A transient stability preventive control measure generation method and related system

Abstract

The application discloses a transient stability prevention control measure generation method and a related system, including the following steps: in a predicted fault scenario, a machine learning model is used to predict the transient stability of the current operation mode of a power system, the input of the machine learning model including various electrical characteristics related to the transient stability of the power system; the operation mode predicted as unstable is taken as an explained sample, a model explanation method is used to determine the influence degree of each electrical characteristic in the explained sample on the transient stability; candidate control units are determined according to the influence degree, and the candidate control units are adjusted, and the machine learning model is used to predict the transient stability of the adjusted operation mode until an operation mode predicted as stable is obtained, so that transient stability prevention control measures are generated. The method improves the explainability of the data-driven transient stability analysis model, the overall scheme requires a short time, and is suitable for online application.

Description

Technical Field

[0001] This invention belongs to the field of transient stability prevention and control in power systems, and more specifically, relates to a method for generating transient stability prevention and control measures and a related system. Background Technology

[0002] As a crucial component of the energy supply system, the power system holds strategic significance for national economic development and social activities. The safe and stable operation of the power system is vital for ensuring efficient power supply and maintaining normal social functioning. Transient stability prevention and control play a vital role in ensuring the safe and stable operation of the power system. Numerous major power outages demonstrate that once the transient stability of the power system is disrupted, it can easily trigger a major power outage, causing severe economic losses to society.

[0003] With the development of power systems, their structure and dynamic characteristics have become increasingly complex. Faced with the complex and ever-changing operating modes of power systems, traditional preventative control models based on offline simulation and key anticipated fault sets are inadequate. Traditional transient stability preventative control typically needs to consider all possible operating modes over a long period, and its control must simultaneously satisfy safety requirements under all operating modes, significantly limiting the power system's transmission capacity. Furthermore, traditional transient stability preventative control cannot respond in real time to more severe operating modes during actual operation, is prone to missing faults, and has weak response capabilities to multiple faults.

[0004] Machine learning, as one of the most sought-after artificial intelligence technologies, possesses the ability to efficiently mine data information and provide real-time predictions. Thanks to WAMS and SCADA systems, power systems generate massive amounts of operational data during operation, which provides strong data support for the application of machine learning technology in power system analysis.

[0005] Transient stability prevention and control faces two major challenges: (1) the problem of controlling unit limitations. Traditional offline transient stability prevention and control includes all units in the control scope to achieve network-wide scheduling control, but this scheduling method is neither economical nor convenient for transient stability prevention and control. (2) the problem of time consumption. Traditional methods obtain prevention and control measures by taking into account the Transient stability constraint optimal power flow (TSCOPF). The time consumption of TSCOPF solution mainly comes from the determination of transient stability constraints. In addition, TSCOPF solution is an iterative optimization process, and the increase of constraints increases the difficulty of solution. Therefore, traditional TSCOPF solution is relatively time-consuming. Summary of the Invention

[0006] To address the shortcomings and improvement needs of existing technologies, this invention provides a method and related system for generating transient stability prevention and control measures, aiming to solve the problem of long iterative optimization processes caused by network-wide scheduling control and solving for optimal power flow that takes into account transient stability constraints in existing methods.

[0007] To achieve the above objectives, in a first aspect, the present invention provides a method for generating transient stability prevention and control measures, comprising the following steps:

[0008] In anticipated fault scenarios, a machine learning model is used to predict the transient stability of the current operation mode of the power system. The input of the machine learning model includes various electrical features related to the transient stability of the power system.

[0009] The predicted unstable operating modes are taken as samples to be explained, and the influence of each electrical feature in the samples to be explained on transient stability is determined by the model interpretation method.

[0010] Candidate control units are determined based on the degree of impact, and the candidate control units are adjusted. Then, the machine learning model is used to predict the transient stability of the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures.

[0011] Furthermore, the machine learning model is trained in the following manner:

[0012] Collect actual power system operation data or artificially generated operation data; annotate the operation data based on time-domain simulation under anticipated faults; and train machine learning models using the annotated operation data.

[0013] Furthermore, the determination of the degree of influence of each electrical feature in the sample to be explained on transient stability using the model interpretation method includes:

[0014] Based on the sample to be explained, a perturbation sample is generated, and the perturbation sample is labeled using the machine learning model; the weight of the perturbation sample is calculated, and a linear explanatory model is obtained by fitting the weighted perturbation sample. The weight coefficients of the linear explanatory model characterize the degree of influence of each electrical feature in the sample to be explained on transient stability.

[0015] Further, generating perturbation samples based on the samples to be explained includes:

[0016] Centered on the sample to be explained, a perturbation randomly generated from a standard normal distribution is superimposed to obtain a perturbation sample set within the neighborhood of the sample to be explained.

[0017] Furthermore, the step of determining candidate control units based on the degree of influence includes:

[0018] Based on the degree of impact, the electrical characteristics are ranked, and at least one electrical characteristic with the greatest positive and negative impact on the transient stability prediction results is obtained. The corresponding unit is then identified as a candidate control unit.

[0019] Furthermore, the adjustment of the candidate control unit includes:

[0020] The active power of the candidate control unit is adjusted in a step-by-step manner, and the machine learning model is used to make transient stability predictions on the adjusted operating mode. If the adjusted active power exceeds the output constraint of the candidate control unit, the range of candidate control units is expanded according to the degree of influence until a stable operating mode is obtained.

[0021] Secondly, the present invention provides a transient stability prevention and control measure generation system, comprising:

[0022] The prediction module is used to perform transient stability prediction of the current operating mode of the power system under anticipated fault scenarios using a machine learning model. The input of the machine learning model includes various electrical features related to the transient stability of the power system.

[0023] The interpretation module is used to take the predicted unstable operating mode as the sample to be interpreted and use the model interpretation method to determine the degree of influence of each electrical feature in the sample to be interpreted on transient stability.

[0024] The generation module is used to determine candidate control units based on the degree of impact, adjust the candidate control units, and then use the machine learning model to make transient stability predictions on the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures.

[0025] Furthermore, the explanation module is also used to generate perturbation samples based on the sample to be explained, and to label the perturbation samples using the machine learning model; calculate the weights of the perturbation samples, and to fit a linear explanation model using the weighted perturbation samples, wherein the weight coefficients of the linear explanation model characterize the degree of influence of each electrical feature in the sample to be explained on transient stability.

[0026] Thirdly, the present invention provides an electronic device, comprising:

[0027] processor;

[0028] A memory storing a computer-executable program, which, when executed by the processor, causes the processor to perform the transient stability prevention and control measure generation method as described in the first aspect.

[0029] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, wherein when a processor executes the computer-executable instructions, the transient stability prevention and control measure generation method as described in the first aspect is implemented.

[0030] In summary, the above-described technical solutions conceived in this invention can achieve the following beneficial effects:

[0031] (1) This invention adopts a data-driven approach, utilizes machine learning models to perform rapid transient stability analysis, and generates rapid preventive control measures based on model interpretation. It has the advantages of achieving rapid transient stability analysis and generating transient stability preventive control measures, meeting the requirements of online application and security. In terms of time consumption in the online stage, it is superior to other existing global method solutions.

[0032] (2) The model interpretation method can perform feature importance analysis on unsafe samples monitored in real time, improving the interpretability of the transient stability analysis model. The model interpretation method effectively identifies the electrical features that have the greatest impact on the sample prediction results, realizing the identification of candidate control units, thereby narrowing the selection range of control units in transient stability prevention and control measures and effectively reducing the complexity of prevention and control. At the same time, this invention clearly provides the identification method and adjustment method of candidate control units, providing more accurate prevention and control measures when the power system has the risk of instability under anticipated faults. Attached Figure Description

[0033] Figure 1 A flowchart illustrating a method for generating transient stability prevention and control measures according to an embodiment of the present invention;

[0034] Figure 2 A schematic diagram of a general process for transient stability analysis based on machine learning provided in an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram illustrating the LIME interpretation for transient stability analysis provided in an embodiment of the present invention;

[0036] Figure 4 A flowchart illustrating the LIME interpretation process provided in this embodiment of the invention;

[0037] Figure 5 This is a schematic diagram of the continuous prediction method provided in an embodiment of the present invention;

[0038] Figure 6 This is a LIME interpretation result for an unsafe sample provided in an embodiment of the present invention;

[0039] Figure 7 This is a schematic diagram illustrating the continuous prediction results of the transient stability analysis model when generating preventive and control measures provided in this embodiment of the invention.

[0040] Figure 8 The diagram shows the generator rotor angle and maximum relative rotor angle trajectory before and after preventive control, provided in an embodiment of the present invention. (a) is before preventive control, and (b) is after preventive control.

[0041] Figure 9 A block diagram of a transient stability prevention and control measure generation system provided in an embodiment of the present invention;

[0042] Figure 10 This is a block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0044] In this invention, the terms "first," "second," etc. (if present) in the invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0045] Example 1

[0046] See Figure 1 , combined Figures 2 to 8 The present invention provides a method for generating transient stability prevention and control measures, including operations S1 to S3.

[0047] Operation S1 involves using a machine learning model to predict the transient stability of the current power system operation under a anticipated fault scenario. The inputs of the machine learning model include various electrical features related to the transient stability of the power system.

[0048] Specifically, the machine learning model building process is as follows: Figure 2 As shown, a dataset is constructed by collecting actual power system operation data or manually generating operation data based on domain knowledge. This operation data includes various electrical characteristics (i.e., input features) related to the transient stability of the power system, such as generator output power and node voltage amplitude. Time-domain simulations under anticipated fault conditions are performed on the dataset, and stability criteria are used to add classification labels to the operation data. After preprocessing, the dataset can be used to train and test machine learning models. During the online evaluation phase, the state variables of the current operating mode are collected in real time, and the machine learning model is used to perform rapid transient stability analysis on the current power system operation mode, evaluating the transient safety of the current operating mode under anticipated fault scenarios.

[0049] Among them, the aforementioned machine learning models can employ models such as support vector machines and neural networks.

[0050] Operation S2 involves taking the predicted unstable operating mode as a sample to be explained and using the model interpretation method to determine the degree of influence of each electrical feature in the sample to be explained on transient stability.

[0051] Specifically, the model interpretation method can be selected from those used to explain sample predictions, such as Local Interpretable Model-agnostic (LIME) or Shapley Additive Explanation (SHAP). For unsafe current operating modes detected by the machine learning model in step S1, the model interpretation method is activated, and the prediction results of the unsafe operating modes are analyzed using this method to quantify the influence of input features on the unsafe prediction results.

[0052] In this embodiment, for the unsafe current operating mode detected by the machine learning model in step S1, the model interpretation method is activated, and the LIME interpretation method is used to analyze the prediction result of the unsafe operating mode, quantifying the degree of influence of input features on the unsafe prediction result.

[0053] The core idea of ​​LIME interpretation is to fit a linear interpretation model to a local region near the sample to be interpreted using a weighted perturbation set of samples, thereby obtaining the contribution of each feature of the sample to the model's judgment, i.e., the local interpretation result. A schematic diagram of LIME interpretation is shown below. Figure 3 As shown (selecting two characteristic generators in the transient stability analysis model, generator 1 active power P), G1 And generator 2 active power P G2 (as variables for the horizontal and vertical axes).

[0054] The linear model g is defined as follows, where w0 and w i These represent the intercept and weight coefficients of the model, respectively; x represents the sample to be explained, x i This represents the i-th feature of the sample to be explained.

[0055]

[0056] The explanatory model ξ(x) of the sample to be explained is fitted by minimizing the difference between the linear model g and the original transient stability analysis model (i.e. the machine learning model in step S1) f, as shown below.

[0057]

[0058]

[0059]

[0060] Where G represents the set of candidate explanatory models; Ω(g) represents the complexity of model g, that is, the number of features of model g constrained in a regularized manner; Let Z represent the loss function representing the "difference"; Z represents the perturbed sample set; W represents the loss function representing the "difference" ... W represents the loss function representing the "difference"; W represents the loss function representing the "difference"; W represents the loss function representing the "difference"; W represents the loss x (z) represents the weight of the perturbation sample z, defined as the distance from the perturbation sample z based on the Gaussian kernel to the sample x to be explained; σ represents the width of the Gaussian kernel. The perturbation sample z is generated by superimposing a perturbation randomly generated from the standard normal distribution around the sample x to be explained.

[0061] Specifically, the LIME interpretation process is as follows: Figure 4 As shown, first, the samples to be explained are determined; then, running samples are generated, and perturbation samples are labeled using the original transient stability analysis model; next, the weights of the perturbation samples are calculated, and the explanatory model is trained using the weighted perturbation samples; finally, the weight coefficients of the linear model are returned, which is the model explanation result.

[0062] Operation S3 involves determining candidate control units based on the degree of impact, adjusting the candidate control units, and then using the machine learning model to predict the transient stability of the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures.

[0063] Specifically, based on the degree of influence, the electrical characteristics are ranked, and at least one electrical characteristic with the greatest positive and negative impact on the transient stability prediction result is obtained. The corresponding unit is then identified as a candidate control unit. The active power of the candidate control unit is adjusted using a step-by-step adjustment method, while the machine learning model is used to perform transient stability prediction on the adjusted operating mode. If the adjusted power exceeds the output constraint of the candidate control unit, the range of candidate control units is expanded according to the degree of influence until a stable operating mode is obtained.

[0064] In this embodiment, considering the balance of power regulation in the power system, a paired control unit approach is adopted. The paired control units include positive control units and negative control units, with equal regulation power and opposite directions. It should be noted that the paired control units must include both positive and negative control units, but the number of positive and negative control units is not required to be the same.

[0065] After identifying candidate control units, a continuous prediction method is further used to search for safe operating points to determine adjustment parameters and generate candidate operating modes, as shown in the diagram below. Figure 5As shown, when using the continuous prediction method to search for a safe operating point, the power system is considered safe once the transient stability analysis model predicts a safety probability exceeding a threshold. In practical engineering applications, it is necessary to ensure the conservative operation of the power system; therefore, the safety probability threshold corresponding to the conservative operating point (search stopping point) should be reasonably increased.

[0066] Furthermore, transient stability prediction is performed on the new operating mode. If the initially generated new operating mode still does not meet the safety constraints, step-by-step regulation is implemented based on the aforementioned regulation rate. If the regulated active power exceeds the output constraints of the candidate control units, the range of candidate control units is expanded according to the aforementioned model interpretation results. This process is iterated until the generator units to be adjusted and their output that meet the safety constraints are determined, thus generating preventive control measures.

[0067] To further illustrate the model-interpretation-based method for generating transient stability prevention and control measures in power systems provided by this invention, an unsafe sample from the New England 10-unit 39-bus system was selected as a real-world unsafe scenario under a anticipated fault to validate the proposed method. The LIME interpretation results for this unsafe sample are as follows: Figure 6 As shown. The interpretation results indicate that feature P G2(31) Feature Vm Bus34 The model prediction (safety probability) is negatively correlated with the feature P. G9(38) Feature P G8(37) The features P are positively correlated with the model predictions. G2(31) The corresponding weight coefficient has the largest absolute value, indicating that it has the greatest impact on the insecurity prediction of the transient stability analysis model. Therefore, generator G2 is the optimal regulating generator set. Furthermore, considering the need to maintain power system power balance, generators G9 and G2 are selected as the regulating generator pair, performing positive and negative regulation respectively.

[0068] The safety probability threshold corresponding to the final operating point (search stop point) was set to 0.95. Generator sets G9 and G2 were continuously adjusted, and continuous predictions were performed using a trained transient stability analysis model. The results are as follows: Figure 7 As shown, the planned adjustment of the active power output of generator units to G9 and G2 at the final operating point is 338.05MW.

[0069] Furthermore, the initial and final operating points before and after preventive control were determined through power flow calculations, and the results are shown in Table 1. In the power flow calculation results, the actual regulation of the main frequency regulating unit (i.e., the balancing machine) G2 was 308.23MW, which deviated from the planned regulation by 29.82MW.

[0070] Table 1. Results of initial and final operating points before and after prevention and control measures.

[0071]

[0072] *Note: G2 and G9 are units with active power regulation.

[0073] The generator rotor angular trajectory before and after prevention and control is as follows: Figure 8 As shown, before the preventive control measures were implemented, the system was at risk of instability after the anticipated fault, and generator G2 deviated from the system, making the operating scenario unsafe. After the preventive control measures were implemented, the maximum relative rotor angle of the system was 114.8° during the simulation time, which is less than the safety threshold of 360°, and the system was in a transient stable state, indicating that the preventive control measures were effective.

[0074] The time consumption of the prevention and control measure generation method in this invention is shown in Table 2. The time consumption is divided into offline and online stages. The offline stage time mainly depends on data preprocessing and model building; the online stage time includes online transient stability analysis, model interpretation, and the search for the final (safe) operating point. The proposed method searches for the safe operating point at a speed of seconds, meeting the requirements for online applications in terms of time consumption.

[0075] Table 2. Time Consumption of Methods for Generating Prevention and Control Measures

[0076]

[0077] Traditional methods for generating transient stability preventive control measures involve multiple iterative alternations of time-domain simulation and optimal power flow calculation, which typically take several times longer than the sum of the time taken for time-domain simulation and optimal power flow calculation. Taking the New England 10-machine 39-bus system as an example, a 10-second time-domain simulation based on PSAT (with a step size of 0.01 seconds) takes 9.114 seconds, while optimal power flow takes 2.697 seconds. Therefore, compared to traditional transient stability preventive control, introducing a machine learning-based transient stability analysis model to replace time-domain simulation significantly improves the speed of online evaluation. The preventive control measure generation method in this invention has a shorter online execution time.

[0078] Further comparative analysis was conducted with other existing transient stability prevention and control methods in terms of online computation time, including particle swarm optimization, differential evolution, DC power flow differential evolution, and transient sensitivity methods, as shown in Table 3. The results show that, compared with other methods, the model-interpreted power system transient stability prevention and control measure generation method of this invention has a significant advantage in terms of online computation time.

[0079] Table 3 Comparison of online calculation time for various transient stability prevention and control methods

[0080]

[0081] Example 2

[0082] Figure 9 This is a block diagram of a transient stability prevention and control measure generation system provided in an embodiment of the present invention. (See also...) Figure 9 The transient stability prevention and control measures generation system 900 includes a prediction module 910, an interpretation module 920, and a generation module 930.

[0083] For example, the prediction module 910 performs operation S1 to perform transient stability prediction of the current operating mode of the power system under the anticipated fault scenario using a machine learning model. The input of the machine learning model includes various electrical features related to the transient stability of the power system.

[0084] The interpretation module 920, for example, performs operation S2, which uses the predicted unstable operating mode as a sample to be interpreted and uses the model interpretation method to determine the degree of influence of each electrical feature in the sample to be interpreted on transient stability.

[0085] The generation module 930, for example, performs operation S3 to determine candidate control units based on the degree of influence, adjust the candidate control units, and then use the machine learning model to make transient stability predictions on the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures.

[0086] Transient stability prevention and control measures generation system 900 is used to execute the above. Figures 1-8 The method for generating transient stability prevention and control measures in the illustrated embodiment. For details not covered in this embodiment, please refer to the foregoing. Figures 1-8 The method for generating transient stability prevention and control measures in the illustrated embodiment will not be described in detail here.

[0087] Example 3

[0088] This invention also provides an electronic device, such as... Figure 10 As shown, the electronic device 100 includes a processor 110 and a readable storage medium 120. The electronic device 100 can perform the above-described... Figures 1-8 The method for generating transient stability prevention and control measures described in the paper.

[0089] Specifically, processor 110 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. Processor 110 may also include onboard memory for caching purposes. Processor 110 may be used for executing reference... Figures 1-8 The method flow described according to embodiments of this disclosure refers to a single processing unit or multiple processing units performing different actions.

[0090] The readable storage medium 120 can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, the readable storage medium can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, apparatuses, or propagation media. Specific examples of readable storage media include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); memories such as random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0091] Readable storage medium 120 may include computer program 121, which may include code / computer-executable instructions that, when executed by processor 110, cause processor 110 to perform, for example, the above-described combination. Figures 1-8 The described method and any variations thereof.

[0092] Computer program 121 can be configured to have computer program code, for example, including computer program modules. For example, in an exemplary embodiment, the code in computer program 121 may include one or more program modules, such as 121A, 121B, ... It should be noted that the division and number of modules are not fixed. Those skilled in the art can use appropriate program modules or combinations of program modules according to the actual situation. When these combinations of program modules are executed by processor 110, processor 110 can perform, for example, the above-described combinations... Figures 1-8 The described method and any variations thereof.

[0093] Example 4

[0094] This invention also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement... Figures 1-8 The method for generating transient stability prevention and control measures is shown.

[0095] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating transient stability prevention and control measures, characterized in that, Includes the following steps: In anticipated fault scenarios, a machine learning model is used to predict the transient stability of the current operation mode of the power system. The input of the machine learning model includes various electrical features related to the transient stability of the power system. The predicted unstable operating mode is used as the sample to be explained; The method of determining the influence of each electrical feature in the sample to be explained on transient stability using a model interpretation method includes: generating perturbation samples based on the sample to be explained, and labeling the perturbation samples using the machine learning model; calculating the weights of the perturbation samples, and fitting a linear interpretation model using the weighted perturbation samples, wherein the weight coefficients of the linear interpretation model characterize the influence of each electrical feature in the sample to be explained on transient stability; Candidate control units are determined based on the degree of impact, and the candidate control units are adjusted. Then, the machine learning model is used to make transient stability predictions on the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures. The step of determining candidate control units based on the degree of influence includes: sorting each electrical feature according to the degree of influence, obtaining at least one electrical feature that has the greatest positive and negative impact on the transient stability prediction result, and determining the corresponding unit as a candidate control unit. The adjustment of the candidate control unit includes: adjusting the active power of the candidate control unit in a step-by-step manner, and using the machine learning model to make transient stability predictions on the adjusted operating mode; if the adjusted active power exceeds the output constraint of the candidate control unit, the range of candidate control units is expanded according to the degree of influence until a stable operating mode is obtained.

2. The method for generating transient stability prevention and control measures according to claim 1, characterized in that, The machine learning model was trained in the following way: Collect actual power system operation data or generate operation data manually; Based on time-domain simulation under anticipated faults, the running data is labeled; Train a machine learning model using labeled runtime data.

3. The method for generating transient stability prevention and control measures according to claim 1, characterized in that, The generation of perturbation samples based on the sample to be explained includes: Centered on the sample to be explained, a perturbation randomly generated from a standard normal distribution is superimposed to obtain a perturbation sample set within the neighborhood of the sample to be explained.

4. A transient stability prevention and control measure generation system, characterized in that, The method for generating transient stability prevention and control measures according to any one of claims 1-3 includes: The prediction module is used to perform transient stability prediction of the current operating mode of the power system under anticipated fault scenarios using a machine learning model. The input of the machine learning model includes various electrical features related to the transient stability of the power system. The interpretation module is used to take the predicted unstable operating mode as the sample to be interpreted and use the model interpretation method to determine the degree of influence of each electrical feature in the sample to be interpreted on transient stability. The generation module is used to determine candidate control units based on the degree of impact, adjust the candidate control units, and then use the machine learning model to make transient stability predictions on the adjusted operating mode until a stable operating mode is obtained, thereby generating transient stability prevention and control measures. The explanation module is further configured to generate perturbation samples based on the sample to be explained, and label the perturbation samples using the machine learning model; calculate the weights of the perturbation samples, and fit a linear explanation model using the weighted perturbation samples, wherein the weight coefficients of the linear explanation model characterize the degree of influence of each electrical feature in the sample to be explained on transient stability.

5. An electronic device, characterized in that, include: processor; A memory storing a computer-executable program, which, when executed by the processor, causes the processor to perform the transient stability prevention and control measure generation method as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by the processor, implement the transient stability prevention and control measure generation method as described in any one of claims 1 to 3.

Images