Method and device for identifying modal continuity of multi-modal oscillation data

Abstract

The invention discloses a modal continuity identification method and device for multi-modal oscillation data. The method comprises the following steps: carrying out feature extraction on multi-frame broadband oscillation data to obtain respective modal features of a plurality of modals of each frame of broadband oscillation data; clustering is carried out according to the modal characteristics, a plurality of clusters of each frame of broadband oscillation data are obtained, and cluster center change characteristics of each frame of broadband oscillation data are determined; obtaining a modal matching threshold value of the clustering cluster of each frame according to the cluster center change characteristics; and carrying out modal continuity matching by adopting the cluster and the corresponding modal matching threshold to obtain a modal continuity identification result. According to the embodiment of the invention, the intra-frame modal clustering on the time scale is carried out on each modal feature, and the self-adaptive threshold prediction is carried out in combination with the threshold prediction model of the preset hyper-parameter, so that the real-time and accurate modal continuity identification of frequency-intensive, strong-time-varying and multi-modal broadband oscillation data is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method and device for identifying modal continuity of multi-modal oscillation data. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the large-scale integration of renewable energy sources such as wind and photovoltaic power, the grid operating environment has become increasingly complex. Due to the intermittent and uncertain nature of renewable energy generation, oscillations in the power system have become more common, exhibiting characteristics such as high frequency density, strong time variation, and the coexistence of multiple modes. For example, the integration of photovoltaic inverters and wind turbines can introduce multiple oscillation modes, such as subsynchronous and supersynchronous oscillations, making the dynamic characteristics of grid signals even more complex.

[0004] In actual measurement, broadband measurement devices can acquire oscillation signals from power systems at different time frames. However, due to noise interference, measurement errors, and signal aliasing, the modal characteristics of different time frames, such as frequency, amplitude, and phase, can vary slightly. This makes it difficult to directly determine whether the modes between different time frames are the same. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a modal continuity identification method, system method and device for modal oscillation data, aiming to realize real-time and accurate modal continuity identification of oscillation measurement data with frequency-intensive, strong time-varying and multi-modal characteristics.

[0006] To achieve the above object, according to a first aspect of an embodiment of the present invention, a method for identifying modal continuity of multimodal oscillation data is provided, comprising: Performing feature extraction on multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of the broadband oscillation data; Clustering the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data of each frame, and determining cluster center change features of the broadband oscillation data of each frame; Performing adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame; Modal continuity matching is performed using the clustering clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

[0007] Furthermore, the modal characteristics of the mode include at least a frequency, a 0th-order amplitude, a 1st-order amplitude, a 0th-order phase, a 1st-order phase, a 0th-order damping ratio coefficient, and a 1st-order damping ratio coefficient corresponding to the mode; Clustering the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data of each frame, and determining cluster center change features of the broadband oscillation data of each frame, including: Acquire modal features of the broadband oscillation data of the current frame and the previous frame, and construct a corresponding feature vector of the modal according to the modal features; Determining the clusters of the broadband oscillation data of the current frame and the previous frame and the modes contained therein by using the characteristic vectors, and determining cluster centers according to the characteristic vectors of the modes contained in the clusters; The cluster center change feature of the current frame is determined according to the cluster centers of the current frame and the previous frame.

[0008] Furthermore, adaptive threshold prediction is performed based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame, including: Construct a prediction input vector based on the cluster center change characteristics of each frame in the preset time window, The prediction input vector is input into a pre-trained threshold prediction model, and the threshold prediction model outputs a modality matching threshold of the cluster.

[0009] Further, The wide frequency oscillation data of the frame The eigenvector of the mode for: , in, 、 、 、 Respectively Frame width oscillation data The frequency of the mode, Order amplitude, Phase and The damping ratio coefficient of order, , , For the The number of modes of the frame-wide frequency oscillation data; Determining the clusters of the broadband oscillation data of the current frame and the previous frame and the modes contained therein by using the characteristic vectors, and determining the cluster centers according to the characteristic vectors of the modes contained in the clusters, including: The improved K-means clustering algorithm is used to The characteristic vector of each mode of the frame wide frequency oscillation data Perform clustering and obtain The clusters and their respective cluster centers ,in, , For the The set of all the modes contained in the clusters, .

[0010] Furthermore, the cluster center change characteristics include the The cluster center weighted displacement of the frame broadband oscillation data The weighted displacement of cluster centers of clusters , ; in, , 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 and The frequency and Weight coefficients for order amplitude, phase and damping ratio.

[0011] Furthermore, adaptive threshold prediction is performed based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame, including: Construct a prediction input vector based on the cluster center change characteristics of each frame in the preset time window , the predicted input vector ,in, Indicates the preset time window size, ; The prediction input vector A pre-trained LSTM model is input to perform adaptive threshold prediction, and the LSTM model outputs the modality matching threshold of the clustering cluster.

[0012] Furthermore, the predicted The wide frequency oscillation data of the frame modal The modality matching threshold of the cth cluster is ; Adjacent frame modal continuity matching is performed using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result, including: Calculate the The wide frequency oscillation data of the frame modal Hedi The wide frequency oscillation data of the frame modal The weighted distance ; Compare the weighted distances and the modality matching threshold is size; In response to the weighted distance is less than the modality matching threshold , then the mode With the modal continuous.

[0013] According to a second aspect of an embodiment of the present invention, there is provided a device for identifying modal continuity of multimodal oscillation data, comprising: a feature extraction module, configured to extract features from multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of broadband oscillation data; a clustering processing module, configured to cluster the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data for each frame, and determine cluster center variation features of the broadband oscillation data for each frame; A threshold prediction module is used to perform adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame; The continuity judgment module is used to perform modal continuity matching using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

[0014] According to a third aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which a program is stored. When the program is executed by a processor, the steps in the modal continuity identification method of multimodal oscillation data as described in the first aspect of the present invention are implemented.

[0015] According to a fourth aspect of an embodiment of the present invention, a terminal device is provided, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for identifying modal continuity of multimodal oscillation data as described in the first aspect of the present invention are implemented.

[0016] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps in the method for identifying modal continuity of multimodal oscillation data as described in the first aspect of the present invention.

[0017] The embodiments of the present invention have the following advantages or beneficial effects: The embodiments of the present invention address the limitations of current modal identification and matching methods. By performing intra-frame modal clustering on a time scale for each modality and combining it with a threshold prediction model with preset hyperparameters for adaptive threshold prediction, compared to the method of using a fixed frequency deviation threshold to determine whether two modes are continuous, this method achieves real-time and accurate modal continuity identification for broadband oscillation data with frequency-intensive, strongly time-varying, and multi-modal characteristics.

[0018] The embodiment of the present invention clusters each frame of broadband oscillation data and obtains the typical cluster change characteristics of the cluster cluster for threshold prediction, thereby reducing the computational complexity and amount of the threshold prediction model, and can meet the real-time and accuracy requirements of modal continuity identification of broadband oscillation data with dense frequency, strong time variation, and multimodal characteristics.

[0019] The embodiment of the present invention uses cluster center change characteristics to construct a prediction input vector for adaptive threshold prediction, which solves the continuity misjudgment problem caused by using a single fixed threshold that is too strict or loose due to the dynamic change characteristics of power grid oscillation.

[0020] The embodiment of the present invention uses an improved k-means clustering algorithm that can adapt to the frequency-intensive, strongly time-varying, and multimodal characteristics of broadband oscillation signals in power grids. It has good recognition accuracy in high-dimensional space and can process newly emerging oscillation modes with less training data. It has low computational complexity and is suitable for low-latency real-time processing scenarios in actual power grid monitoring.

[0021] The embodiments of the present invention address the problems faced by traditional modal recognition methods when faced with frequency-intensive, highly time-varying, and multi-modal oscillation data, such as insufficient time correlation modeling capabilities, difficulty in fuzzy recognition of modal transitions, and weak robustness to disturbances. Based on the strong time series modeling and nonlinear mapping capabilities of the pre-trained LSTM model, the modal evolution characteristics are automatically learned through the LSTM model, effectively improving the accuracy and stability of modal continuity recognition. Therefore, the method and device of the embodiments of the present invention can be applied to modal recognition in complex dynamic power systems.

[0022] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. Throughout the drawings, the same reference symbols denote the same components. Figure 1 Schematic diagram of the main process of the method for identifying modal continuity of multimodal oscillation data according to an embodiment of the present invention; Figure 2 Schematic diagram of main modules of a device for identifying modal continuity of multimodal oscillation data according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the composition of a terminal device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0025] Example 1

[0026] Figure 1 FIG. 1 is a schematic diagram of the main process of a method for identifying modal continuity of multimodal oscillation data according to an embodiment of the present invention. Figure 1 As shown, the modal continuity identification method for multi-modal oscillation data in this embodiment of the present invention includes the following steps S101 to S104.

[0027] Step S101 : performing feature extraction on multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of the broadband oscillation data.

[0028] Step S102 : clustering the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data of each frame, and determining cluster center variation features of the broadband oscillation data of each frame.

[0029] Step S103 : performing adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the clustering clusters of each frame.

[0030] Step S104 : performing modal continuity matching using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

[0031] It can be understood that the broadband oscillation data in the embodiment of the present invention is data obtained by collecting the broadband oscillation signal of the power system according to a preset time scale, and the broadband oscillation data has frequency-intensive, strong time-varying, and multi-modal characteristics. The method of the embodiment of the present invention performs continuity identification on multi-modal oscillation data, which refers to whether the modes in two adjacent frames of broadband oscillation data obtained by measurement belong to the same mode. Each frame of broadband oscillation data is obtained by intercepting and sampling the broadband oscillation signal according to a preset time length. The broadband oscillation signal includes typical electrical quantities such as voltage phasors and current phasors. The mode is a different dynamic response mode of a certain voltage or current, and the modal characteristics of the mode include frequency, amplitude and its rate of change, phase and its rate of change, and damping ratio coefficient and its rate of change. In some embodiments of the present invention, the extracted modal features include at least the frequency, amplitude (0th order amplitude) and the first-order rate of change of amplitude (1st order amplitude), phase (0th order phase) and the first-order rate of change of phase (1st order phase), damping ratio coefficient (0th order damping ratio coefficient) and the first-order rate of change of damping ratio coefficient (1st order damping ratio coefficient) corresponding to the mode.

[0032] Specifically, in step S101 of this embodiment, The modal characteristics of the frame broadband oscillation data include Frame No. The frequency of the mode , No. Frame No. The amplitude of the mode (0th order amplitude), amplitude primary rate of change (1st order amplitude), quadratic rate of change of amplitude (2nd order amplitude), amplitude cubic rate of change (3rd order amplitude), Frame No. The phase of the mode (0th order phase), phase primary rate of change (1st-order phase), quadratic rate of phase change (2nd order phase), phase cubic rate of change (3rd order phase), Frame No. The damping ratio coefficient of each mode (0th order damping ratio coefficient), damping ratio coefficient first rate of change (1st order damping ratio coefficient), quadratic change rate of damping ratio coefficient (2nd order damping ratio coefficient), the third rate of change of the damping ratio coefficient (0th order damping ratio coefficient), , For the The number of modalities detected for the frame.

[0033] It will be appreciated that clustering in step S102 is an unsupervised learning method used to divide a dataset into multiple subsets (clusters) such that data objects within the same cluster have high similarity, while data objects between different clusters have low similarity. Specifically, in this embodiment and some embodiments of the present invention, step S102 includes step S102a and step S102c.

[0034] Step S102a: obtaining modal features of the broadband oscillation data of the current frame and the previous frame, and constructing a corresponding feature vector of the modal according to the modal features.

[0035] Specifically, in the embodiment of the present invention, The wide frequency oscillation data of the frame The eigenvector of the mode The frequency amplitude and phase of the mode are the basic characteristics of the mode. In order to distinguish the modes with dense frequency, strong time variation and quantity in broadband oscillation data, the embodiment of the present invention also selects the damping ratio coefficient and the first-order and higher-order change rates of the amplitude / phase / damping ratio coefficient as characteristic components. The first-order change rate of the amplitude / phase / damping ratio coefficient is used to represent the instantaneous change trend, such as the sudden drop in amplitude caused by the wind turbine being disconnected from the grid; the higher-order change rate of the amplitude / phase / damping ratio coefficient is used to identify nonlinear dynamic characteristics, such as the nonlinear amplitude fluctuation caused by the limiting of the photovoltaic inverter.

[0036] In this embodiment, The wide frequency oscillation data of the frame The eigenvector of the mode , according to Frame The mode can be obtained Multimodal feature matrix of the frame .

[0037] ; Among them, the eigenvector is the multimodal feature matrix No. data points, .

[0038] Step S102b: using the feature vector to determine the clusters of the broadband oscillation data of the current frame and the previous frame and the modes contained therein, and determining the cluster centers according to the feature vectors of the modes contained in the clusters.

[0039] When selecting a clustering method, the embodiment of the present invention considers: DBSCAN (density-based spatial clustering method) can be used for modal classification, but the Euclidean distance it relies on may lead to a decrease in recognition accuracy in high-dimensional feature space, and its parameters (such as neighborhood radius) are difficult to adapt to time-varying characteristics. Some studies use support vector machines SVM, deep learning (such as LSTM, CNN) and other methods for modal classification, but these methods usually require a large amount of training data and have difficulty in processing newly emerging oscillation modes. In addition, the computational complexity of machine learning methods is high, and it is difficult to achieve low-latency real-time processing in actual power grid monitoring. Therefore, specifically, in this embodiment and some embodiments of the present invention, step S102b includes: using an improved k-means clustering algorithm to cluster the first The characteristic vector of each mode of the frame wide frequency oscillation data Perform clustering and obtain The clusters and their respective cluster centers ,in, , For the The set of all the modes contained in the clusters, .

[0040] It is understandable that the improved k-means clustering algorithm is used to find the nearest cluster center for each modal data point, and then classify the data point into the corresponding cluster. Further, in this embodiment, step S102b includes steps S102b1 to S102b4: Step S102b1: Determine a set of cluster centers, specifically including: Step S102b11: Randomly select the first cluster center as the selected cluster center.

[0041] In this embodiment and some embodiments of the present invention, from the multimodal feature matrix Randomly select a data point As the first cluster center, in order to distinguish the selected cluster center from the non-selected cluster center, record ,in, For the Frame the first cluster center, For the Frame No. The eigenvectors of the modes.

[0042] Step S102b12: Calculate the weighted distance from the data point to the selected cluster center.

[0043] In this embodiment and some embodiments of the present invention, the minimum distance from all data points to the selected cluster center is calculated: Each data point in the non-selected cluster center , calculate the data points Minimum weighted distance to the nearest chosen cluster center: , , ; in, Represents the selected Cluster centers , recorded as , , Representative data points The minimum weighted distance to the selected nearest cluster center, , 、 、 and The frequency and The weight coefficients of the order amplitude, phase and damping ratio coefficient represent the influence of different features on the modal distance within the frame. The weights are related to the application scenario of modal continuity recognition. Usually, the weights in the pre-training stage and the weights in the prediction stage can use the same value. In some embodiments of the present invention, 、 、 and .

[0044] Step S102b13: Calculate the probability of each data point becoming the next cluster center based on the minimum weighted distance calculation result to obtain a probability distribution.

[0045] In this embodiment and some embodiments of the present invention, the minimum distance from each data point to the selected cluster center is , calculate and select this point The probability of being the next cluster center: , , ; in, For data points The probability of being selected as the next cluster center, For the Frame No. The eigenvectors of each mode make it more likely that data points farther away from the existing center point will be selected as the new center point, thus ensuring uniform distribution among different center points.

[0046] Step S102b14: Select a new cluster center as the new selected cluster center according to the probability distribution.

[0047] In this embodiment and some embodiments of the present invention, the probability distribution calculated in the previous step (step S102b13) is used to select a new cluster center. .

[0048] ; in, is the newly selected cluster center, To use the cumulative distribution function according to the probability Sample the selected data points.

[0049] Step S102b15: Return to step S102b12 to calculate the new selected cluster center until a set of selected cluster centers is obtained. Specifically, repeat S102b12 to S102b14 until a set of selected cluster centers is obtained. Cluster centers , It is the preset number of cluster centers. Generally, the number of cluster centers is smaller than the number of modes. It is used to reduce the dimension of adaptive threshold prediction and improve the speed of continuity judgment.

[0050] Step S102b2: Assign data points to the nearest cluster center.

[0051] In this embodiment and some embodiments of the present invention, the data points are allocated The formula to the nearest cluster center is: ; in, for The label of the cluster center, Indicates finding the minimum weighted distance ,Right now The closest cluster center , Indicates the data points and the selected cluster centers distance.

[0052] Step S102b3: After each data point is assigned, the new center point of each cluster is updated. The calculation method is: ; in, is the updated feature mean, For the The set of all data points contained in the cluster to which the cluster center belongs, Represents the number of modes in the set.

[0053] .

[0054] Step S102b4: Determine whether the cluster assignment labels of all data points will change. If the cluster assignment labels of all data points no longer change, stop the iteration and output the final cluster center. As the cluster center, otherwise continue to execute steps S102b2 and S102b3 until the following conditions are met: ; in, Represents the latest cluster label number of all modalities in the frame, Represents the last cluster label number of all modalities in the frame.

[0055] Step S102c: determining the cluster center change feature of the current frame based on the cluster centers of the current frame and the previous frame.

[0056] Furthermore, the cluster center change characteristics include the Frame width oscillation data The weighted displacement of cluster centers , , ; in, , 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 and The frequency and Weight coefficients for order amplitude, phase and damping ratio.

[0057] Specifically, in this embodiment and the embodiments of the present invention, Clustering of frames and Frames are clustered into equal numbers.

[0058] Specifically, in this embodiment and some embodiments of the present invention, step S103 includes step S103a and step S103b.

[0059] Step S103a, constructing a prediction input vector based on the cluster center change characteristics of each frame in the preset time window. Specifically, constructing a prediction input vector based on the cluster center change characteristics of each frame in the preset time window , the predicted input vector ,in, Indicates the preset time window size, .

[0060] Step S103b, input the predicted input vector into a pre-trained threshold prediction model, and the threshold prediction model outputs the modality matching threshold of the cluster. The pre-trained LSTM model is input to perform adaptive threshold prediction. The modality matching threshold of the cluster output by the LSTM model is the first Frame width oscillation data Adaptive prediction threshold for clusters .

[0061] In this embodiment, the LSTM model threshold optimization model is adopted because the LSTM model has strong time series modeling capabilities and nonlinear mapping capabilities, can automatically learn modal evolution characteristics, and effectively improve the accuracy and stability of modal continuity identification. It is particularly suitable for the modal intelligent identification scenario in the complex dynamic power system in the embodiment of the present invention.

[0062] Specifically, in this embodiment and some embodiments of the present invention, the pre-training of the LSTM model is to use the LSTM model as a threshold optimization model, and perform hyperparameter optimization and training on the LSTM model, including steps S103b10 to S103b14.

[0063] Step S103b10: Define the hyperparameter search space of the LSTM model and initialize the hyperparameter sample.

[0064] In this embodiment and some embodiments of the present invention, the prediction effect of the LSTM model depends on multiple hyperparameters, and the parameters of the LSTM model need to be split in the parameter space. Optimize, where is the learning rate, which is used to affect the speed of gradient descent. is the LSTM hidden layer size, used to control the model capacity, is the Dropout rate, used to prevent overfitting, is the batch size, which affects the training stability. is the number of training rounds, which is used to influence the convergence effect.

[0065] Step S103b12: Use the hyperparameter sample to train the LSTM model and calculate the loss.

[0066] In this embodiment and some embodiments of the present invention, random sampling is performed in the search space. Group hyperparameters , expressed as: ; in, For the The learning rate of the group hyperparameters, For the The hidden layer size of the LSTM model of the group hyperparameters, For the Dropout rate of group hyperparameters, For the The batch size of the group hyperparameters, For the The number of training rounds for the set hyperparameters.

[0067] When determining the hyperparameters of the LSTM model, for each hyperparameter , used for modal continuity identification broadband oscillation data samples, each input data is a feature vector constructed using a broadband vibration data sample, specifically broadband vibration data sample Clusters in the time window The feature vector composed of the weighted displacement sequence of the cluster center of each time frame , expressed as: ; in, For the broadband vibration data sample The cluster center is The weighted displacement of cluster centers of frames, For the broadband vibration data sample The clusters are in Frame and The weighted displacement change of the cluster center of the frame, Indicates the time window size, representing the number of historical frames used for prediction, .

[0068] Further, ; ; in, 、 、 、 Respectively broadband vibration data sample Frame No. The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 、 Respectively broadband vibration data sample Frame No. The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 and The frequency and Weight coefficients for order amplitude, phase and damping ratio.

[0069] In this embodiment and some embodiments of the present invention, the optimal adaptive threshold is predicted based on an LSTM model. Specifically, the LSTM model uses a gating unit for prediction, including an input gate, a forget gate, an output gate, and a memory unit with the same shape as the hidden state.

[0070] (1) Input Gate for: ; in, is the activation value of the input gate, which determines the degree of update of the current information. is the weight matrix of the input gate, For the The hidden state of the frame, is the prediction input vector input during training, is the bias term of the input gate, It is the Sigmoid activation function, which controls the amount of information flowing in.

[0071] (2) Forget Gate for: ; in, The activation value of the forget gate determines the information that needs to be forgotten. is the weight matrix of the forget gate, Represents the bias term of the forget gate.

[0072] (3) Memory unit update for: ; in, is the current state of the memory unit, storing long-term information, For the The memory cell state of the frame, represents element-wise multiplication, is the weight matrix that controls the update of the memory unit, A bias term that controls the update of memory cells, is the hyperbolic tangent function, which controls the information update.

[0073] (4) Output Gate for: ; in, is the activation value of the output gate, which determines the proportion of current information output. is the weight matrix of the output gate, is the bias term of the output gate.

[0074] (5) LSTM model calculation The predicted value of the adaptive threshold is calculated according to the following formula: ; in, To use Group hyperparameters The predicted The first sample The first cluster The modality matching threshold of the frame, is the LSTM model output layer weight matrix, is the bias term of the output layer of the LSTM model, is the hidden state at the current moment, .

[0075] Specifically, in this embodiment and some embodiments of the present invention, the first Group hyperparameters Train the LSTM model and calculate the modality matching threshold loss function according to the following formula , ; in, For the Group hyperparameters The modality matching threshold loss function is: is the number of broadband oscillation data samples for modal continuity identification, To use Group hyperparameters The predicted broadband oscillation data sample Cluster The modality matching threshold of the frame, To pass the The first sample of broadband oscillation data is obtained Cluster The true threshold for the frame.

[0076] As can be understood, the true threshold for clustering is based on physical modeling and simulation, generating synthetic data with known modal evolutionary characteristics under controlled conditions. In this process, the system simulates its dynamic response in continuous time frames by injecting oscillating sources with known frequency, amplitude, and phase characteristics. Because the modal parameters set during data generation are clearly unique and time-continuous, the corresponding relationships between the same mode in different time frames can be automatically annotated, thereby constructing true continuous modal pairs.

[0077] Specifically, first use It is known that the first The feature vectors of different frames of the same modality of clusters are used to calculate the weighted distance between the feature vectors of adjacent time frames of the same modality: ; in, For the The first of the broadband oscillation data samples Cluster Modal Frame and The true weighted distance of the frame; and Respectively The first of the broadband oscillation data samples Cluster The mode in Frame and Frequency parameters of the frame; and Respectively The first of the broadband oscillation data samples Cluster The mode in Frame and The b-th order amplitude parameter of the frame; and Respectively The first of the broadband oscillation data samples Cluster The mode in Frame and The b-th order phase parameter of the frame; and Respectively The first of the broadband oscillation data samples Cluster The mode in Frame and The b-th order damping ratio coefficient parameter of the frame; 、 、 and The frequency and Weight coefficients for order amplitude, phase and damping ratio.

[0078] The previous step counted the total number of The first of the broadband oscillation data samples In the clusters Then, the weighted distances of these true matching pairs are averaged as the first The first of the broadband oscillation data samples True matching threshold for clusters: .

[0079] Step S103b13: Determine the better and worse hyperparameter distributions and select new hyperparameters.

[0080] In this embodiment and some embodiments of the present invention, the hyperparameters There are two categories: The first category is the better hyperparameter distribution with better performance, that is, Modality matching threshold loss function for clusters Less than : .

[0081] The second category is poor hyperparameter distribution and poor performance, that is, Modality matching threshold loss function for clusters Greater than : .

[0082] New hyperparameters are chosen by minimizing the ratio of the worse distribution to the better distribution: ; in, is the hyperparameter search space, that is, the range of all possible hyperparameter values, For dynamically adjusted The top modal cluster centers in historical experiments %The maximum loss function value in the result, is the modality matching threshold loss function value after LSTM training, is the optimal hyperparameter.

[0083] In this embodiment, γ is a threshold for taking the top 10% of the results in the historical experiments. All hyperparameter sets in the search space that satisfy the loss function value less than this threshold after LSTM training are considered to be better hyperparameter distributions, otherwise they are considered to be worse hyperparameter distributions.

[0084] Step S103b14: Set the optimal hyperparameters Substitute into LSTM model training.

[0085] In this embodiment and some embodiments of the present invention, the new loss function : ; like Less than the set threshold , end the optimization and set the final value of the hyperparameter , and get the corresponding optimal hyperparameters The weight matrices and bias terms of the LSTM model are obtained, that is, the optimal hyperparameters LSTM model; otherwise, return to step S103b12 and continue to sample new hyperparameters for optimization until the iterative convergence condition is met. .

[0086] Specifically, in this embodiment and some embodiments of the present invention, in step S103b, the optimal hyperparameters obtained by using the pre-trained LSTM model are , the predicted input vector Input corresponds to the optimal hyperparameters LSTM model to obtain the predicted modality matching threshold Furthermore, in this embodiment and some embodiments of the present invention, the optimal hyperparameters are used. The LSTM model predicts the adaptive modality matching threshold using a gating unit, including: (1) The 0-input gate is: ; in, is the activation value of the input gate, which determines the degree of update of the current information. is the weight matrix of the input gate, For the The hidden state of the frame, For the Frame No. The predicted input vector of clusters, is the bias term of the input gate, It is the Sigmoid activation function, which controls the amount of information flowing in.

[0087] (2) The forget gate is: ; in, The activation value of the forget gate determines the information that needs to be forgotten. is the weight matrix of the forget gate, Represents the bias term of the forget gate.

[0088] (3) The memory unit update formula is: ; in, is the current state of the memory unit, storing long-term information, For the The memory cell state of the frame, is the weight matrix that controls the update of the memory unit, A bias term that controls the update of memory cells, is the hyperbolic tangent function, which controls the information update.

[0089] (4) The output gate is: ; in, is the activation value of the output gate, which determines the proportion of current information output. is the weight matrix of the output gate, is the bias term of the output gate.

[0090] (5) Calculate the adaptive modal matching threshold: ; in, is the predicted frame t+1 The adaptive modality matching threshold of the modality clustering center, is the weight matrix of the output layer, is the bias term of the output layer, is the hidden state at the current moment, .

[0091] Specifically, in this embodiment and some embodiments of the present invention, step S104 includes steps S104a to S104c: Step S104a, calculate the The wide frequency oscillation data of the frame modal Hedi The wide frequency oscillation data of the frame modal The weighted distance ; , ; Step S104b, comparing the weighted distances and the modality matching threshold is size; Step S104c, in response to the weighted distance is less than the modality matching threshold , then the mode With the modal Continuous; otherwise, the mode is noise or a new mode.

[0092] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0093] Example 2

[0094] According to another aspect of the embodiment of the present invention, Figure 2 As shown, a modal continuity identification device for multimodal oscillation data is provided, comprising: a feature extraction module, configured to extract features from multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of broadband oscillation data; a clustering processing module, configured to cluster the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data for each frame, and determine cluster center variation features of the broadband oscillation data for each frame; A threshold prediction module is used to perform adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame; The continuity judgment module is used to perform modal continuity matching using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

[0095] Example 3

[0096] like Figure 3 As shown, embodiment three of the present invention provides a terminal device, including at least one processor, and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps in the modal continuity identification method of multimodal oscillation data as described in the first aspect of the present invention.

[0097] The memory and processor are connected using a bus, which can include any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits through interfaces, which are all well known in the art. The interface provides an interface between the bus and the transceiver, such as a communication interface or a user interface. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by the processor is transmitted on a wireless medium via an antenna. Furthermore, the antenna also receives data and transmits the data to the processor.

[0098] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0099] Example 4

[0100] A fourth embodiment of the present invention provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the program implements the steps of the method for identifying modal continuity of multimodal oscillation data as described in the first aspect of the present invention.

[0101] Those skilled in the art will understand from the above description that all or part of the steps in the above-described method embodiments can be implemented by instructing the relevant hardware through a program. The program is stored in a storage medium and includes a number of instructions for causing a device (such as a microcontroller or chip) or a processor to execute all or part of the steps in the method embodiments of this application. The aforementioned storage medium includes, but is not limited to, various media capable of storing program code, such as USB flash drives, mobile hard drives, magnetic storage devices, and optical storage devices.

[0102] Example 5

[0103] A fifth embodiment of the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method for identifying modal continuity of multimodal oscillation data as described in the first aspect of the present invention.

[0104] The embodiments of the present invention address certain limitations of current modal identification and matching methods. By performing intra-frame modal clustering on a time scale for each modality and combining it with a threshold prediction model with preset hyperparameters for adaptive threshold prediction, compared to the method of using a fixed frequency deviation threshold to determine whether two modes are continuous, this method achieves real-time and accurate modal continuity identification for broadband oscillation data with frequency-intensive, strongly time-varying, and multi-modal characteristics.

[0105] The embodiment of the present invention clusters each frame of broadband oscillation data and obtains the typical cluster change characteristics of the cluster cluster for threshold prediction, thereby reducing the computational complexity and amount of the threshold prediction model, and can meet the real-time and accuracy requirements of modal continuity identification of broadband oscillation data with dense frequency, strong time variation, and multimodal characteristics.

[0106] The embodiment of the present invention uses cluster center change characteristics to construct a prediction input vector for adaptive threshold prediction, which solves the continuity misjudgment problem caused by using a single fixed threshold that is too strict or loose due to the dynamic change characteristics of power grid oscillation.

[0107] The embodiment of the present invention uses an improved k-means clustering algorithm that can adapt to the frequency-intensive, strongly time-varying, and multimodal characteristics of broadband oscillation signals in power grids. It has good recognition accuracy in high-dimensional space and can process newly emerging oscillation modes with less training data. It has low computational complexity and is suitable for low-latency real-time processing scenarios in actual power grid monitoring.

[0108] The embodiments of the present invention address the problems faced by traditional modal recognition methods when faced with frequency-intensive, highly time-varying, and multi-modal oscillation data, such as insufficient time correlation modeling capabilities, difficulty in fuzzy recognition of modal transitions, and weak robustness to disturbances. Based on the strong time series modeling and nonlinear mapping capabilities of the pre-trained LSTM model, the modal evolution characteristics are automatically learned through the LSTM model, effectively improving the accuracy and stability of modal continuity recognition. Therefore, the method and device of the embodiments of the present invention can be applied to modal recognition in complex dynamic power systems.

[0109] Therefore, the embodiments of the present invention achieve intelligent matching and continuity identification of oscillation modes in different time frames, which is of great significance for power grid security monitoring, stability control, and oscillation source tracing analysis.

[0110] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for identifying modal continuity of multimodal oscillation data, characterized in that: include: Performing feature extraction on multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of the broadband oscillation data; Clustering the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data of each frame, and determining cluster center change features of the broadband oscillation data of each frame; Performing adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame; Modal continuity matching is performed using the clustering clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

2. The method according to claim 1, characterized in that The modal characteristics of the mode include at least a frequency, a 0th-order amplitude, a 1st-order amplitude, a 0th-order phase, a 1st-order phase, a 0th-order damping ratio coefficient, and a 1st-order damping ratio coefficient corresponding to the mode; Clustering the multiple modes according to the modal features to obtain multiple clusters of the broadband oscillation data of each frame, and determining cluster center change features of the broadband oscillation data of each frame, including: Acquire modal features of the broadband oscillation data of the current frame and the previous frame, and construct a corresponding feature vector of the modal according to the modal features; Determining the clusters of the broadband oscillation data of the current frame and the previous frame and the modes contained therein by using the characteristic vectors, and determining cluster centers according to the characteristic vectors of the modes contained in the clusters; Determining the cluster center change feature of the current frame according to the cluster centers of the current frame and the previous frame; and / or, Adaptive threshold prediction is performed based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame, including: Construct a prediction input vector based on the cluster center change characteristics of each frame in the preset time window, The prediction input vector is input into a pre-trained threshold prediction model, and the threshold prediction model outputs a modality matching threshold of the cluster.

3. The method according to claim 2, characterized in that No. The wide frequency oscillation data of the frame The eigenvector of the mode for ， in, 、 、 、 Respectively Frame width oscillation data The frequency of the mode, Order amplitude, Phase and The damping ratio coefficient of order, , , For the The number of modes of the frame-wide frequency oscillation data; Determining the clusters of the broadband oscillation data of the current frame and the previous frame and the modes contained therein by using the characteristic vectors, and determining the cluster centers according to the characteristic vectors of the modes contained in the clusters, including: The improved K-means clustering algorithm is used to The characteristic vector of each mode of the frame wide frequency oscillation data Perform clustering and obtain The clusters and their respective cluster centers ,in, , For the The set of all the modes contained in the clusters, .

4. The method according to claim 3, characterized in that The cluster center change characteristics include Frame width oscillation data The weighted displacement of cluster centers of clusters , , ； in, , 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 、 Respectively Frame width oscillation data The frequency of the cluster center is Order amplitude, phase and damping ratio coefficient, 、 、 and The frequency and Weight coefficients for order amplitude, phase and damping ratio.

5. The method according to claim 4, characterized in that Adaptive threshold prediction is performed based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame, including: Construct a prediction input vector based on the cluster center change characteristics of each frame in the preset time window , the predicted input vector ,in, Indicates the preset time window size, ; The prediction input vector A pre-trained LSTM model is input to perform adaptive threshold prediction, and the LSTM model outputs the modality matching threshold of the clustering cluster.

6. The method according to any one of claims 1 to 5, characterized in that The predicted The wide frequency oscillation data of the frame modal The modality matching threshold of the cth cluster is ; Adjacent frame modal continuity matching is performed using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result, including: Calculate the The wide frequency oscillation data of the frame modal Hedi The wide frequency oscillation data of the frame modal The weighted distance ; Compare the weighted distances and the modality matching threshold is size; In response to the weighted distance is less than the modality matching threshold , then the mode With the modal continuous.

7. A modal continuity identification device for multimodal oscillation data, characterized in that: include: a feature extraction module, configured to extract features from multiple frames of broadband oscillation data to obtain modal features of multiple modes of each frame of broadband oscillation data; a clustering processing module, configured to cluster the multiple modes according to the modal features, obtain multiple clusters of the broadband oscillation data for each frame, and determine cluster center variation features of the broadband oscillation data for each frame; A threshold prediction module is used to perform adaptive threshold prediction based on the cluster center change characteristics to obtain the modality matching threshold of the cluster cluster of each frame; The continuity judgment module is used to perform modal continuity matching using the clusters and the corresponding modal matching thresholds to obtain a modal continuity recognition result.

8. A computer-readable storage medium, characterized in that A program is stored thereon, and when the program is executed by a processor, the method for identifying modal continuity of multimodal oscillation data according to any one of claims 1 to 6 is implemented.

9. A terminal device, characterized in that: The invention comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for identifying modal continuity of multimodal oscillation data according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for identifying modal continuity of multimodal oscillation data according to any one of claims 1 to 6 is implemented.

Images