Abnormal prediction method and device for wind power generation unit, and storage medium

Abstract

The present invention provides a method and apparatus for predicting abnormalities in a wind power generation unit, and a storage medium. 【Solution means】 The abnormality prediction method includes steps of acquiring terrain data and meteorological data of the mechanical position of the wind power generation unit, determining wind parameter sequence data by large eddy simulation based on the terrain data and the meteorological data, determining unit load data based on the wind parameter sequence data, and predicting whether an abnormality has occurred in the wind power generation unit based on the unit load data. The present disclosure can achieve reliable abnormality prediction.

Description

Technical Field

[0001] The present disclosure relates to the field of fans, and more specifically, to a method and apparatus for predicting abnormalities of a wind power generation unit, and a storage medium.

Background Art

[0002] Wind parameters have an important influence on the operating state of a wind power generation unit. Strong winds, high wind speeds and wind shear, strong updrafts and unstable turbulence caused by extreme wind conditions pose a significant threat to the safe operation of the wind power generation unit, and are a major challenge to the robustness and load capacity of the unit. Abnormal loads and abnormal vibrations of the wind power generation unit lead to the ultimate load of the unit. If left unaddressed, this can lead to operating failures such as fatigue loads of the unit, and further accidents such as blade and tower sweeps, and tower collapses, causing significant economic losses and even casualties. Therefore, predicting the operating state of a wind power generation unit under different weather conditions for immediate response has important safety and economic significance.

[0003] In related technologies, the analysis of unit abnormalities and failures generally relies on investigation information of component damage sites, operating data of the unit, design parameters of components, etc. Currently, there is little research on unit abnormalities caused by wind parameters, and the prediction and early warning capabilities for fan abnormalities under extreme wind conditions are very limited.

Summary of the Invention

Problems to be Solved by the Invention

[0004] Therefore, how to accurately simulate wind parameters and predict unit abnormalities based on them is extremely important for the reliable prediction of unit abnormalities caused by extreme weather.

Means for Solving the Problems

[0005] Generally, a method for predicting abnormalities in a wind power generation unit is provided, including the steps of obtaining topographical data and meteorological data of the mechanical position of the wind power generation unit, determining wind parameter sequence data by large eddy simulation based on the topographical data and the meteorological data, determining unit load data based on the wind parameter sequence data, and predicting whether an abnormality has occurred in the wind power generation unit based on the unit load data.

[0006] Generally, an apparatus for predicting abnormalities in a wind power generation unit is provided, including an acquisition means arranged to obtain topographical data and meteorological data of the mechanical position of the wind power generation unit, a calculation means arranged to determine wind parameter sequence data by large eddy simulation based on the topographical data and the meteorological data, the calculation means further including a calculation means arranged to determine unit load data based on the wind parameter sequence data, and a prediction means arranged to predict whether an abnormality has occurred in the wind power generation unit based on the unit load data.

[0007] Generally, a computer-readable storage medium is provided, and when instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is caused to execute the abnormality prediction method described above.

[0008] Generally, a computer device is provided, including at least one processor and at least one memory storing computer-executable instructions, and when the computer-executable instructions are executed by the at least one processor, the at least one processor is caused to execute the abnormality prediction method described above.

Advantages of the Invention

[0009] In the present disclosure, using a large eddy simulation method capable of simulating turbulent flows with different scales, topographic data and meteorological data are combined to simulate accurate and detailed wind parameter sequence data, determine accurate and reliable unit load data, identify the root causes of unit abnormalities, and achieve reliable abnormality prediction.

[0010] Here, the above general description and the following detailed description do not limit the present disclosure, but are merely exemplary and explanatory.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Modes for Carrying Out the Invention

[0012] By providing the following specific embodiments, readers can obtain a comprehensive understanding of the methods, devices, and / or systems described herein. It should be noted that various modifications, corrections, and equivalents of the methods, devices, and / or systems described herein are clear upon understanding the disclosure of this application. For example, the operation sequences described herein are merely exemplary and are not limited to the sequences described herein. It is clear upon understanding the disclosure of this application that, except for operations that must occur in a specific order, they can be changed. Also, for the sake of clarity and conciseness, descriptions of known features in the art may be omitted.

[0013] The features described herein are not limited to the examples described in this specification and may be implemented in different forms. Conversely, the examples described herein show only some of the possible forms among many possible forms for implementing the methods, devices, and / or systems described herein, and when the disclosure of this application is understood, said many possible forms will become clear.

[0014] For example, the term "and / or" as used herein includes any one, two, or any combination of multiple of the related items.

[0015] Here, terms such as "first", "second", and "third" can be used to describe each component, component, region, layer, or part, but these components, components, regions, layers, or parts are not limited to these terms. Conversely, these terms are merely used to distinguish one component, component, region, layer, or part from another component, component, region, layer, or part. Therefore, without departing from the teachings of the examples, the first component, the first component, the first region, the first layer, or the first part in the examples described herein may be referred to as the second component, the second component, the second region, the second layer, or the second part.

[0016] In the specification, when an element (such as a layer, region, or substrate) is described as "located" on, "connected" to, or "coupled" to another element, the element is "located" directly on, "connected" directly to, or "coupled" directly to the other element, or one or more other elements may exist in between. Conversely, when an element is described as "directly located" on, "directly connected" to, or "directly coupled" to another element, no other element may exist in between.

[0017] The terms used herein do not limit the disclosure but are merely used to describe each example. Unless the context clearly indicates otherwise, the singular form includes the plural form. The terms "comprising," "including," and "having" indicate the presence of the described features, numbers, operations, components, elements, and / or combinations thereof, but do not exclude the presence or addition of one or more other features, numbers, operations, components, elements, and / or combinations thereof.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Unless explicitly defined herein, terms (such as those defined in a general dictionary) should not be interpreted ideally or overly formally, but rather should be interpreted to have a meaning consistent with the meaning in the context of the relevant art and this disclosure.

[0019] Also, in the exemplary description, if a detailed description of a known related structure or function would lead to an ambiguous explanation of this disclosure, such a detailed description is omitted.

[0020] FIG. 1 is a flowchart of a method for predicting anomalies of a wind power generation unit according to an embodiment of this disclosure.

[0021] Referring to FIG. 1, in step S101, topographic data and meteorological data of the mechanical position of the wind power generation unit are acquired. By acquiring the topographic data and meteorological data, in subsequent steps, it is possible to perform a simulation of the airflow flow in the space where the mechanical position is located.

[0022] Preferably, the spatial resolution of the terrain data and the meteorological data is equal to or higher than a predetermined spatial resolution, and the temporal resolution of the meteorological data is equal to or higher than a predetermined temporal resolution. The spatial resolution is reflected by the minimum distance between two adjacent spatial points where the corresponding data (i.e., terrain data or meteorological data) can be displayed. The smaller the value, the higher the spatial resolution. Similarly, the temporal resolution is reflected by the minimum time interval between two adjacent time points where the corresponding data can be displayed. The smaller the value, the higher the temporal resolution. The minimum distance corresponding to the spatial resolution of the terrain data and the meteorological data should be less than or equal to the minimum distance corresponding to the predetermined spatial resolution, and the minimum time interval corresponding to the temporal resolution of the meteorological data should be less than or equal to the minimum time interval corresponding to the predetermined temporal resolution, thereby realizing a high spatio-temporal resolution. By obtaining terrain data and meteorological data with a high spatio-temporal resolution, high-precision basic data can be provided for subsequent simulations, high-precision simulation results can be obtained, and the accuracy of anomaly prediction can be improved. Here, the period targeted for simulation is short, generally only a few hours or a few days. During this period, the characteristics of the underlying surface are likely to remain unchanged. Therefore, only static high-spatial-resolution terrain data needs to be obtained, that is, the terrain data is not required to have a high temporal resolution. For meteorological data, in order to reflect meteorological changes, it is necessary to obtain dynamic high-spatio-temporal-resolution meteorological data.

[0023] As an example, the terrain data is elevation terrain data of SRTM (Shuttle Radar Topography Mission), for example, SRTM3 data with a resolution of 90 meters, and is used as terrain information of the ground surface. The weather data is obtained by combining ground weather observation data and weather forecast data. The former is actual observation data, and the latter is forecast data. The weather forecast data may use GFS (Global Forecast System) data or FNL (Final Reanalysis Data). FNL is a complete reanalysis data set obtained after performing quality control and assimilation processing on observation data from various sources (such as the ground, ships, radiosondes, rawinsondes, aircraft, satellites, etc.). It can reflect the actual atmospheric situation more accurately than GFS data, but is more delayed than GFS data. In practice, it may be selected according to different needs. As an example, when predicting extreme weather, it is necessary to predict abnormal predictions as quickly as possible, so GFS data is used. When reproducing extreme weather that has occurred in various places, FNL reanalysis data is used.

[0024] Preferably, step S101 further includes the following steps: First, based on the mechanical position of the wind power generation unit, determine the space to be simulated. This step determines the obvious space for the simulation and clarifies the spatial range of the terrain data and meteorological data to be acquired. To ensure the simulation effect, the height of the space to be simulated generally needs to exceed the troposphere and reach, for example, the stratosphere. Specifically, with reference to atmospheric pressure, extend the space to be simulated upward so that the atmospheric pressure reaches 10 hPa (hectopascal) to 50 hPa. The coverage range of the space to be simulated in the horizontal plane is determined based on the mechanical position of the wind power generation unit to be predicted. As an example, the mechanical position of the wind power generation unit is the coordinates of the wind power generation unit, indicated by longitude and latitude. When determining the coverage range of the space to be simulated in the horizontal plane, first, confirm the wind power generation units to be predicted and their mechanical positions, and based on the criterion of covering all the wind power generation units to be predicted, determine the horizontal range of the space to be simulated. To facilitate the calculation, generally select a rectangular or square area for the horizontal range of the space to be simulated, and to meet different scale prediction needs, the side length level of the area reaches several kilometers to hundreds of kilometers.

[0025] In addition, for the space to be simulated, acquire terrain data with a spatial resolution equal to or higher than a predetermined spatial resolution. The minimum distance corresponding to the predetermined spatial resolution is, for example, 50 meters or 80 meters, and the terrain data is the above-mentioned SRTM3 data.

[0026] Finally, obtain the ground meteorological observation data and meteorological prediction data for the target period of the space to be simulated, and perform four-dimensional assimilation processing on the ground meteorological observation data according to the target time resolution and target spatial resolution. For example, use the Nudging four-dimensional assimilation technology to obtain meteorological data. Note that the target time resolution is equal to or higher than a predetermined time resolution, and the target spatial resolution is equal to or higher than a predetermined spatial resolution. The minimum time interval corresponding to the predetermined time resolution may be set according to needs. For example, it may be 1 minute, 3 minutes, or 5 minutes at the "minute" level, or 20 seconds or 30 seconds at the "second" level. By determining the meteorological data based on the ground meteorological observation data and meteorological prediction data actually observed within the target period, predictive simulations can be performed using the meteorological prediction data, and the observed data can be fused to improve the accuracy of the simulation. Through four-dimensional assimilation processing, the ground meteorological observation data is assimilated into a standardized three-dimensional mesh (four-dimensional in addition to the time dimension), which is convenient for subsequent simulation calculations.

[0027] Here, after determining the space to be simulated, the execution order of the following two steps, namely, obtaining terrain data and obtaining meteorological data, does not have to be limited, and the target spatial resolution may be equal to the spatial resolution of the terrain data.

[0028] In step S102, based on the terrain data and meteorological data, large eddy simulation is used to determine the wind parameter sequence data. In the fan field, the application of large eddy simulation mainly targets wind speed and wind power prediction, wake vortex analysis, blade optimization, etc. Such a simulation method is a spatial average of turbulent pulsations (or turbulent vortices), that is, a certain filter function is used to separate large-scale turbulence from small-scale turbulence, directly simulate the large-scale turbulence, block the small-scale turbulence by a model, and obtain a more detailed description of the turbulence through calculations with a finite number of meshes. The impeller surface of a wind power unit is greatly affected by local terrain, and small-scale turbulence has a significant impact on the load and vibration of the unit. Compared with the intermediate-scale simulation method that cannot simulate small-scale turbulence, large eddy simulation can reliably simulate different scales of turbulence, and further obtain wind parameter sequence data at different scales of turbulence, improve the accuracy of the wind parameter sequence data, reflect the influence of different scales of turbulence on the wind power unit, include small-scale turbulence within the simulation range, reduce the risk of underestimating small-scale turbulence, and improve the accuracy of anomaly prediction. The large eddy simulation executed by combining meteorological data is called WRF-LES (Weather Research and Forecasting Model-Large Eddy Simulation).

[0029] Here, the magnitude of the time resolution of the simulation output result by the conventional WRF is adjustable, and the minimum is the integration step size. For large eddy simulation, the output time resolution can reach 1 s.

[0030] Preferably, step S102 includes: constructing a simulation mesh for the space to be simulated, and increasing the number of meshes according to a predetermined rule within the target height range of the simulation mesh; and performing large-eddy simulation calculations on the simulation mesh based on terrain data and meteorological data to obtain wind parameter sequence data. Specifically, through simulation calculations, the wind parameter sequence data of each node in the simulation mesh is obtained. By increasing the number of meshes within the target height range, that is, increasing the mesh density within the target height range, i.e., improving the resolution, the wind parameter sequence data at multiple heights within the target height range is obtained, making the simulation results more detailed. As an example, the target height range is the height range where the impeller surface of the wind power generation unit is located, and the predetermined rule is the mesh size. This ensures that as much vertical wind field information as possible can be obtained within the impeller surface range, facilitating subsequent analysis of the load distribution situation on the impeller surface. For example, the height range of 20m to 200m can be used as the target height range, but it is not limited to this. For every 10m, one mesh is constructed, that is, nodes are constructed at heights such as 20m, 30m, 40m,..., 190m, 200m. For the height range above 200m, the default mesh size can be used. Here, within the target height range, the meshes in the horizontal plane can be appropriately refined, but the mesh size in the horizontal plane can also be larger than the mesh size in the height range. For example, in an embodiment where the mesh size in the above height range is 10m, to avoid excessive computational complexity, the mesh size in the horizontal plane can be 100m.

[0031] Here, the simulation mesh is partitioned into nested regions, where the mesh density (resolution) of the outer layer region is small and the mesh density (resolution) of the inner layer region is large. For example, four or five nested regions are partitioned, and from the outside to the inside, the mesh density of each nested region gradually increases, which can reduce the computational amount of non-critical regions of interest. However, based on the meteorological data obtained in step S101, the boundary conditions of each inner layer's nested region cannot be obtained, and only the initial conditions of each layer's nested region and the boundary conditions of the outermost layer's nested region can be determined. At this time, the Ndown technology is used to determine the boundary conditions of each inner layer's nested region from the simulation results of the outermost layer's nested region, and the initial conditions of each inner layer's nested region can be updated to ensure data consistency. Specifically, the Ndown technology first simulates the outermost layer's nested region (denoted as d01 for example), performs Ndown downscaling using the simulation results, generates the boundary conditions of the inner layer's nested region (denoted as d02 for example, and continues to denote the other inner nested regions in this order), updates the initial conditions of the d02 region based on the simulation results, then simulates the d02 region, performs Ndown downscaling again using the simulation results of the d02 region, generates the boundary conditions of the d03 region, updates the initial conditions of the d03 region, and continues until the simulation of all nested regions is completed.

[0032] Preferably, the wind parameter sequence data includes sequence data of target parameters at a plurality of predetermined heights within the impeller range, and the target parameters include at least one of horizontal wind speed, wind direction, vertical velocity, wind shear, and turbulence intensity. By obtaining the sequence data of the above target parameters, the wind conditions at the mechanical position of the wind power generation unit can be comprehensively reflected, and furthermore, the abnormal prediction of the wind power generation unit can be realized.

[0033] In step S103, unit load data is determined based on wind parameter sequence data. In extreme weather, abnormalities occur in the wind power generation unit. As the root cause, due to extreme wind conditions, the unit receives abnormal loads. Based on accurate and detailed wind parameter sequence data obtained through large eddy simulation, accurate and reliable unit load data is determined to clarify the root cause of unit abnormalities and achieve reliable abnormality prediction.

[0034] Preferably, step S103 includes the step of inputting wind parameter sequence data into a computational fluid dynamics model and a computational structural dynamics model to obtain unit load data. The computational fluid dynamics model (CFD) solves aerodynamic and hydrodynamic forces by the finite volume method of the unsteady Navier-Stokes equations based on wind parameter sequence data to initially obtain unit load data. The computational structural dynamics model (CSD) uses finite element numerical values to solve structural responses and obtains the structural response displacements of the wind power generation unit under the action of aerodynamic and hydrodynamic forces. Since the change in displacement affects the unit load, the structural response displacements are interpolated to the CFD mesh points to obtain the deformed mesh of CFD. The above process is continuously repeated, and the final unit load data is obtained with the iterative error of displacement or load as the convergence criterion for coupling.

[0035] Preferably, the unit load data includes at least one of blade load data and tower frame load data. In abnormal prediction, since the blade and the tower frame are the main structures to be focused on, the load data at these two locations are mainly simulated to achieve a highly targeted calculation. The blade load data includes at least one of a blade simulating bending moment and a blade rotating bending moment, and the tower frame load data includes at least one of a tower top pitch bending moment, a tower top overturning bending moment, a tower top torque, a tower bottom pitch bending moment, and a tower bottom overturning bending moment. Specifically, by obtaining the above load data, the load changes of the blade and the tower frame in the main abnormal situations can be effectively covered, providing a basis for reliable unit abnormal prediction.

[0036] In step S104, based on the unit load data, it is predicted whether an abnormality has occurred in the wind power generation unit. As an example, in addition to the unit load data, the wind parameter sequence data obtained in step S102 is added and jointly used for prediction, thereby increasing the basic data for prediction and improving the accuracy of prediction.

[0037] Preferably, step S104 includes a step of predicting whether an abnormality has occurred in the wind power generation unit by inputting the unit load data into a pre-trained abnormality prediction model. By using the pre-trained abnormality prediction model to predict abnormalities, during the pre-training process, the relationship between the unit load data and the abnormalities of the wind power generation unit is learned, and the prediction result can be quickly obtained in the prediction stage, ensuring the timeliness and accuracy of the prediction. As described above, the wind parameter sequence data may also be input into the pre-trained abnormality prediction model together.

[0038] Preferably, the pre-trained anomaly prediction model is obtained by training in the following steps: obtaining historical anomaly data of a reference wind power generation unit, corresponding reference terrain data and historical meteorological data for the historical anomaly data; determining historical wind parameter sequence data by large eddy simulation based on the reference terrain data and the historical meteorological data; determining historical unit load data based on the historical wind parameter sequence data; inputting the historical unit load data into an anomaly prediction model to be trained to obtain historical anomaly prediction data; and adjusting parameters of the anomaly prediction model to be trained based on the historical anomaly data and the historical anomaly prediction data to obtain a pre-trained anomaly prediction model. In short, the anomaly prediction model is trained by a supervised learning method, and the relationship between the unit load data and the anomalies of the wind power generation unit can be effectively learned, improving the learning efficiency. Here, the historical anomaly data is data recorded during the daily operation and maintenance of the wind power generation unit, and includes, for example, anomaly data such as mechanical system failure information, electrical system failure information, and control system failure information. Each piece of anomaly data includes, for example, the failure type and failure location of the unit, and further includes sensor data (such as data on tilt angle, acceleration, strain, and temperature), and data on the tilt, vibration, load, and deformation of the fan tower can be obtained. Therefore, model training can be realized using conventional data, improving the possibility of solutions. Correspondingly, the pre-trained anomaly prediction model outputs anomaly data in the same format as prediction anomaly data. If the output prediction anomaly data indicates that there is an anomaly in a certain part of the unit, the prediction result in step S104 is considered to be that an anomaly has occurred in the wind power generation unit, and the specific output prediction anomaly data further guides the operator to make a corresponding plan in advance. If no prediction anomaly data is output, or the output prediction anomaly data is empty, the prediction result in step S104 is considered to be that no anomaly has occurred in the wind power generation unit.As an example, the model is trained using the XGBoost machine learning algorithm to capture the non-linear changes in the unit load data and the wind parameter sequence data, thereby improving the prediction effect of the model. Specifically, first, data cleaning and missing value processing are performed to ensure the validity of the data. Then, the data is partitioned into a training set and a test set according to a certain ratio (for example, 4:1). Finally, a multi-factor prediction model is established, and the model object is processed using tune_model to adjust its hyperparameters and optimize the model.

[0039] Figure 2 is a flow schematic diagram of the abnormal prediction method for a wind power unit according to an embodiment of the present disclosure.

[0040] Generally, according to the abnormal prediction method for a wind power unit according to an embodiment of the present disclosure, in the model establishment stage, examples of different abnormal types (abnormal load, vortex-induced vibration, blade stall) are selected to simulate the process. The FNL reanalysis data is used as the driving data for the meteorological simulation mode, and the historical wind parameter sequence data output from WRF-LES includes sequence data of horizontal wind speed, wind direction, vertical speed, wind shear, and turbulence intensity at multiple heights within the impeller range. To ensure the accuracy of the simulation and obtain wind parameters with high spatio-temporal resolution, the main technical points used in the simulation process by WRF-LES include the following: 1. As the underlying surface terrain information, the SRTM3 elevation terrain data is introduced. 2. The ground meteorological observation data is assimilated using the Nudging four-dimensional assimilation technique. 3. The Ndown technique is used to provide boundary conditions for the internal high-resolution nested region by the meteorological field in the outer region. 4. By refining and outputting the lower layer of the boundary layer, it is ensured that there is as much vertical wind field information as possible within the impeller surface range.

[0041] Using the impeller surface sequence wind parameters output from WRF-LES as input, CFD / CSD is used to simulate the aerodynamic elastic load distribution of the unit. CFD solves the aerodynamic forces using the unsteady Navier-Stokes equations finite volume method, and CSD solves the structural response using the finite element numerical method. The structural response displacement is interpolated to the CFD mesh points to obtain the deformed mesh of CFD. The above process is continuously repeated, and the iterative error of displacement or load is used as the convergence criterion for coupling. Finally, the aerodynamic elastic load is obtained.

[0042] Then, based on different abnormal types, using the XGBoost machine learning algorithm, an abnormal prediction model with the wind parameter sequence data and unit load data as input (corresponding to the wide arrow on the left side of Figure 2, that is, inputting the wind parameter sequence data output from WRF-LES and the unit load data output from CFD / CSD together into the "XGBoost algorithm") is established, or an abnormal prediction model with the unit load data as input (corresponding to the wide arrow on the right side of Figure 2, that is, not inputting the wind parameter sequence data output from WRF-LES into the "XGBoost algorithm") is established to adjust the model. XGBoost captures the non-linear changes of meteorological elements, and the correction effect is better than that of ordinary statistical methods. The process of establishing the abnormal prediction model is as follows: 1. Perform data cleaning and missing value processing, and partition the data into a training set and a test set according to a ratio of 4:1. 2. Establish a multi-factor prediction model, use tune_model to process the model object, adjust its hyperparameters, and optimize the model.

[0043] After the model establishment is completed, taking the situation where abnormal prediction is performed as quickly as possible when predicting extreme weather as an example, the process of prediction by the abnormal prediction model is as follows: 1. Based on the normal weather forecast, pay attention in advance to the potential occurrence of extreme wind conditions, such as cold wave gales and convective gales by short-term forecasts (3 days). 2. The GFS forecast is used as the WRF-LES driving data to simulate the wind parameter sequence data on the impeller surface and is used for the load distribution simulation by the CFD / CSD model. The wind parameter sequence data and unit load data are input into the anomaly prediction model to predict the unit operation state and anomaly type.

[0044] In the present disclosure, the simulation of high-resolution wind parameter sequence data is realized by WRF-LES, coupled with the CFD / CSD model to calculate the unit load distribution, establish a prediction model for different types of unit anomalies, and on the premise of clarifying the physical mechanism of unit anomalies, establish an anomaly prediction model, with the prediction effect being significantly improved. Based on the weather forecast potential, the unit operation state is predicted in advance, and the root cause analysis of unit failures caused by severe weather and the prediction of unit operation state are realized.

[0045] Furthermore, after the anomaly prediction model outputs the predicted anomaly data, the characteristics of the wind parameters and the anomaly type are combined to guide the pitch angle, yaw, start / stop control policies, thereby dealing with possible anomalies, reducing the load and vibration of the wind power unit, reducing the risk of component damage, and ensuring the safe operation of the unit. Here, for the specific control policy, the operator may determine it manually or the computer may determine it. In the latter case, for example, a control model may be established in advance, and the predicted anomaly data and the wind parameter sequence data on the impeller surface are used as the input of the control model, and the corresponding control policy is used as the output of the control model, but it is not limited thereto. The control model may calculate the control policy by theoretical calculation or refer to the anomaly prediction model and train the control model by the XGBoost machine learning algorithm, and the present disclosure does not limit this.

[0046] As an example, generally, in a wind power plant, there are multiple wind power generation units, and each wind power generation unit is equipped with a stand-alone control system to ensure stable operation. A field-level control system is further arranged in the wind power plant to achieve centralized and unified control of the entire wind power plant. According to the abnormal prediction method of the embodiments of the present disclosure, the data used when simulating the wind parameter sequence data is only the terrain data and meteorological data of the mechanical position of the wind power generation unit. The mechanical position is manually input, and the meteorological data is related to the ground meteorological observation data and meteorological prediction data. Both the terrain data and the meteorological prediction data are open data and can be directly obtained. For example, as the terrain data, download SRTM data from the EarthExplorer website, and the ground meteorological observation data is obtained from the local meteorological bureau or the wind condition observation tower. When determining the unit load data, it is further necessary to use the structural parameters of the unit, but the structural parameters are static parameters and can also be manually input. When performing abnormal prediction based on the determined unit load data, if a pre-trained abnormal prediction model is used, regarding the training and inference of the abnormal prediction model, in the training stage, it is necessary to use the historical abnormal data of the reference wind power generation unit stored in the stand-alone control system or the field-level control system, and in the inference stage, it is not necessary to use it. In this way, it is not necessary to use the real-time control data of the unit during the abnormal prediction process. In the training stage of the abnormal prediction model, although it is necessary to obtain data from the stand-alone control system or the field-level control system, the obtained data is the historical abnormal data accumulated historically, so there is no need to interact frequently with the stand-alone control system or the field-level control system, and the abnormal prediction system can be developed separately, and the abnormal prediction method of the present disclosure can be executed by the abnormal prediction system.In addition, since the computational complexity of the prediction process is large and requires significant computing power, an independent anomaly prediction system is developed separately from the stand-alone control system or the field-level control system to reduce the computing load on the stand-alone control system or the field-level control system and fully ensure the reliable operation of the stand-alone control system or the field-level control system. During the prediction process, the anomaly prediction system obtains the terrain data, meteorological data of the mechanical position of the specified wind power generation unit, and the structural parameters of the specified wind power generation unit, and further realizes the prediction of abnormal operation of the specified wind power generation unit. Regarding the training of the anomaly prediction model, since it is necessary to determine the historical wind parameter sequence data and the historical unit load data during training, the anomaly prediction system executes model training to realize the integration of the model training and the execution entity of the inference, and immediately updates and trains the anomaly prediction model. Here, historical anomaly data of the reference wind power generation unit is obtained from the stand-alone control system or the field-level control system, and reference terrain data and historical meteorological data of the mechanical position of the reference wind power generation unit are obtained from the outside.

[0047] Of course, the anomaly prediction method according to the embodiments of the present disclosure may be installed in a stand-alone control system or a field-level control system, and the prediction may be executed by the stand-alone control system or the field-level control system. In this case, the mechanical position and unit structure parameters stored in the stand-alone control system or the field-level control system may be used, or the mechanical position and unit structure parameters may be obtained by manual input. In addition, it is necessary to obtain the terrain data and meteorological data of the mechanical position from the outside. When using a pre-trained anomaly prediction model, the historical anomaly data of the reference wind power generation unit stored in the stand-alone control system or the field-level control system is directly used to obtain the reference terrain data and historical meteorological data of the mechanical position of the reference wind power generation unit from the outside.

[0048] When an independent anomaly prediction system is developed for the situation of outputting a control policy using a control model, the control model is placed in the anomaly prediction system. The anomaly prediction system independently completes anomaly prediction and determination of the control policy, and transmits the determined control policy to a stand-alone control system or a field-level control system. Thereby, the stand-alone control system or the field-level control system controls the operation of the specified wind power generation unit according to the determined control policy. The control model may be placed in a stand-alone control system or a field-level control system. In this case, the anomaly prediction system transmits the acquired anomaly prediction data and wind parameter sequence data to the stand-alone control system or the field-level control system. Thereby, the stand-alone control system or the field-level control system operates the control model to determine a control policy and controls the operation of the specified wind power generation unit according to the determined control policy. However, the present disclosure is not limited thereto.

[0049] FIG. 3 is a block diagram of an anomaly prediction device for a wind power generation unit according to one embodiment of the present disclosure.

[0050] Referring to FIG. 3, the anomaly prediction device 300 for a wind power generation unit includes an acquisition means 301, a calculation means 302, and a prediction means 303.

[0051] The acquisition means 301 acquires topographic data and meteorological data of the mechanical position of the wind power generation unit. By acquiring the topographic data and the meteorological data, in subsequent steps, a speculation simulation of the airflow flow in the space where the mechanical position is located is performed.

[0052] Preferably, the spatial resolution of the terrain data and the meteorological data is equal to or higher than a predetermined spatial resolution, and the temporal resolution of the meteorological data is equal to or higher than a predetermined temporal resolution. The spatial resolution is reflected by the minimum distance between two adjacent spatial points where the corresponding data (i.e., terrain data or meteorological data) can be displayed. The smaller the value, the higher the spatial resolution. Similarly, the temporal resolution is reflected by the minimum time interval between two adjacent time points where the corresponding data can be displayed. The smaller the value, the higher the temporal resolution. The minimum distance corresponding to the spatial resolution of the terrain data and the meteorological data should be less than or equal to the minimum distance corresponding to the predetermined spatial resolution, and the minimum time interval corresponding to the temporal resolution of the meteorological data should be less than or equal to the minimum time interval corresponding to the predetermined temporal resolution, thereby realizing a high spatio-temporal resolution. By acquiring terrain data and meteorological data with a high spatio-temporal resolution, high-precision basic data is provided for subsequent simulations, high-precision simulation results are obtained, and the accuracy of anomaly prediction is improved. Here, the period for which the simulation is performed is short, generally only a few hours or a few days. During this period, the characteristics of the underlying surface are likely to remain unchanged. Therefore, it is only necessary to acquire static high-spatial-resolution terrain data, that is, there is no requirement for the terrain data to have a high temporal resolution. For meteorological data, in order to reflect meteorological changes, it is necessary to acquire dynamic high-spatio-temporal-resolution meteorological data.

[0053] Preferably, the acquisition means 301 further executes the following steps.

[0054] First, based on the mechanical position of the wind power generation unit, the space to be simulated is determined. This step determines the space that is obvious for the simulation and clarifies the spatial range of the terrain data and the meteorological data to be acquired.

[0055] Then, for the space to be simulated, terrain data with a spatial resolution equal to or higher than a predetermined spatial resolution is acquired. The minimum distance corresponding to the predetermined spatial resolution is, for example, 50 meters or 80 meters, and the terrain data is the above-mentioned SRTM3 data.

[0056] Finally, obtain the ground meteorological observation data and meteorological prediction data for the target period of the space to be simulated, and perform four-dimensional assimilation processing on the ground meteorological observation data according to the target time resolution and target spatial resolution. For example, use the Nudging four-dimensional assimilation technology to obtain meteorological data. The target time resolution is not less than a predetermined time resolution, and the target spatial resolution is not less than a predetermined spatial resolution. The minimum time interval corresponding to the predetermined time resolution may be set according to needs. For example, it may be 1 minute, 3 minutes, or 5 minutes at the "minute" level, or 20 seconds or 30 seconds at the "second" level. Within the target period, by determining the meteorological data based on the actually observed ground meteorological observation data and meteorological prediction data, predictive simulation can be performed using the meteorological prediction data, and the observed data can be fused to improve the accuracy of the simulation. Through four-dimensional assimilation processing, the ground meteorological observation data is assimilated into a standardized three-dimensional mesh (four-dimensional including the time dimension), which is convenient for subsequent simulation calculations.

[0057] Here, after determining the space to be simulated, the execution order of the following two steps of obtaining subsequent terrain data and obtaining meteorological data does not have to be limited, and the target spatial resolution may be equal to the spatial resolution of the terrain data.

[0058] The calculation means 302 determines wind parameter sequence data by large eddy simulation based on terrain data and meteorological data. In the fan field, the application of large eddy simulation mainly targets wind speed and wind power prediction, wake vortex analysis, blade optimization, etc. Such a simulation method is the average of time and space for turbulent pulsation (or turbulent vortex), that is, a certain filter function is used to separate large-scale turbulence from small-scale turbulence, directly simulate the large-scale turbulence, block the small-scale turbulence by a model, and obtain a more detailed description of the turbulence through calculations with a finite number of meshes. The impeller surface of a wind power unit is greatly affected by local terrain, and small-scale turbulence has a significant impact on the load and vibration of the unit. Compared with the intermediate-scale simulation method that cannot simulate small-scale turbulence, large eddy simulation can reliably simulate different scales of turbulence, and further obtain wind parameter sequence data for different scales of turbulence, improve the accuracy of the wind parameter sequence data, reflect the influence of different scales of turbulence on the wind power unit, include small-scale turbulence within the simulation range, reduce the risk of underestimating small-scale turbulence, and improve the accuracy of anomaly prediction.

[0059] Here, the magnitude of the time resolution of the simulation output result by the conventional WRF can be set, and the minimum is the integration step size. For large eddy simulation, the output time resolution can reach 1 s.

[0060] Preferably, the calculation means 302 further constructs a simulation mesh for the space to be simulated, increases the number of meshes according to a predetermined rule within the target height range of the simulation mesh, and performs large eddy simulation calculations on the simulation mesh based on the terrain data and the meteorological data to obtain wind parameter sequence data. Specifically, the simulation calculation obtains the wind parameter sequence data of each node in the simulation mesh. By increasing the number of meshes within the target height range, that is, increasing the mesh density within the target height range, that is, improving the resolution, the wind parameter sequence data at multiple heights within the target height range can be obtained, and the simulation result becomes more detailed. As an example, the target height range is the height range where the impeller surface of the wind power generation unit is located, and the predetermined rule is the size of the mesh, whereby it is possible to ensure that as much vertical wind field information as possible is obtained within the impeller surface range, and subsequently, the load distribution situation of the impeller surface can be easily analyzed.

[0061] Preferably, the wind parameter sequence data includes sequence data of target parameters at a plurality of predetermined heights within the impeller range, and the target parameters include at least one of horizontal wind speed, wind direction, vertical speed, wind shear, and turbulence intensity. By obtaining the sequence data of the above target parameters, the wind conditions at the mechanical position of the wind power generation unit can be comprehensively reflected, and the abnormal prediction of the wind power generation unit can be realized.

[0062] The calculation means 302 further determines unit load data based on the wind parameter sequence data. Abnormalities occur in the wind power generation unit in extreme weather. As the root cause, due to extreme wind conditions, the unit receives abnormal loads. Based on the accurate and detailed wind parameter sequence data obtained by large eddy simulation, accurate and reliable unit load data is determined, the root cause of the unit abnormality is clarified, and reliable abnormal prediction is realized.

[0063] Preferably, the calculation means 302 further inputs the wind parameter sequence data into the numerical fluid dynamics model and the numerical structural mechanics model to obtain the unit load data. The numerical fluid dynamics model (CFD) solves the aerodynamic and hydrodynamic forces by the unsteady N-S equation finite volume method based on the wind parameter sequence data to initially obtain the unit load data. The numerical structural mechanics model (CSD) solves the structural response using the finite element numerical method to obtain the structural response displacement of the wind power unit under the action of aerodynamic and hydrodynamic forces. Since the change in displacement affects the unit load, the structural response displacement is interpolated to the CFD mesh points to obtain the deformed mesh of CFD. The above process is continuously repeated, and the iterative error of displacement or load is used as the convergence criterion for coupling to finally obtain the unit load data.

[0064] Preferably, the unit load data includes at least one of the blade load data and the tower frame load data. In abnormal prediction, since the blade and the tower frame are the main structures to be focused on, the load data at these two locations are mainly simulated to achieve highly targeted calculations. The blade load data includes at least one of the blade flap bending moment and the blade edgewise bending moment, and the tower frame load data includes at least one of the tower top pitch bending moment, the tower top overturning bending moment, the tower top torque, the tower bottom pitch bending moment, and the tower bottom overturning bending moment. Specifically, by obtaining the above load data, the load changes of the blade and the tower frame in the main abnormal situations can be effectively covered, providing a basis for reliable unit abnormal prediction.

[0065] The prediction means 303 predicts whether an abnormality has occurred in the wind power unit based on the unit load data. As an example, in addition to the unit load data, the wind parameter sequence data by the calculation means 302 is added and jointly used for prediction, thereby increasing the basic data for prediction and improving the accuracy of prediction.

[0066] Preferably, the prediction means 303 further inputs the unit load data into a pre-trained anomaly prediction model to predict whether an anomaly has occurred in the wind power unit. By predicting anomalies using the pre-trained anomaly prediction model, the relationship between the unit load data and the anomalies of the wind power unit is learned during the pre-training process, and further, the prediction results can be quickly obtained at the prediction stage to ensure the timeliness and accuracy of the prediction. As described above, the wind parameter sequence data may be input together into the pre-trained anomaly prediction model.

[0067] Preferably, the pre-trained anomaly prediction model is obtained by training in the following steps: obtaining historical anomaly data of a reference wind power unit, corresponding reference terrain data, and historical meteorological data for the historical anomaly data; determining historical wind parameter sequence data by large eddy simulation based on the reference terrain data and the historical meteorological data; determining historical unit load data based on the historical wind parameter sequence data; inputting the historical unit load data into an anomaly prediction model to be trained to obtain historical anomaly prediction data; and adjusting the parameters of the anomaly prediction model to be trained based on the historical anomaly data and the historical anomaly prediction data to obtain a pre-trained anomaly prediction model. In short, the anomaly prediction model is trained by a supervised learning method, which can effectively learn the relationship between unit load data and anomalies of wind power units, improving the learning efficiency. Here, the historical anomaly data is data recorded during the daily operation and maintenance of wind power units, including, for example, anomaly data such as mechanical system failure information, electrical system failure information, and control system failure information. Each piece of anomaly data includes, for example, the failure type and location of the unit, and further includes sensor data (such as data on tilt angle, acceleration, strain, and temperature), enabling the acquisition of data on the tilt, vibration, load, and deformation of the fan tower. Therefore, model training is realized using conventional data, improving the possibility of solutions. Correspondingly, the pre-trained anomaly prediction model may output anomaly data in the same format as prediction anomaly data.

[0068] The method for predicting anomalies of a wind power generation unit according to an embodiment of the present disclosure is described as a computer program and stored in a computer-readable storage medium. When instructions corresponding to the computer program are executed by a processor, the method for predicting anomalies of the wind power generation unit described above can be realized. Examples of computer-readable storage media include read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-RLTH, BD-RE, blue or optical disc memory, hard disk drive (HDD), solid state drive (SSD), memory card (e.g., multimedia card, secure digital (SD) card or extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid disk and any other device, and the any other device non-temporarily stores a computer program, any related data, data file and data structure and provides the computer program, any related data, data file and data structure to a processor or a computer, so that the processor or the computer can execute the computer program. In one example, the computer program, any related data, data file and data structure are distributed to a computer system connected to a network, so that the computer program, any related data, data file and data structure are distributedly stored, accessed and executed by one or more processors or computers.

[0069] FIG. 4 is a block diagram of a computer device according to an embodiment of the present disclosure.

[0070] Referring to FIG. 4, computer device 400 includes at least one memory 401 and at least one processor 402. A set of computer-executable instructions is stored in the at least one memory 401. When the set of computer-executable instructions is executed by the at least one processor 402, an abnormal prediction method for a wind power generation unit according to an exemplary embodiment of the present disclosure is executed.

[0071] By way of example, computer device 400 may be a PC computer, a tablet device, a personal digital assistant, a smartphone, or other device capable of executing the above instruction set. Here, computer device 400 does not necessarily have to be an individual electronic device, and may be any device or integrated circuit capable of executing the above instructions (or instruction set) individually or jointly. Computer device 400 may further be a part of an integrated control system or a system manager, or may be arranged as a portable electronic device connected to a local or remote device via an interface (e.g., via wireless transmission).

[0072] In computer device 400, processor 402 includes a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, the processor may include a simulation processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, and the like.

[0073] Processor 402 causes instructions or code stored in memory 401 to be executed, and data may be stored in memory 401. The instructions and data are further transmitted and received by a network via a network interface device, and the network interface device may use any known transport protocol.

[0074] The memory 401 and the processor 402 may be integrated together. For example, a RAM or a flash memory may be disposed inside an integrated circuit microprocessor or the like. Further, the memory 401 may include an independent device, for example, an external disk drive, a storage array, or any other storage device that can be used by a database system. The memory 401 and the processor 402 may be coupled operatively, or may communicate with each other via, for example, an I / O port, a network connection, etc., whereby the processor 402 can read a file stored in the memory.

[0075] In addition, the computer device 400 may further include a video display (for example, a liquid crystal display) and a user interaction interface (for example, a keyboard, a mouse, a touch input device, etc.). All components of the computer device 400 are connected to each other via a bus and / or a network.

[0076] In the present disclosure, a simulation of high-resolution wind parameter sequence data is realized by WRF-LES, coupled with a CFD / CSD model to calculate the unit load distribution, establish a prediction model for different types of unit anomalies, and on the premise of clarifying the physical mechanism of unit anomalies, establish an anomaly prediction model, the prediction effect is significantly improved, and based on the weather forecast potential, the unit operating state is predicted in advance, and the root cause analysis of unit failures caused by strong weather and the prediction of unit operating states are realized.

[0077] As described above, specific embodiments of the present disclosure have been described in detail and several examples have been shown and described. However, as can be understood by those skilled in the art, corrections and modifications may be made to these examples without departing from the principles and spirit of the present disclosure limited by the claims and their equivalents, and these corrections and modifications should also fall within the protection scope of the claims of the present disclosure.

Claims

1. A method for predicting anomalies in a wind power generation unit, comprising: obtaining topographic data and meteorological data of the mechanical position of the wind power generation unit; determining wind parameter sequence data by large eddy simulation based on the topographic data and the meteorological data; determining unit load data based on the wind parameter sequence data; predicting whether an anomaly has occurred in the wind power generation unit based on the unit load data.

2. The anomaly prediction method according to claim 1, wherein the spatial resolution of the topographic data and the meteorological data is equal to or higher than a predetermined spatial resolution, and the temporal resolution of the meteorological data is equal to or higher than a predetermined temporal resolution.

3. The step of obtaining topographic data and meteorological data of the mechanical position of the wind power generation unit comprises: determining a space to be simulated based on the mechanical position of the wind power generation unit; obtaining the topographic data with a spatial resolution equal to or higher than the predetermined spatial resolution for the space to be simulated; obtaining ground meteorological observation data and meteorological prediction data for a target period of the space to be simulated, and performing four-dimensional assimilation processing on the ground meteorological observation data according to a target temporal resolution and a target spatial resolution to obtain the meteorological data, wherein the target temporal resolution is equal to or higher than the predetermined temporal resolution, and the target spatial resolution is equal to or higher than the predetermined spatial resolution.

4. The step of determining wind parameter sequence data by large eddy simulation based on the topographic data and the meteorological data comprises: constructing a simulation mesh for the space to be simulated and increasing the number of meshes according to a predetermined rule within a target height range of the simulation mesh; performing large eddy simulation calculation on the simulation mesh based on the topographic data and the meteorological data to obtain the wind parameter sequence data.

5. The step of determining unit load data based on the wind parameter sequence data is: The method for predicting abnormality according to any one of claims 1 to 4, including the step of inputting the wind parameter sequence data into a numerical fluid dynamics model and a numerical structural mechanics model to obtain the unit load data.

6. The step of predicting whether an abnormality has occurred in the wind power unit based on the unit load data is The method for predicting abnormality according to any one of claims 1 to 4, including the step of predicting whether an abnormality has occurred in the wind power unit by inputting the unit load data into an abnormality prediction model trained in advance.

7. The abnormality prediction model trained in advance is obtained by training through the following steps: The step of obtaining historical abnormality data of a reference wind power unit, and reference terrain data and historical meteorological data corresponding to the historical abnormality data; The step of determining historical wind parameter sequence data by large eddy simulation based on the reference terrain data and the historical meteorological data; The step of determining historical unit load data based on the historical wind parameter sequence data; The step of inputting the historical unit load data into an abnormality prediction model to be trained to obtain historical abnormality prediction data; The method for predicting abnormality according to claim 6, including the step of adjusting the parameters of the abnormality prediction model to be trained based on the historical abnormality data and the historical abnormality prediction data to obtain the abnormality prediction model trained in advance.

8. The wind parameter sequence data includes sequence data of target parameters at a plurality of predetermined heights within the impeller range, and the target parameters include at least one of horizontal wind speed, wind direction, vertical velocity, wind shear, and turbulence intensity, and / or The unit load data includes at least one of blade load data and tower frame load data. The blade load data includes at least one of blade simi bending moment and blade torsional bending moment. The tower frame load data includes at least one of tower top pitch bending moment, tower top overturning bending moment, tower top torque, tower bottom pitch bending moment, and tower bottom overturning bending moment. The method for predicting abnormality according to any one of claims 1 to 4.

9. An abnormal prediction device for a wind power generation unit, an acquisition means arranged to acquire topographic data and meteorological data of the mechanical position of the wind power generation unit; a calculation means arranged to determine wind parameter sequence data by large eddy simulation based on the topographic data and the meteorological data, wherein the calculation means is further arranged to determine unit load data based on the wind parameter sequence data; an abnormal prediction device including a prediction means arranged to predict whether an abnormality has occurred in the wind power generation unit based on the unit load data.

10. The abnormal prediction device according to claim 9, wherein the spatial resolution of the topographic data and the meteorological data is equal to or higher than a predetermined spatial resolution, and the temporal resolution of the meteorological data is equal to or higher than a predetermined temporal resolution.

11. The acquisition means further determines a space to be simulated based on the mechanical position of the wind power generation unit, acquires the topographic data having a spatial resolution equal to or higher than the predetermined spatial resolution for the space to be simulated, acquires ground meteorological observation data and meteorological prediction data for a target period of the space to be simulated, and performs four-dimensional data assimilation processing on the ground meteorological observation data according to a target temporal resolution and a target spatial resolution to acquire the meteorological data, wherein the target temporal resolution is equal to or higher than the predetermined temporal resolution, and the target spatial resolution is equal to or higher than the predetermined spatial resolution. The abnormal prediction device according to claim 10.

12. The calculation means further constructs a simulation mesh of the space to be simulated, and increases the number of meshes according to a predetermined rule within a target height range of the simulation mesh, and is arranged to perform large eddy simulation calculation on the simulation mesh based on the topographic data and the meteorological data to acquire the wind parameter sequence data. The abnormal prediction device according to claim 11.

13. The calculation means is further arranged to input the wind parameter sequence data into a numerical fluid dynamics model and a numerical structural mechanics model to acquire the unit load data. The abnormal prediction device according to any one of claims 9 to 12.

14. The prediction means is further arranged to predict whether an abnormality has occurred in the wind power generation unit by inputting the unit load data into an abnormality prediction model that has been pre-trained, according to any one of claims 9 to 12.

15. The pre-trained abnormality prediction model is trained and obtained through the following steps: Obtaining historical abnormality data of a reference wind power generation unit, and reference terrain data and historical meteorological data corresponding to the historical abnormality data; Determining historical wind parameter sequence data by large eddy simulation based on the reference terrain data and the historical meteorological data; Determining historical unit load data based on the historical wind parameter sequence data; Inputting the historical unit load data into an abnormality prediction model to be trained to obtain historical abnormality prediction data; Adjusting the parameters of the abnormality prediction model to be trained based on the historical abnormality data and the historical abnormality prediction data to obtain the pre-trained abnormality prediction model, according to claim 14.

16. The wind parameter sequence data includes sequence data of target parameters at a plurality of predetermined heights within the impeller range, the target parameters include at least one of horizontal wind speed, wind direction, vertical velocity, wind shear, and turbulence intensity, and / or The unit load data includes at least one of blade load data and tower frame load data, the blade load data includes at least one of blade bending moment and blade torsional bending moment, and the tower frame load data includes at least one of tower top pitch bending moment, tower top overturning bending moment, tower top torque, tower bottom pitch bending moment, and tower bottom overturning bending moment, according to any one of claims 9 to 12.

17. A computer-readable storage medium, wherein when instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is caused to execute the abnormality prediction method according to any one of claims 1 to 8.

18. A computer device, including at least one processor and at least one memory storing computer-executable instructions, wherein when the computer-executable instructions are executed by the at least one processor, the at least one processor is caused to execute the anomaly prediction method according to any one of claims 1 to 8.

Images