessment task of this application, it is connected between the third convolutional pooling layer and the first fully connected layer. The overall loss function L of TSA-EFE is:

[0108]

[0109] Among them, L TSA To evaluate the classification loss of the model for samples; ||·|| F is the F norm; α and β are weight parameters.

[0110] This embodiment also provides a computer device, which is suitable for the case of a method for transient stability assessment of an electric power system based on embedded feature enhancement, and includes a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method for transient stability assessment of an electric power system based on embedded feature enhancement as proposed in the above embodiment.

[0111] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. 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 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 achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0112] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for evaluating transient stability of a power system based on embedded feature enhancement as proposed in the above embodiment is implemented.

[0113] In summary, the TSA-EFE method proposed in the present invention is improved on the basis of 1D-CNN. By introducing an embedded enhancement module and utilizing orthogonality constraints and cosine similarity constraints, the influence of useless components on feature expression is effectively eliminated, thereby enhancing the feature expression ability of the model.

[0114] Example 2

[0115] This embodiment provides a power system transient stability assessment system based on embedded feature enhancement, comprising:

[0116] A trajectory cluster building module is used to treat node voltages of multiple stages as trajectory clusters and construct input features based on corresponding geometric properties;

[0117] Feature extractor module, which is used to build an evaluation model feature extractor based on deep learning and using a convolutional neural network model;

[0118] The embedded feature enhancement module is used to design and establish the embedded feature enhancement module and integrate it into the evaluation model feature extractor.

[0119] Example 3

[0120] Reference Figure 1 and Figure 2 , which is the second embodiment of the present invention, and this embodiment provides a method for transient stability assessment of a power system based on embedded feature enhancement. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0121] Exemplarily, the evaluation performance of TSA-EFE can be tested and analyzed by the following steps: In order to verify the performance of TSA-EFE for transient stability assessment tasks, four shallow machine learning models (SVM, DT, RF, KNN) and a deep learning model 1D-CNN are selected for comparative analysis.

[0122] Among them, the kernel function of SVM adopts radial basis kernel function; DT adopts C5.0 algorithm; the number of base classifiers of RF is 25; the number of nearest neighbors in KNN is 15; 1D-CNN contains 3 convolutional pooling layers and 3 fully connected layers, the learning rate is 0.001, and the cross entropy loss function and Adam optimization algorithm are adopted.

[0123] The feature extractor structure and parameters of TSA-EFE proposed in this application are consistent with those of 1D-CNN, in which the embedded feature enhancement module is embedded after the third convolutional pooling layer and before the first fully connected layer of 1D-CNN. cc , stable sample recognition rate T sr , unstable sample recognition rate T ur And comprehensive performance index G meanThe performance of the model is comprehensively evaluated using four evaluation indicators. The specific test results are shown in Table 2:

[0124] Table 2

[0125]

[0126]

[0127] The test results reveal the limitations of shallow machine learning models in processing high-dimensional transient stability data, which is mainly attributed to its lack of feature extraction capabilities, resulting in its failure to surpass 1D-CNN in various evaluation indicators. The reason why 1D-CNN can achieve a high prediction accuracy (98.25%) is due to its excellent nonlinear feature aggregation ability, which effectively makes up for the shortcomings of shallow models. Furthermore, the TSA-EFE method proposed in this study is improved on the basis of 1D-CNN. By introducing an embedded enhancement module and using orthogonality constraints and cosine similarity constraints, the influence of useless components on feature expression is effectively eliminated, thereby enhancing the feature expression ability of the model. The results show that the prediction accuracy of the TSA-EFE method is further improved to 98.50%, which is better than all the aforementioned schemes, fully verifying the effectiveness and superiority of the method proposed in this paper.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for transient stability assessment of power systems based on embedded feature enhancement, characterized in that: include, The node voltages at multiple stages are considered as trajectory clusters, and input features are constructed based on the corresponding geometric properties; Based on deep learning, a convolutional neural network model is used to build a transient stability assessment model feature extractor, and an embedded feature enhancement module is designed and established; The embedded feature enhancement module is integrated into the transient stability assessment model feature extractor.

2. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 1, characterized in that: The multiple stages include three stages: before the fault, during the fault and after the fault; The geometric properties include basic characteristics, rate of change, curvature and acceleration.

3. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 2, characterized in that: The transient stability assessment model feature extractor includes m1 convolutional pooling layers and m2 fully connected layers, where m1 and m2 are constants.

4. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 3, characterized in that: The design and establishment of the embedded feature enhancement module includes: Design orthogonality constraints; Design cosine similarity constraints to push useless components to the zero vector and completely eliminate useless components from feature representation; Constructing an embedded feature enhancement module, and comprehensively applying the orthogonality constraint and the cosine similarity constraint; The embedded feature enhancement module is connected to the transient stability assessment model to assess the transient stability of the power system.

5. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 4, characterized in that: The orthogonality constraints include: The feature extractor converts the power system measurement data into a set of feature vectors x, which are decomposed into the decisive component x1 and the useless component x0; If x1 is transformed into the form of x by linear transformation: x1=A(x1+x0)=Ax1+0 Where A is a matrix; The decisive component x1 is the eigenvector of A, and the useless component x0 is located in the null space of A, that is, Ax0 = 0. If x1 ⊥ x0, it is most beneficial for feature extraction; The row space and null space of matrix A are used to extract decision components and useless components respectively; Using matrix decomposition technology, A is decomposed into the product of two non-full rank matrices by introducing a relaxation factor, namely: R A =D1×E1 N A =D0×E0 Among them, R A and N A are the row space and null space of A respectively; x1∈R A , x0∈N A ; D1 and D0 correspond to the subspaces of the row space and null space of matrix A respectively; When the row space is orthogonal to the null space, x1 is orthogonal to x0.

6. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 5, characterized in that: The cosine similarity constraint is: cos(x,N A x)=1。 7. The method for evaluating transient stability of a power system based on embedded feature enhancement according to claim 6, characterized in that: Integrating the embedded feature enhancement module into the evaluation model feature extractor includes: Connecting the embedded feature enhancement module between the third convolutional pooling layer and the first fully connected layer; The overall loss function L is: Among them, L TSA To evaluate the classification loss of the model for samples; ||·|| F is the F norm; α and β are weight parameters.

8. A power system transient stability assessment system based on embedded feature enhancement, based on the power system transient stability assessment method based on embedded feature enhancement according to any one of claims 1 to 7, characterized in that: Also includes, A trajectory cluster building module is used to treat node voltages of multiple stages as trajectory clusters and construct input features based on corresponding geometric properties; Feature extractor module, which is used to build an evaluation model feature extractor based on deep learning and using a convolutional neural network model; The embedded feature enhancement module is used to design and establish the embedded feature enhancement module and integrate it into the evaluation model feature extractor.

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 for transient stability assessment of a power system based on embedded feature enhancement as described in any one of claims 1 to 7 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 for transient stability assessment of a power system based on embedded feature enhancement as described in any one of claims 1 to 7 are implemented.

Images