Method, device and equipment for evaluating transient stability of power system and medium

Abstract

The invention discloses an evaluation method, device and equipment for transient stability of an electric power system and a medium. The evaluation method for transient stability of the electric power system comprises the following steps: acquiring real-time operation data of the electric power system; respectively inputting the real-time operation data into a preset positive classifier and a preset negative classifier to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier; when the first evaluation result is different from the second evaluation result, inputting the real-time operation data into a preset fair classifier so as to evaluate the transient stability of the power system; the justice classifier is a classifier obtained by training based on a hard sample of the power system. According to the method, sample data, namely hard samples, which are difficult to be correctly classified and evaluated in the power system can be accurately classified, so that the accuracy of transient stability evaluation of the power system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system automation, and in particular to a method, device, equipment and medium for evaluating transient stability of a power system. Background Art

[0002] Power angle stability is crucial to the safe operation of the power system. Once the power system is disturbed, if the power angle stability is evaluated as soon as possible and emergency control measures are taken in time, the stability of the system can be protected. In the face of challenges such as the increasing access to new energy sources and system load fluctuations, ensuring the power angle stability of the power system has become an urgent task. In order to solve this problem, establishing a power angle stability analysis model is a crucial step.

[0003] In current research at home and abroad, a variety of methods have been explored to evaluate the transient stability of power systems. However, existing technologies usually misclassify hard samples, thus affecting the accuracy of transient stability evaluation of power systems. Summary of the invention

[0004] The present invention provides a method, device, equipment and medium for evaluating transient stability of a power system to solve the above-mentioned technical problems and effectively improve the accuracy of transient stability evaluation of a power system.

[0005] In order to solve the above technical problems, the present invention provides a method for evaluating transient stability of a power system, comprising:

[0006] Obtain real-time operation data of the power system;

[0007] Inputting the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier;

[0008] When the first evaluation result and the second evaluation result are different, the real-time operation data is input into a preset fair classifier to evaluate the transient stability of the power system; the fair classifier is a classifier trained based on hard samples of the power system.

[0009] The beneficial effects of the present invention are:

[0010] The present invention inputs the real-time operation data of the power system into two classifiers with opposite evaluation tendencies respectively to obtain two evaluation results. When the two evaluation results are different, the real-time operation data of the power system is input into a fair classifier trained based on hard samples of the power system to evaluate the transient stability of the power system. In this way, sample data in the power system that is difficult to be correctly classified and evaluated, that is, hard samples, can be accurately classified, thereby improving the accuracy of transient stability evaluation of the power system.

[0011] As a preferred solution, the loss function of the preset positive classifier is specifically:

[0012] G p =-y (l) ln y' (l) γ(1,0)-[1-y (l) ]ln[1-y (l) ]γ(0,1)

[0013] γ(0,1)=1

[0014] γ(1,0)=Ns / Nuns

[0015] The loss function of the preset anti-classifier is:

[0016] G n =-y (l) ln y' (l) θ(1,0)-β[1-y (l) ]ln[1-y (l) ]θ(0,1)

[0017] θ(0,1)=Ns / Nuns

[0018] θ(1,0)=1

[0019] In the above formula, y (l) is the true label value of the lth sample in the input sample; y' (l) is the probability that the lth sample in the input sample is a stable sample; Ns is the number of stable samples in the input sample; Nuns is the number of unstable samples in the input sample; β is the correction coefficient.

[0020] For training set samples in different scenarios, the ratio of unstable samples to stable samples is often different, which makes the discrimination results of the model trained under unbalanced samples have a certain tendency. The preferred solution introduces parameters γ and θ into the loss function of the classifier, which can correct the imbalance in the number of various samples in each scenario. The preferred solution also introduces a correction coefficient β into the loss function of the anti-classifier, which can make the anti-classifier tend to evaluate the input samples as unstable, thereby making the evaluation tendency of the positive classifier opposite to the evaluation tendency of the anti-classifier.

[0021] As a preferred solution, the method for obtaining the preset fair classifier includes:

[0022] Conduct time domain simulation on the power system to obtain power system simulation data;

[0023] obtaining a plurality of initial data sets based on the power system simulation data;

[0024] Inputting a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set;

[0025] From a number of initial data sets, an initial data set with different positive classifier evaluation results and negative classifier evaluation results is selected to obtain a critical data set;

[0026] A preset initial fair classifier is trained based on a critical data set to obtain a preset fair classifier.

[0027] This preferred scheme selects an initial data set with different positive classifier evaluation results and negative classifier evaluation results from several initial data sets to obtain a critical data set, and trains a preset initial fair classifier based on the critical data set to obtain a preset fair classifier, so that the fair classifier has the ability to learn the characteristics of the critical data set well, and can accurately classify sample data that are difficult to be correctly classified and evaluated in the power system, that is, hard samples, thereby improving the accuracy of transient stability assessment of the power system.

[0028] As a preferred solution, after selecting the initial data sets with different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain the critical data set, the method further includes:

[0029] Perform data augmentation on the critical dataset to update the critical dataset.

[0030] If the collected critical data sets are too few, they cannot be used to train the model. However, if the number of critical data sets is sufficient and an appropriate model is selected, that is, a model that can learn the characteristics of the critical data sets well, then the model will obtain a strong ability to correctly identify the critical data set categories through training with a sufficient number of critical data sets. This preferred solution performs data enhancement on the critical data sets, which can enrich the quantity and quality of the critical sample sets. In addition, the preset initial fair classifier is trained with the enhanced critical data sets, which can improve the classification and recognition ability of the fair classifier and improve the accuracy of the transient stability assessment of the power system.

[0031] As a preferred solution, the method for evaluating transient stability of a power system further includes: if the first evaluation result and the second evaluation result are the same, outputting the first evaluation result to complete the evaluation of transient stability of the power system.

[0032] Accordingly, in order to solve the above technical problems, the present invention also provides a device for evaluating transient stability of a power system, comprising: a data acquisition module, a first evaluation module and a second evaluation module;

[0033] Wherein, the data acquisition module is used to acquire real-time operation data of the power system;

[0034] The first evaluation module is used to input the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier;

[0035] The second evaluation module is used to input the real-time operation data into a preset fair classifier to evaluate the transient stability of the power system when the first evaluation result and the second evaluation result are different; the fair classifier is a classifier trained based on hard samples of the power system.

[0036] As a preferred solution, the power system transient stability assessment device further includes: a model acquisition module; the model acquisition module includes: a first acquisition unit, a second acquisition unit and a model acquisition unit;

[0037] Wherein, the first acquisition unit is used to perform time domain simulation on the power system to obtain power system simulation data; and obtain a plurality of initial data sets based on the power system simulation data;

[0038] The second acquisition unit is used to input a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set; and to screen out initial data sets having different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain a critical data set;

[0039] The model acquisition unit is used to train a preset initial fair classifier based on a critical data set to obtain a preset fair classifier.

[0040] As a preferred embodiment, the device for evaluating transient stability of a power system further includes: a data enhancement module; the data enhancement module is used to perform data enhancement on the critical data set after the initial data set is respectively input into a preset positive classifier and a preset negative classifier to obtain a critical data set, so as to update the critical data set.

[0041] The present invention also provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for evaluating transient stability of an electric power system as described in any one of the above items are implemented.

[0042] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for evaluating transient stability of a power system as described in any one of the above items are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 : A flow chart of an embodiment of a method for evaluating transient stability of a power system provided by the present invention;

[0044] Figure 2 : A structural schematic diagram of an embodiment of an electric power system transient stability assessment device provided by the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Embodiment 1

[0047] In order to solve the above technical problems, the present invention provides a method, device, equipment and medium for evaluating transient stability of a power system to solve the above technical problems and effectively improve the accuracy of transient stability evaluation of a power system.

[0048] Please refer to Figure 1 , is a flow chart of an embodiment of a method for evaluating transient stability of a power system provided by an embodiment of the present invention. Figure 1As shown, the method for evaluating transient stability of a power system includes steps 101 to 103, specifically:

[0049] Step 101: Acquire real-time operation data of the power system;

[0050] Step 102: inputting the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier;

[0051] Step 103: When the first evaluation result and the second evaluation result are different, the real-time operation data is input into a preset fair classifier to evaluate the transient stability of the power system; the fair classifier is a classifier trained based on hard samples of the power system.

[0052] The present invention inputs the real-time operation data of the power system into two classifiers with opposite evaluation tendencies respectively to obtain two evaluation results. When the two evaluation results are different, the real-time operation data of the power system is input into a fair classifier trained based on hard samples of the power system to evaluate the transient stability of the power system. In this way, sample data in the power system that is difficult to be correctly classified and evaluated, that is, hard samples, can be accurately classified, thereby improving the accuracy of transient stability evaluation of the power system.

[0053] In this embodiment, one implementation method is to integrate a preset positive classifier, a preset negative classifier and a preset fair classifier into an integrated TSA model through an integration strategy. When it is necessary to evaluate the transient stability of the power system, the integrated TSA model is used to evaluate the transient stability of the power system.

[0054] Furthermore, the loss function of the preset positive classifier is specifically:

[0055] G p =-y (l) ln y' (l) γ(1,0)-[1-y (l) ]ln[1-y (l) ]γ(0,1),

[0056] γ(0,1)=1,

[0057] γ(1,0)=Ns / Nuns,

[0058] The loss function of the preset anti-classifier is:

[0059] G n =-y (l) ln y' (l) θ(1,0)-β[1-y(l) ]ln[1-y (l) ]θ(0,1),

[0060] θ(0,1)=Ns / Nuns,

[0061] θ(1,0)=1,

[0062] In the above formula, y (l) is the true label value of the lth sample in the input sample; y' (l) is the probability that the lth sample in the input sample is a stable sample; Ns is the number of stable samples in the input sample; Nuns is the number of unstable samples in the input sample; β is the correction coefficient.

[0063] For the transient stability assessment problem of power system, the classifier parameter modification weight coefficient can be expressed in Table 1.

[0064] Table 1 Model parameter correction weight coefficient matrix

[0065]

[0066] Among them, C(1,1)=C(0,0)=0. C(1,0) and C(0,1) represent the weight coefficients for correcting the model parameters when predicting a stable situation as unstable and predicting an unstable situation as stable, respectively. Usually, C(1,0)=C(0,1)=1.

[0067] In this embodiment, cost sensitivity is applied to the selected primary classifier, so that a model that originally had no preference can have a tendency, that is, a positive model that tends to evaluate the input sample as stable, and a negative model that tends to evaluate the input sample as unstable.

[0068] For training set samples in different scenarios, the ratio of unstable samples to stable samples is often different, which makes the discrimination results of the model trained under unbalanced samples have a certain tendency. The present invention introduces parameters γ and θ in the loss function of the primary classifier, which can correct the imbalance in the number of various samples in each scenario. By changing the primary classifier parameter correction process to adjust the fitting of the unbalanced samples, the loss function of the preset positive classifier is obtained accordingly. After cost-sensitive application, the positive classifier tends to evaluate the input sample as a stable sample.

[0069] Because the original classifier model is affected by sample imbalance, it has a tendency to evaluate the sample as stable. Therefore, in order to make the classifier model tend to evaluate the input sample as unstable, it is necessary to introduce a correction coefficient β. Therefore, the present invention also introduces a correction coefficient β in the loss function of the anti-classifier, which can make the anti-classifier tend to evaluate the input sample as an unstable sample, and finally make the evaluation tendency of the positive classifier opposite to the evaluation tendency of the anti-classifier.

[0070] In this embodiment, for the correction coefficient β, when β=1, the anti-classifier only corrects the tendency problem caused by the unbalanced training samples; when β>1, the anti-classifier improves the fitting degree of unstable class samples. The larger the value, the smaller the probability of misclassification of unstable class samples.

[0071] Furthermore, the method for obtaining the preset fair classifier includes:

[0072] Conduct time domain simulation on the power system to obtain power system simulation data;

[0073] obtaining a plurality of initial data sets based on the power system simulation data;

[0074] Inputting a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set;

[0075] From a number of initial data sets, an initial data set with different positive classifier evaluation results and negative classifier evaluation results is selected to obtain a critical data set;

[0076] A preset initial fair classifier is trained based on a critical data set to obtain a preset fair classifier.

[0077] In this embodiment, performing time domain simulation on the power system to obtain power system simulation data includes:

[0078] The corresponding power system is divided into 4 areas according to the adjacent busbars. Considering 4 different compliance levels of 60%, 70%, 120%, and 125%, different load levels are assigned to the four areas respectively, and a total of 256 different flows, that is, 256 different operating modes, are obtained, and the generator output is adjusted accordingly to ensure the convergence of the flow calculation. The fault is set as a three-phase short circuit fault, and the system is scanned N-1, and faults are set on different lines one after another. The fault duration is set to 0.2s. On the line without a transformer, the fault point is set at 50% of the line, and on the line with a transformer, the fault point is set at the first section of the line. Thus, a total of tens of thousands of samples are generated to obtain power system simulation data.

[0079] In this embodiment, before inputting several initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set, it also includes: constructing a loss function of the positive classifier and the negative classifier, and after the loss function of the positive classifier and the negative classifier is constructed, the initial data sets are respectively input into the positive classifier and the negative classifier for training.

[0080] The present invention inputs an initial data set into a positive and negative classifier for evaluation, collects data with inconsistent evaluation results of the positive and negative classifiers, sorts them out to obtain a critical data set, and trains a preset initial fair classifier based on the critical data set to obtain a preset fair classifier, so that the fair classifier has the ability to learn the characteristics of the critical data set well, and can accurately classify sample data that is difficult to be correctly classified and evaluated in the power system, that is, hard samples, thereby improving the accuracy of transient stability evaluation of the power system.

[0081] Furthermore, after selecting the initial data sets with different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain the critical data set, the method further includes:

[0082] Perform data augmentation on the critical dataset to update the critical dataset.

[0083] If the collected critical data sets are too few, they cannot be used to train the model. However, if the number of critical data sets is sufficient and an appropriate model is selected, that is, a model that can well learn the characteristics of the critical data sets, then the model will obtain a strong ability to correctly identify the critical data set categories through training with a sufficient number of critical data sets. The present invention performs data enhancement on the critical data sets, which can enrich the quantity and quality of the critical sample sets. In addition, the preset initial fair classifier can be trained with the enhanced critical data sets, which can improve the classification and recognition ability of the fair classifier and improve the accuracy of the transient stability assessment of the power system.

[0084] In this embodiment, the critical sample set can be enhanced by using a GAN model, but is not limited to it. Specifically, the obtained critical data set is input into the GAN model, the discriminator and the generator form an adversarial network, and the distribution characteristics of the critical data set are learned through alternating training, and critical data that conforms to the real distribution is generated in batches, thereby updating the critical data set to obtain an enhanced critical sample set.

[0085] In this embodiment, the fair classifier is used to classify hard samples, that is, samples that cannot be correctly classified by the initial simple classification. Then, the sample set used to train the fair classifier is the enhanced critical sample set, and the present invention also takes this factor into consideration when selecting the model of the fair classifier.

[0086] When the preset positive classifier and the preset negative classifier cannot correctly classify the real-time operation data of the power system, a fair classifier is needed to evaluate the transient stability of the power system. In order to improve the accuracy of the evaluation of the transient stability of the power system, the output result of the fair classifier needs to be high in accuracy and persuasive. Secondly, the sample set used to train the fair classifier is a critical data set, and the data information of the critical sample set is very rich and the information entropy is very high. A suitable model is needed to extract the feature information of this part.

[0087] For ELM model training, ideal samples should be mainly data close to the stability boundary, so as to obtain refined stability rules, which can enable a good learning of the critical area and thus have a strong ability to correctly distinguish the category of hard samples.

[0088] Based on the above considerations, the embodiment of the present invention selects the ELM model as a fair classifier. Because the distribution characteristics of the critical data set just meet the requirements of ELM model training, if the ELM training samples are samples in the critical area, then the ELM model after training will have a strong correct classification ability for the critical data set.

[0089] Furthermore, the method for evaluating transient stability of a power system further includes: if the first evaluation result and the second evaluation result are the same, outputting the first evaluation result to complete the evaluation of transient stability of the power system.

[0090] In an embodiment of the present invention, the process of evaluating the transient stability of the power system is divided into two stages, wherein a preset positive classifier and a preset negative classifier are first-stage evaluation models, and the evaluation result output by the first-stage evaluation model is the first-stage evaluation result; the preset fair classifier is the second-stage evaluation model, and the evaluation result output by the second-stage evaluation model is the second-stage evaluation result; the first-stage evaluation model and the second-stage evaluation model are integrated into an integrated TSA model.

[0091] The embodiment of the present invention designs a score for the evaluation result of the first stage, and defines λ, λ1 and λ2, wherein λ=λ1+λ2, λ is the total score of the first stage, and λ1 and λ2 are the positive and negative classifier scores, respectively.

[0092] In order to greatly improve the running speed, λ1 and λ2 are linked to the output results of the positive and negative classifiers. If the output result of the positive classifier is stable (label is 1), then λ1 is 1 at this time, otherwise it is 0.

[0093] The relationship between λ, λ1 and λ2 is shown in Table 2.

[0094] Table 2 λ score table

[0095] λ λ2=0 λ2=1 λ1=0 0 1 λ1=1 1 2

[0096] Only when the output results of the positive and negative classifiers in the first stage evaluation are consistent, the results are credible and the evaluation results are directly output, that is, when λ is 0 or 2, the first stage evaluation results are credible and directly output, and there is no need to perform the second stage evaluation. When the output results of the positive and negative classifiers in the first stage evaluation are inconsistent, that is, the value of λ is 1, the first stage evaluation results are unreliable, and the second stage evaluation is entered. In the embodiment of the present invention, the overall evaluation process is shown in Table 3.

[0097] Table 3 Evaluation process

[0098]

[0099] As shown in Table 3, when the total evaluation score is 1, the output of the second-stage evaluation result is the evaluation result of the integrated TSA model; when the total evaluation score is 2, that is, λ is 1, the evaluation result output by the first-stage evaluation model is credible, and the output of the first-stage evaluation result is the evaluation result of the integrated TSA model.

[0100] Therefore, the overall process of the transient stability assessment of the power system in the embodiment of the present invention is that when the real-time acquired data of the power system is evaluated, the real-time acquired data is input into the positive and negative classifiers for evaluation. When the evaluation results are consistent, the evaluation results are output to complete the evaluation; when the evaluation results are inconsistent, the real-time acquired data is input into the fair classifier, the evaluation results of the fair classifier are output to complete the evaluation.

[0101] Accordingly, in order to solve the above technical problems, the present invention also provides an evaluation device for transient stability of a power system, please refer to Figure 2 , which is a structural schematic diagram of an embodiment of an electric power system transient stability evaluation device provided by the present invention.

[0102] like Figure 2 As shown, the power system transient stability evaluation device 20 includes: a data acquisition module 201, a first evaluation module 202 and a second evaluation module 203;

[0103] Wherein, the data acquisition module 201 is used to acquire real-time operation data of the power system;

[0104] The first evaluation module 202 is used to input the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier;

[0105] The second evaluation module 203 is used to input the real-time operation data into a preset fair classifier to evaluate the transient stability of the power system when the first evaluation result and the second evaluation result are different; the fair classifier is a classifier trained based on hard samples of the power system.

[0106] Furthermore, the power system transient stability assessment device further includes: a model acquisition module; the model acquisition module includes: a first acquisition unit, a second acquisition unit and a model acquisition unit;

[0107] Wherein, the first acquisition unit is used to perform time domain simulation on the power system to obtain power system simulation data; and obtain a plurality of initial data sets based on the power system simulation data;

[0108] The second acquisition unit is used to input a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set; and to screen out initial data sets having different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain a critical data set;

[0109] The model acquisition unit is used to train a preset initial fair classifier based on a critical data set to obtain a preset fair classifier.

[0110] Furthermore, the device for evaluating transient stability of a power system also includes: a data enhancement module; the data enhancement module is used to perform data enhancement on the critical data set after the initial data set is respectively input into a preset positive classifier and a preset negative classifier to obtain a critical data set, so as to update the critical data set.

[0111] The present invention also provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for evaluating transient stability of an electric power system as described in any one of the above items are implemented.

[0112] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for evaluating transient stability of a power system as described in any one of the above items are implemented.

[0113] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0114] The present invention combines cost sensitivity and integrated learning to realize a two-stage transient stability assessment of a power system. Compared with the existing methods, the present invention enhances hard samples, selects a suitable model as a fair classifier, and all samples used to train the fair classifier are critical samples, which fully meets the training requirements of an extreme learning machine (ELM). At the same time, the credibility of the classification results of the fair classifier is greatly improved. Considering the imbalance of samples, the classification ability and persuasiveness of the fair classifier for critical samples are enhanced, so that the model proposed in the present invention can not only quickly classify ordinary samples, but also reliably classify hard samples. Therefore, the present invention improves the efficiency, precision and accuracy of transient stability assessment of power systems, and has the advantages of high credibility, strong real-time performance, high precision, and low possibility of misclassification.

[0115] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating transient stability of a power system, characterized in that: include: Obtain real-time operation data of the power system; Inputting the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier; When the first evaluation result and the second evaluation result are different, the real-time operation data is input into a preset fair classifier to evaluate the transient stability of the power system; the fair classifier is a classifier trained based on hard samples of the power system.

2. A method for evaluating transient stability of a power system according to claim 1, characterized in that: The loss function of the preset positive classifier is: G p =s (l) flax ’(l) γ(1,0)-[1-y (l) ]ln[1-y (l) ]γ(0,1) γ(0,1)=1 γ(1,0)=Ns / Nuns The loss function of the preset anti-classifier is: G n =-y (l) lny ’(l) θ(1,0)-β[1-y (l) ]ln[1-y (l) ]θ(0,1) θ(0,1)=Ns / Nuns θ(1,0)=1 In the above formula, y (l) is the true label value of the lth sample in the input sample; y' (l) is the probability that the lth sample in the input sample is a stable sample; Ns is the number of stable samples in the input sample; Nuns is the number of unstable samples in the input sample; β is the correction coefficient.

3. A method for evaluating transient stability of a power system according to claim 1, characterized in that: The method for obtaining the preset fair classifier includes: Conduct time domain simulation on the power system to obtain power system simulation data; obtaining a plurality of initial data sets based on the power system simulation data; Inputting a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set; From a number of initial data sets, an initial data set with different positive classifier evaluation results and negative classifier evaluation results is selected to obtain a critical data set; A preset initial fair classifier is trained based on a critical data set to obtain a preset fair classifier.

4. A method for evaluating transient stability of a power system as claimed in claim 3, characterized in that: After selecting the initial data sets with different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain the critical data set, the method further includes: Perform data augmentation on the critical dataset to update the critical dataset.

5. A method for evaluating transient stability of a power system according to claim 1, characterized in that: Also includes: If the first evaluation result and the second evaluation result are the same, the first evaluation result is output to complete the evaluation of the transient stability of the power system.

6. A device for evaluating transient stability of a power system, characterized in that: include: A data acquisition module, a first evaluation module and a second evaluation module; Wherein, the data acquisition module is used to acquire real-time operation data of the power system; The first evaluation module is used to input the real-time operation data into a preset positive classifier and a preset negative classifier respectively to obtain a first evaluation result and a second evaluation result; wherein the evaluation tendency of the preset positive classifier is opposite to the evaluation tendency of the preset negative classifier; The second evaluation module is used to input the real-time operation data into a preset fair classifier to evaluate the transient stability of the power system when the first evaluation result and the second evaluation result are different; the fair classifier is a classifier trained based on hard samples of the power system.

7. The device for evaluating transient stability of a power system according to claim 6, characterized in that: Also includes: Model acquisition module; The model acquisition module includes: a first acquisition unit, a second acquisition unit and a model acquisition unit; Wherein, the first acquisition unit is used to perform time domain simulation on the power system to obtain power system simulation data; and obtain a plurality of initial data sets based on the power system simulation data; The second acquisition unit is used to input a plurality of initial data sets into a preset positive classifier and a preset negative classifier to obtain a positive classifier evaluation result and a negative classifier evaluation result corresponding to each initial data set; and to screen out initial data sets having different positive classifier evaluation results and negative classifier evaluation results from the plurality of initial data sets to obtain a critical data set; The model acquisition unit is used to train a preset initial fair classifier based on a critical data set to obtain a preset fair classifier.

8. The device for evaluating transient stability of a power system according to claim 6, characterized in that: Also includes: Data augmentation module; The data enhancement module is used to perform data enhancement on the critical data set to update the critical data set after the initial data set is respectively input into a preset positive classifier and a preset negative classifier to obtain a critical data set.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for evaluating transient stability of a power system as claimed in any one of claims 1 to 5 are implemented.

10. A 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 evaluating transient stability of a power system as claimed in any one of claims 1 to 5 are implemented.

Images