Fan controller oscillation risk prediction method and system based on hardware-in-the-loop and svm
By combining hardware-in-the-loop and SVM methods, and using RT-LAB controller simulation data to train an SVM model, the complexity of modeling and the difficulty of prediction of converter oscillation problems in new energy units are solved, and the risk of subsynchronous oscillation is accurately identified and prevented.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-10
- Publication Date
- 2026-03-24
AI Technical Summary
Existing technologies are insufficient to effectively solve the resonance or oscillation problems caused by converters in new energy units, leading to equipment damage and grid instability. Furthermore, the complex modeling makes it difficult to accurately predict the risk of subsynchronous oscillations.
A hardware-in-the-loop and support vector machine (SVM) approach is adopted. Data is obtained through RT-LAB controller semi-physical simulation test, and the wind turbine controller is trained by combining the SVM algorithm to construct a four-level support vector machine model for subsynchronous oscillation risk prediction.
It simplifies the modeling process, improves prediction accuracy, reduces testing costs, and can accurately identify subsynchronous oscillation risks, ensuring the safe grid connection of new energy sources.
Smart Images
Figure CN115425641B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of new energy power system technology, and in particular relates to a method and system for predicting the oscillation risk of wind turbine controllers based on hardware-in-the-loop and SVM. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] Unlike traditional rotating generators, new energy power units mostly use power electronic converters to connect to the grid. The interaction between the converter and the grid can cause resonance or oscillation problems, leading to unit tripping or even equipment damage, endangering the safe and stable operation of the power grid. With the large-scale grid connection of new energy sources and the widespread application of other large-capacity power electronic devices, oscillation stability issues are increasingly becoming a huge potential risk. If timely, systematic and comprehensive research and solutions are not conducted, it may cause huge economic losses to equipment manufacturers, power generation companies and grid companies, and may even cause large-scale stability accidents, endangering the safe and reliable operation of the power grid.
[0004] In recent years, various forms of oscillation stability issues related to renewable energy generating units have occurred worldwide. Subsynchronous oscillations seriously threaten the safety of power grid equipment, system stability, and power quality, and restrict the large-scale integration and consumption of renewable energy sources such as wind and solar power.
[0005] To address the oscillation problem in direct-drive wind farms, domestic and international research primarily focuses on modeling and analyzing the grid-connected inverter system of wind turbines. Methods include frequency scanning, eigenvalue analysis, time-domain simulation, and impedance analysis. While the first three methods can solve some oscillation problems, the converters in direct-drive wind farms contain numerous power electronic devices, making them typical nonlinear systems with exceptionally complex modeling. Impedance analysis establishes a small-signal, wide-frequency impedance model of the power electronic equipment and uses the Nyquist stability criterion to analyze system stability. Impedance analysis includes dq-axis models and positive / negative sequence models; however, in three-phase systems, the d and q axes are mutually coupled and computationally complex, making direct measurement impossible. Summary of the Invention
[0006] To address at least one of the technical problems mentioned above, this invention provides a method and system for predicting the oscillation risk of wind turbine controllers based on hardware-in-the-loop and SVM. It utilizes data obtained from RT-LAB controller hardware-in-the-loop simulation tests, combined with the Support Vector Machine (SVM) algorithm which has good applications in binary classification, to train the SVM on the grid-connected hardware-in-the-loop simulation data of the wind turbine controller, and uses the trained model to complete the prediction of non-test operating conditions.
[0007] To achieve the above objectives, the present invention adopts the following technical solution:
[0008] The first aspect of the present invention provides a method for predicting the oscillation risk of wind turbine controllers based on hardware-in-the-loop and SVM, comprising the following steps:
[0009] Obtain disturbance feedback signals from the wind turbine controller;
[0010] Subsynchronous oscillation risk prediction is performed based on the perturbation feedback signal and a four-level support vector machine to obtain the oscillation risk prediction result; wherein, the construction process of the four-level support vector machine includes:
[0011] The optimal hyperplane is constructed based on the minimum cost function. Relaxation variables are introduced, and the relationship between signal data points and the optimal hyperplane is analyzed based on the relaxation variables and thresholds. The binary classification result is obtained based on this relationship.
[0012] Based on the oscillation conditions of the disturbance feedback signal and the corresponding SVM classifier, the classification is performed step by step. When the oscillation conditions are met, the final classification result is output. The final classification result is compared with the binary classification result to obtain the oscillation risk prediction result.
[0013] A second aspect of the present invention provides a wind turbine controller oscillation risk prediction system based on hardware-in-the-loop and SVM, comprising:
[0014] The signal acquisition module is used to acquire disturbance feedback signals from the wind turbine controller;
[0015] An oscillation risk prediction module is used to predict subsynchronous oscillation risk based on disturbance feedback signals and a four-level support vector machine, and to obtain the oscillation risk prediction result; wherein, the construction process of the four-level support vector machine includes:
[0016] The optimal hyperplane is constructed based on the minimum cost function. Relaxation variables are introduced, and the relationship between signal data points and the optimal hyperplane is analyzed based on the relaxation variables and thresholds. The binary classification result is obtained based on this relationship.
[0017] Based on the oscillation conditions of the disturbance feedback signal and the corresponding SVM classifier, the classification is performed step by step. When the oscillation conditions are met, the final classification result is output. The final classification result is compared with the binary classification result to obtain the oscillation risk prediction result.
[0018] A third aspect of the present invention provides a computer-readable storage medium.
[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the wind turbine controller oscillation risk prediction method described above based on hardware-in-the-loop and SVM.
[0020] A fourth aspect of the present invention provides a computer device.
[0021] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described above.
[0022] Compared with the prior art, the beneficial effects of the present invention are:
[0023] This invention utilizes data obtained from RT-LAB controller hardware-in-the-loop simulation tests, combined with the Support Vector Machine (SVM) algorithm, which has good applications in binary classification, to train the SVM on the hardware-in-the-loop simulation data of the wind turbine controller. The trained model then completes the prediction of non-test operating conditions. This SVM-based classification and recognition method has low requirements for the number of training samples, good real-time performance, and can be well applied to the prediction of subsynchronous oscillation risks in wind turbine controllers, providing a good algorithmic foundation for ensuring the safe grid connection of new energy and preventing subsynchronous oscillation risks.
[0024] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0026] Figure 1 This is a schematic diagram of the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM according to an embodiment of the present invention;
[0027] Figure 2 This is a hardware-in-the-loop schematic diagram of an embodiment of the present invention;
[0028] Figure 3 This is a schematic diagram of the step-by-step classification of the SVM classifier in an embodiment of the present invention. Detailed Implementation
[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0030] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0031] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0032] Typically, full-condition HIL testing is very time-consuming. With the rise of intelligent algorithms, full-condition prediction can be carried out based on limited conditions. The analysis of subsynchronous oscillation can essentially be regarded as a classification or pattern recognition problem of disturbance feedback signals. This invention uses data obtained from the hardware-in-the-loop simulation test of the RT-LAB controller, combined with the Support Vector Machine (SVM) algorithm which has good applications in the field of binary classification, to train the SVM on the grid-connected hardware-in-the-loop simulation data of the wind turbine controller, and completes the prediction of non-test conditions through the trained model.
[0033] Example 1
[0034] like Figure 1 As shown, this embodiment provides a method for predicting the oscillation risk of wind turbine controllers based on hardware-in-the-loop and SVM, including...
[0035] The following steps are required:
[0036] Step 1: Obtain the disturbance feedback signal of the wind turbine controller based on the constructed hardware-in-the-loop model;
[0037] like Figure 2 As shown, the process of building the hardware-in-the-loop model includes:
[0038] A controllable voltage source digital model of the wind turbine circuit is built in the RT-LAB simulation platform according to the actual parameter model. The wind turbine controller under test is connected to the RT-LAB simulation platform through a data cable. The disturbance signal generator is connected to the RT-LAB simulation platform through a physical interface. The disturbance signal is injected into the controllable voltage source of the controllable voltage source digital model. The disturbance feedback signal is collected through a data acquisition card connected to the physical interface.
[0039] The disturbance signal is a three-phase harmonic AC voltage source h with a small amplitude injected between the AC power grid and the target power equipment. ha h hb h hc The perturbation signal matrix with different frequencies and amplitudes is generated, and its analytical expression is:
[0040]
[0041] Among them, h ha hhb h hc Indicates the harmonic voltage or current signal to be injected; A hp and A hu Indicates the amplitude of harmonic voltage; ω s and ω c Represents the injected harmonic angular frequency, where ω s =2πf s ω c =2πf c f s +f c =2f1, where f1 is the power frequency of 50Hz.
[0042] The advantages of the above scheme are:
[0043] (1) Usually, the control strategy of the wind turbine controller is an important factor affecting oscillation. However, due to the technical confidentiality of the control strategy by the controller manufacturer, the construction of the controller model becomes very complicated during the simulation modeling process. In order to obtain the control strategy consistent with the field, this method uses RT-LAB hardware-in-the-loop modeling of the actual wind turbine controller of the new energy power station to make the simulation model more accurate and consistent with the performance of the controller of the actual new energy power station. Thus, it is not necessary to build a complex wind turbine controller model, which greatly reduces the workload of obtaining oscillation feedback signals, simplifies the modeling process, and improves the test accuracy.
[0044] (2) By injecting the disturbance signal matrix into the RT-LAB hardware-in-the-loop simulation model through the controllable voltage source module, the coupling relationship between the disturbance signal and the power signal between the grid aggregation equipment can be accurately simulated. At the same time, the amplitude, frequency and phase of the disturbance signal can be adjusted arbitrarily, so that this scheme can simulate more complex disturbance situations. The disturbance signal coverage is wide, and the actual disturbances that occur in the grid can be simulated more accurately.
[0045] Step 2: Preprocess the disturbance feedback signal;
[0046] To acquire perturbation feedback signals via the RT-LAB physical interface and obtain suitable training samples, it is necessary to label the perturbation feedback signals and assign labels to them, which represent the expected values of the training sample data.
[0047] Because different frequencies f are injected into the power grid in the RT-LAB hardware-in-the-loop simulation model. i The disturbance signal (i = 1, 2, 3, ..., 100) is followed by the feedback power signal P after passing through the wind turbine controller. i No longer a constant value, through P i The trajectory of P is used to determine the oscillation state. i An oscillation is considered triggered if the following conditions are met:
[0048] The first oscillation condition is: the absolute value of the difference between the instantaneous sampling power of the disturbance feedback signal and the average power within the sampling window is greater than the subsynchronous oscillation threshold.
[0049] The second oscillation condition is: the absolute value of the difference between the maximum and minimum power of the subsynchronous oscillation period caused by the disturbance feedback signal is greater than the power amplitude threshold of the subsynchronous oscillation.
[0050] The third oscillation condition is: the subsynchronous oscillation frequency caused by the disturbance feedback signal is between the upper limit and the lower limit of the subsynchronous oscillation frequency;
[0051] The fourth oscillation condition is: the number of oscillations of the current subsynchronous oscillation caused by the disturbance feedback signal is greater than the threshold of the number of oscillations of the subsynchronous oscillation.
[0052] Its expression is as follows:
[0053] Condition 1: |P i -P m |>P T
[0054] Condition 2: |P imax -P imin |>P A
[0055] Condition 3: f min <f ic <f max
[0056] Condition 4:n i >n set
[0057] Among them, P i For instantaneous sampling power, P m P is the average power during the sampling window. T This is the threshold value for subsynchronous oscillation. P imax f i The maximum power of the subsynchronous oscillation period caused by frequency disturbance, P imin f i The minimum power, P, of the subsynchronous oscillation period caused by frequency disturbance. A f is the power amplitude threshold for subsynchronous oscillation. ic f i The subsynchronous oscillation frequency caused by frequency disturbance, f max Upper limit of subsynchronous oscillation frequency, f min n is the lower limit of the subsynchronous oscillation frequency. i f i The number of oscillations, n, caused by the frequency disturbance in the current subsynchronous oscillation. setThis is the threshold for the number of oscillations in subsynchronous oscillations.
[0058] When the above constraints are met, it is considered that a subsynchronous oscillation has occurred.
[0059] For different frequencies f i The disturbance feedback signal is labeled with a data tag, i.e., its expected value tag, according to the satisfaction of conditions 1 to 4. If the conditions are met, the corresponding disturbance signal f i The expected value is 1. If the condition is not met, then the corresponding disturbance signal f i The expected value is -1.
[0060] The disturbance signal is normalized using the following mapping:
[0061]
[0062] Where x represents the perturbation signal, and y is the normalized result, x min x max These represent the minimum and maximum values of the disturbance signal, respectively. The normalized raw data will be rounded to the range [0, 1].
[0063] The normalized data is used to obtain the expected value at the response frequency based on the oscillation condition discrimination result of the perturbation feedback signal, thereby constructing complete support vector machine (SVM) training and testing samples. These samples contain perturbation response signals of different amplitudes.
[0064] Step 3: Construct the optimal hyperplane based on the minimum cost function, introduce slack variables, analyze the relationship between the signal data points and the optimal hyperplane based on the slack variables and threshold, and obtain the binary classification result based on this relationship; specifically including:
[0065] Support Vector Machines (SVMs) can be used for pattern classification and nonlinear regression. The main idea is to construct a classification hyperplane as a decision surface that maximizes the separation margin between positive and negative examples. An SVM is constructed from a small subset of the training data extracted by the algorithm. SVMs can construct functions on a wide range of function sets, are robust, require no fine-tuning, are computationally simple, and their implementation only requires simple optimization techniques, making them theoretically quite robust.
[0066] For P sample data {(X 1 d 1 ), (X 2 d 2 ), ..., (X p d p ), ..., (X P d PFor any input sample X p Its expected output is d p = ±1, representing the category identifiers for satisfying and not satisfying the oscillation condition, respectively. The hyperplane used for classification is:
[0067] W T X+b=0
[0068] In the formula, X is the input sample vector, W is the weight vector, and b is the bias, then we have:
[0069] W T X p +b>0, when d p =1
[0070] W T X p +b<0, when d p =-1
[0071] It can also be expressed as:
[0072] d p (W T X p +b)≥1p=1,2,...,P
[0073] Therefore, the weights W0 and bias b0 of the optimal hyperplane should be unique. The equation of the optimal hyperplane is:
[0074] W T X0+b0=0
[0075] Constructing the optimal hyperplane is equivalent to finding the optimal weight vector W, which, for a given training sample, minimizes the cost function:
[0076]
[0077] Introducing the Lagrange function:
[0078]
[0079] In the formula, α p For the Lagrange coefficients, solving the Lagrange function yields:
[0080]
[0081] thereby:
[0082]
[0083] Let Q(α) = L(W,b,α), then the problem is transformed into finding a set of constraints given training samples: αp Lagrange coefficients {α1, α2, ... α} for ≥0, p=1,2,...,P p , ..., α P}, so that Q(α) reaches its maximum value.
[0084] Let the optimal solution of Q(α) be {α} 01 α 02 ,...α 0p , ..., α 0P}, then the optimal weight vector is:
[0085]
[0086] The superscript s is the label of the special sample closest to the optimal hyperplane, and SV is the support vector.
[0087] The optimal classification discriminant function is:
[0088]
[0089] To enhance the generalization ability of the discriminant function, slack variables are introduced to appropriately relax the constraints. Let:
[0090] d p (W T X p +b)≥1-ξ p p = 1, 2, ..., P
[0091] Wherein, slack variable ξ p ≥0, p=1,2,...,P, when 0≤ξ p When ξ ≤ 1, the data point falls inside the separation region and on the correct side of the classification hyperplane. p When ξ > 1, the data point enters the incorrect side of the classification hyperplane; when ξ < 1, the data point enters the incorrect side of the classification hyperplane. p When = 0, the corresponding data point is the support vector X of the exact hyperplane. s .
[0092] The category of the training sample X can be determined by f(X) being 1 or -1.
[0093] Step 4: Classify the disturbance feedback signal step by step according to the oscillation conditions and the corresponding SVM classifier. When the oscillation conditions are met, output the final classification result. Compare the final classification result with the binary classification result to obtain the oscillation risk prediction result.
[0094] like Figure 3 As shown, this embodiment uses four SVMs for four-level classification training and recognition.
[0095] First, a first-level SVM classifier is used for first-level classification training. When the difference between the instantaneous sampled power of the perturbation feedback signal and the average power within the sampling window satisfies the first oscillation condition, the output is 1; otherwise, the output is -1. Samples with an output of 1 are further input into a second-level SVM classifier to further train whether the relationship between the maximum and minimum power of the perturbation feedback signal satisfies the second oscillation condition. If it does, the output is 1; otherwise, the output is -1. Samples with an output of 1 are then input into a third-level SVM classifier to train whether the frequency of the perturbation feedback signal satisfies the third oscillation condition. If it does, the output is 1; otherwise, the output is -1.
[0096] Finally, the samples filled with the above three-level SVMs are further input into the four-level SVM classifier to determine whether the number of oscillations meets the fourth oscillation condition. If it does, the output is 1; otherwise, the output is -1. After training with the four-level SVM, the subsynchronous oscillation prediction model can be obtained.
[0097] The technical advantages of the above solution are:
[0098] (1) The SVM algorithm that determines whether subsynchronous oscillation occurs has a final decision function determined by only a few support vectors. The computational complexity depends on the number of support vectors rather than the dimension of the sample space, thus avoiding the "curse of dimensionality". It can make full use of key samples and eliminate a large number of redundant samples. This method is not only simple in algorithm, but also has low sensitivity to unnecessary or abnormal samples, and has good "robustness". This can reduce the dimension of RT-LAB hardware-in-the-loop testing, simplify the process of obtaining the required disturbance feedback signal, and reduce the number of test samples, which is of great significance for improving test efficiency.
[0099] (2) The oscillation feedback signal is subdivided into four aspects, namely power, amplitude, frequency and number of oscillations, through a four-level SVM to determine whether subsynchronous oscillation has occurred. This can effectively avoid misjudgment of the oscillation signal caused by power grid load fluctuations, power shifts, power flow redistribution caused by faults, and can more accurately distinguish between subsynchronous oscillation signals and non-subsynchronous oscillation signals or pseudo-subsynchronous oscillation signals, making the classification of disturbance feedback signals used to determine whether the controller will cause subsynchronous oscillation more reliable.
[0100] Example 2
[0101] This embodiment provides a wind turbine controller oscillation risk prediction system based on hardware-in-the-loop and SVM, including:
[0102] The signal acquisition module is used to acquire disturbance feedback signals from the wind turbine controller;
[0103] An oscillation risk prediction module is used to predict subsynchronous oscillation risk based on disturbance feedback signals and a four-level support vector machine, and to obtain the oscillation risk prediction result; wherein, the construction process of the four-level support vector machine includes:
[0104] The optimal hyperplane is constructed based on the minimum cost function. Relaxation variables are introduced, and the relationship between signal data points and the optimal hyperplane is analyzed based on the relaxation variables and thresholds. The binary classification result is obtained based on this relationship.
[0105] Based on the oscillation conditions of the disturbance feedback signal and the corresponding SVM classifier, the classification is performed step by step. When the oscillation conditions are met, the final classification result is output. The final classification result is compared with the binary classification result to obtain the oscillation risk prediction result.
[0106] Example 3
[0107] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described above.
[0108] Example 4
[0109] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described above.
[0110] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.
[0111] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.
[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0114] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0115] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM, characterized in that, Includes the following steps: Obtain disturbance feedback signals from the wind turbine controller; Subsynchronous oscillation risk prediction is performed based on the perturbation feedback signal and a four-level support vector machine to obtain the oscillation risk prediction result; wherein, the construction process of the four-level support vector machine includes: The optimal hyperplane is constructed based on the minimum cost function. Relaxation variables are introduced, and the relationship between signal data points and the optimal hyperplane is analyzed based on the relaxation variables and thresholds. The binary classification result is obtained based on this relationship. Based on the oscillation conditions of the perturbation feedback signal and the corresponding SVM classifier, the classification is performed step by step. When the oscillation conditions are met, the final classification result is output, which includes: A first-level SVM classifier is used for first-level classification training. The perturbation feedback signal that meets the first oscillation condition is input into a second-level SVM classifier for second-level classification training. When the perturbation feedback signal that meets the second oscillation condition is input into a third-level SVM classifier for third-level classification training, and when the perturbation feedback signal that meets the third oscillation condition is input into a fourth-level SVM classifier for fourth-level classification training, the classification result is output when the fourth oscillation condition is met. The final classification result is compared with the binary classification result to obtain the oscillation risk prediction result. The first oscillation condition is: the absolute value of the difference between the instantaneous sampled power of the disturbance feedback signal and the average power within the sampling window is greater than the subsynchronous oscillation threshold; the second oscillation condition is: the absolute value of the difference between the maximum power and the minimum power of the subsynchronous oscillation period caused by the disturbance feedback signal is greater than the power amplitude threshold of the subsynchronous oscillation; the third oscillation condition is: the subsynchronous oscillation frequency caused by the disturbance feedback signal is between the upper limit and the lower limit of the subsynchronous oscillation frequency; the fourth oscillation condition is: the number of oscillations of the current subsynchronous oscillation caused by the disturbance feedback signal is greater than the number of oscillations threshold of the subsynchronous oscillation.
2. The wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in claim 1, characterized in that, The disturbance feedback signal of the wind turbine controller is obtained through a hardware-in-the-loop model, and the construction of the hardware-in-the-loop model includes: The circuit of the wind turbine is built into a controllable voltage source digital model in the simulation platform according to the actual parameter model. The wind turbine controller under test is connected to the simulation platform through a data cable. The disturbance signal generator is connected to the simulation platform through a physical interface. The disturbance signal is injected into the controllable voltage source of the controllable voltage source digital model. The disturbance feedback signal is collected through a data acquisition card connected to the physical interface.
3. The wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in claim 1, characterized in that, The construction of the optimal hyperplane based on the minimum cost function includes: Based on the minimum cost function, the Lagrange function is introduced to find the optimal weight vector; Construct the optimal hyperplane based on the optimal weight vector and bias.
4. The wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in claim 1, characterized in that, The introduction of relaxation variables, the analysis of the relationship between signal data points and the optimal hyperplane based on relaxation variables and thresholds, and the obtaining of binary classification results based on this relationship include: when When a data point falls inside the separation region and is on the correct side of the classification hyperplane, When, the data point enters the incorrect side of the classification hyperplane; when At that time, the corresponding data points are the support vectors of the precise hyperplane.
5. The wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in claim 1, characterized in that, After acquiring the disturbance feedback signal from the wind turbine controller, the disturbance feedback signal is labeled and a label is set for the disturbance feedback signal to train the expectation of the sample data.
6. A wind turbine controller oscillation risk prediction system based on hardware-in-the-loop and SVM, characterized in that, include: The signal acquisition module is used to acquire disturbance feedback signals from the wind turbine controller; An oscillation risk prediction module is used to predict subsynchronous oscillation risk based on disturbance feedback signals and a four-level support vector machine, and to obtain the oscillation risk prediction result; wherein, the construction process of the four-level support vector machine includes: The optimal hyperplane is constructed based on the minimum cost function. Relaxation variables are introduced, and the relationship between signal data points and the optimal hyperplane is analyzed based on the relaxation variables and thresholds. The binary classification result is obtained based on this relationship. Based on the oscillation conditions of the perturbation feedback signal and the corresponding SVM classifier, the classification is performed step by step. When the oscillation conditions are met, the final classification result is output, which includes: A first-level SVM classifier is used for first-level classification training. The perturbation feedback signal that meets the first oscillation condition is input into a second-level SVM classifier for second-level classification training. When the perturbation feedback signal that meets the second oscillation condition is input into a third-level SVM classifier for third-level classification training, and when the perturbation feedback signal that meets the third oscillation condition is input into a fourth-level SVM classifier for fourth-level classification training, the classification result is output when the fourth oscillation condition is met. The final classification result is compared with the binary classification result to obtain the oscillation risk prediction result. The first oscillation condition is: the absolute value of the difference between the instantaneous sampled power of the disturbance feedback signal and the average power within the sampling window is greater than the subsynchronous oscillation threshold; the second oscillation condition is: the absolute value of the difference between the maximum power and the minimum power of the subsynchronous oscillation period caused by the disturbance feedback signal is greater than the power amplitude threshold of the subsynchronous oscillation; the third oscillation condition is: the subsynchronous oscillation frequency caused by the disturbance feedback signal is between the upper limit and the lower limit of the subsynchronous oscillation frequency; the fourth oscillation condition is: the number of oscillations of the current subsynchronous oscillation caused by the disturbance feedback signal is greater than the number of oscillations threshold of the subsynchronous oscillation.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in any one of claims 1-5.
8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the wind turbine controller oscillation risk prediction method based on hardware-in-the-loop and SVM as described in any one of claims 1-5.