Wind speed reduction method and device for free flow at front end of cabin of wind turbine generator and medium

By preprocessing Scada data from wind turbines and training convolutional neural networks, a wind speed restoration model was constructed, which solved the problem of wind speed measurement error caused by wind energy absorption by wind turbine blades and improved the accuracy of wind power characteristic testing and performance evaluation.

CN118965166APending Publication Date: 2024-11-15CSIC HAIZHUANG WINDPOWER CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411037498.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In existing technologies, because the blades of wind turbines absorb some wind energy during rotation, the wind speed measured by the anemometer is less than the actual wind speed, resulting in errors in wind power characteristic testing and performance evaluation.

Method used

By acquiring Scada data from the wind turbine, performing data preprocessing and outlier detection, and using a convolutional neural network to train a free-flow wind speed restoration model at the front end of the nacelle, a 16-layer convolutional neural network model was constructed and trained using feature factors and standardization to restore the wind speed.

Benefits of technology

It significantly improves the accuracy of wind power characteristic testing and performance evaluation, and reduces errors caused by wind speed differences due to wind energy absorption by the blades.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965166A_ABST
    Figure CN118965166A_ABST
Patent Text Reader

Abstract

The invention provides a wind turbine generator cabin front-end free flow wind speed reduction model method, which comprises the following steps: S1, acquiring fan Scada data and carrying out data preprocessing to obtain a Scada data set; s2, based on the Scada data set, training a cabin front-end free flow wind speed reduction model by using a convolutional neural network; and S3, acquiring real-time data of the fan in the wind field, and inputting the real-time data into the free flow wind speed reduction model at the front end of the cabin to obtain a reduced wind speed value at the front end of the fan. According to the invention, the problems in the prior art that the wind speed measured by an anemometer is usually smaller than the actually transmitted wind speed due to the fact that blades of the wind generating set absorb a part of wind energy in the rotating process, and the wind speed difference is caused can be solved; therefore, large errors exist in the subsequent wind power characteristic test, model establishment and performance evaluation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of wind speed prediction, and in particular to a method, a device and a medium for restoring the free flow wind speed at the front end of a nacelle of a wind turbine generator set. Background Art

[0002] In the field of wind power generation, wind power generation equipment is a key facility for converting wind energy into electrical energy. It is mainly composed of wind turbines, towers, generators, control systems and other parts. As the core component, the wind turbine captures wind energy through its blades and converts it into mechanical energy, which in turn drives the generator to generate electricity. Among them, the free-stream wind speed at the front of the nacelle refers to the wind speed of the wind turbine at the front of the nacelle (that is, before the wind enters the nacelle) when it is not affected by the wind turbine itself (such as blades, towers, etc.). In the field of wind power generation, this parameter is crucial for accurately evaluating the performance of wind turbines, establishing wind power characteristic tests, and constructing accurate mathematical and physical models. Therefore, in the operation of wind turbines, the free-stream wind speed at the front of the nacelle is a crucial parameter.

[0003] In the prior art, an anemometer (such as the Scada system) is usually installed above the nacelle to monitor the wind speed in order to monitor and adjust the operation of the wind turbine. However, since the blades of the wind turbine absorb part of the wind energy during rotation, the wind speed measured by the anemometer is often lower than the actual incoming wind speed, that is, the free stream wind speed at the front of the nacelle. This difference may introduce large errors in subsequent wind power characteristic tests, model building, and performance evaluation.

[0004] In view of this, the inventor invented a method, device and medium for restoring the free flow wind speed at the front end of a nacelle of a wind turbine. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a method for restoring the free flow wind speed at the front end of the nacelle of a wind turbine generator set, so as to solve the technical problem in the prior art that the wind speed measured by the anemometer is often lower than the actual transmitted wind speed due to the fact that the blades of the wind turbine generator set absorb part of the wind energy during the rotation process, resulting in a wind speed difference, which leads to large errors in subsequent wind power characteristic testing, model building and performance evaluation.

[0006] In a first aspect, the present invention provides a method for restoring a free flow wind speed model at the front end of a nacelle of a wind turbine generator set, comprising:

[0007] S1, obtaining wind turbine Scada data and performing data preprocessing to obtain a Scada data set;

[0008] S2, based on the Scada dataset, a convolutional neural network is used to train the free stream wind speed restoration model at the front end of the cabin;

[0009] S3. Acquire the real-time data of the wind turbines in the wind farm, input the data into the free-flow wind speed restoration model at the front end of the nacelle, and obtain the wind speed value at the front end of the wind turbine after restoration.

[0010] Furthermore, the wind turbine Scada data is obtained and preprocessed to obtain a Scada data set, and the wind turbine Scada data is tested for accuracy and outliers, and missing values ​​are filled.

[0011] Furthermore, the outlier test includes:

[0012] Eliminate wind speed, fan speed, power where any one of them is less than or equal to 0;

[0013] When the screening wind speed is less than the cut-in wind speed, the power is greater than 0, and when the wind speed is greater than the cut-out wind speed, the power is greater than 0;

[0014] Eliminate data with power greater than 1.2 times the rated power.

[0015] Furthermore, the missing value filling includes:

[0016] Determine whether the data integrity rate is less than 95%;

[0017] If the data completeness rate is less than 95%, the missing data will be filled using the neighboring interpolation filling method;

[0018] After completing the interpolated data, perform outlier detection. If there are no outliers in the test results, complete the missing value filling.

[0019] Furthermore, the method of using a convolutional neural network to train a free flow wind speed restoration model at the front end of the cabin based on the Scada data set includes:

[0020] S21, extract characteristic factors based on Scada data set;

[0021] S22, taking the characteristic factors as input, and using a convolutional neural network to construct a free stream wind speed restoration model at the front end of the cabin;

[0022] S23. Divide the Scada data set into a training set and a test set, and train to obtain a free stream wind speed restoration model at the front end of the cabin.

[0023] Furthermore, the characteristic factors include: NFT wind speed results, unit Scada wind speed, wind direction, pitch angle, generator power, impeller diameter, hub height, terrain complexity steepness index, air temperature, air pressure and air density.

[0024] Furthermore, the free stream wind speed restoration model at the front end of the cabin is trained using a convolutional neural network, a gradient descent algorithm is used for learning, and a standardization process is performed.

[0025] Furthermore, the specific formula for the standardization and normalization process is as follows:

[0026]

[0027] Among them, μ B is the average value of characteristic factors, σ 2 is the characteristic factor variance, is the normalized value.

[0028] Furthermore, the free stream wind speed restoration model at the front end of the cabin adopts 16 convolution layers, 1※3 convolution kernels, the pooling layer adopts average pooling, the activation function adopts the Reul function, the loss function adopts the root mean square error, and the number of training iterations is 50 rounds.

[0029] In a second aspect, the present invention provides an electronic device, comprising a processor and a memory connected to the processor for storing executable instructions for the processor, wherein the processor is used to execute the free flow wind speed restoration model method for the front end of the cabin proposed in the first aspect.

[0030] In a third aspect, the present invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the free flow wind speed restoration model method for the front end of the cabin described in the first aspect.

[0031] It can be seen from the above technical solution that the beneficial technical effects of the present invention are as follows:

[0032] The present invention proposes a method, device and medium for restoring a free-stream wind speed model at the front end of a nacelle of a wind turbine generator set, wherein the method acquires and preprocesses Scada data of the wind turbine, and constructs and trains a free-stream wind speed restoration model at the front end of the nacelle in combination with a convolutional neural network model, thereby accurately restoring the actual wind speed value at the front end of the nacelle, thereby significantly improving the accuracy and reliability of subsequent wind power characteristic tests, so as to solve the technical problem in the prior art that the wind speed measured by the anemometer is often less than the actual incoming wind speed due to the fact that the blades of the wind turbine generator set absorb part of the wind energy during rotation, resulting in a wind speed difference, which causes large errors in subsequent wind power characteristic tests, model establishment and performance evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the prior art description. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0034] Figure 1This is a schematic diagram of the overall process of a method for restoring a free-flow wind speed model at the front end of a nacelle of a wind turbine according to Embodiment 1 of the present invention;

[0035] Figure 2 This is a schematic diagram of a model building process proposed by a method for restoring a free-flow wind speed model at the front end of a nacelle of a wind turbine according to Embodiment 1 of the present invention;

[0036] Figure 3 This is a schematic diagram of the model structure flow chart of a method for restoring a free-flow wind speed model at the front end of a nacelle of a wind turbine according to Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0037] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0038] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the invention belongs.

[0039] In order to solve the technical problem in the prior art that the wind speed measured by the anemometer is often lower than the actual wind speed due to the absorption of part of the wind energy by the blades of the wind turbine during rotation, the subsequent wind power characteristic test, model establishment and performance evaluation will have large errors, such as Figure 1 As shown, this embodiment provides a method for restoring a free flow wind speed model at the front end of a nacelle of a wind turbine generator set, comprising:

[0040] S1. Obtain wind turbine Scada data and perform data preprocessing to obtain a Scada data set. In this embodiment, the data preprocessing process preferably includes accuracy testing and outlier testing of the obtained wind turbine Scada data, and filling in missing values, wherein the outlier testing includes:

[0041] Eliminate data where any one of the wind speed, fan speed, and power is less than or equal to 0. In this step, by eliminating data where any one of the wind speed, fan speed, and power is less than or equal to 0, records that are obviously inconsistent with physical laws or measurement errors can be directly removed. These outliers may be caused by sensor failure, data transmission errors, or improper data recording, and their existence will seriously affect the accuracy and reliability of subsequent data analysis.

[0042] Screening out data where the power is greater than 0 when the wind speed is less than the cut-in wind speed, and the power is greater than 0 when the wind speed is greater than the cut-out wind speed, and screening out data where the wind speed is less than the cut-in wind speed and the power is greater than 0, and the wind speed is greater than the cut-out wind speed and the power is greater than 0, helps prevent the model from learning unreasonable input-output relationships. The cut-in wind speed and the cut-out wind speed are key parameters in the design of wind turbines, representing the wind speed thresholds for the unit to start and stop generating electricity, respectively. When the wind speed is lower than the cut-in wind speed, the unit should theoretically not generate electricity (the power should be 0); similarly, when the wind speed is higher than the cut-out wind speed, in order to protect the unit from damage at high wind speeds, the unit should stop generating electricity.

[0043] Data with power greater than 1.2 times the rated power are screened out to ensure that the model training process will not be affected by extreme operating conditions. When the output power of a wind turbine exceeds a certain proportion of its rated power, it may cause damage to the unit or shorten its service life. Eliminating these outliers helps protect the safety of the unit and ensures that the wind speed value predicted by the model is within a reasonable range that the unit can withstand.

[0044] Missing value filling includes:

[0045] Determine whether the data integrity rate is less than 95%. First, by determining whether the data integrity rate is less than 95%, we can identify those data sets with many missing values. If these data sets are used directly for model training without processing, the performance and prediction accuracy of the model will be seriously affected.

[0046] If the data completeness rate is less than 95%, the missing data are filled by the neighboring interpolation method. In this setting, the missing values ​​are estimated based on the existing neighboring values ​​in the data set, thereby retaining the original distribution and trend of the data to a certain extent. Compared with simple mean filling or zero value filling methods, it can better maintain the authenticity and accuracy of the data.

[0047] After completing the interpolated data, outlier detection is performed. If there are no outliers in the test results, the missing value filling is completed. Since some unreasonable values ​​may be introduced during the interpolation process, or unrecognized outliers in the original data may be amplified after interpolation, by performing outlier detection again, it can be ensured that the filled data set does not contain any outliers, thereby further improving the data quality.

[0048] S2. Based on the Scada data set, a convolutional neural network is used to train a free flow wind speed restoration model at the front end of the cabin. Specifically, the following steps are included:

[0049] S21. Extract characteristic factors based on the Scada data set. By extracting characteristic factors from the Scada data set, the automation and standardization of feature extraction are achieved, reducing human intervention and errors.

[0050] Preferably, in this embodiment, the characteristic factors include: NFT wind speed results, unit Scada wind speed, wind direction, pitch angle, generator power, impeller diameter, hub height, terrain complexity steepness index, air temperature, air pressure and air density. By incorporating more characteristic factors related to wind speed restoration, the model can capture more key factors affecting wind speed restoration, thereby improving the accuracy and reliability of the prediction.

[0051] S22. Taking the characteristic factors as input, a convolutional neural network is used to build a free flow wind speed restoration model at the front end of the cabin. The convolutional neural network is used to build the model, which fully utilizes the advantages of CNN in processing complex data structures and extracting high-dimensional features, so that the model can learn more complex and accurate input-output relationships. Among them, the gradient descent algorithm is used for learning because the gradient descent algorithm is a widely used optimization algorithm that can adjust the model parameters by continuous iteration to minimize the loss function, which helps the model converge quickly to the optimal solution.

[0052] As a preferred solution of this embodiment, the data is standardized and normalized to eliminate the dimension differences and numerical range differences between different feature factors, so that the model training process is more stable, and it also helps to improve the convergence speed and prediction accuracy of the model. The specific formula is as follows:

[0053]

[0054] Among them, μ B is the average value of characteristic factors, σ 2 is the characteristic factor variance, is the normalized value.

[0055] S23. Divide the Scada data set into a training set and a test set, and train to obtain a free stream wind speed restoration model at the front end of the cabin.

[0056] As a preferred solution of this embodiment, Figure 3 As shown in the figure, the free flow wind speed restoration model at the front end of the cabin uses 16 convolution layers, 1※3 convolution kernels, average pooling, Reul function, root mean square error, and 50 training iterations. Specifically, 16 convolution layers are used with 1×3 convolution kernels. This network architecture can effectively extract high-dimensional features from input data while maintaining computational efficiency. An appropriate increase in the number of convolution layers helps the model learn more complex and sophisticated feature representations, while a 1×3 convolution kernel can expand the receptive field and capture more contextual information without adding too many parameters.

[0057] Furthermore, using average pooling as the pooling layer helps reduce the dimension of the feature map and the computational complexity while retaining important features. Average pooling can smooth the feature map, reduce noise and redundant information, and improve the robustness of the model by taking the average value of the local area.

[0058] Furthermore, using the ReLU (Rectified Linear Unit) function as an activation function can introduce nonlinear factors and enhance the expressiveness of the model. The ReLU function has the advantages of simple calculation, fast convergence speed, and the ability to alleviate gradient disappearance, which helps the model to quickly converge to the optimal solution during training.

[0059] Furthermore, the root mean square error (RMSE) is used as the loss function, which can directly reflect the deviation between the model prediction value and the true value. RMSE is more sensitive to outliers, which helps the model pay more attention to samples with large prediction deviations during training, thereby improving the overall prediction accuracy.

[0060] Furthermore, setting the number of training iterations to 50 rounds is a reasonable value obtained after experimental verification to avoid overfitting problems caused by too many iterations.

[0061] S3. Acquire the real-time data of the wind turbines in the wind farm, input the data into the free-flow wind speed restoration model at the front end of the nacelle, and obtain the wind speed value at the front end of the wind turbine after restoration.

[0062] That is, the present invention proposes a method, device and medium for restoring a free-stream wind speed model at the front end of a nacelle of a wind turbine generator set, wherein the method acquires and preprocesses Scada data of the wind turbine, and constructs and trains a free-stream wind speed restoration model at the front end of the nacelle in combination with a convolutional neural network model, so as to accurately restore the actual wind speed value at the front end of the nacelle, thereby significantly improving the accuracy and reliability of subsequent wind power characteristic tests, so as to solve the technical problem in the prior art that the wind speed measured by the anemometer is often lower than the actual incoming wind speed due to the fact that the blades of the wind turbine generator set absorb part of the wind energy during rotation, resulting in a wind speed difference, which causes large errors in subsequent wind power characteristic tests, model establishment and performance evaluation.

[0063] This embodiment also provides an electronic device, including a processor and a memory that is communicatively connected to the processor and is used to store executable instructions of the processor, wherein the processor is used to execute the free flow wind speed restoration model method at the front end of the cabin proposed in the first aspect.

[0064] This embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for restoring the free flow wind speed model at the front end of the cabin described in the first aspect.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A free-flow wind speed restoration model method at the front end of the wind turbine nacelle. It is characterized in that include: S1, obtaining wind turbine Scada data and performing data preprocessing to obtain a Scada data set; S2, based on the Scada dataset, a convolutional neural network is used to train the free stream wind speed restoration model at the front end of the cabin; S3. Acquire the real-time data of the wind turbines in the wind farm, input the data into the free-flow wind speed restoration model at the front end of the nacelle, and obtain the wind speed value at the front end of the wind turbine after restoration.

2. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 1, It is characterized in that The wind turbine Scada data is obtained and preprocessed to obtain a Scada data set, and the wind turbine Scada data is tested for accuracy and outliers, and missing values ​​are filled.

3. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 2, It is characterized in that The outlier test includes: Eliminate wind speed, fan speed, power where any one of them is less than or equal to 0; When the screening wind speed is less than the cut-in wind speed, the power is greater than 0, and when the wind speed is greater than the cut-out wind speed, the power is greater than 0; Eliminate data with power greater than 1.2 times the rated power; The missing values ​​are filled in by: Determine whether the data integrity rate is less than 95%; If the data completeness rate is less than 95%, the missing data will be filled using the neighboring interpolation filling method; After completing the interpolated data, perform outlier detection. If there are no outliers in the test results, complete the missing value filling.

4. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 1, It is characterized in that The method of using a convolutional neural network to train a free flow wind speed restoration model at the front end of the cabin based on the Scada data set includes: S21, extract characteristic factors based on Scada data set; S22, taking the characteristic factors as input, and using a convolutional neural network to construct a free stream wind speed restoration model at the front end of the cabin; S23. Divide the Scada data set into a training set and a test set, and train to obtain a free stream wind speed restoration model at the front end of the cabin.

5. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 4, It is characterized in that The characteristic factors include: NFT wind speed results, unit Scada wind speed, wind direction, blade pitch angle, generator power, impeller diameter, hub height, terrain complexity steepness index, air temperature, air pressure and air density.

6. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 1, It is characterized in that The free stream wind speed restoration model at the front end of the cabin is trained by using a convolutional neural network, a gradient descent algorithm is used for learning, and a standardization process is performed.

7. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 6, It is characterized in that The specific formula for the standardization and normalization process is as follows: Among them, μ B is the average value of characteristic factors, σ 2 is the characteristic factor variance, is the normalized value.

8. The free flow wind speed restoration model method at the front end of the wind turbine nacelle according to claim 1, It is characterized in that The free stream wind speed restoration model at the front end of the cabin adopts 16 convolution layers, 1※3 convolution kernels, average pooling is adopted in the pooling layer, Reul function is adopted in the activation function, root mean square error is adopted in the loss function, and the number of training iterations is 50 rounds.

9. An electronic device comprising a processor and a memory connected to the processor for storing instructions executable by the processor, It is characterized in that The processor is used to execute the free stream wind speed restoration model method at the front end of the nacelle as described in any one of claims 1-8 above.

10. A computer-readable storage medium storing a computer program, It is characterized in that When the computer program is executed by a processor, the method for restoring the free stream wind speed model at the front end of the nacelle according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Wind turbine generator bearing temperature early warning method

    CN112598172A

  • Wind turbine generator power characteristic analysis method

    CN114298512A

  • Method for acquiring effective incoming flow wind speed in front of wind wheel of wind turbine generator

    CN115419557A

  • Wind turbine generator Bladed modeling method based on SCADA data extraction

    CN115495883A

  • Methods of operating a wind turbine, wind turbines and wind parks

    EP2878809A1