A method and system for determining static voltage stability of a power system

Abstract

The application relates to a power system static voltage stability determination method and system, which comprises the following steps: generating characteristic data of different prediction points by using a continuous power flow method; determining the label of each characteristic data based on a static voltage stability threshold; constructing a sample data set according to the characteristic data of different prediction points and the label of each characteristic data; performing random oversampling processing on the sample data set to balance the positive sample data and the negative sample data in the sample data set; training a support vector machine model by using the sample data set after random oversampling to obtain a trained support vector machine model; and determining the static voltage stability of the power system by using the trained support vector machine model. The application generates characteristic data by using the continuous power flow method to increase sample data, and balances the sample data by using the random oversampling method, thereby improving the accuracy of the determination of the static voltage stability of the power system and reducing misclassification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system monitoring, in particular to a method and system for determining static voltage stability of a power system. BACKGROUND

[0002] There are two kinds of misclassification in static voltage stability assessment: one is to judge the unsafe state as the safe state, and the second is to judge the safe state as the unsafe state. There are fewer samples that can be directly obtained in the process of static voltage stability assessment, and the samples in the unstable state are much lower than those in the stable state, which causes a serious imbalance problem, resulting in a high proportion of misclassification in the existing classification method. SUMMARY

[0003] Therefore, the present application provides a method and system for determining static voltage stability of a power system to improve the accuracy of determining static voltage stability of a power system and reduce misclassification.

[0004] To achieve the above object, the present application provides the following solutions:

[0005] A method for determining static voltage stability of a power system, comprising the following steps:

[0006] Generating feature data of different prediction points by using the continuous power flow method; the feature data includes voltage amplitude and phase angle of nodes in the power system;

[0007] Determining the label of each feature data based on the static voltage stability threshold; when the voltage amplitude in the feature data is greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power system is unstable, and when the voltage amplitude in the feature data is not greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power system is stable;

[0008] According to the feature data of different prediction points and the label of each feature data, a sample data set is constructed;

[0009] Randomly oversampling the sample data set to balance the positive sample data and the negative sample data in the sample data set, and obtaining a sample data set after random oversampling; the positive sample data represents the sample with the label of static voltage stability of the power system, and the negative sample data represents the sample with the label of static voltage instability of the power system;

[0010] Training a support vector machine model by using the sample data set after random oversampling to obtain a trained support vector machine model;

[0011] Determining the static voltage stability of the power system by using the trained support vector machine model.

[0012] Optionally, the generating the characteristic data of different prediction points by the continuous power flow method specifically comprises:

[0013] A continuously increasing load is applied to the power system, and the characteristic data of the power system under the action of the continuously increasing load at each prediction point is sequentially predicted by the continuous power flow method.

[0014] Optionally, the static voltage stability threshold is 1.15V cr ; wherein V cr represents the limit voltage.

[0015] Optionally, the training the support vector machine model using the sample data set after random oversampling to obtain the trained support vector machine model specifically comprises:

[0016] Divide the sample data set after random oversampling into a training set and a test set;

[0017] Initialize a search region and a search step;

[0018] In the search region, according to the training set and the search step, the hyperparameters of the support vector machine model are optimized by using the grid search method and the multiple cross-validation method to obtain the trained support vector machine model; the hyperparameters include a penalty parameter and a Gaussian kernel function;

[0019] The trained support vector machine model is verified by using the test set;

[0020] When the verification is passed, the trained support vector machine model is output as the trained support vector machine model;

[0021] When the verification fails, a preset range centered on the hyperparameters of the trained support vector machine model is taken as the search region, and the search step is reduced, and the step of "in the search region, according to the training set, the hyperparameters of the support vector machine model are optimized by using the grid search method and the multiple cross-validation method to obtain the trained support vector machine model" is returned.

[0022] Optionally, the training the support vector machine model using the sample data set after random oversampling to obtain the trained support vector machine model specifically comprises:

[0023] In the search region, according to the search step, a two-dimensional hyperparameter network is generated; each network node in the two-dimensional hyperparameter network represents a hyperparameter;

[0024] According to the training set, multiple cross-validations are performed on the support vector machine model under each hyperparameter in the two-dimensional hyperparameter network to obtain a classification accuracy of the support vector machine model under each hyperparameter;

[0025] According to the classification accuracy of the support vector machine model under each hyperparameter, a classification accuracy contour is constructed.

[0026] According to the classification accuracy contour, the support vector machine model under the hyperparameter corresponding to the highest classification accuracy is determined as the trained support vector machine model.

[0027] A power system static voltage stability determination system, the system comprising:

[0028] A feature data generation module is configured to generate feature data of different prediction points by using a continuation power flow method; the feature data comprises voltage amplitude and phase angle of nodes in the power system;

[0029] A label classification module is configured to determine a label of each feature data based on a static voltage stability threshold; when the voltage amplitude in the feature data is greater than the static voltage stability threshold, the label of the feature data is used to represent that the power system static voltage is unstable; when the voltage amplitude in the feature data is not greater than the static voltage stability threshold, the label of the feature data is used to represent that the power system static voltage is stable.

[0030] A sample data set construction module is configured to construct a sample data set according to the feature data of different prediction points and the label of each feature data.

[0031] A random oversampling module is configured to perform random oversampling processing on the sample data set to balance positive sample data and negative sample data in the sample data set, and obtain a sample data set after random oversampling; the positive sample data represents a sample with a label of power system static voltage stability, and the negative sample data represents a sample with a label of power system static voltage instability.

[0032] A training module is configured to train a support vector machine model by using the sample data set after random oversampling to obtain a trained support vector machine model.

[0033] A power system static voltage stability determination module is configured to determine power system static voltage stability by using the trained support vector machine model.

[0034] Optionally, the feature data generation module specifically comprises:

[0035] A feature data generation sub-module is configured to simulate the application of continuously increasing load to the power system, and sequentially predict feature data of the power system at each prediction point under the action of the continuously increasing load by using the continuation power flow method.

[0036] Optionally, the static voltage stability threshold is 1.15V cr ; wherein V cr represents the limit voltage.

[0037] Optionally, the training module specifically comprises:

[0038] a sample data set division sub-module, configured to divide the sample data set after random oversampling into a training set and a test set;

[0039] an initialization sub-module, configured to initialize a search region and a search step;

[0040] a training sub-module, configured to optimize hyperparameters of a support vector machine model in the search region according to the training set and the search step, using a grid search method and a multiple cross-validation method, to obtain a trained support vector machine model; the hyperparameters include a penalty parameter and a Gaussian kernel function;

[0041] a verification sub-module, configured to verify the trained support vector machine model using the test set;

[0042] a model output sub-module, configured to output the trained support vector machine model as a trained support vector machine model when the verification passes;

[0043] a return sub-module, configured to, when the verification fails, take a preset range centered on the hyperparameters of the trained support vector machine model as the search region, and reduce the search step, and return to the step of optimizing the hyperparameters of the support vector machine model in the search region according to the training set, using the grid search method and the multiple cross-validation method, to obtain the trained support vector machine model.

[0044] Optionally, the training sub-module specifically comprises:

[0045] a two-dimensional hyperparameter network generation unit, configured to generate a two-dimensional hyperparameter network in the search region according to the search step; each network node in the two-dimensional hyperparameter network represents a hyperparameter;

[0046] a multiple cross-validation unit, configured to perform multiple cross-validations on support vector machine models under each hyperparameter in the two-dimensional hyperparameter network according to the training set, to obtain classification accuracy rates of the support vector machine models under each hyperparameter;

[0047] a classification accuracy rate contour line construction unit, configured to construct a classification accuracy rate contour line according to the classification accuracy rates of the support vector machine models under each hyperparameter;

[0048] The training unit is configured to determine, according to the classification accuracy contour, a support vector machine model corresponding to the highest classification accuracy as the trained support vector machine model.

[0049] According to the specific embodiments of the present application, the following technical effects are disclosed:

[0050] The application discloses a power system static voltage stability determination method and system, and the method comprises the following steps: generating characteristic data of different prediction points by using a continuation power flow method; determining the label of each characteristic data based on a static voltage stability threshold; constructing a sample data set according to the characteristic data of different prediction points and the label of each characteristic data; performing random oversampling processing on the sample data set to balance the positive sample data and the negative sample data in the sample data set, and obtaining a sample data set after random oversampling; training a support vector machine model by using the sample data set after random oversampling, and obtaining a trained support vector machine model; and determining the power system static voltage stability by using the trained support vector machine model. The characteristic data is generated by using the continuation power flow method, so that the sample data is increased, the random oversampling method is used to balance the sample, the accuracy of the power system static voltage stability determination is improved, and the misclassification is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0052] Figure 1 A flow chart of a power system static voltage stability determination method provided for the embodiment 1 of the present application is provided.

[0053] Figure 2 A principle diagram of a power system static voltage stability determination method provided for the embodiment 1 of the present application is provided.

[0054] Figure 3 A flow chart of a grid search method provided for the embodiment 1 of the present application is provided. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0056] The application aims to provide a power system static voltage stability determination method and system to improve the accuracy of power system static voltage stability determination and reduce misclassification.

[0057] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the application will be further described in detail below with reference to the drawings and specific embodiments.

[0058] Embodiment 1

[0059] As shown in Figure 1 and 2 , the application embodiment 1 provides a power system static voltage stability determination method, which comprises the following steps:

[0060] Step 101: generating feature data of different prediction points by using the continuation power flow method; the feature data includes the voltage amplitude and phase angle of nodes in the power system.

[0061] Continuation power flow method: a method for predicting the stability limit by applying the continuation method to the power system. The next solution of the power flow is predicted by giving an initial point on the curve and correcting the next point on the curve. For example: there are n equations g(x) = 0, x ∈ R. Add an equation containing λ to the equation group, predict x by changing λ, and get the equation f(x, y) = 0 with n+1 dimensions. The operation process is as follows: first, predict the value of x j , λ j ) by the current value (x j+1 , λ j+1 ). Finally, the value of x T×N , λ node ) at the next time is obtained by correction.

[0062] Selecting feature attributes: when the load level of the power system is stable within a normal level, the voltage bearing capacity of the power system is large, while the change of the node voltage and phase angle of the power system is relatively small. On the contrary, when the power system operates close to the limit point of stability, even a small change in load will cause a huge change in node voltage and phase angle. In short, the node voltage amplitude and phase angle are closely related to static voltage stability. Therefore, the amplitude and phase angle of the node voltage are selected as the feature attributes, that is, the input data of the misclassification constraint static voltage stability evaluation is studied. The input attribute matrix can be obtained as follows: X T×N = [V node , θ node ], X T×N is the input data of T time sections (prediction points) N feature attributes, V node is the voltage matrix of all nodes, N node is the phase angle matrix of all nodes.

[0063] The step 101 of generating the characteristic data of different prediction points by the continuous power flow method specifically comprises: simulating the application of continuously increasing load in the power system, and predicting the characteristic data of the power system at each prediction point under the action of continuously increasing load by the continuous power flow method. The simulation of the application of continuously increasing load in the power system means that the actual continuously increasing load is not added to the actual power system, and based on the simulation model or mathematical model of the power system, the continuously increasing load is applied to the simulation model or mathematical model of the power system in the form of function or data as a variable, and then the simulation result of the simulation model or the calculation result of the mathematical model is obtained, and the characteristic data is further obtained from the simulation result or the calculation result.

[0064] The step 102 of determining the label of each characteristic data based on the static voltage stability threshold value; when the voltage amplitude in the characteristic data is greater than the static voltage stability threshold value, the label of the characteristic data is used to represent that the static voltage of the power system is unstable, and when the voltage amplitude in the characteristic data is not greater than the static voltage stability threshold value, the label of the characteristic data is used to represent that the static voltage of the power system is stable.

[0065] According to the definition of the static voltage stability index of the power system, the sample obtained by the continuous power flow method can be classified by labels. The voltage V cr of the selected limit point has a margin of 0.15, i.e. the point (point on the time section) lower than 1.15V cr is static voltage stable, and the point exceeding it is static voltage unstable.

[0066] The quantitative index of voltage stability: the static voltage instability problem is often caused by the continuously increasing load parameter λ(t). When the load power changes, the voltage of the node will also change, and the P-V curve is generated by the continuous power flow method under the continuously increasing load level. The value of the load parameter λ takes the maximum value at the limit of the voltage stability curve. When the load change of the system is between λ0-λ max , the system is in a voltage stable state, and when it is not in this interval, the system is in a voltage unstable state. λ0is the initial load level, and λ max is the critical point load level.

[0067] Regarding the power load margin, the static voltage stability evaluation standard is described as the regional active power load margin K p % and the bus load reactive power margin K q %. In the normal operation or maintenance mode, K p % and K q % are required to be greater than 8%; and in the N-1 fault, K p % and K q % are required to be greater than 5%.

[0068]

[0069] wherein P max is the active power of the limit operating point, P is the active power of the stable operation; Q max is the reactive power of the limit operating point, Q is the reactive power of the stable operation.

[0070] Regarding the voltage stability margin, the voltage stability margin is greater than 7% when the power system is in normal operation, the voltage stability margin is greater than 5% when N-1 fault occurs, and the voltage stability margin is greater than 2.5% when N-2 fault occurs. The parameters of the voltage stability margin can be written as:

[0071]

[0072] wherein V is the load node voltage under the operation mode; V cr is the limit voltage of the point, and the minimum margin of the normal operation voltage is not less than 15%.

[0073] Considering various factors, the limit voltage is selected as an index for measuring whether the voltage is stable or not in the application, when the voltage is greater than or equal to the limit voltage, the system is determined to be stable, and when the voltage is less than the limit voltage, the system is determined to be unstable.

[0074] Step 103, constructing a sample data set according to the characteristic data of different prediction points and the label of each characteristic data.

[0075] Step 104, performing random oversampling processing on the sample data set, balancing the positive sample data and the negative sample data in the sample data set, and obtaining a sample data set after random oversampling; the positive sample data represents a sample with a label of power system static voltage stability, and the negative sample data represents a sample with a label of power system static voltage instability.

[0076] The random oversampling is used to constrain the sample imbalance, and since the number of safe samples and unsafe samples is not equal, the classification accuracy of the classifier may be affected, so the random oversampling is used to constrain the misclassification, so that the safe and unsafe samples are controlled to be consistent, thereby reducing the misclassification caused by the class imbalance.

[0077] Random oversampling: there are two methods to improve the learning ability of the classifier in the unbalanced data set. One is the optimization of unbalanced data classification algorithm; the other is the optimization of unbalanced data sampling algorithm. The optimization of unbalanced data classification is due to the current classification algorithm assumes that the class of data set is balanced, so the algorithm will cause the inaccuracy of classification result when optimizing the class imbalance, and the recognition rate of minority class is very low. The optimization of unbalanced data sampling algorithm is because the sampling algorithm can effectively make the positive and negative samples balanced by changing the data, and it is relatively easy to optimize the design, which can effectively improve the accuracy of classification. This scheme mainly optimizes the unbalanced data sampling algorithm.

[0078] Since random oversampling is the most convenient way to deal with class imbalance data set, this scheme will use random oversampling to copy the data of minority samples to balance the samples of minority class and majority class. The accuracy of the classification algorithm is effectively improved, and the safe and unsafe states of the power system static voltage stability are reasonably classified, and the misclassification problem is effectively constrained.

[0079] Step 105, using the sample data set after random oversampling to train the support vector machine model, and obtaining the trained support vector machine model.

[0080] According to the proportion, the sample data set is divided into training set and test set, the optimal parameters of support vector machine are selected by using grid search and cross validation, and the selected optimal parameters are used for machine learning to obtain a support vector machine classifier which can constrain the misclassification problem.

[0081] Support vector machine: it is a machine learning method based on statistical learning theory, and also an implementation method of structural risk minimization principle. The basic principle of support vector machine is to map the linearly inseparable points in low-dimensional space to high-dimensional space, so that it becomes linearly separable, and then the method of constructing optimal hyperplane is used to complete the classification by using kernel function to simply use the hyperplane of high-dimensional space.

[0082] Parameter optimization based on grid search algorithm: the penalty factor and the parameter in kernel function are important parameters affecting the level of support vector machine. Researchers have proposed to use optimization algorithm to automatically adjust the value of penalty factor and kernel function parameter in support vector machine through long-term research. The commonly used optimization algorithms are genetic algorithm, particle swarm algorithm and grid search algorithm.

[0083] The penalty parameter C and Gaussian kernel function g in support vector machine are optimized by using grid search method. For example, Figure 3The specific way is shown as follows: first, the range of C and g is set, a search step is set, a two-dimensional network is established on the coordinates of C and g, and the nodes between the networks are the parameter pairs of C and g; then, the classification accuracy R of the K-fold cross-validation of the support vector machine is calculated for each group of C and g; then, the R corresponding to each group of C and g values of C and g is drawn using contour lines, and thus the contour line graph is obtained, so as to determine the optimal C and g.

[0084] The completed training classifier is verified using the test set to check the classification accuracy. If the classification accuracy does not meet the target requirement, a relatively small search area can be selected on the basis of the original contour, and a more refined search is performed by reducing the step size, and finally the C and g values are determined.

[0085] Step 105 described using the sample data set after random oversampling to train the support vector machine model, obtain the trained support vector machine model, specifically including: dividing the sample data set after random oversampling into training set and test set; initializing the search area and search step; in the search area, according to the training set and the search step, using grid search method and multiple cross validation method, optimizing the hyperparameters of support vector machine model, obtaining the trained support vector machine model; the hyperparameters include penalty parameter and Gaussian kernel function; using the test set to verify the trained support vector machine model; when the verification passes, output the trained support vector machine model as the trained support vector machine model; when the verification does not pass, the preset range with the hyperparameters of the trained support vector machine model as the center is taken as the search area, and the search step is reduced, and the step of "in the search area, according to the training set, using grid search method and multiple cross validation method, optimizing the hyperparameters of support vector machine model, obtaining the trained support vector machine model" is returned.

[0086] In the search area, according to the training set and the search step, the grid search method and the multiple cross validation method are used to optimize the hyperparameters of the support vector machine model, and the trained support vector machine model is obtained, specifically including: in the search area, according to the search step, a two-dimensional hyperparameter network is generated; each network node in the two-dimensional hyperparameter network represents a hyperparameter; according to the training set, the support vector machine model under each hyperparameter in the two-dimensional hyperparameter network is subjected to multiple cross validation, and the classification accuracy of the support vector machine model under each hyperparameter is obtained; according to the classification accuracy of the support vector machine model under each hyperparameter, a classification accuracy contour line is constructed; according to the classification accuracy contour line, the support vector machine model under the hyperparameters corresponding to the highest classification accuracy is determined as the trained support vector machine model.

[0087] In step 106, the trained support vector machine model is used to determine the static voltage stability of the power system.

[0088] Embodiment 2

[0089] Embodiment 2 of the present application provides a system for determining the static voltage stability of a power system, which comprises:

[0090] A feature data generation module is configured to generate feature data of different prediction points by using the continuation power flow method; the feature data comprises voltage amplitude and phase angle of nodes in the power system.

[0091] The feature data generation module specifically comprises a feature data generation submodule configured to simulate the application of continuously increasing load to the power system, and predict the feature data of the power system at each prediction point under the action of the continuously increasing load by using the continuation power flow method.

[0092] A label classification module is configured to determine the label of each feature data based on a static voltage stability threshold; when the voltage amplitude in the feature data is greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power system is unstable; when the voltage amplitude in the feature data is not greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power system is stable. The static voltage stability threshold is 1.15V cr ; wherein V cr represents the limit voltage.

[0093] A sample data set construction module is configured to construct a sample data set according to the feature data of different prediction points and the label of each feature data.

[0094] A random oversampling module is configured to perform random oversampling processing on the sample data set, so as to balance the positive sample data and the negative sample data in the sample data set, and obtain a sample data set after random oversampling; the positive sample data represents a sample with a label of static voltage stability of the power system, and the negative sample data represents a sample with a label of static voltage instability of the power system.

[0095] A training module is configured to train a support vector machine model by using the sample data set after random oversampling, and obtain a trained support vector machine model.

[0096] The training module specifically comprises: a sample data set division submodule for dividing the sample data set after random oversampling into a training set and a test set; an initialization submodule for initializing a search region and a search step; a training submodule for optimizing hyperparameters of a support vector machine model in the search region according to the training set and the search step by using a grid search method and a multiple cross-validation method, to obtain a trained support vector machine model; the hyperparameters include a penalty parameter and a Gaussian kernel function; a verification submodule for verifying the trained support vector machine model by using the test set; a model output submodule for outputting the trained support vector machine model as a trained support vector machine model when the verification is passed; and a return submodule for taking a preset range centered on the hyperparameters of the trained support vector machine model as the search region and reducing the search step when the verification is not passed, and returning to the step of optimizing the hyperparameters of the support vector machine model in the search region according to the training set by using the grid search method and the multiple cross-validation method, to obtain the trained support vector machine model.

[0097] The training submodule specifically comprises: a two-dimensional hyperparameter network generation unit for generating a two-dimensional hyperparameter network in the search region according to the search step; each network node in the two-dimensional hyperparameter network represents a hyperparameter; a multiple cross-validation unit for performing multiple cross-validation on a support vector machine model under each hyperparameter in the two-dimensional hyperparameter network according to the training set, to obtain a classification accuracy of the support vector machine model under each hyperparameter; a classification accuracy contour construction unit for constructing a classification accuracy contour according to the classification accuracy of the support vector machine model under each hyperparameter; and a training unit for determining, according to the classification accuracy contour, a support vector machine model under a hyperparameter corresponding to the highest classification accuracy as the trained support vector machine model.

[0098] The power system static voltage stability determination module is configured to determine power system static voltage stability by using the trained support vector machine model.

[0099] According to the specific embodiments of the present application, the following technical effects are achieved:

[0100] The present application discloses a power system static voltage stability determination method and system, which reasonably constrains misclassification in power system static voltage stability evaluation. The data-driven method is used to select the feature attributes with high correlation degree with voltage stability, and then the static load active power margin is used to set the threshold to construct the classification label. Then, the random oversampling method is used to constrain the imbalance of the categories. Finally, the support vector machine classifier model based on random oversampling is used to classify the data and analyze the results. The stable evaluation of the power system static voltage is realized.

[0101] The various embodiments described in this specification are presented for the purpose of illustrating the principles of the present application and its best mode of operation. Each of the embodiments described in this specification has been provided for the purpose of illustration and not for the purpose of limitation. The same parts and principles are applied to the same way in different embodiments, and the same parts and principles are applied to the same way in different embodiments. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant part is described in the method part.

[0102] The principles and implementation modes of the present application are described by applying specific examples in this specification. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In summary, the content of this specification should not be understood as a limitation of the present application.

Claims

1. A method of determining power grid system static voltage stability, characterized by, The method comprises the following steps: Generating feature data of different prediction points by using a continuous power flow method; the feature data comprises voltage amplitude and phase angle of nodes in the power grid system; Determining a label of each feature data based on a static voltage stability threshold value; When the voltage amplitude in the feature data is greater than the static voltage stability threshold value, the label of the feature data is used to represent that the static voltage of the power grid system is unstable, and when the voltage amplitude in the feature data is not greater than the static voltage stability threshold value, the label of the feature data is used to represent that the static voltage of the power grid system is stable; Constructing a sample data set according to the feature data of different prediction points and the label of each feature data; Balancing positive sample data and negative sample data in the sample data set by performing random oversampling processing on the sample data set, to obtain a sample data set after random oversampling; the positive sample data represents a sample with a label of static voltage stability of the power grid system, and the negative sample data represents a sample with a label of static voltage instability of the power grid system; Training a support vector machine model by using the sample data set after random oversampling, to obtain a trained support vector machine model; Determining the static voltage stability of the power grid system by using the trained support vector machine model; The training of the support vector machine model by using the sample data set after random oversampling to obtain the trained support vector machine model specifically comprises the following steps: Dividing the sample data set after random oversampling into a training set and a test set; Initializing a search region and a search step; In the search region, optimizing hyperparameters of the support vector machine model by using a grid search method and a multiple cross-validation method according to the training set and the search step, to obtain a trained support vector machine model; the hyperparameters comprise a penalty parameter and a Gaussian kernel function; Verifying the trained support vector machine model by using the test set; When the verification is passed, outputting the trained support vector machine model as the trained support vector machine model; When the verification is not passed, taking a preset range centered on the hyperparameters of the trained support vector machine model as the search region, and reducing the search step, and returning to the step of optimizing the hyperparameters of the support vector machine model by using the grid search method and the multiple cross-validation method according to the training set in the search region, to obtain the trained support vector machine model.

2. The power grid system static voltage stability determination method of claim 1, wherein, The generation of the feature data of different prediction points by using the continuous power flow method specifically comprises the following steps: Imitating the application of continuously increasing loads to the power grid system, and sequentially predicting, by using the continuous power flow method, the feature data of the power grid system at each prediction point under the action of the continuously increasing loads.

3. The method of determining grid system static voltage stability according to claim 1, characterized in that, The static voltage stability threshold is ; wherein, represents the limiting voltage.

4. The method of determining grid system static voltage stability according to claim 1, wherein, In the search region, optimizing the hyperparameters of the support vector machine model by using the grid search method and the multiple cross-validation method according to the training set and the search step, to obtain the trained support vector machine model, specifically comprises the following steps: Generating a two-dimensional hyperparameter network according to the search step in the search region; each network node in the two-dimensional hyperparameter network represents a hyperparameter; According to the training set, multiple cross-validations are performed on the support vector machine model under each hyperparameter in the two-dimensional hyperparameter network to obtain a classification accuracy of the support vector machine model under each hyperparameter; According to the classification accuracy of the support vector machine model under each hyperparameter, a classification accuracy contour line is constructed; According to the classification accuracy contour line, the support vector machine model under the hyperparameter corresponding to the highest classification accuracy is determined as the trained support vector machine model.

5. A power grid system static voltage stability determination system, characterized by, The system comprises: A feature data generation module is configured to generate feature data of different prediction points by using a continuous power flow method; the feature data comprises voltage amplitude and phase angle of nodes in the power grid system; A label classification module is configured to determine a label of each feature data based on a static voltage stability threshold; when the voltage amplitude in the feature data is greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power grid system is unstable; when the voltage amplitude in the feature data is not greater than the static voltage stability threshold, the label of the feature data is used to represent that the static voltage of the power grid system is stable; A sample data set construction module is configured to construct a sample data set according to the feature data of different prediction points and the label of each feature data; A random oversampling module is configured to perform random oversampling processing on the sample data set to balance positive sample data and negative sample data in the sample data set, and obtain a sample data set after random oversampling; the positive sample data represents a sample with a label of static voltage stability of the power grid system, and the negative sample data represents a sample with a label of static voltage instability of the power grid system; A training module is configured to train a support vector machine model by using the sample data set after random oversampling, and obtain a trained support vector machine model; A power grid system static voltage stability determination module is configured to determine the static voltage stability of the power grid system by using the trained support vector machine model. The training module specifically comprises: A sample data set division sub-module is configured to divide the sample data set after random oversampling into a training set and a test set; An initialization sub-module is configured to initialize a search region and a search step; A training sub-module is configured to optimize hyperparameters of the support vector machine model in the search region according to the training set and the search step by using a grid search method and a multiple cross-validation method, and obtain a trained support vector machine model; the hyperparameters comprise a penalty parameter and a Gaussian kernel function; A verification sub-module is configured to verify the trained support vector machine model by using the test set; A model output sub-module is configured to output the trained support vector machine model as the trained support vector machine model when the verification is passed; A return sub-module is configured to take a preset range with the hyperparameters of the trained support vector machine model as the center as the search region and reduce the search step when the verification fails, and return to the step of optimizing the hyperparameters of the support vector machine model in the search region according to the training set by using the grid search method and the multiple cross-validation method, and obtaining the trained support vector machine model.

6. The power grid system static voltage stability determination system of claim 5, wherein, The feature data generation module specifically comprises: The feature data generation submodule is configured to apply a continuously increasing load to the power grid system, and predict feature data of the power grid system at each prediction point under the continuously increasing load by using a continuous flow method.

7. The power grid system static voltage stability determination system of claim 5, wherein, The static voltage stability threshold is ; wherein, represents the limiting voltage.

8. The power grid system static voltage stability determination system of claim 5, wherein, The training submodule specifically comprises: A two-dimensional hyperparameter network generation unit is configured to generate a two-dimensional hyperparameter network in the search region according to the search step length, and each network node in the two-dimensional hyperparameter network represents a hyperparameter. A multiple cross-validation unit is configured to perform multiple cross-validation on a support vector machine model under each hyperparameter in the two-dimensional hyperparameter network according to the training set, and obtain a classification accuracy of the support vector machine model under each hyperparameter. A classification accuracy contour construction unit is configured to construct a classification accuracy contour according to the classification accuracy of the support vector machine model under each hyperparameter. A training unit is configured to determine, according to the classification accuracy contour, a support vector machine model under a hyperparameter corresponding to the highest classification accuracy as a trained support vector machine model.

Images