Fan vibration suppression method, device, equipment and computer readable storage medium

Abstract

This application discloses a vibration suppression method, apparatus, device, and computer-readable storage medium for wind turbines, relating to the field of wind turbine control technology. The method includes: acquiring initial operating condition data and initial vibration response data of the target wind turbine during its initial operating cycle after commissioning; extracting structural dynamics feature vectors of the target wind turbine during the initial operating cycle from the initial vibration response data, using these as initial structural dynamics feature vectors; inputting the initial operating condition data and initial structural dynamics feature vectors into a pre-trained meta-learning model to obtain target vibration suppression parameters corresponding to the target wind turbine; wherein the meta-learning model is pre-trained using historical operating condition data and historical structural dynamics feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels; and tuning the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to suppress the vibration of the target wind turbine.

Description

Technical Field

[0001] This application relates to the field of wind turbine control technology, and in particular to a method, apparatus, equipment and computer-readable storage medium for suppressing the vibration of a wind turbine. Background Technology

[0002] Wind turbines are subjected to complex alternating loads during operation, and vibration problems in key components such as the tower, blades, and drive train directly affect the safety and service life of the turbine. To suppress harmful vibrations, active vibration control strategies are commonly used. This involves adding a damping control loop to the pitch controller or torque controller, adjusting operating parameters (such as pitch angle and generator torque) based on real-time monitored vibration signals to generate a damping force opposite to the vibration direction, thereby reducing the vibration amplitude.

[0003] The performance of this active vibration suppression strategy is highly dependent on the tuning quality of controller parameters (such as gain coefficient, time constant, and filter center frequency). Currently, the industry typically performs independent parameter tuning for each wind turbine to be put into operation. A common practice is to collect a large amount of operational data in the initial stage of wind turbine grid connection, and based on the wind turbine dynamics model or data-driven model, use intelligent optimization algorithms such as genetic algorithms and particle swarm optimization to repeatedly search within the parameter space. By comparing the vibration response performance under different parameter combinations, a set of optimal vibration suppression parameters is finally determined and fixed for subsequent operation of the wind turbine.

[0004] However, during parameter tuning, wind turbines are prone to unnecessarily additional fatigue loads due to suboptimal parameters. Furthermore, under the aforementioned parameter tuning methods, each wind turbine must undergo an independent tuning process "starting from scratch," with a commissioning cycle typically lasting several weeks or even months. This increases the turbine's operation and maintenance costs and the risk of early failures. Summary of the Invention

[0005] The main objective of this application is to provide a method, apparatus, device, and computer-readable storage medium for suppressing the vibration of a wind turbine, aiming to shorten the commissioning cycle of the vibration suppression parameters, so as to effectively suppress the vibration of the wind turbine while reducing the operation and maintenance costs and early failure risks of the wind turbine.

[0006] This application provides a method for suppressing the vibration of a wind turbine, the method comprising: During the initial operating cycle after the target wind turbine is put into operation, the initial operating condition data and initial vibration response data of the target wind turbine are acquired; The structural dynamic feature vector of the target wind turbine in the initial operating cycle is extracted from the initial vibration response data and used as the initial structural dynamic feature vector. The initial operating condition data and the initial structural dynamic feature vector are input into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine. The meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples and vibration suppression parameters corresponding to the input samples as output labels. The parameters of the vibration suppression model of the target wind turbine are tuned to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned. The vibration suppression model, after tuning the vibration suppression parameters, suppresses the vibration of the target wind turbine.

[0007] In one embodiment, the step of extracting the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data includes: The initial vibration response data is subjected to time-frequency analysis or modal parameter identification to extract at least one natural frequency of the target wind turbine during the initial operating cycle and the modal damping ratio corresponding to the natural frequency, thereby forming the structural dynamic feature vector of the target wind turbine during the initial operating cycle.

[0008] In one embodiment, after the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned with the vibration suppression parameters, the method further includes: Dynamically acquire real-time operating condition data and real-time vibration response data of the target wind turbine during its operation; Extract the real-time structural dynamics feature vector of the target wind turbine from the real-time vibration response data; Determine whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector; If so, the real-time operating data and the real-time structural dynamics feature vector are input into the meta-learning model to obtain the updated vibration suppression parameters corresponding to the target wind turbine; The parameters of the vibration suppression model are tuned to the updated vibration suppression parameters.

[0009] In one embodiment, the step of determining whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector includes: Determine the drift index value between the real-time structural dynamics feature vector and the initial structural dynamics feature vector; If the drift index value is greater than the preset drift index threshold, then it is determined that there is a feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector.

[0010] In one embodiment, after the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned with the vibration suppression parameters, the method further includes: Obtain the actual vibration amplitude of the target wind turbine after the vibration suppression model has been run for a preset verification time with the target vibration suppression parameters; If the actual vibration amplitude is greater than the preset amplitude threshold, then the target vibration suppression parameter is used as the initial search point, and an iterative search is performed within the parameter adjustment range corresponding to the target vibration suppression parameter to obtain the fine-tuned target vibration suppression parameter. The parameters of the vibration suppression model are updated to the fine-tuned target vibration suppression parameters.

[0011] In one embodiment, after the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned with the vibration suppression parameters, the method further includes: Obtain the current operating condition data of the target wind turbine and determine whether the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model; If so, the parameters of the vibration suppression model are tuned to preset safe vibration suppression parameters. The safe vibration suppression parameters refer to the vibration suppression parameters that are tuned offline for a wind turbine of the same model as the target wind turbine under operating conditions that exceed the operating condition distribution range.

[0012] In one embodiment, the step of determining whether the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model includes: Calculate the Mahalanobis distance between the current working condition data and the working condition distribution center covered during the training of the meta-learning model; If the Mahalanobis distance is greater than a preset distance threshold, it is determined that the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model.

[0013] Furthermore, to achieve the above objectives, this application also provides a vibration suppression device for a fan, the device comprising: The data acquisition module is used to acquire the initial operating condition data and initial vibration response data of the target wind turbine during the initial operating cycle after the target wind turbine is put into operation. The feature extraction module is used to extract the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data, and use it as the initial structural dynamic feature vector; The parameter tuning module is used to input the initial operating condition data and the initial structural dynamic feature vector into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine. The meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels. The module tunes the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to obtain the vibration suppression model after parameter tuning. The vibration suppression module is used to suppress the vibration of the target fan based on the vibration suppression model after the vibration suppression parameters are tuned.

[0014] In addition, to achieve the above objectives, this application also provides an electronic device, the electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the vibration suppression method for a fan as described above.

[0015] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the vibration suppression method for a fan as described above.

[0016] This application provides a vibration suppression method for wind turbines, comprising: acquiring initial operating condition data and initial vibration response data of the target wind turbine during the initial operating cycle after its commissioning; extracting structural dynamic feature vectors of the target wind turbine during the initial operating cycle from the initial vibration response data, as initial structural dynamic feature vectors; inputting the initial operating condition data and initial structural dynamic feature vectors into a pre-trained meta-learning model to obtain target vibration suppression parameters corresponding to the target wind turbine; wherein, the meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels; tuning the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to obtain a vibration suppression model after vibration suppression parameter tuning; and suppressing the vibration of the target wind turbine based on the vibration suppression model after vibration parameter tuning.

[0017] Therefore, the technical solution provided in this application pre-employs a model-independent meta-learning algorithm. It uses historical operating data and historical structural dynamic feature vectors from multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels to train a meta-learning model. This meta-learning model internalizes the mapping law between the same model of wind turbine under different operating conditions and structural states and the optimal vibration suppression parameters. Thus, during the initial operating cycle after the target wind turbine is put into operation, only its initial operating data and initial vibration response data need to be obtained, and the initial structural dynamic feature vector is extracted from them. By inputting the initial operating data and initial structural dynamic feature vector into the meta-learning model, the target vibration suppression parameters applicable to the target wind turbine can be directly obtained. The parameters of the vibration suppression model of the target wind turbine are then tuned to these target vibration suppression parameters, and the vibration of the target wind turbine is suppressed based on the tuned vibration suppression model. Therefore, the target wind turbine does not need to undergo a parameter search and trial-and-error process starting from scratch and lasting for weeks or even months. The vibration suppression parameters can be tuned and an effective vibration suppression state established within the initial operating cycle, significantly shortening the debugging cycle of the vibration suppression parameters. Furthermore, the shortened commissioning cycle allows the target wind turbine to operate with parameters that provide good vibration damping earlier, reducing the additional fatigue load caused by non-optimal parameters during the commissioning period. This enables effective suppression of wind turbine vibration while reducing wind turbine operation and maintenance costs and early failure risks.

[0018] In summary, the technical solution provided in this application can shorten the commissioning cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A schematic flowchart of the vibration suppression method for a fan provided in the first embodiment of this application; Figure 2 This is a schematic flowchart of the vibration suppression method for a fan provided in the second embodiment of this application; Figure 3 A schematic flowchart of the vibration suppression method for a fan provided in the third embodiment of this application; Figure 4 This is a schematic flowchart of the vibration suppression method for a fan provided in the fourth embodiment of this application; Figure 5 A schematic diagram of the module structure of the vibration suppression device for a fan provided in an embodiment of this application; Figure 6 This is a schematic diagram of the hardware operating environment involved in the embodiments of this application.

[0022] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] Wind turbines are subjected to complex alternating loads during operation, and vibration problems in key components such as the tower, blades, and drive train directly affect the safety and service life of the turbine. To suppress harmful vibrations, active vibration control strategies are commonly used. This involves adding a damping control loop to the pitch controller or torque controller, adjusting operating parameters (such as pitch angle and generator torque) based on real-time monitored vibration signals to generate a damping force opposite to the vibration direction, thereby reducing the vibration amplitude.

[0026] The performance of this active vibration suppression strategy is highly dependent on the tuning quality of controller parameters (such as gain coefficient, time constant, and filter center frequency). Currently, the industry typically performs independent parameter tuning for each wind turbine to be put into operation. A common practice is to collect a large amount of operational data in the initial stage of wind turbine grid connection, and based on the wind turbine dynamics model or data-driven model, use intelligent optimization algorithms such as genetic algorithms and particle swarm optimization to repeatedly search within the parameter space. By comparing the vibration response performance under different parameter combinations, a set of optimal vibration suppression parameters is finally determined and fixed for subsequent operation of the wind turbine.

[0027] However, during parameter tuning, wind turbines are prone to unnecessarily additional fatigue loads due to suboptimal parameters. Furthermore, under the aforementioned parameter tuning methods, each wind turbine must undergo an independent tuning process "starting from scratch," with a commissioning cycle typically lasting several weeks or even months. This increases the turbine's operation and maintenance costs and the risk of early failures.

[0028] Based on this, this application provides a vibration suppression method for wind turbines, comprising: acquiring initial operating condition data and initial vibration response data of the target wind turbine during the initial operating cycle after the target wind turbine is put into operation; extracting the structural dynamics feature vector of the target wind turbine during the initial operating cycle from the initial vibration response data, as the initial structural dynamics feature vector; inputting the initial operating condition data and the initial structural dynamics feature vector into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine; wherein, the meta-learning model is pre-trained using a model-independent meta-learning algorithm, using historical operating condition data and historical structural dynamics feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and using the vibration suppression parameters corresponding to the input samples as output labels; tuning the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned; and suppressing the vibration of the target wind turbine based on the vibration suppression model after the vibration suppression parameters are tuned.

[0029] Therefore, the technical solution provided in this application pre-employs a model-independent meta-learning algorithm. It uses historical operating data and historical structural dynamic feature vectors from multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels to train a meta-learning model. This meta-learning model internalizes the mapping law between the same model of wind turbine under different operating conditions and structural states and the optimal vibration suppression parameters. Thus, during the initial operating cycle after the target wind turbine is put into operation, only its initial operating data and initial vibration response data need to be obtained, and the initial structural dynamic feature vector is extracted from them. By inputting the initial operating data and initial structural dynamic feature vector into the meta-learning model, the target vibration suppression parameters applicable to the target wind turbine can be directly obtained. The parameters of the vibration suppression model of the target wind turbine are then tuned to these target vibration suppression parameters, and the vibration of the target wind turbine is suppressed based on the tuned vibration suppression model. Therefore, the target wind turbine does not need to undergo a parameter search and trial-and-error process starting from scratch and lasting for weeks or even months. The vibration suppression parameters can be tuned and an effective vibration suppression state established within the initial operating cycle, significantly shortening the debugging cycle of the vibration suppression parameters. Furthermore, the shortened commissioning cycle allows the target wind turbine to operate with parameters that provide good vibration damping earlier, reducing the additional fatigue load caused by non-optimal parameters during the commissioning period. This enables effective suppression of wind turbine vibration while reducing wind turbine operation and maintenance costs and early failure risks.

[0030] In summary, the technical solution provided in this application can shorten the commissioning cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine.

[0031] The subject executing the vibration suppression method for wind turbines in this application can be an electronic device with data processing, network communication, and program execution functions, such as a mobile phone, computer, cloud server, etc., or a control system or control circuit capable of realizing the above functions. This embodiment does not specifically limit this.

[0032] The following description uses an electronic device as the execution subject to illustrate the various embodiments.

[0033] This application presents a vibration suppression method for a wind turbine according to a first embodiment. Please refer to [link / reference]. Figure 1 The vibration suppression method for the fan may include steps S10 to S50: Step S10: During the initial operating cycle after the target wind turbine is put into operation, acquire the initial operating condition data and initial vibration response data of the target wind turbine. It should be noted that the target wind turbine refers to the wind turbine whose vibration suppression parameters need to be tuned. This is typically a newly commissioned turbine, but can also be an existing turbine that requires parameter readjustment. This embodiment does not impose specific limitations on this. Commissioning refers to the wind turbine completing installation and commissioning and officially connecting to the grid for power generation. The initial operating period refers to a preset time interval after the target wind turbine is commissioned, used to collect initial data to complete the initial parameter tuning. This initial operating period is usually much shorter than the time required for parameter tuning "from zero" in existing technologies. Initial operating condition data refers to the external conditions and control state parameters collected during the initial operating period that describe the operating state of the target wind turbine. These may include, but are not limited to, wind speed, turbulence intensity, rotor speed, blade pitch angle, and / or generator torque. This embodiment does not impose specific limitations on these parameters. Initial vibration response data refers to the time-series signals collected during the initial operating period that reflect the vibration state of key components of the target wind turbine. These may include, but are not limited to, nacelle vibration acceleration signals, tower vibration acceleration signals, and / or drive train torsional vibration signals. This embodiment does not impose specific limitations on these parameters.

[0034] Initial operating condition data can be obtained through the wind turbine's data acquisition and monitoring control system, and initial vibration response data can be obtained through vibration sensors installed in the nacelle, tower, and drive train.

[0035] Step S20: Extract the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data, and use it as the initial structural dynamic feature vector; It should be noted that the structural dynamics eigenvector is a set of parameters used to quantitatively describe the dynamic characteristics of the wind turbine structure. It is the "vibration fingerprint" of the wind turbine, which may include, but is not limited to, some or all of the wind turbine's natural frequencies and the modal damping ratios corresponding to each natural frequency. This embodiment does not make specific limitations on this.

[0036] When extracting the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data, time-frequency analysis or modal parameter identification processing can be performed on the initial vibration response data to extract at least one natural frequency of the target wind turbine in the initial operating cycle and the modal damping ratio corresponding to the natural frequency, thus forming the structural dynamic feature vector of the target wind turbine in the initial operating cycle.

[0037] Among them, natural frequencies refer to the vibration frequencies corresponding to each principal vibration of a structure in a free vibration state. They are inherent properties of the structure and are independent of external excitation. For wind turbines, the first-order front-to-back natural frequencies, the first-order left-to-right natural frequencies of the tower, and the first-order torsional natural frequency of the transmission chain are usually of concern. Modal damping ratio refers to the ratio of actual damping to critical damping under a certain modal vibration of the structure, reflecting the rate of energy dissipation of that modal vibration. Time-frequency analysis processing refers to methods for joint time-frequency domain analysis of vibration response data, such as short-time Fourier transform, wavelet transform, Hilbert-Huang transform, etc. Modal parameter identification processing refers to methods for identifying structural modal parameters from vibration response data, such as random subspace identification, characteristic system realization algorithms, least squares complex frequency domain methods, etc.

[0038] Understandably, by utilizing time-frequency analysis or modal parameter identification, the natural frequencies and modal damping ratios reflecting the structural characteristics of the wind turbine can be accurately extracted from the initial vibration response data collected during the initial operation of the target wind turbine, and an initial structural dynamics feature vector can be constructed accordingly. This allows the constructed initial structural dynamics feature vector to effectively eliminate the interference of external operating conditions such as wind speed and turbulence on the vibration signal, truly reflecting the "vibration fingerprint" of the target wind turbine in a healthy state. Compared to directly using the raw vibration response data as model input, the low-dimensional structural dynamics feature vector obtained after feature extraction has stronger operating condition invariance and wind turbine individual characterization capabilities, providing a high-quality input foundation for the subsequent meta-learning model to accurately infer the target vibration suppression parameters, thereby ensuring the accuracy and reliability of cross-wind turbine experience transfer.

[0039] When using Hilbert-Huang transform to process initial vibration response data to extract at least one natural frequency of the target wind turbine during its initial operating cycle and the corresponding modal damping ratio, the initial vibration response data can first be processed by empirical mode decomposition to obtain several intrinsic mode functions (EMFs). Then, a Hilbert transform is performed on the EMF component with the largest amplitude to obtain the instantaneous frequency and instantaneous amplitude. Statistical analysis of the instantaneous frequency sequence determines the first natural frequency of the target wind turbine. Exponential fitting of the decay envelope of the instantaneous amplitude determines the corresponding modal damping ratio. If the vibration signal energy is sufficient, the second natural frequency and corresponding modal damping ratio can be further extracted from the EMF component with the second largest amplitude.

[0040] When using stochastic subspace identification to process initial vibration response data to extract at least one natural frequency and corresponding modal damping ratio of the target wind turbine during the initial operating cycle, operating segments that meet preset steady-state conditions can be selected from the initial operating cycle, such as operating segments with wind speed fluctuations less than 1 m / s, rotor speed fluctuations less than 0.5 rpm, and pitch angle changes less than 0.5°. Then, the vibration acceleration signals in the nacelle forward and backward directions and the tower left and right directions corresponding to the operating segments are processed using a covariance-driven stochastic subspace identification method to construct a state-space model of the vibration response. Finally, by performing eigenvalue decomposition on the state-space model, at least one natural frequency and corresponding modal damping ratio can be extracted.

[0041] Step S30: Input the initial operating condition data and the initial structural dynamic feature vector into the pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine; wherein, the meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels. It should be noted that the meta-learning model in this embodiment has been trained using historical wind turbine data from multiple wind turbines of the same model as the target wind turbine, thus acquiring experience in configuring vibration suppression parameters across wind turbines. Model-Agnostic Meta-Learning (MAML) is a typical meta-learning algorithm. Its core idea is to find a set of initial model parameters sensitive to task changes through training on a large number of similar tasks, enabling the model to quickly converge to the optimal solution with only a small number of samples and a few gradient updates when facing a new task. The target vibration suppression parameters are the vibration suppression parameter configurations output by the meta-learning model for the current state of the target wind turbine, applicable to that wind turbine. These can include, but are not limited to, the gain coefficient and time constant of the pitch damping controller, or the center frequency and notch depth of the notch filter, etc. This embodiment does not specifically limit these parameters.

[0042] The pre-training of the meta-learning model can be completed offline before steps S10 to S50 are executed. The training process may include: First, acquiring historical operating condition data and historical vibration response data of multiple wind turbines of the same model as the target wind turbine in their respective historical operating cycles, as well as the optimal vibration suppression parameters of each wind turbine of the same model confirmed by offline optimization; wherein, offline optimization can be carried out using methods such as genetic algorithm, particle swarm optimization or Bayesian optimization, and the optimal vibration suppression parameters of the same model wind turbine confirmed by offline optimization are obtained through sufficient search in the early stage of operation of the same model wind turbine and verified to have good vibration suppression effect. Then, corresponding historical structural dynamic feature vectors are extracted from the historical vibration response data of each wind turbine of the same model. Next, using the historical operating condition data and historical structural dynamic feature vectors of each wind turbine of the same model as input samples, and the corresponding optimal vibration suppression parameters as output labels, a meta-training dataset is constructed (treating the data of each wind turbine of the same model as an independent meta-task). Finally, based on the constructed meta-training dataset, a model-independent meta-learning algorithm is used to meta-train the initial neural network. During the meta-training process, the inner loop performs a small number of gradient descent steps on each meta-task to adapt to the task, while the outer loop summarizes the adaptation effects on each task to update the initial parameters of the model, ultimately obtaining the trained meta-learning model. This meta-learning model internalizes the universal mapping law across wind turbines of "operating condition-structural feature-optimal vibration suppression parameter".

[0043] Step S40: The parameters of the vibration suppression model of the target wind turbine are tuned to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned. It should be noted that the vibration suppression model is the vibration suppression control algorithm module running in the wind turbine controller. It is used to generate additional control commands based on the real-time monitored vibration signals to suppress wind turbine vibration. Specifically, it can be an additional damping control model or a notch filter model, etc. This embodiment does not make specific limitations on it.

[0044] When the target wind turbine adopts an additional damping control strategy, the vibration suppression model can be an additional damping control model. This additional damping control model is usually connected in series in the main control loop of the pitch controller or torque controller, and it can generate additional pitch angle commands or torque commands based on the vibration signal. In this case, the target vibration suppression parameters can include the gain coefficient and time constant of the additional damping controller.

[0045] When the target wind turbine employs a notch filter strategy, the vibration suppression model can be a notch filter model. This notch filter model is typically connected in series in the controller's measurement signal loop or command output loop to filter out signal components within a specific frequency range, preventing the controller from generating excitation near the resonant frequency. In this case, the target vibration suppression parameters can include the center frequency and notch depth of the notch filter.

[0046] Step S50: Suppress the vibration of the target fan based on the vibration suppression model after the vibration suppression parameters are tuned.

[0047] It should be noted that the vibration suppression model based on the vibration suppression parameters suppresses the vibration of the target wind turbine. In essence, it adjusts the wind turbine's operating parameters (such as pitch angle and generator torque) through additional control commands output by the vibration suppression model to generate a damping force or torque opposite to the current vibration direction, thereby reducing the vibration amplitude of the key components of the target wind turbine.

[0048] Based on the above, the technical solution provided in this embodiment employs a model-independent meta-learning algorithm. It uses historical operating data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels to train a meta-learning model. This meta-learning model internalizes the mapping relationship between the same model of wind turbine under different operating conditions and structural states and the optimal vibration suppression parameters. Therefore, during the initial operating cycle after the target wind turbine is put into operation, only its initial operating data and initial vibration response data need to be obtained, and the initial structural dynamic feature vector can be extracted from them. By inputting the initial operating data and initial structural dynamic feature vector into the meta-learning model, the target vibration suppression parameters applicable to the target wind turbine can be directly obtained. The parameters of the vibration suppression model of the target wind turbine are then tuned to these target vibration suppression parameters, and the vibration of the target wind turbine is suppressed based on the tuned vibration suppression model. Therefore, the target wind turbine does not need to undergo a parameter search and trial-and-error process starting from scratch and lasting for weeks or even months. The vibration suppression parameters can be tuned and an effective vibration suppression state established within the initial operating cycle, significantly shortening the commissioning cycle. Furthermore, the shortened commissioning cycle allows the target wind turbine to operate with parameter configurations that provide good vibration suppression earlier, reducing the additional fatigue loads caused by suboptimal parameters during the commissioning period. This enables effective suppression of wind turbine vibration while reducing maintenance costs and the risk of early failures.

[0049] In summary, the technical solution provided in this embodiment can shorten the debugging cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine.

[0050] Based on the first embodiment described above, a second embodiment of the vibration suppression method for the fan of this application is proposed. For the second embodiment, please refer to... Figure 2 After step S50, the vibration suppression method for the fan may further include steps S51 to S55: Step S51: Dynamically acquire real-time operating condition data and real-time vibration response data of the target wind turbine during operation; It should be noted that real-time operating condition data refers to the external conditions and control parameters describing the operating status of the target wind turbine, which are collected in real time during continuous operation. The content is the same as the initial operating condition data, but the data is collected throughout the entire service life of the wind turbine. Real-time vibration response data refers to the time-series signals reflecting the vibration status of key components of the target wind turbine, which are collected in real time during continuous operation. The acquisition location and signal type are the same as the initial vibration response data, but the data is collected throughout the entire service life of the wind turbine.

[0051] When dynamically acquiring real-time operating condition data and real-time vibration response data during the operation of the target wind turbine, the acquisition can be done in real time or periodically at certain time intervals. This embodiment does not impose any specific limitations on this.

[0052] Step S52: Extract the real-time structural dynamics feature vector of the target wind turbine from the real-time vibration response data; It should be noted that the real-time structural dynamics feature vector is the feature vector extracted from the real-time vibration response data that reflects the structural dynamic characteristics of the target wind turbine. Its composition dimension is the same as that of the initial structural dynamics feature vector, but its value may drift due to the structural degradation of the target wind turbine.

[0053] Step S53: Determine whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector; It should be noted that feature drift refers to the deviation of the real-time structural dynamics feature vector from the initial structural dynamics feature vector, exceeding a preset allowable range. This indicates that the structural dynamic characteristics of the target wind turbine have undergone significant changes. Feature drift is typically caused by structural degradation factors such as gearbox wear, blade erosion, and foundation settlement.

[0054] In one feasible implementation, step S53 may include steps S531 to S532: Step S531: Determine the drift index value between the real-time structural dynamics eigenvector and the initial structural dynamics eigenvector; Step S532: If the drift index value is greater than the preset drift index threshold, it is determined that there is a feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector.

[0055] The drift index is a numerical indicator used to quantify the degree of difference between the real-time structural dynamics eigenvector and the initial structural dynamics eigenvector. The preset drift index threshold is a pre-set critical value for the drift index, which is used to distinguish between "normal fluctuations" and "feature drift". It can be a default value or can be flexibly set by the user according to the actual situation. This embodiment does not make specific limitations on this.

[0056] The Euclidean distance between the real-time structural dynamics feature vector and the initial structural dynamics feature vector can be calculated (the specific calculation process can be referred to Formula 1 below) as the drift index value; alternatively, the Modal Assurance Criterion (MAC) can be calculated first between the real-time structural dynamics feature vector and the initial structural dynamics feature vector to obtain a similarity value; then the difference between 1 and the similarity value is used as the drift index value. This embodiment does not impose specific limitations on this.

[0057] Formula 1; Where D is the Euclidean distance between the real-time structural dynamics feature vector and the initial structural dynamics feature vector, x_(t,i) is the i-th dimension value of the real-time structural dynamics feature vector, and x_(0,i) is the i-th dimension value of the initial structural dynamics feature vector.

[0058] In this embodiment, when determining whether feature drift exists, the drift index value between the real-time structural dynamics feature vector and the initial structural dynamics feature vector is first calculated to quantify the abstract difference into a comparable value. Then, this drift index value is compared with a preset drift index threshold. When the drift index value is greater than the preset drift index threshold, feature drift is determined to exist. Therefore, the judgment of feature drift no longer relies on human experience or qualitative observation, but is based on the comparison result of a quantifiable numerical index and an objective threshold, thereby effectively improving the accuracy of drift judgment and the consistency of the judgment results.

[0059] This embodiment does not specifically limit the implementation of step S53. For example, in other feasible implementations, to avoid misjudgment caused by transient interference, it can be limited that the drift index value between each real-time structural dynamics feature vector extracted multiple times and the initial structural dynamics feature vector must all be greater than a preset drift index threshold before feature drift is determined to exist.

[0060] Step S54: If so, input the real-time operating data and the real-time structural dynamics feature vector into the meta-learning model to obtain the updated vibration suppression parameters corresponding to the target wind turbine. It should be noted that updating the vibration suppression parameters refers to the vibration suppression parameter configuration output by the meta-learning model based on the current operating conditions and structural state of the target wind turbine, which is applicable to updating the vibration suppression model.

[0061] Step S55: Tune the parameters of the vibration suppression model to update the vibration suppression parameters.

[0062] In this embodiment, the target wind turbine continuously acquires real-time operating condition data and real-time vibration response data during operation, and extracts real-time structural dynamics feature vectors from the real-time vibration response data. By determining whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector, timely detection is achieved when the structural dynamic characteristics of the target wind turbine change. When feature drift is determined to exist, the real-time operating condition data and the real-time structural dynamics feature vector are input into the meta-learning model to obtain updated vibration suppression parameters, and the parameters of the vibration suppression model are tuned to these updated vibration suppression parameters. Therefore, the parameters of the vibration suppression model can be adjusted accordingly to follow the actual changes in the structural dynamic characteristics of the target wind turbine, ensuring that the vibration suppression parameters remain adapted to the actual structural state of the wind turbine at different operating stages, thereby guaranteeing the continuous and stable vibration suppression effect.

[0063] Based on the first and / or second embodiments described above, a third embodiment of the vibration suppression method for wind turbines of this application is proposed. In this third embodiment, please refer to... Figure 3 After step S50, the vibration suppression method for the fan may further include steps S60 to S80: Step S60: Obtain the actual vibration amplitude of the target wind turbine after the vibration suppression model has been run for a preset verification time with the target vibration suppression parameters; It should be noted that the preset verification duration is the pre-set time for verifying the vibration suppression effect after the target fan is running with the target vibration suppression parameters. It can be a default value or can be flexibly set by the user according to the actual situation. This embodiment does not impose specific limitations on it. The actual vibration amplitude is the amplitude statistics of the vibration signal collected from the key detection points of the target fan. It is used to quantify the vibration intensity and can be the root mean square value or peak value of the vibration acceleration, etc.

[0064] When obtaining the actual vibration amplitude of the target wind turbine after running the vibration suppression model with the target vibration suppression parameters for a preset verification time, the vibration acceleration signal of the nacelle in the forward and backward directions of the target wind turbine during the preset verification time of running the vibration suppression model with the target vibration suppression parameters can be obtained first, and then the root mean square value or peak value of the vibration acceleration signal can be used as the actual vibration amplitude.

[0065] Step S70: If the actual vibration amplitude is greater than the preset amplitude threshold, then take the target vibration suppression parameter as the initial search point and perform an iterative search within the parameter adjustment range corresponding to the target vibration suppression parameter to obtain the fine-tuned target vibration suppression parameter. It should be noted that the preset amplitude threshold is the upper limit of the pre-set vibration amplitude, used to determine whether the vibration suppression effect meets expectations. It can be a default value or flexibly set by the user according to actual conditions; this embodiment does not impose specific limitations on it. The initial search point is the starting point of the iterative search. The parameter adjustment range corresponding to the target vibration suppression parameter is the parameter search neighborhood range defined with the target vibration suppression parameter as the center, which limits the spatial boundary of the fine-tuning search. Iterative search is the process of trying different parameter combinations sequentially within the parameter adjustment range corresponding to the target vibration suppression parameter according to a preset search strategy, gradually approaching a better parameter by comparing the vibration response performance. The fine-tuned target vibration suppression parameter is the updated parameter obtained after iterative search, which has a better vibration suppression effect than the original target vibration suppression parameter.

[0066] In one feasible implementation, step S70 may include: within the parameter adjustment range corresponding to the target vibration suppression parameter, setting a search step size and number of search points for each parameter to be adjusted in the target vibration suppression parameter, with the target vibration suppression parameter as the center. For example, for the gain coefficient Kp, the search range is set to [Kp0×(1-α), Kp0×(1+α)], where Kp0 is the original gain value in the target vibration suppression parameter, and α is the adjustment ratio coefficient (e.g., 0.2). Within this range, M search points are selected at equal intervals. For the time constant, the search range is defined in the same way, and N search points are selected to form an M×N parameter combination grid. Each parameter combination is loaded into the vibration suppression model in sequence and run for a preset test duration to obtain the corresponding actual vibration amplitude. The actual vibration amplitude under each parameter combination is compared, and the parameter combination with the smallest actual vibration amplitude is selected as the fine-tuned target vibration suppression parameter.

[0067] In another feasible implementation, step S70 may include: taking the target vibration suppression parameter as the initial point and the vibration amplitude as the objective function; in each iteration, applying a small perturbation to the current parameter, testing the actual vibration amplitude change under positive and negative perturbations respectively, and estimating the gradient direction of the objective function at the current point; updating the parameter along the gradient direction that causes the actual vibration amplitude to decrease by a preset step size to obtain a new parameter point. Repeating the above process until the actual vibration amplitude is less than or equal to a preset amplitude threshold or the number of iterations reaches a preset maximum number of iterations, the finally obtained parameter is used as the fine-tuned target vibration suppression parameter. Compared to grid search, this implementation has higher search efficiency and is suitable for scenarios with high parameter dimensions or a large search range.

[0068] Step S80: Update the parameters of the vibration suppression model to the fine-tuned target vibration suppression parameters.

[0069] In this embodiment, after the target wind turbine has run for a preset verification period using the target vibration suppression parameters output by the meta-learning model, its actual vibration amplitude is obtained and compared with a preset amplitude threshold. When the actual vibration amplitude is greater than the preset amplitude threshold, it indicates that the vibration suppression effect of the target vibration suppression parameters in actual operation has not met expectations. At this time, using the target vibration suppression parameters as the initial search point, an iterative search is performed within the parameter adjustment range corresponding to the target vibration suppression parameters to obtain finely adjusted target vibration suppression parameters with lower actual vibration amplitudes. The parameters of the vibration suppression model are then updated to these finely adjusted target vibration suppression parameters. Thus, the vibration suppression effect is further guaranteed after the one-time inference result output by the meta-learning model is verified by actual operation and optimized by local fine-tuning, avoiding the situation where the vibration suppression parameters are inapplicable due to inference bias of the meta-learning model or individual differences of the target wind turbine.

[0070] Based on the first, second, and / or third embodiments described above, a fourth embodiment of the vibration suppression method for wind turbines of this application is proposed. In this fourth embodiment, please refer to... Figure 4 After step S50, the vibration suppression method for the fan may further include steps S01 to S02: Step S01: Obtain the current operating condition data of the target wind turbine and determine whether the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model. It should be noted that the current operating condition data refers to the external conditions and control state parameters collected in real time at the current operating moment of the target wind turbine, used to describe the current operating state of the target wind turbine. The parameter types are the same as those in the initial operating condition data. The operating condition distribution range refers to the operating condition space covered by the historical operating condition data of multiple wind turbines of the same model as the target wind turbine used during the training phase of the meta-learning model. It represents the boundary of the operating conditions within which the meta-learning model has reliable inference capabilities. If the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model, that is, if the current operating condition data falls into an operating condition region not covered by the training samples of the meta-learning model, it indicates that the parameter inference of the meta-learning model under this operating condition may have significant uncertainty.

[0071] In one feasible implementation, step S01 may include: calculating the Mahalanobis distance between the current working condition data and the working condition distribution center covered during the training of the meta-learning model; if the Mahalanobis distance is greater than a preset distance threshold, it is determined that the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model.

[0072] The operating condition distribution center is the mean vector of the historical operating condition data sample set of multiple wind turbines of the same model as the target wind turbine used in the meta-learning model training. It represents the central location of the training operating condition space. The Mahalanobis distance between the current operating condition data and the operating condition distribution center is used to quantify the degree of deviation between the current operating condition data and the operating condition distribution center.

[0073] When calculating the Mahalanobis distance between the current operating condition data and the center of the operating condition distribution covered during the training of the meta-learning model, we can first obtain and summarize the historical operating condition data of multiple wind turbines of the same model as the target wind turbine used during the training of the meta-learning model to obtain the operating condition data sample set; then calculate the mean vector μ_cond and covariance matrix S_cond of the operating condition data sample set; then use the following formula 2 to calculate the Mahalanobis distance D_cond between the current operating condition data x_cond and the center of the operating condition distribution.

[0074] Formula 2; Where T is the matrix transpose operator.

[0075] In this embodiment, when determining whether the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model, the Mahalanobis distance between the current operating condition data and the center of the training operating condition distribution is calculated, and then compared with a preset distance threshold. The Mahalanobis distance considers the correlation and dimensional differences between various operating condition parameters. Compared to a simple judgment method based on the boundary values ​​of each parameter, it can more accurately reflect the overall deviation of the current operating condition from the training operating condition space. When the Mahalanobis distance is greater than the preset distance threshold, it is determined that the current operating condition data exceeds the operating condition distribution range. This determination serves as an objective basis for triggering the switching of safety vibration suppression parameters, effectively avoiding misjudgments caused by occasional out-of-bounds exceedances of single parameters or normal fluctuations in operating conditions, thus improving the accuracy of extreme operating condition identification and the reliability of the safety switching mechanism.

[0076] This embodiment does not specifically limit the implementation of step S01. For example, in other feasible implementations, the value range of each working condition parameter in the historical working condition data used during the training of the meta-learning model can also be statistically analyzed. Thus, it is possible to check one by one whether each working condition parameter in the current working condition data falls within the corresponding value range. If there is a working condition parameter in the current working condition data whose value is outside the corresponding value range, it is determined that the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model.

[0077] Step S02, if yes, then the parameters of the vibration suppression model are tuned to the preset safe vibration suppression parameters. The safe vibration suppression parameters refer to the vibration suppression parameters that are tuned offline for a fan of the same model as the target fan under operating conditions that exceed the operating condition distribution range.

[0078] It should be noted that offline tuning is the parameter configuration process determined by manual debugging or offline optimization algorithms based on the operating data or simulation data of the same model of wind turbine under the same or similar extreme operating conditions before the target wind turbine is actually put into operation.

[0079] In one feasible implementation, the process for determining the safety vibration suppression parameters may include: collecting operational data from multiple wind turbines of the same model as the target turbine during their historical operation when they encountered extreme operating conditions exceeding the range of operating conditions. For example, this may include vibration response data from high wind speed sections, strong turbulence sections, or power grid fault crossing sections. Based on this data, offline optimization methods such as genetic algorithms or particle swarm optimization can be used to tune a set of safety vibration suppression parameters with the joint optimization objectives of minimizing vibration amplitude and ensuring controller output stability.

[0080] In another feasible implementation, the process for determining the safe vibration suppression parameters may include: constructing a simulation model of a wind turbine of the same model as the target wind turbine using wind turbine dynamics simulation software; setting extreme operating conditions beyond the range of operating conditions in the simulation environment, such as setting the wind speed to exceed 120% of the rated wind speed, setting the turbulence intensity to reach the upper limit of Class A turbulence level, and setting a sudden drop in grid voltage; and running an offline optimization algorithm in the simulation environment, using the constraint that the vibration amplitude does not exceed the safety limit, to tune and obtain the safe vibration suppression parameters.

[0081] In this embodiment, during the operation of the vibration suppression model of the target wind turbine after the vibration suppression parameters are tuned, the current operating condition data is acquired in real time, and it is determined whether the current operating condition data exceeds the operating condition distribution range covered by the meta-learning model during training. When it is determined that the current operating condition data exceeds the distribution range, the parameters of the vibration suppression model are tuned to preset safe vibration suppression parameters. These safe vibration suppression parameters are vibration suppression parameters that were tuned offline for wind turbines of the same model as the target wind turbine under operating conditions exceeding the distribution range. Therefore, when the target wind turbine encounters extreme operating conditions that the meta-learning model has not been trained on, it can automatically switch to the pre-prepared conservative safe parameters, avoiding a decrease in vibration suppression effect or an increase in vibration due to inference bias of the meta-learning model under such operating conditions, thereby ensuring the operational safety and reliability of the wind turbine under extreme operating conditions.

[0082] This application also provides a vibration suppression device for a fan, please refer to... Figure 5 The vibration suppression device for the fan may include: The data acquisition module 10 is used to acquire the initial operating condition data and initial vibration response data of the target wind turbine during the initial operating cycle after the target wind turbine is put into operation. Feature extraction module 20 is used to extract the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data, as the initial structural dynamic feature vector; The parameter tuning module 30 is used to input the initial operating condition data and the initial structural dynamic feature vector into the pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine. The meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels. The parameters of the vibration suppression model of the target wind turbine are tuned to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned. The vibration suppression module 40 is used to suppress the vibration of the target fan based on the vibration suppression model after the vibration suppression parameters are tuned.

[0083] In one embodiment, the feature extraction module 20 is further configured to: The initial vibration response data is processed by time-frequency analysis or modal parameter identification to extract at least one natural frequency of the target wind turbine in the initial operating cycle and the modal damping ratio corresponding to the natural frequency, thus forming the structural dynamic characteristic vector of the target wind turbine in the initial operating cycle.

[0084] In one embodiment, the parameter tuning module 30 is further configured to: Dynamically acquire real-time operating condition data and real-time vibration response data of the target wind turbine during operation; Extract the real-time structural dynamics feature vector of the target wind turbine from the real-time vibration response data; Determine whether there is feature drift between the real-time structural dynamics eigenvectors and the initial structural dynamics eigenvectors; If so, the real-time operating data and real-time structural dynamics feature vectors are input into the meta-learning model to obtain the updated vibration suppression parameters corresponding to the target wind turbine; The parameters of the vibration suppression model are tuned to update the vibration suppression parameters.

[0085] In one embodiment, the parameter tuning module 30 is further configured to: Determine the drift index value between the real-time structural dynamics eigenvector and the initial structural dynamics eigenvector; If the drift index value is greater than the preset drift index threshold, it is determined that there is a feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector.

[0086] In one embodiment, the parameter tuning module 30 is further configured to: Obtain the actual vibration amplitude of the target wind turbine after the vibration suppression model is run with the target vibration suppression parameters for a preset verification time; If the actual vibration amplitude is greater than the preset amplitude threshold, the target vibration suppression parameter is used as the initial search point, and an iterative search is performed within the parameter adjustment range corresponding to the target vibration suppression parameter to obtain the fine-tuned target vibration suppression parameter. Update the parameters of the vibration suppression model to the fine-tuned target vibration suppression parameters.

[0087] In one embodiment, the parameter tuning module 30 is further configured to: Obtain the current operating condition data of the target wind turbine and determine whether the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model; If so, the parameters of the vibration suppression model will be tuned to the preset safe vibration suppression parameters. The safe vibration suppression parameters refer to the vibration suppression parameters that have been tuned offline for a fan of the same model as the target fan under operating conditions that exceed the operating condition distribution range.

[0088] In one embodiment, the parameter tuning module 30 is further configured to: Calculate the Mahalanobis distance between the current working condition data and the working condition distribution center covered during the training of the meta-learning model; If the Mahalanobis distance is greater than the preset distance threshold, it is determined that the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model.

[0089] The vibration suppression device for wind turbines provided in this application can shorten the debugging cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine. Compared with the prior art, the beneficial effects of the vibration suppression device for wind turbines provided in this application are the same as those of the vibration suppression method for wind turbines provided in the above embodiments, and other technical features in this vibration suppression device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0090] This application also provides an electronic device, which may include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the vibration suppression method for the fan in the above embodiments.

[0091] The following is for reference. Figure 6 It shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of this application. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0092] like Figure 6As shown, the electronic device may include a processing unit 101 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory 102 or a program loaded from storage device 103 into random access memory 104. Random access memory 104 also stores various programs and data required for the operation of the electronic device. The processing unit 101, read-only memory 102, and random access memory 104 are interconnected via bus 105. Input / output interface 106 is also connected to bus 105. Typically, the following systems can be connected to input / output interface 106: input devices 107 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 108 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 103 including, for example, magnetic tape, hard disks, etc.; and communication devices 109. Communication device 109 allows the electronic device to communicate wirelessly or wiredly with other devices to exchange data. Although the diagrams show electronic devices with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented alternatively.

[0093] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 103, or installed from read-only memory 102. When the computer program is executed by processing device 101, it performs the functions defined in the methods of the embodiments of this application.

[0094] The electronic device provided in this application, employing the vibration suppression method for wind turbines described in the above embodiments, can shorten the debugging cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the vibration suppression method for wind turbines provided in the above embodiments, and other technical features of the electronic device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0095] It should be understood that various parts of the embodiments of this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0096] The above description is merely a specific implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of the embodiments of this application. Therefore, the protection scope of the embodiments of this application should be determined by the protection scope of the above claims.

[0097] This application also provides a computer-readable storage medium storing a computer program that can run on a processor, the computer program being used to execute the fan vibration suppression method in the above embodiments.

[0098] The computer-readable storage medium provided in this application embodiment may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0099] The aforementioned computer-readable storage medium may be included in an electronic device or may exist independently without being assembled into an electronic device.

[0100] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by an electronic device, the electronic device causes the following actions: During the initial operating cycle after the target wind turbine is put into operation, it acquires initial operating condition data and initial vibration response data of the target wind turbine; extracts the structural dynamics feature vector of the target wind turbine during the initial operating cycle from the initial vibration response data, using it as the initial structural dynamics feature vector; inputs the initial operating condition data and the initial structural dynamics feature vector into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine; wherein the meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamics feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels; tunes the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned; and suppresses the vibration of the target wind turbine based on the vibration suppression model after the vibration suppression parameters are tuned.

[0101] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0102] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0103] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0104] The computer-readable storage medium provided in this application embodiment stores computer-readable program instructions for executing the vibration suppression method of the above-described wind turbine. This shortens the debugging cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application embodiment are the same as those of the vibration suppression method of the wind turbine provided in the above-described embodiments, and will not be repeated here.

[0105] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the vibration suppression method for a fan as described above.

[0106] The computer program product provided in this application can shorten the debugging cycle of vibration suppression parameters, thereby effectively suppressing wind turbine vibration while reducing the operation and maintenance costs and early failure risks of the wind turbine. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the wind turbine vibration suppression method provided in the above embodiments, and will not be repeated here.

[0107] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A method for suppressing vibration of a fan, characterized in that, The method includes: During the initial operating cycle after the target wind turbine is put into operation, the initial operating condition data and initial vibration response data of the target wind turbine are acquired; The structural dynamic feature vector of the target wind turbine in the initial operating cycle is extracted from the initial vibration response data and used as the initial structural dynamic feature vector. The initial operating condition data and the initial structural dynamic feature vector are input into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine. The meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples and vibration suppression parameters corresponding to the input samples as output labels. The parameters of the vibration suppression model of the target wind turbine are tuned to the target vibration suppression parameters to obtain the vibration suppression model after the vibration suppression parameters are tuned. The vibration suppression model, after tuning the vibration suppression parameters, suppresses the vibration of the target wind turbine.

2. The method as described in claim 1, characterized in that, The step of extracting the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data includes: The initial vibration response data is subjected to time-frequency analysis or modal parameter identification to extract at least one natural frequency of the target wind turbine during the initial operating cycle and the modal damping ratio corresponding to the natural frequency, thereby forming the structural dynamic feature vector of the target wind turbine during the initial operating cycle.

3. The method according to claim 1, characterized in that, After the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned by the vibration suppression parameters, the method further includes: Dynamically acquire real-time operating condition data and real-time vibration response data of the target wind turbine during its operation; Extract the real-time structural dynamics feature vector of the target wind turbine from the real-time vibration response data; Determine whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector; If so, the real-time operating data and the real-time structural dynamics feature vector are input into the meta-learning model to obtain the updated vibration suppression parameters corresponding to the target wind turbine; The parameters of the vibration suppression model are tuned to the updated vibration suppression parameters.

4. The method according to claim 3, characterized in that, The step of determining whether there is feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector includes: Determine the drift index value between the real-time structural dynamics feature vector and the initial structural dynamics feature vector; If the drift index value is greater than the preset drift index threshold, then it is determined that there is a feature drift between the real-time structural dynamics feature vector and the initial structural dynamics feature vector.

5. The method according to claim 1, characterized in that, After the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned by the vibration suppression parameters, the method further includes: Obtain the actual vibration amplitude of the target wind turbine after the vibration suppression model has been run for a preset verification time with the target vibration suppression parameters; If the actual vibration amplitude is greater than the preset amplitude threshold, then the target vibration suppression parameter is used as the initial search point, and an iterative search is performed within the parameter adjustment range corresponding to the target vibration suppression parameter to obtain the fine-tuned target vibration suppression parameter. The parameters of the vibration suppression model are updated to the fine-tuned target vibration suppression parameters.

6. The method according to claim 1, characterized in that, After the step of suppressing the vibration of the target wind turbine based on the vibration suppression model tuned by the vibration suppression parameters, the method further includes: Obtain the current operating condition data of the target wind turbine and determine whether the current operating condition data exceeds the operating condition distribution range covered during the training of the meta-learning model; If so, the parameters of the vibration suppression model are tuned to preset safe vibration suppression parameters. The safe vibration suppression parameters refer to the vibration suppression parameters that are tuned offline for a wind turbine of the same model as the target wind turbine under operating conditions that exceed the operating condition distribution range.

7. The method as described in claim 6, characterized in that, The step of determining whether the current working condition data exceeds the range of working condition distributions covered during the training of the meta-learning model includes: Calculate the Mahalanobis distance between the current working condition data and the working condition distribution center covered during the training of the meta-learning model; If the Mahalanobis distance is greater than a preset distance threshold, it is determined that the current working condition data exceeds the working condition distribution range covered during the training of the meta-learning model.

8. A vibration suppression device for a fan, characterized in that, The device includes: The data acquisition module is used to acquire the initial operating condition data and initial vibration response data of the target wind turbine during the initial operating cycle after the target wind turbine is put into operation. The feature extraction module is used to extract the structural dynamic feature vector of the target wind turbine in the initial operating cycle from the initial vibration response data, and use it as the initial structural dynamic feature vector; The parameter tuning module is used to input the initial operating condition data and the initial structural dynamic feature vector into a pre-trained meta-learning model to obtain the target vibration suppression parameters corresponding to the target wind turbine. The meta-learning model is pre-trained using a model-independent meta-learning algorithm, with historical operating condition data and historical structural dynamic feature vectors of multiple wind turbines of the same model as the target wind turbine as input samples, and vibration suppression parameters corresponding to the input samples as output labels. The module tunes the parameters of the vibration suppression model of the target wind turbine to the target vibration suppression parameters to obtain the vibration suppression model after parameter tuning. The vibration suppression module is used to suppress the vibration of the target fan based on the vibration suppression model after the vibration suppression parameters are tuned.

9. An electronic device, characterized in that, The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the vibration suppression method for a wind turbine as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the vibration suppression method for a wind turbine as described in any one of claims 1 to 7.

Images