Method, device and electronic equipment for performance analysis of offshore wind turbine

Abstract

The application provides a performance analysis method and device of an offshore wind turbine and electronic equipment, and relates to the technical field of wind turbines, which collects environmental data through multiple sensors deployed in a target offshore wind field area, and performs time reconstruction and multi-dimensional analysis on the data to identify wind field working condition types; then, wind field prediction models are respectively constructed for different working conditions, wind field parameter prediction results are output, and high-quality wind field data sets are screened out; finally, load analysis is performed based on the data sets to generate load data for evaluating the performance of the unit. The application solves the technical problems of inaccurate wind field feature extraction and poor accuracy of wind turbine performance analysis, and achieves the technical effects of improving the accuracy of wind field feature extraction and improving the accuracy of wind turbine performance analysis.

Description

Technical Field

[0001] This invention relates to the field of wind turbine technology, and more specifically, to a performance analysis method, apparatus, and electronic equipment for offshore wind turbines. Background Technology

[0002] Existing performance analyses of offshore wind turbines typically rely on simplified data acquisition and model development. A common practice is to install single-point wind measurement devices (such as anemometers and wind vanes) near the turbine to obtain limited wind field information, and then use empirical formulas or steady-state flow field models to estimate aerodynamic loads. For structural safety assessments, static load calculations and fatigue analysis methods based on simplified assumptions are frequently employed; these approaches are still applicable in relatively stable onshore wind fields.

[0003] However, in the complex marine environment, these traditional analytical methods reveal significant limitations. On the one hand, single-point measurements struggle to depict the true changes and turbulent structures of the wind field in the spatiotemporal dimensions, leading to substantial uncertainties in the input conditions themselves. On the other hand, simplified models fail to accurately simulate the complex aerodynamic loads faced by offshore wind turbines, and cannot fully reflect the conditions under extreme wind and wave action, resulting in insufficient load prediction accuracy. More critically, existing methods do not adequately consider the stochastic and time-varying characteristics of loads during long-term operation, making it difficult to accurately assess the cumulative fatigue damage of the turbine under alternating loads, and also unable to reliably predict the extreme loads it may encounter. These limitations collectively lead to discrepancies between existing analytical results and the actual operating conditions of the turbines, hindering the safety assessment and operation and maintenance decisions of offshore wind turbines.

[0004] Based on this, existing technologies suffer from problems such as insufficient accuracy in wind field feature extraction and poor accuracy in performance analysis of offshore wind turbines. Summary of the Invention

[0005] The purpose of this invention is to provide a performance analysis method, apparatus, and electronic device for offshore wind turbines, so as to alleviate the technical problems of insufficient accuracy in wind field feature extraction and poor accuracy in wind turbine performance analysis in the prior art.

[0006] In a first aspect, embodiments of the present invention provide a performance analysis method for offshore wind turbines, comprising: acquiring an initial environmental dataset collected by a multi-sensor network in a target offshore wind field area; performing time reconstruction and multi-scale feature extraction on the initial environmental dataset to obtain target wind field features and a set of wind field operating condition types; the set of wind field operating condition types includes at least one wind field operating condition type; for each wind field operating condition type, inputting the target wind field features into a pre-constructed wind field parameter prediction model corresponding to the wind field operating condition type, outputting the prediction result of the wind field parameters, and filtering the initial environmental dataset according to the prediction result to obtain a wind field environmental dataset; and performing performance load analysis based on the wind field environmental dataset to obtain load analysis data for evaluating the performance of offshore wind turbines.

[0007] In some optional implementations, the aforementioned multi-sensor network includes: wind speed sensor, wind direction sensor, and turbulence intensity sensor; the aforementioned method also includes: dividing the aforementioned target offshore wind field area into grids to obtain wind field grid data; and determining the placement and number of each type of sensor based on the aforementioned wind field grid data and historical data.

[0008] In some optional implementations, the initial environmental dataset is temporally reconstructed and multi-scale feature extracted to obtain target wind field features and a set of wind field operating conditions. This includes: calibrating the sensor network using a time synchronization protocol to obtain synchronized sampling timestamps; reconstructing the initial environmental dataset temporally based on the synchronized sampling timestamps to obtain a time-aligned wind field data matrix; performing multi-scale decomposition and principal component analysis on the wind field data matrix to obtain target wind field features; and performing cluster analysis on the target wind field features to obtain a set of wind field operating conditions.

[0009] In some optional implementations, for each wind field condition type, the aforementioned target wind field features are input into a pre-constructed wind field parameter prediction model corresponding to the wind field condition type, and the predicted wind field parameters are output. Based on these prediction results, the initial environmental dataset is filtered to generate a wind field environment dataset. This includes: modeling the wind field parameters under each wind field condition using a support vector regression algorithm to obtain a wind field parameter prediction model; the input to the wind field parameter prediction model is the aforementioned target wind field features, and the output is the predicted values ​​of the wind field parameters; the predicted values ​​of the wind field parameters include: predicted wind speed, predicted wind direction, and predicted turbulence intensity; parameter optimization and uncertainty analysis are performed on the wind field parameter prediction model to obtain the confidence intervals for each predicted wind field parameter value; and data filtering is performed on the initial environmental dataset based on the confidence intervals to obtain the aforementioned wind field environment dataset.

[0010] In some optional implementations, before the step of performing performance load analysis based on the above wind field environment dataset, the method further includes: preprocessing the above wind field environment dataset to obtain a preprocessed wind field environment dataset; the preprocessed wind field environment dataset includes: a set of Weibull distribution parameters for wind speed, a set of characteristics for wind direction distribution, and turbulence intensity distribution data.

[0011] In some optional implementations, performance load analysis is performed based on the aforementioned wind field environment dataset, including: performing hydrodynamic analysis based on the preprocessed wind field environment dataset to obtain three-dimensional wind field flow field data; performing aerodynamic load simulation analysis based on the aforementioned three-dimensional wind field flow field data to obtain dynamic response data of the target offshore wind field area; performing strain tests on the key components of the aforementioned wind turbine based on the aforementioned dynamic response data to obtain load time series; performing rainflow counting analysis on the aforementioned load time series to generate a fatigue load spectrum; performing extreme value analysis on the aforementioned fatigue load spectrum to obtain an estimate of the ultimate load; and generating performance analysis results based on the aforementioned fatigue load spectrum and the aforementioned estimate of the ultimate load.

[0012] In some optional implementations, rainflow counting analysis is performed on the aforementioned load time series to generate a fatigue load spectrum, including: preprocessing and counting the aforementioned load time series to extract the amplitude and mean of stress cycles and generate a fatigue load spectrum; performing extreme value analysis on the aforementioned fatigue load spectrum to obtain an estimate of the ultimate load, including: performing mean stress correction and cumulative damage calculation based on the aforementioned fatigue load spectrum to obtain a linear cumulative damage value; fitting an extreme value distribution and extrapolating it based on the aforementioned linear cumulative damage value to obtain a long-term ultimate load prediction value and the corresponding confidence interval; estimating fatigue life based on the aforementioned long-term ultimate load prediction value and the corresponding confidence interval to obtain fatigue life prediction data and an estimate of the ultimate load for the target component; and generating performance analysis results based on the aforementioned fatigue load spectrum and the aforementioned ultimate load estimate, including: generating a performance analysis report for the wind turbine based on the aforementioned fatigue life prediction data and the aforementioned ultimate load estimate for the target component.

[0013] Secondly, embodiments of the present invention provide a performance analysis device for offshore wind turbines, comprising: a multi-sensor network acquisition module for acquiring an initial environmental dataset acquired by a multi-sensor network in a target offshore wind field area; a feature extraction module for performing time reconstruction and multi-scale feature extraction on the initial environmental dataset to obtain target wind field features and a wind field operating condition type set; the wind field operating condition type set includes at least one wind field operating condition type; a wind field parameter prediction module for inputting the target wind field features into a pre-constructed wind field parameter prediction model corresponding to the wind field operating condition type for each wind field operating condition type, outputting the prediction result of the wind field parameters, and filtering the initial environmental dataset according to the result to obtain a wind field environmental dataset; and a performance load analysis module for performing performance load analysis based on the wind field environmental dataset to obtain load analysis data for evaluating the performance of offshore wind turbines.

[0014] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the steps of the method described in any of the first aspects above.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to perform the method described in any of the first aspects above.

[0016] This invention provides a method, apparatus, and electronic device for performance analysis of offshore wind turbines. The method collects environmental data through multiple sensors deployed in the target offshore wind farm area, performs time reconstruction and multi-dimensional analysis on the data to identify wind farm operating conditions. Then, it constructs wind farm prediction models for different operating conditions, outputs wind farm parameter prediction results, and uses these to select high-quality wind farm datasets. Finally, it performs load analysis based on this dataset to generate load data for evaluating turbine performance. This invention solves the technical problems of insufficient accuracy in wind farm feature extraction and poor accuracy in wind turbine performance analysis, achieving the technical effect of improving the accuracy of wind farm feature extraction and enhancing the accuracy of wind turbine performance analysis. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a performance analysis method for offshore wind turbines provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another performance analysis method for offshore wind turbines provided in an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a performance analysis device for offshore wind turbines provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Existing methods for analyzing the performance of offshore wind turbines typically employ traditional data acquisition and analysis techniques. These methods rely heavily on single-point measurements and simplified models, such as using a single anemometer and wind vane to collect wind field data, and using empirical formulas or simplified hydrodynamic models to estimate the turbine's performance. In load analysis, commonly used methods include static load calculations and simple fatigue analysis, which struggle to accurately reflect the actual operating conditions of wind turbines in complex offshore environments. However, these traditional methods have several significant shortcomings, especially in complex offshore environments. First, single-point measurements cannot fully capture the complexity and variability of offshore wind fields, resulting in inaccurate descriptions of wind field characteristics. Second, simplified hydrodynamic models struggle to accurately simulate the complex aerodynamic loads faced by offshore wind turbines, potentially underestimating or overestimating the actual performance of the turbine. Furthermore, traditional load analysis methods often neglect the dynamic characteristics and long-term cumulative effects of the offshore environment, failing to accurately predict the turbine's fatigue life and ultimate load.

[0021] Based on this, the present invention provides a method, apparatus and electronic device for performance analysis of offshore wind turbines, in order to solve the technical problems of insufficient accuracy in wind field feature extraction and poor accuracy in wind turbine performance analysis in the prior art.

[0022] To facilitate understanding of this embodiment, a performance analysis method for offshore wind turbines disclosed in this invention will first be described in detail. It is understood that the executing entity of this invention can be a performance analysis device for offshore wind turbines, or it can be a terminal or a server; no specific limitation is made here. This embodiment of the invention will use a server as an example for illustration.

[0023] See Figure 1 The diagram shows a flow chart of a performance analysis method for offshore wind turbines. This method can be executed by electronic equipment and mainly includes the following steps S102 to S108: Step S102: Obtain the initial environmental dataset collected by a multi-sensor network in the target offshore wind field area; Before collecting environmental data on the target offshore wind field area, a multi-sensor network can be deployed in the target wind turbine units in the target offshore wind field area. Preferably, the multi-sensor network can include: wind speed sensor, wind direction sensor and turbulence intensity sensor; the number and location of each sensor in the target offshore wind field area are determined according to the grid division results.

[0024] Specifically, in implementing the performance analysis method for offshore wind turbines, a sensor network is first deployed for the target wind turbines in the target offshore wind farm area. This network includes wind speed sensors, wind direction sensors, and turbulence intensity sensors, which are used to measure wind speed, wind direction, and turbulence intensity, respectively.

[0025] For example, around a typical 5MW offshore wind turbine, multiple ultrasonic anemometers, wind vanes, and 3D ultrasonic anemometers can be installed at different heights and locations. These sensors are connected through a data acquisition system to collect data synchronously at high frequency, forming a wind field environmental dataset.

[0026] In one embodiment, the deployment of the aforementioned multi-sensor network includes: dividing the target offshore wind field area into grids to obtain wind field grid data; and determining the placement location and quantity of each type of sensor based on the wind field grid data and historical data.

[0027] When implementing the performance analysis method for offshore wind turbines, the target offshore wind field area is first divided into grids to obtain wind field grid data. For example, a 10 square kilometer offshore wind field can be divided into a 100x100 grid, with each grid cell being 100 meters x 100 meters. Based on this wind field grid data, the optimal placement of sensors is determined.

[0028] Preferably, sensors can be deployed at the four corners of the wind farm, the center point, and around each wind turbine to form a comprehensive sensor network. Wind speed sensors, wind direction sensors, and turbulence intensity sensors are then installed and configured. Wind speed sensors typically use ultrasonic anemometers, which can accurately measure three-dimensional wind speed; wind direction sensors use wind vanes, which can continuously measure wind direction angles; and turbulence intensity sensors may be three-dimensional ultrasonic anemometers, capable of measuring high-frequency wind speed fluctuations. These sensors constitute a sensor network through which data is collected to obtain an initial environmental dataset.

[0029] Step S104: Perform time reconstruction and multi-scale feature extraction on the initial environmental dataset to obtain target wind field features and a wind field condition type set; the wind field condition type set includes at least one wind field condition type. Among them, the wind farm operating condition type can be a typical state category based on the dynamic evolution characteristics of wind farm environmental data in the time domain and scale domain, and with significant distinguishability to the aerodynamic load response of wind turbine units; for example: strong wind stability, weak wind turbulence, sudden wind direction change, enhanced shear, wake dominance, low frequency oscillation, etc.

[0030] In one embodiment, the initial environmental dataset is temporally reconstructed and multi-scale feature extracted to obtain target wind field features and a set of wind field operating conditions. This includes: performing time calibration on the sensor network using a time synchronization protocol to obtain synchronized sampling timestamps; reconstructing the initial environmental dataset temporally based on the synchronized sampling timestamps to obtain a time-aligned wind field data matrix; performing multi-scale decomposition and principal component analysis on the wind field data matrix to obtain target wind field features; and performing cluster analysis on the target wind field features to obtain a set of wind field operating conditions.

[0031] Specifically, the sensor network can be time-calibrated using a time synchronization protocol to obtain a synchronized sampling timestamp; the initial environmental dataset can then be reconstructed into a time series based on the synchronized sampling timestamp to obtain a time-aligned wind field data matrix.

[0032] Then, the wind field data matrix is ​​decomposed into multiple scales using wavelet transform to obtain wind field feature sets of different frequency components; the wind field feature sets are then dimensionality-reduced using principal component analysis to obtain target wind field feature vectors; and the target wind field feature vectors are then clustered using a Gaussian mixture model to obtain a set of typical wind field operating conditions. In step S106, for each wind field operating condition type, the target wind field features are input into a pre-constructed wind field parameter prediction model corresponding to the wind field operating condition type, the predicted wind field parameters are output, and the initial environmental dataset is filtered based on the prediction results to obtain a wind field environmental dataset. In one embodiment, for each wind field condition type, the target wind field features are input into a pre-constructed wind field parameter prediction model corresponding to the wind field condition type, the predicted wind field parameters are output, and the initial environmental dataset is filtered based on the prediction results to generate a wind field environment dataset. This includes: modeling the wind field parameters for each wind field condition type in the wind field condition type set using a support vector regression algorithm to obtain the corresponding wind field parameter prediction model; the input of the wind field parameter prediction model is the target wind field features, and the output is the predicted values ​​of the wind field parameters; the predicted values ​​of the wind field parameters include: predicted wind speed, predicted wind direction, and predicted turbulence intensity; parameter optimization and uncertainty analysis are performed on the wind field parameter prediction model to obtain the confidence interval of each predicted wind field parameter value; and the initial environmental dataset is filtered based on the confidence intervals to obtain the wind field environment dataset.

[0033] Specifically, the wind field parameters under typical operating conditions of each wind field can be modeled using the support vector regression algorithm to obtain a wind field parameter prediction model; the parameters of the wind field parameter prediction model can be optimized using the particle swarm optimization algorithm to obtain an optimized wind field parameter prediction model.

[0034] Furthermore, the Monte Carlo method can be used to perform uncertainty analysis on the optimized wind field parameter prediction model to obtain the confidence interval of the wind field parameter prediction, and the initial environmental dataset can be filtered using the confidence interval to obtain the wind field environmental dataset.

[0035] In one embodiment, the above method may further include: preprocessing the wind field environment dataset to obtain a preprocessed wind field environment dataset; the preprocessed wind field environment dataset includes: a set of Weibull distribution parameters for wind speed, a set of wind direction distribution features, and turbulence intensity distribution data.

[0036] Preferably, the above preprocessing steps may specifically include: cleaning the wind field environment dataset to obtain a cleaned dataset; performing time series interpolation on the cleaned dataset to obtain an interpolated time series dataset, wherein the interpolated time series dataset includes: interpolated wind speed data, interpolated wind direction data, and interpolated turbulence intensity data.

[0037] Furthermore, the interpolated wind speed data is fitted with a Weibull distribution parameter using the maximum likelihood estimation method to obtain the Weibull distribution parameter set for wind speed, wherein the Weibull distribution parameter set for wind speed includes wind speed scale parameters and wind speed shape parameters. The interpolated wind direction data is analyzed using a cyclic statistical algorithm to obtain the wind direction distribution feature set, wherein the wind direction distribution feature set includes the dominant wind direction and wind direction dispersion.

[0038] Furthermore, the interpolated turbulence intensity data is analyzed to obtain turbulence intensity distribution data, wherein the turbulence intensity distribution data includes turbulence intensity and turbulence scale.

[0039] Step S108: Perform performance load analysis based on the wind farm environment dataset to obtain load analysis data for evaluating the performance of offshore wind turbines.

[0040] The aforementioned performance load analysis typically includes hydrodynamic analysis, aerodynamic load simulation analysis, rainflow counting analysis, and extreme value analysis.

[0041] In one embodiment, performance load analysis based on a wind field environment dataset includes: performing hydrodynamic analysis based on a preprocessed wind field environment dataset to obtain three-dimensional wind field flow field data; performing aerodynamic load simulation analysis based on the three-dimensional wind field flow field data to obtain dynamic response data of the target offshore wind field area; performing strain tests on key components of the wind turbine based on the dynamic response data to obtain a load time series; performing rainflow counting analysis on the load time series to generate a fatigue load spectrum; performing extreme value analysis on the fatigue load spectrum to obtain an estimate of the ultimate load; and generating performance analysis results based on the fatigue load spectrum and the estimate of the ultimate load.

[0042] As a concrete example, performing hydrodynamic analysis based on a preprocessed wind field environment dataset to obtain three-dimensional wind field flow field data can specifically include: constructing a flow field from the preprocessed dataset to obtain a target flow field; performing flow field solution analysis on the target flow field to obtain a target flow field solution, and generating flow field data based on the target flow field solution; performing flow field vortex structure analysis on the flow field data to obtain flow field vortex structure distribution data; performing modal decomposition on the flow field vortex structure distribution data to obtain a flow field mode set; extracting dynamic features from the flow field mode set to obtain flow field dynamic feature data; and performing order reduction processing on the flow field dynamic feature data to obtain the three-dimensional wind field flow field data.

[0043] As a specific example, aerodynamic load simulation modeling based on the three-dimensional wind field data to obtain dynamic response data of the target offshore wind field region can include: First, the three-dimensional flow field data of the wind field is interpolated to obtain the flow field data of the turbine blades of the target wind turbine; then, the flow field data of the turbine blades is analyzed to obtain the blade angle of attack and relative wind speed of the target wind turbine. Then, based on the pre-collected blade airfoil characteristic data, the blade angle of attack and relative wind speed are analyzed by looking up a table to obtain the aerodynamic coefficients; the aerodynamic coefficients are accumulated along the spanwise direction of the blade of the target wind turbine to obtain the overall aerodynamic force and moment of the blade; the overall aerodynamic force and moment of the blade are transformed into a coordinate system to obtain the wind turbine thrust and wind turbine torque corresponding to the blade of the target wind turbine. Furthermore, a transmission analysis is performed on the wind turbine thrust and the wind turbine torque to obtain the main shaft load and the generator torque; a spectral analysis is performed on the main shaft load and the generator torque to obtain the load frequency components and the load amplitude distribution; a structural dynamics analysis is performed on the load frequency components to obtain the component stress and deformation response data. Finally, statistical analysis is performed on the load and deformation response data to obtain the extreme value distribution and fatigue load spectrum; data fitting is performed on the extreme value distribution and fatigue load spectrum to obtain the dynamic response data.

[0044] As a specific example, strain testing of key components of a wind turbine based on dynamic response data may include: performing finite element analysis on the key components of the target wind turbine to obtain stress concentration regions; determining the strain gauge placement positions based on the stress concentration regions to obtain a strain test point distribution map; performing surface treatment and strain gauge attachment on the strain test points according to the strain test point distribution map to obtain a test strain measurement system; acquiring signals from the strain measurement system using a Wheatstone bridge to obtain raw strain signals; performing low-pass filtering on the raw strain signals to obtain denoised strain data; performing spectral analysis on the denoised strain data using Fourier transform to obtain strain frequency characteristics; and performing stress conversion on the strain frequency characteristics based on pre-acquired material constitutive relations to obtain a stress time series.

[0045] Next, the stress time series is processed to obtain the principal stress direction and magnitude; statistical analysis is performed on the principal stress direction and magnitude to obtain stress amplitude probability distribution data; fatigue life is estimated based on the stress amplitude probability distribution data to obtain component fatigue life prediction data and load time series.

[0046] In one embodiment, rainflow counting analysis is performed on the load time series to generate a fatigue load spectrum, including: preprocessing the load time series and counting rainflows, extracting the amplitude and mean of stress cycles, and generating a fatigue load spectrum.

[0047] Extreme value analysis is performed on the fatigue load spectrum to obtain the ultimate load estimate, including: based on the fatigue load spectrum, mean stress correction and cumulative damage calculation are performed to obtain the linear cumulative damage value.

[0048] Based on the linear cumulative damage value, the extreme value distribution is fitted and extrapolated to obtain the long-term ultimate load prediction value and the corresponding confidence interval; fatigue life is estimated based on the long-term ultimate load prediction value and the corresponding confidence interval to obtain the fatigue life prediction data and ultimate load estimate of the target component.

[0049] Based on the fatigue load spectrum and the estimated ultimate load, performance analysis results are generated, including: a performance analysis report of the wind turbine unit based on the fatigue life prediction data and the estimated ultimate load of the target component.

[0050] This invention provides a performance analysis method for offshore wind turbines. The method collects environmental data using multiple sensors deployed in the target offshore wind farm area, performs time reconstruction and multi-dimensional analysis on the data to identify wind farm operating conditions. Then, it constructs wind farm prediction models for different operating conditions, outputs wind farm parameter prediction results, and uses these to select high-quality wind farm datasets. Finally, it performs load analysis based on this dataset to generate load data for evaluating turbine performance. This invention solves the technical problems of insufficient accuracy in wind farm feature extraction and poor accuracy in wind turbine performance analysis, achieving the technical effect of improving the accuracy of wind farm feature extraction and enhancing the accuracy of wind turbine performance analysis.

[0051] This invention also provides an application example of a performance analysis method for offshore wind turbines, see [link to relevant documentation]. Figure 2 The flowchart shown is another performance analysis method for offshore wind turbines. This method mainly includes the following steps S201 to S206: Step S201: Deploy a sensor network on the target wind turbines in the target offshore wind farm area, and collect data through the sensor network to obtain a wind farm environment dataset.

[0052] The sensor network includes wind speed sensors, wind direction sensors, and turbulence intensity sensors. Specifically, when implementing the performance analysis method for offshore wind turbines, a sensor network is first deployed on the target wind turbines in the target offshore wind farm area. This network can include wind speed sensors, wind direction sensors, and turbulence intensity sensors, which are used to measure wind speed, wind direction, and turbulence intensity, respectively.

[0053] For example, around a typical 5MW offshore wind turbine, multiple ultrasonic anemometers, wind vanes, and 3D ultrasonic anemometers can be installed at different heights and locations. These sensors are connected through a data acquisition system to synchronously collect data at high frequencies (such as 10Hz), forming a wind field environmental dataset.

[0054] In one specific embodiment, the process of performing step S201 may specifically include the following steps: (1) Grid the target offshore wind field area to obtain wind field grid data; (2) Determine the sensor placement location based on the wind field grid data, install and configure the wind speed sensor, wind direction sensor and turbulence intensity sensor to obtain the sensor network, and collect data through the sensor network to obtain the initial environmental dataset; (3) Time calibration of the sensor network is performed using a time synchronization protocol to obtain the synchronous sampling timestamp; (4) The initial environmental dataset is reconstructed into a time series based on the synchronous sampling timestamp to obtain a time-aligned wind field data matrix; (5) The wind field data matrix is ​​decomposed into multiple scales by wavelet transform to obtain wind field feature sets of different frequency components; (6) The wind field feature set is reduced in dimensionality by principal component analysis algorithm to obtain the target wind field feature vector; (7) Cluster analysis of the target wind field feature vectors is performed using a Gaussian mixture model to obtain a set of typical wind field operating conditions; (8) The wind field parameters under typical working conditions of each wind field are modeled by the support vector regression algorithm to obtain the wind field parameter prediction model; (9) The wind field parameter prediction model is optimized by using the particle swarm optimization algorithm to obtain the optimized wind field parameter prediction model; (10) The uncertainty analysis of the optimized wind field parameter prediction model is carried out by Monte Carlo method to obtain the confidence interval of the wind field parameter prediction. The initial environmental dataset is then filtered by the confidence interval to obtain the wind field environmental dataset.

[0055] Specifically, when implementing the performance analysis method for offshore wind turbines, the target offshore wind field area is first divided into grids to obtain wind field grid data. For example, a 10-square-kilometer offshore wind field can be divided into a 100x100 grid, with each grid cell being 100 meters x 100 meters. Based on this wind field grid data, the optimal placement of sensors is determined.

[0056] In this example, sensors can be placed at the four corners of the wind farm, the center point, and around each wind turbine to form a fully covered sensor network.

[0057] The wind speed sensor, wind direction sensor, and turbulence intensity sensor are installed and configured. The wind speed sensor is typically an ultrasonic anemometer, which can accurately measure three-dimensional wind speed; the wind direction sensor uses a wind vane, which can continuously measure the wind direction angle; the turbulence intensity sensor may be a three-dimensional ultrasonic anemometer, which can measure high-frequency wind speed fluctuations. These sensors form a sensor network, through which data is acquired to obtain an initial environmental dataset.

[0058] To ensure data temporal consistency, the sensor network is time-calibrated using a time synchronization protocol (such as Network Time Protocol, NTP) to obtain synchronized sampling timestamps. This step is crucial because in large wind farms, there can be microsecond-level time differences between sensors at different locations, which is significant for high-frequency data acquisition. With the synchronized sampling timestamps, the initial environmental dataset can be reconstructed into a time-series aligned wind farm data matrix. For example, with 100 sensors sampling 10 times per second, one hour's worth of data will form a 100x36000 matrix.

[0059] Next, the wind field data matrix is ​​decomposed into multiple scales using wavelet transform to obtain wind field feature sets with different frequency components.

[0060] Wavelet transform is a time-frequency analysis method that can provide information in both the time and frequency domains simultaneously. In this example, Morlet wavelets can be used to decompose features representing long-term trends, diurnal variations, turbulence, and other characteristics at different scales.

[0061] Then, the wind field feature set is dimensionality reduced by principal component analysis (PCA) algorithm to obtain the target wind field feature vector.

[0062] PCA can identify the main directions of change in data, greatly reducing data dimensionality. For example, it can reduce the original hundreds of dimensions of features to 10-20 dimensions of principal components, which capture the main patterns of wind field changes.

[0063] Next, cluster analysis of the target wind field feature vectors was performed using a Gaussian mixture model (GMM) to obtain a set of typical wind field operating conditions.

[0064] GMM is a probabilistic model capable of describing complex data distributions. In this example, typical operating conditions such as "strong wind stability," "weak wind turbulence," and "sudden wind change" can be identified.

[0065] For the various typical wind field conditions identified, the wind field parameters are modeled using the support vector regression (SVR) algorithm to obtain a wind field parameter prediction model.

[0066] SVR is a powerful nonlinear regression method capable of handling high-dimensional data and complex nonlinear relationships. For example, a model can be built to predict wind field parameters for the next 10 minutes based on current wind speed, wind direction, and turbulence intensity.

[0067] To further improve model performance, the wind field parameter prediction model was optimized using the particle swarm optimization (PSO) algorithm, resulting in an optimized wind field parameter prediction model.

[0068] Prediction Search (PSO) is a heuristic optimization algorithm capable of efficiently searching for optimal solutions in high-dimensional spaces. In this example, PSO can optimize the kernel function parameters and regularization parameters of the SVR to minimize the prediction error.

[0069] Finally, uncertainty analysis was performed on the optimized wind field parameter prediction model using the Monte Carlo method to obtain the confidence intervals for the wind field parameter predictions.

[0070] The Monte Carlo method assesses model uncertainty through extensive random sampling. For example, 10,000 simulations can be performed to obtain a 95% confidence interval for wind speed predictions of ±1.5 m / s. Based on this confidence interval, the initial environmental dataset is filtered to obtain the final wind field environmental dataset.

[0071] This step can eliminate outlier data points that do not meet expectations, improving the reliability of subsequent analysis. For example, if the measured wind speed at a certain moment is far outside the prediction confidence interval, it can be marked as an outlier and removed from the dataset.

[0072] Next, the collected wind field environmental dataset will be preprocessed.

[0073] Step S202: Preprocess the wind field environment dataset to obtain a preprocessed dataset, which includes: wind speed Weibull distribution parameter set, wind direction distribution feature set, and turbulence intensity distribution data.

[0074] The above preprocessing process may include data cleaning, interpolation, and feature extraction.

[0075] During the data cleaning phase, outliers, such as wind speed data that exceeds reasonable limits, are removed. Interpolation is then used to fill in any missing data.

[0076] Feature extraction is a key step. The maximum likelihood estimation method is used to fit the wind speed data to a Weibull distribution to obtain the set of Weibull distribution parameters for wind speed, including scale parameters and shape parameters.

[0077] For example, wind speed data from a certain offshore wind field might yield a Weibull distribution with a scale parameter of 10 m / s and a shape parameter of 2. For wind direction data, cyclic statistical methods are used to analyze the wind direction distribution feature set, including the prevailing wind direction and wind direction dispersion. Turbulence intensity distribution data is obtained by calculating the ratio of the standard deviation to the mean wind speed.

[0078] In one specific embodiment, the process of executing step S202 may specifically include the following steps: (1) Clean the wind field environment dataset to obtain a cleaned dataset; (2) Perform time series interpolation on the cleaned dataset to obtain the interpolated time series dataset, which includes: interpolated wind speed data, interpolated wind direction data and interpolated turbulence intensity data; (3) The interpolated wind speed data is fitted with Weibull distribution parameters by the maximum likelihood estimation method to obtain the wind speed Weibull distribution parameter set, which includes wind speed scale parameters and wind speed shape parameters. (4) The interpolated wind direction data is analyzed by a cyclic statistical algorithm to obtain a wind direction distribution feature set, which includes the dominant wind direction and the wind direction dispersion. (5) Analyze the interpolated turbulence intensity data to obtain turbulence intensity distribution data, which includes turbulence intensity and turbulence scale.

[0079] Specifically, when implementing the performance analysis method for offshore wind turbines, the wind farm environment dataset is first cleaned to obtain a cleaned dataset.

[0080] This step aims to remove outliers and illogical data. For example, in a typical offshore wind field, if a wind speed sensor records an instantaneous wind speed exceeding 50 m / s, this is likely a measurement error, as sustained wind speeds rarely exceed this value, even under typhoon conditions. Similarly, if a wind direction sensor records a wind direction change exceeding 720 degrees in a short period, this could also be due to equipment malfunction. By setting reasonable thresholds and using statistical methods (such as the three-standard-deviation method), these outliers can be identified and removed, resulting in a more reliable cleaned dataset.

[0081] Time series interpolation is performed on the cleaned dataset to obtain an interpolated time series dataset. Interpolation is used to fill in any missing values ​​in the data to ensure the continuity of the time series.

[0082] Commonly used interpolation methods include linear interpolation, spline interpolation, and nearest neighbor interpolation. In this example, cubic spline interpolation can be chosen because it preserves the smoothness of the data and is suitable for handling changes in physical quantities such as wind speed.

[0083] The interpolated time-series dataset includes interpolated wind speed data, interpolated wind direction data, and interpolated turbulence intensity data. For example, if the original data is recorded every 10 minutes, interpolation can yield data points every 1 minute, greatly increasing the temporal resolution of the data.

[0084] Then, the interpolated wind speed data is fitted with Weibull distribution parameters using the maximum likelihood estimation method to obtain the Weibull distribution parameter set for wind speed.

[0085] The Weibull distribution is a commonly used model for describing wind speed distribution, and its parameters include wind speed scale parameters and wind speed shape parameters.

[0086] Maximum likelihood estimation is a statistical method that estimates distribution parameters by maximizing the probability of observed data occurring.

[0087] Next, the interpolated wind direction data is analyzed using a cyclic statistical algorithm to obtain a wind direction distribution feature set. Since the wind direction data is cyclic (0 degrees and 360 degrees are the same), a special statistical method is required. The cyclic statistical algorithm maps the wind direction data onto a unit circle and then calculates the average vector to obtain the prevailing wind direction and wind direction dispersion.

[0088] Finally, the interpolated turbulence intensity data were analyzed to obtain turbulence intensity distribution data.

[0089] Turbulence intensity is an important parameter describing short-term fluctuations in wind speed, usually defined as the ratio of the standard deviation of wind speed to the mean wind speed. Turbulence scale reflects the spatial characteristics of turbulence.

[0090] In practical analysis, the power spectral density (PSD) method can be used to analyze turbulence characteristics. For example, by performing a Fast Fourier Transform (FFT) on high-frequency wind speed data (such as a 10 Hz sampling rate), the spectrum of wind speed fluctuations can be obtained. From this spectrum, the turbulence intensity (e.g., 0.15, indicating that the wind speed fluctuation is approximately 15% of the average wind speed) and the turbulence scale (e.g., 100 meters, representing the typical size of maximum energy turbulence) can be calculated.

[0091] Step S203: Perform fluid dynamic analysis based on the preprocessed dataset to obtain three-dimensional wind field data.

[0092] Based on the preprocessed dataset, hydrodynamic analysis is performed to obtain three-dimensional flow field data for the wind field. This step typically involves computational fluid dynamics (CFD) methods. The computational domain can be discretized using the finite volume method, and then the Navier-Stokes equations are solved using the SIMPLE algorithm to obtain the initial flow field solution. Subsequently, a turbulence model (such as the k-ε model) is applied to correct the initial solution, taking turbulence effects into account. Through these steps, detailed three-dimensional flow field data, including the spatial distribution of wind speed, pressure, and turbulence parameters, can be obtained.

[0093] In one specific embodiment, the process of executing step S203 may specifically include the following steps: (1) Construct the flow field from the preprocessed dataset to obtain the target flow field; (2) Perform flow field solution analysis on the target flow field to obtain the target flow field solution, and generate flow field data based on the target flow field solution; (3) Perform flow field vortex structure analysis on the flow field data to obtain flow field vortex structure distribution data; (4) Perform modal decomposition on the vortex structure distribution data of the flow field to obtain the set of flow field modes; (5) Dynamic feature extraction is performed on the flow field mode set to obtain dynamic feature data of the flow field; (6) The dynamic characteristic data of the flow field are reduced in order to obtain the three-dimensional flow field data of the wind field.

[0094] Specifically, in the performance analysis method of offshore wind turbines, constructing the flow field from the preprocessed dataset is a key step.

[0095] First, using the wind speed, wind direction, and turbulence intensity information in the preprocessed dataset, a three-dimensional computational domain is constructed, which is the target flow field.

[0096] For example, for a 5MW offshore wind turbine, a computational domain of 1km*1km*500m can be constructed, with the wind turbine located at the center. Flow field analysis is then performed on the target flow field to obtain the target flow field solution.

[0097] This step typically involves computational fluid dynamics (CFD) methods. Specifically, the computational domain can be discretized into millions of mesh elements using the finite volume method, and then the Navier-Stokes equations can be solved.

[0098] The solution process employs the SIMPLE (semi-implicit pressure coupling equation) algorithm for iterative calculations. For example, at each time step (possibly 0.1 seconds), the coupling problem between the velocity and pressure fields must be addressed.

[0099] In this way, the wind speed vector and pressure value at each grid point can be obtained, which constitute the target flow field solution. Based on the target flow field solution, flow field data is generated. This includes the velocity field, pressure field, and turbulence characteristics (such as turbulent kinetic energy and dissipation rate) in three-dimensional space. For example, at a point 100 meters in front of a wind turbine, data might be obtained showing a wind speed of 10 m / s, a direction of 225 degrees, and a turbulent kinetic energy of 0.5 m² / s².

[0100] Then, vortex structure analysis was performed on the flow field data to obtain the vortex structure distribution data. Vortex structure is an important feature in the flow field and has a significant impact on the performance of wind turbine units.

[0101] The Q criterion or the λ² criterion can be used to identify vortex structures. For example, when using the Q criterion, the second invariant of the velocity gradient tensor is calculated, and when Q > 0, the point is considered to be located within the vortex core.

[0102] This method allows us to obtain the three-dimensional distribution of vortex structures throughout the flow field, including the location, intensity, and size of each vortex.

[0103] Next, modal decomposition is performed on the vortex structure distribution data of the flow field to obtain the set of flow field modes.

[0104] The Intrinsic Orthogonal Decomposition (POD) method can be used here. POD can decompose a complex flow field into a series of orthogonal spatial modes and corresponding time coefficients. For example, for a flow field dataset containing 1000 time steps, POD can extract 100 principal modes, which are ordered from largest to smallest according to their energy contribution. The first mode may capture large-scale flow field structures, such as the wake behind a wind turbine, while subsequent modes may represent smaller-scale turbulent structures.

[0105] Then, dynamic features are extracted from the flow field mode set to obtain dynamic feature data of the flow field.

[0106] This step can utilize the Dynamic Mode Decomposition (DMD) method. DMD can extract the main dynamic features in the flow field, including periodic structures and transient phenomena at various scales. For example, DMD can identify periodic structures related to the frequency of wind turbine blades, or capture the evolution of large-scale turbulent structures in the atmospheric boundary layer.

[0107] Finally, the dynamic characteristic data of the flow field are reduced in order to obtain the three-dimensional flow field data of the wind field.

[0108] The goal of this step is to significantly reduce the amount of data while retaining key information, making subsequent analysis more efficient.

[0109] Truncated singular value decomposition (SVD) can be used to retain the few modes with the highest energy content. For example, selecting the top 20 modes may contain more than 95% of the flow field energy.

[0110] Through the above series of steps, the original preprocessed dataset was transformed into highly informative three-dimensional wind field flow field data.

[0111] These data not only encompass the overall characteristics of the wind field but also capture complex vortex structures and dynamic properties. For example, it clearly shows how the wake generated by the wind turbines evolves over time and how these wakes affect other downstream wind turbines.

[0112] This detailed flow field information is crucial for accurately evaluating wind turbine performance. It helps to better understand the operating status of wind turbines under different wind conditions, predict potential fatigue loads, and provide a basis for optimizing wind farm layout. For example, by analyzing the vortex structure distribution, the spacing between adjacent wind turbines can be optimized to minimize wake interference. By studying the dynamic characteristics of the flow field, the control strategies of wind turbines can be improved, enabling them to better adapt to rapidly changing wind conditions. Overall, this high-precision flow field analysis method provides strong technical support for the performance optimization and reliability improvement of offshore wind turbines.

[0113] Step S204: Perform aerodynamic load simulation analysis based on the three-dimensional flow field data of the wind field to obtain the dynamic response data of the target offshore wind field area.

[0114] With three-dimensional wind field data, aerodynamic load simulation analysis can be performed to obtain dynamic response data of the target wind turbine under different operating conditions. This process typically uses blade element momentum theory and multibody dynamics methods.

[0115] First, the flow field data is interpolated onto the stress surface of the wind turbine blades, and then the angle of attack and relative wind speed of each blade element are calculated. By referring to pre-prepared airfoil aerodynamic characteristic data, the lift coefficient and drag coefficient of each blade element can be obtained. Integrating these forces along the blade yields the aerodynamic forces and moments of the entire blade.

[0116] For example, at a rated wind speed of 12 m / s, a single blade of a 5MW wind turbine may generate approximately 100 kN of thrust and 2000 kN·m of torque.

[0117] In one specific embodiment, the process of executing step S204 may specifically include the following steps: (1) Interpolate the three-dimensional flow field data of the wind field to obtain the flow field data of the turbine blades of the target wind turbine unit; (2) Analyze the flow field data of the turbine blades to obtain the blade angle of attack and relative wind speed of the target wind turbine; (3) Based on the pre-collected airfoil characteristic data, the blade angle of attack and relative wind speed are analyzed by table lookup to obtain the aerodynamic coefficients; (4) The aerodynamic coefficients are summed along the spanwise direction of the blades of the target wind turbine to obtain the overall aerodynamic force and moment of the blades; (5) Perform coordinate system transformation on the overall aerodynamic force and torque of the blades to obtain the wind turbine thrust and wind turbine torque corresponding to the blades of the target wind turbine; (6) Perform transmission analysis on wind turbine thrust and wind turbine torque to obtain main shaft load and generator torque; (7) Perform spectral analysis on the spindle load and generator torque to obtain the load frequency components and load amplitude distribution; (8) Perform structural dynamics analysis on the load frequency components to obtain stress and deformation response data of the components; (9) Perform statistical analysis on the load and deformation response data to obtain the extreme value distribution and fatigue load spectrum; (10) Perform data fitting on the extreme value distribution and fatigue load spectrum to obtain dynamic response data.

[0118] Specifically, in the performance analysis method of offshore wind turbines, the three-dimensional flow field data of the wind field is first interpolated to obtain the flow field data of the turbine blades of the target wind turbine.

[0119] This step uses a three-dimensional interpolation algorithm, such as trilinear interpolation, to map discrete flow field data points onto a grid on the blade surface. For example, for a 5MW wind turbine with a blade length of 61.5 meters, 100 control points can be set on the blade surface, and accurate flow field data can be obtained at each control point.

[0120] By analyzing the flow field data of the turbine blades, the blade angle of attack and relative wind speed of the target wind turbine were obtained.

[0121] This step involves the application of blade element momentum theory. Specifically, it requires considering the vector combination of the blade rotation speed and the inflow wind speed to calculate the relative wind speed and angle of attack at each control point. For example, at a control point at 75% of the blade radius, when the wind speed is 10 m / s and the blade rotation speed is 12 rpm, the relative wind speed might be 70 m / s and the angle of attack might be 5 degrees.

[0122] Then, based on the pre-collected airfoil characteristic data, the blade angle of attack and relative wind speed are analyzed by looking up tables to obtain the aerodynamic coefficients.

[0123] The airfoil characteristic data here are usually obtained through wind tunnel testing or CFD simulation, including lift and drag coefficients at different angles of attack. For example, for the NACA64-618 airfoil, at an angle of attack of 5 degrees, the lift coefficient may be 1.0 and the drag coefficient may be 0.008.

[0124] Next, the aerodynamic coefficients are accumulated along the spanwise direction of the target wind turbine blades to obtain the overall aerodynamic force and moment of the blades.

[0125] This step is actually a numerical integration process. The blade is divided into multiple infinitesimal segments, the aerodynamic force of each segment is calculated, and then the result is integrated along the blade length. For example, a 61.5-meter-long blade may generate approximately 150 kN of thrust and 3000 kN·m of torque at a wind speed of 12 m / s.

[0126] Subsequently, coordinate system transformation was performed on the overall aerodynamic force and torque of the blades to obtain the rotor thrust and rotor torque corresponding to the blades of the target wind turbine.

[0127] This involves a transformation from the blade coordinate system to the rotor coordinate system, which requires consideration of the blade's pitch angle and cone angle. For example, the resultant force of three blades may generate a rotor thrust of 450 kN and a rotor torque of 9000 kN·m.

[0128] Next, the transmission analysis of the wind turbine thrust and torque is performed to obtain the main shaft load and generator torque.

[0129] This step requires consideration of the transmission system's efficiency and gear ratio. For example, for a transmission system with a gear ratio of 1:100, the torque on the generator shaft might be 90 kN·m.

[0130] Then, spectral analysis was performed on the spindle load and generator torque to obtain the load frequency components and load amplitude distribution.

[0131] This step typically uses the Fast Fourier Transform (FFT) algorithm. It can be found that there are major frequency components in the load that are related to the blade passage frequency (1P) and the number of blades multiplied by the passage frequency (3P).

[0132] Next, structural dynamics analysis was performed on the load frequency components to obtain stress and deformation response data for the components.

[0133] This step typically uses the finite element method, considering the geometry, material properties, and load conditions of each component of the wind turbine. For example, it can be found that the maximum stress at the blade root under extreme load is 200 MPa, and the maximum displacement at the top of the tower is 0.5 meters.

[0134] Subsequently, statistical analysis was performed on the load and deformation response data to obtain the extreme value distribution and fatigue load spectrum.

[0135] Extreme value distributions are usually described by the generalized extreme value distribution (GEV), while the fatigue load spectrum is obtained by rainflow counting. For example, it can be concluded that the ultimate load at the blade root, which occurs once every 50 years, is 300 kN·m, while under normal operating conditions, it will experience 107 cycles of stress amplitude of 50 MPa per year.

[0136] Finally, the extreme value distribution and fatigue load spectrum are fitted to obtain dynamic response data.

[0137] This step may involve fitting various mathematical models, such as using the Weibull distribution to fit the fatigue load spectrum and the generalized Pareto distribution to fit the extreme value data. These fitted models can be used to predict the dynamic response of wind turbines under different operating conditions.

[0138] Step S205: Perform strain tests on the key components of the target wind turbine unit based on the dynamic response data to obtain the load time series.

[0139] Next, strain tests are performed on key components of the wind turbine based on the dynamic response data to obtain the load time series. This typically involves installing strain gauges at critical locations such as the blade root, main shaft, and tower. Specifically, a Wheatstone bridge circuit can be used to convert the strain into an electrical signal, which is then amplified and filtered to obtain the strain time series. For example, at the blade root, a periodic load with a peak strain of 500 microstrains can be measured.

[0140] In one specific embodiment, the process of executing step S205 may specifically include the following steps: (1) Finite element analysis was performed on the key components of the target wind turbine to obtain the stress concentration areas; (2) Determine the strain gauge arrangement position based on the stress concentration area to obtain the strain test point distribution map; (3) Based on the strain test point distribution map, the strain test points are surface treated and strain gauges are pasted to obtain the strain measurement system; (4) The strain measurement system is used to acquire signals through a Wheatstone bridge to obtain the original strain signal; (5) Perform low-pass filtering on the original strain signal to obtain denoised strain data; (6) Perform spectral analysis on the denoised strain data using Fourier transform to obtain strain frequency characteristics; (7) Based on the pre-obtained material constitutive relations, stress transformation is performed on the strain frequency characteristics to obtain the stress time series; (8) Process the stress time series to obtain the principal stress direction and magnitude; (9) Statistical analysis of the principal stress direction and magnitude is performed to obtain the stress amplitude probability distribution data; (10) The fatigue life is estimated by the stress amplitude probability distribution data to obtain the component fatigue life prediction data and load time series.

[0141] Specifically, in the performance analysis method of offshore wind turbines, the key components of the target wind turbine are first analyzed by finite element analysis to obtain the stress concentration areas.

[0142] Taking a 5MW wind turbine as an example, we can focus on key components such as the blade root, main shaft, gearbox, and tower. Using finite element software such as ANSYS, we can establish detailed three-dimensional models of these components and apply the load conditions obtained from the aforementioned flow field analysis. The stress distribution of each component can then be obtained by solving the stress-strain equations.

[0143] For example, at the blade root, the maximum von Mises stress may be found to reach 200 MPa, located at the junction of the blade root and the hub. The strain gauge placement is determined based on the stress concentration area, resulting in a strain test point distribution map.

[0144] Select several test points in and around the area of ​​highest stress. For example, select 10 test points at the blade root, 8 test points on the main shaft, and 12 test points at the bottom of the tower. The specific locations and orientations of these test points need to be determined based on the direction of the principal stress to obtain the most sensitive measurement results.

[0145] Next, based on the strain test point distribution map, surface treatment and strain gauge attachment are performed on the strain test points to obtain the test strain measurement system. Surface treatment includes steps such as grinding, cleaning, and applying a primer to ensure that the strain gauges are in close contact with the surface being measured.

[0146] For offshore wind turbines, waterproofing and corrosion protection are also necessary. Suitable epoxy resin adhesive is used to bond strain gauges, followed by curing. For example, HBM's LY41 high-strength strain gauges with a sensitivity factor of 2.05 are used on the blades.

[0147] Then, the strain measurement system is used to acquire signals to obtain the raw strain signal. A Wheatstone bridge is a precision resistance measurement circuit that can convert minute strain changes into measurable voltage signals.

[0148] In this example, a quarter-bridge method can be used, forming a bridge circuit with a single strain gauge and three precision resistors. A 10V excitation voltage is used, and signal acquisition is performed through a high-precision data acquisition system (such as NI's PXI system), with a sampling frequency set to 1000Hz to capture high-frequency dynamic responses. The raw strain signal is low-pass filtered to obtain denoised strain data. A Butterworth low-pass filter is used with a cutoff frequency set to 100Hz to remove high-frequency noise while retaining meaningful signal components. This step effectively improves the signal-to-noise ratio, providing a more reliable data foundation for subsequent analysis. Spectral analysis of the denoised strain data is performed using Fourier transform to obtain strain frequency characteristics. The time-domain signal is converted to a frequency-domain representation using a Fast Fourier Transform (FFT) algorithm. Significant frequency components related to wind turbine operating characteristics can be observed in the spectrum, such as 1P (blade rotation frequency, approximately 0.2Hz) and 3P (blade passage frequency, approximately 0.6Hz). Stress transformation is performed on the strain frequency characteristics based on the pre-obtained material constitutive relation to obtain a stress time series. Constitutive relations describe the stress-strain relationship of a material. For linear elastic materials, Hooke's law (σ=Eε, where E is Young's modulus) can be used simply. For example, for steel, E is approximately 210 GPa, so the stress corresponding to 1000 microstrains is 210 MPa.

[0149] The stress time series is processed to obtain the direction and magnitude of the principal stresses. The principal stresses and their directions are then calculated using stress tensor theory. For example, at the blade root, the maximum principal stress might be found to be 180 MPa, with its direction at a 45-degree angle to the blade axis.

[0150] Statistical analysis was performed on the principal stress direction and magnitude to obtain the stress amplitude probability distribution data.

[0151] The stress time series was cyclically counted using the rainflow counting method to obtain the stress amplitude-cycle number relationship.

[0152] Then, the probability density function of the stress amplitude is obtained through methods such as histogram or kernel density estimation.

[0153] Finally, fatigue life is estimated using the stress amplitude probability distribution data to obtain component fatigue life prediction data and load time series.

[0154] The cumulative damage is calculated using Miner's linear cumulative damage theory, combined with the material's SN curve (stress-cycle curve).

[0155] For example, for the blade root, it can be concluded that under a given load spectrum, the expected fatigue life is 20 years, which is less than the design life of 25 years, so design optimization or enhanced maintenance may be required.

[0156] Step S206: Perform rainflow counting analysis on the load time series to obtain the fatigue load spectrum, and fit the fatigue load spectrum with a generalized extreme value distribution to obtain the ultimate load estimate; generate the performance analysis report of the target wind turbine based on the ultimate load estimate.

[0157] Finally, rainflow counting analysis was performed on the load time series to obtain the fatigue load spectrum.

[0158] Rainflow counting is a widely used fatigue analysis method that can convert complex load time series into a series of complete stress-strain cycles.

[0159] The obtained fatigue load spectrum is fitted with a generalized extreme value distribution to estimate the ultimate load. For example, through analysis, it can be concluded that the blade root of this wind turbine may withstand an ultimate load of 1000kN under extreme wind conditions that occur once every 50 years.

[0160] In one specific embodiment, the process of executing step S206 may specifically include the following steps: (1) Perform data smoothing on the load time series to obtain denoised load data; (2) Perform rainflow counting on the denoised load data to obtain the stress cycle counting matrix; (3) Perform statistical analysis on the stress cycle counting matrix to obtain the stress amplitude probability density distribution; (4) The stress amplitude probability density distribution is corrected by the mean stress effect to obtain the equivalent completely reverse stress spectrum; (5) Perform cumulative damage calculation on the equivalent fully reverse stress spectrum to obtain the linear cumulative damage value; (6) The extreme value distribution parameters are obtained by fitting the equivalent fully reverse stress spectrum based on the linear cumulative damage value; (7) Perform Monte Carlo simulation based on the extreme value distribution parameters to obtain the predicted value of the long-term ultimate load; (8) Perform confidence interval analysis on the predicted long-term ultimate load to obtain the confidence interval of the ultimate load; (9) Estimate the fatigue life of the component by calculating the confidence interval of the ultimate load, and obtain the component fatigue life prediction data and the ultimate load estimate. (10) Generate a performance analysis report of the target wind turbine based on the component fatigue life prediction data and the ultimate load estimate.

[0161] Specifically, in step (1) above, the load time series can be smoothed to obtain denoised load data. This step can use a moving average filter to effectively remove high-frequency noise by averaging the data points within a time window, while preserving the main load variation trend. For the blade root bending moment data of a 5MW wind turbine, this processing can significantly reduce short-term fluctuations and make the long-term trend more obvious.

[0162] Then, rainflow counting is performed on the denoised load data to obtain the stress cycle counting matrix.

[0163] The core idea of ​​rainflow counting is to transform irregular load time series into a series of closed stress-strain cycles. This method simulates the process of rainwater flowing down the load-time curve, effectively identifying and statistically analyzing stress cycles of various amplitudes. Through this step, the complex load history is transformed into a cycle counting matrix that is easy to analyze.

[0164] Furthermore, statistical analysis is performed on the stress cycle count matrix to obtain the stress amplitude probability density distribution. This step uses kernel density estimation to transform the discrete cycle count data into a continuous probability density function. Kernel density estimation estimates the probability density by placing a kernel function (usually a Gaussian function) around each data point and superimposing these functions. This method can smoothly describe the distribution characteristics of stress amplitude, which is helpful for subsequent fatigue analysis.

[0165] Next, the stress amplitude probability density distribution is corrected for mean stress effect to obtain an equivalent fully inverse stress spectrum. This step considers the influence of mean stress on fatigue life, converting stress cycles under different mean stresses into equivalent fully inverse stress cycles. This conversion allows fatigue damage under different stress states to be compared and accumulated on the same benchmark.

[0166] Subsequently, cumulative damage calculations are performed on the equivalent fully inverse stress spectrum to obtain linear cumulative damage values. This step is based on the linear cumulative damage theory, which linearly superimposes fatigue damage under different stress levels. This method simplifies the actual damage accumulation process in some cases and provides a practical and widely accepted method for estimating fatigue life.

[0167] The equivalent fully inverse stress spectrum is fitted based on the linear cumulative damage values ​​to obtain the extreme value distribution parameters. Here, a generalized extreme value distribution is used for fitting, which can describe the probabilistic characteristics of extreme events. The fitting process typically employs the maximum likelihood estimation method to obtain the distribution parameters that best describe the observed data.

[0168] Monte Carlo simulations are performed based on extreme value distribution parameters to obtain long-term ultimate load predictions. The Monte Carlo method simulates the behavior of complex systems through extensive random sampling. Here, it is used to generate a large number of extreme event samples, thereby predicting the maximum load that may occur in the long term (e.g., 50 years).

[0169] Confidence interval analysis is performed on the long-term ultimate load predictions to obtain the confidence intervals for the ultimate loads. This step uses the Bootstrap method, which repeatedly samples the original data and calculates statistics to obtain the uncertainty range of the predictions. This method does not rely on specific distribution assumptions and can provide robust confidence interval estimates.

[0170] Then, fatigue life is estimated based on the confidence interval of the ultimate load, yielding component fatigue life prediction data and ultimate load estimates. This step combines the material's fatigue characteristics (SN curve) and cumulative damage theory to predict the component's expected life under a given load spectrum.

[0171] At the same time, the safety of the components under extreme conditions is evaluated through ultimate load analysis.

[0172] Finally, based on the component fatigue life prediction data and ultimate load estimates, a performance analysis report for the target wind turbine is generated. This report integrates all the aforementioned analysis results, providing a comprehensive assessment of the wind turbine's performance and reliability. It includes not only quantitative life predictions and load estimates, but also maintenance recommendations and risk assessments based on this data.

[0173] Based on this, the performance analysis method for offshore wind turbines provided in this embodiment of the invention achieves comprehensive and detailed monitoring of the wind field environment by deploying a multi-sensor network including wind speed sensors, wind direction sensors, and turbulence intensity sensors in the target offshore wind field area. This overcomes the limitations of traditional single-point measurement and provides a rich and accurate raw data foundation for subsequent analysis.

[0174] The collected wind field environmental dataset was preprocessed to obtain a preprocessed dataset including the Weibull distribution parameter set of wind speed, the feature set of wind direction distribution, and the turbulence intensity distribution data. This step not only improved the data quality but also extracted the key features of the wind field, laying a solid foundation for subsequent hydrodynamic analysis.

[0175] Fluid dynamic analysis based on preprocessed datasets yields three-dimensional wind field data. This process fully considers the complexity and variability of offshore wind fields, significantly improving the accuracy and reliability of wind field simulation.

[0176] By using the obtained three-dimensional wind field data to perform aerodynamic load simulation analysis, dynamic response data of the target offshore wind field area is obtained. This step realizes the accurate mapping from wind field characteristics to wind turbine response, providing a more realistic and reliable basis for unit performance evaluation.

[0177] Strain tests were conducted on key components of the target wind turbine based on dynamic response data to obtain load time series. This process directly reflects the stress state of the unit under actual operating conditions, providing first-hand data for subsequent fatigue analysis and ultimate load estimation.

[0178] Rainflow counting analysis was performed on the load time series to obtain the fatigue load spectrum, and the ultimate load estimate was obtained by fitting the generalized extreme value distribution. These steps not only considered the long-term fatigue effect of the unit, but also evaluated the ultimate load under extreme conditions, comprehensively improving the accuracy of unit life prediction and safety assessment. Finally, a performance analysis report of the target wind turbine was generated based on the above analysis results, providing a scientific and comprehensive basis for operation decision-making and maintenance strategy formulation.

[0179] Based on the same inventive concept, this invention also provides a performance analysis device for offshore wind turbines, which mainly includes the following parts: The multi-sensor network acquisition module is used to acquire the initial environmental dataset collected by the multi-sensor network in the target offshore wind field area; The feature extraction module is used to perform temporal reconstruction and multi-scale feature extraction on the initial environmental dataset to obtain the target wind field features and the wind field condition type set; the wind field condition type set contains at least one wind field condition type. The wind field parameter prediction module is used to input the target wind field features into a pre-built wind field parameter prediction model corresponding to the wind field condition type for each wind field condition type, output the prediction results of the wind field parameters, and filter the initial environmental dataset based on the results to obtain the wind field environmental dataset. The performance load analysis module is used to perform performance load analysis based on wind farm environment datasets to obtain load analysis data for evaluating the performance of offshore wind turbines.

[0180] The performance analysis device for offshore wind turbines provided in this embodiment of the invention can be specific hardware on the equipment or software or firmware installed on the equipment. The implementation principle and technical effects of the device provided in this embodiment of the invention are the same as those in the foregoing method embodiments. For the sake of brevity, any parts not mentioned in the device embodiments can be referred to the corresponding content in the foregoing method embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0181] As a specific example, this invention also provides an application example of a performance analysis device for offshore wind turbines, see [link to relevant documentation]. Figure 3 As shown, the device includes: Deployment module 301 is used to deploy a sensor network in the target offshore wind field area. The sensor network includes wind speed sensors, wind direction sensors and turbulence intensity sensors, and collects data through the sensor network to obtain a wind field environment dataset. Processing module 302 is used to preprocess the wind field environment dataset to obtain a preprocessed dataset, which includes: wind speed Weibull distribution parameter set, wind direction distribution feature set and turbulence intensity distribution data; Analysis module 303 is used to perform fluid dynamic analysis based on the preprocessed dataset to obtain three-dimensional wind field data; Modeling module 304 is used to perform aerodynamic load simulation analysis based on three-dimensional wind field data to obtain dynamic response data of the target offshore wind field region. Test module 305 is used to perform strain tests on key components of the target wind turbine based on dynamic response data to obtain load time series; The fitting module 306 is used to perform rainflow counting analysis on the load time series to obtain the fatigue load spectrum, and to fit the fatigue load spectrum with a generalized extreme value distribution to obtain the ultimate load estimate. Based on the ultimate load estimate, a performance analysis report of the target wind turbine is generated.

[0182] Based on the same inventive concept, embodiments of the present invention also provide an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program, when run by the processor, executes the method described in any of the above embodiments.

[0183] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 400 includes: a processor 410, a memory 420, a communication interface 430, and a bus 440. The memory 420 stores machine-readable instructions that can be executed by the processor 410. When the electronic device is running, the processor 410 communicates with the memory 420 through the bus 440. The processor 410 executes the machine-readable instructions to perform the steps of the method described above.

[0184] Specifically, the memory 420 and processor 410 can be general-purpose memory and processor, without any specific limitations. When the processor 410 runs the computer program stored in the memory 420, it can execute the above method.

[0185] Processor 410 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 410 or by instructions in software form. The processor 410 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 420, and processor 410 reads the information from memory 420 and, in conjunction with its hardware, completes the steps of the above method.

[0186] Corresponding to the above method, this embodiment of the invention also provides a computer-readable storage medium storing machine-executable instructions. When the machine-executable instructions are called and run by a processor, the machine-executable instructions cause the processor to perform the steps of the above method.

[0187] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0188] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0189] Furthermore, in the various embodiments of the present invention, the functional modules can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0190] It should be noted that if the functionality is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0191] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0192] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A performance analysis method for offshore wind turbines, characterized in that, include: Obtain the initial environmental dataset collected by a multi-sensor network in the target offshore wind field region; Temporal reconstruction and multi-scale feature extraction are performed on the initial environmental dataset to obtain target wind field features and a set of wind field operating condition types; the set of wind field operating condition types includes at least one wind field operating condition type. For each wind field condition type, the target wind field features are input into a pre-constructed wind field parameter prediction model corresponding to the wind field condition type, the prediction results of the wind field parameters are output, and the initial environmental dataset is filtered according to the prediction results to obtain a wind field environmental dataset. Performance load analysis is performed based on the wind farm environment dataset to obtain load analysis data for evaluating the performance of offshore wind turbines.

2. The method according to claim 1, characterized in that, The multi-sensor network includes: a wind speed sensor, a wind direction sensor, and a turbulence intensity sensor; the method further includes: The target offshore wind field area is divided into grids to obtain wind field grid data; Based on the wind field grid data and historical data, the placement and quantity of each type of sensor are determined.

3. The method according to claim 1, characterized in that, The initial environmental dataset is reconstructed over time and multi-scale features are extracted to obtain target wind field features and a set of wind field operating condition types, including: The sensor network is time-calibrated using a time synchronization protocol to obtain a synchronized sampling timestamp. The initial environmental dataset is reconstructed temporally based on the synchronous sampling timestamp to obtain a time-aligned wind field data matrix; Multi-scale decomposition and principal component analysis are performed on the wind field data matrix to obtain the target wind field characteristics; Cluster analysis is performed on the target wind field characteristics to obtain a set of wind field operating condition types.

4. The method according to claim 1, characterized in that, For each wind field condition type, the target wind field features are input into a pre-constructed wind field parameter prediction model corresponding to the wind field condition type, and the predicted wind field parameters are output. Based on the prediction results, the initial environmental dataset is filtered to generate a wind field environmental dataset, including: The wind field parameters for each wind field condition type in the wind field condition type set are modeled using the support vector regression algorithm to obtain the corresponding wind field parameter prediction model. The input of the wind field parameter prediction model is the target wind field feature, and the output is the predicted value of the wind field parameter. The predicted value of the wind field parameter includes: wind speed prediction value, wind direction prediction value, and turbulence intensity prediction value. The wind field parameter prediction model is optimized and uncertainty analysis is performed to obtain the confidence intervals of the predicted values ​​of each wind field parameter. The initial environmental dataset is filtered based on the confidence interval to obtain the wind field environmental dataset.

5. The method according to claim 1, characterized in that, Prior to the step of performing performance load analysis based on the wind field environmental dataset, the method further includes: The wind field environment dataset is preprocessed to obtain a preprocessed wind field environment dataset; the preprocessed wind field environment dataset includes: wind speed Weibull distribution parameter set, wind direction distribution feature set, and turbulence intensity distribution data.

6. The method according to claim 5, characterized in that, Performance load analysis is performed based on the aforementioned wind field environment dataset, including: Based on the preprocessed wind field environment dataset, hydrodynamic analysis was performed to obtain three-dimensional wind field flow field data. Based on the three-dimensional flow field data of the wind field, aerodynamic load simulation analysis is performed to obtain the dynamic response data of the target offshore wind field area; Based on the dynamic response data, strain tests were performed on the key components of the wind turbine to obtain a load time series. Rainflow counting analysis was performed on the load time series to generate a fatigue load spectrum; Extreme value analysis was performed on the fatigue load spectrum to obtain the estimated value of the ultimate load; Based on the fatigue load spectrum and the estimated ultimate load, performance analysis results are generated.

7. The method according to claim 6, characterized in that, Rainflow counting analysis is performed on the load time series to generate a fatigue load spectrum, including: preprocessing the load time series and counting rainflows, extracting the amplitude and mean of stress cycles, and generating a fatigue load spectrum; Extreme value analysis is performed on the fatigue load spectrum to obtain the estimated ultimate load, including: Based on the fatigue load spectrum, mean stress correction and cumulative damage calculation are performed to obtain the linear cumulative damage value; Based on the linear cumulative damage value, the extreme value distribution is fitted and extrapolated to obtain the long-term ultimate load prediction value and the corresponding confidence interval. Based on the long-term ultimate load prediction value and the corresponding confidence interval, fatigue life is estimated to obtain fatigue life prediction data and ultimate load estimate of the target component. Based on the fatigue load spectrum and the estimated ultimate load, a performance analysis result is generated, including: generating a performance analysis report for the wind turbine based on the fatigue life prediction data of the target component and the estimated ultimate load.

8. A performance analysis device for offshore wind turbines, characterized in that, include: The multi-sensor network acquisition module is used to acquire the initial environmental dataset collected by the multi-sensor network in the target offshore wind field area; The feature extraction module is used to perform temporal reconstruction and multi-scale feature extraction on the initial environmental dataset to obtain target wind field features and a set of wind field operating condition types; the set of wind field operating condition types includes at least one wind field operating condition type. The wind field parameter prediction module is used to input the target wind field features into a pre-constructed wind field parameter prediction model corresponding to the wind field condition type for each wind field condition type, output the prediction results of the wind field parameters, and filter the initial environmental dataset according to the results to obtain a wind field environmental dataset. The performance load analysis module is used to perform performance load analysis based on the wind field environment dataset to obtain load analysis data for evaluating the performance of offshore wind turbines.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images