Laser radar wind measurement data interpolation method and device based on machine learning

Through the machine learning-based lidar wind measurement data interpolation method, the problem of low efficiency of lidar long-distance layer data is solved, the availability of wind measurement data is improved, and more accurate data is provided for the performance optimization and control of wind turbines support.

CN119247395APending Publication Date: 2025-01-03DATANG INTERNATIONAL DULAN NEW ENERGY CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411198168.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2025-01-03

AI Technical Summary

Technical Problem

Lidar is susceptible to environmental factors during wind measurement, and the long-distance layer data is inefficient, making it difficult to meet the engineering needs of wind turbines.

Method used

Using machine learning-based lidar wind measurement data interpolation method, we can obtain and preprocess data from multiple distance layers, calculate the target correlation coefficient, determine the target distance layer, and perform data interpolation based on the machine learning model to improve the long distance layer The efficiency of data.

Benefits of technology

The availability rate of lidar wind measurement data is improved, and more accurate flowing wind conditions data is provided to meet the performance optimization and overall control needs of wind turbine units.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119247395A_ABST
    Figure CN119247395A_ABST
Patent Text Reader

Abstract

The invention relates to a laser radar wind measurement data interpolation method and device based on machine learning, and the method comprises the steps: obtaining laser radar data, measured by a laser radar, of all distance layers, and carrying out the preprocessing, thereby obtaining a data rule of each distance layer, calculating target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rule of each distance layer; determining a target distance layer from other distance layers based on the target correlation coefficient, and performing model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model; and acquiring real-time laser radar data of the target distance layer, and inputting the real-time laser radar data into the laser radar data interpolation model to obtain laser radar data of the farthest distance layer. Therefore, the data interpolation model is established based on the machine learning algorithm to perform interpolation on the farthest distance layer wind measurement data measured by the laser radar, and the accuracy of the wind measurement data of the laser radar is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the technical field of renewable energy power generation, and in particular to a method and device for interpolating laser radar wind measurement data based on machine learning. Background Art

[0002] At present, wind turbines mostly use anemometers and wind vanes to monitor wind power, which have a lag in capturing the wind conditions of the turbines. The wind measurement data is affected by the wake of the turbines and cannot accurately describe the wind conditions of the turbines. As an emerging wind monitoring device, laser radar can directly measure the wind speed and direction data of the wind flow in front of the wind rotor. Compared with traditional mechanical wind measuring instruments, it has certain advantages and is currently widely used in the performance optimization and overall control of wind farms. However, laser radar uses the Doppler shift effect to measure wind and is easily affected by environmental factors (topography, weather conditions). The effectiveness of its wind measurement data is affected by the concentration of aerosol particles in the air. The long-distance layer data of laser radar is less affected by the wind rotor blocking effect and is closer to the wind conditions of the wind flow. It is suitable as the wind flow data for wind farm performance optimization or overall control, but the long-distance layer data has low efficiency and is difficult to meet engineering needs. Summary of the invention

[0003] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a laser radar wind measurement data interpolation method and device based on machine learning.

[0004] The present disclosure provides a laser radar wind measurement data interpolation method based on machine learning, comprising:

[0005] Acquire all range layer lidar data measured by the lidar, and pre-process all range layer lidar data to obtain data rules of each range layer;

[0006] Calculate the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer;

[0007] Determine a target distance layer from the other distance layers based on the target correlation coefficient, and perform model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model;

[0008] Real-time laser radar data of the target distance layer is acquired, and the real-time laser radar data is input into the laser radar data interpolation model to obtain the laser radar data of the farthest distance layer.

[0009] In an optional embodiment of the present disclosure, preprocessing the laser radar data of all distance layers to obtain the data regularity of each distance layer includes:

[0010] Performing time resolution reduction processing on all the distance layer lidar data to obtain lidar data of the target measurement period;

[0011] Performing outlier analysis on the laser radar data of the target measurement period, and screening and eliminating the laser radar data of the target measurement period based on the analysis result to obtain candidate laser radar data;

[0012] The candidate laser radar data is subjected to data regularity analysis based on different distance layers, and the numerical distribution law and data validity law of each distance layer are obtained as the data regularity.

[0013] In an optional embodiment of the present disclosure, the calculating the target correlation coefficient between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data regularity of each distance layer includes:

[0014] Obtaining wind speed or wind direction data of the other distance layers and wind speed or wind direction data of the farthest distance layer;

[0015] Calculating based on a plurality of preset correlation algorithms, the wind speed data or wind direction data of the other distance layers, and the wind speed data or wind direction data of the farthest distance layer, to obtain a plurality of correlation coefficients between the other distance layers and the farthest distance layer;

[0016] A target correlation coefficient between the other distance layers and the farthest distance layer is determined based on the data regularity of the respective distance layers and the plurality of correlation coefficients.

[0017] In an optional embodiment of the present disclosure, determining the target distance layer from the other distance layers based on the target correlation coefficient includes:

[0018] The target correlation coefficients between the other distance layers and the farthest distance layer are sorted, and the distance layer corresponding to the maximum target correlation coefficient is determined as the target distance layer.

[0019] In an optional embodiment of the present disclosure, the model training is performed based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model, including:

[0020] Preprocessing the laser radar data of the target distance layer and the laser radar data of the farthest distance layer, and dividing them into a training set and a test set according to a certain ratio;

[0021] Based on the machine learning model, the lidar data interpolation model is constructed with the lidar data of the target distance layer as input and the lidar data of the farthest distance layer as output; wherein the model is trained by the training set and tested by the test set.

[0022] The present disclosure provides a laser radar wind measurement data interpolation device based on machine learning, comprising:

[0023] An acquisition module is used to acquire lidar data of all range layers measured by the lidar;

[0024] A preprocessing module, used to preprocess the laser radar data of all distance layers to obtain the data regularity of each distance layer;

[0025] A calculation module, used for calculating the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer;

[0026] A determination module, configured to determine a target distance layer from the other distance layers based on the target correlation coefficient;

[0027] A training module, used for performing model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model;

[0028] The model application module is used to obtain real-time laser radar data of the target distance layer, input the real-time laser radar data into the laser radar data interpolation model, and obtain the laser radar data of the farthest distance layer.

[0029] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0030] A laser radar wind measurement data interpolation method based on machine learning is proposed, which includes: obtaining laser radar data of all distance layers measured by the laser radar, preprocessing the laser radar data of all distance layers, obtaining the data rules of each distance layer, calculating the target correlation coefficient between other distance layers and the farthest distance layer based on the preset correlation algorithm and the data rules of each distance layer, determining the target distance layer from other distance layers based on the target correlation coefficient, and training the model based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and the preset machine learning model to obtain the laser radar data interpolation model, obtaining the real-time laser radar data of the target distance layer, inputting the real-time laser radar data into the laser radar data interpolation model, and obtaining the laser radar data of the farthest distance layer. Therefore, by comprehensively considering multiple correlation coefficients and data rules, analyzing the target distance layer with the strongest comprehensive correlation with the farthest distance layer, and establishing a data interpolation model based on the machine learning algorithm with the target distance layer data as input and the farthest distance layer as output, the farthest layer wind measurement data in the wind turbine cabin laser radar monitoring data can be interpolated, which effectively improves the availability of laser radar wind measurement data. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0033] Figure 1 A flowchart of a method for interpolating wind data using a laser radar based on machine learning provided in an embodiment of the present disclosure;

[0034] Figure 2 A flowchart of another method for interpolating wind data using a laser radar based on machine learning provided in an embodiment of the present disclosure;

[0035] Figure 3 A schematic diagram of a laser radar wind measurement data interpolation method based on machine learning provided in an embodiment of the present disclosure;

[0036] Figure 4 This is a structural example diagram of a laser radar wind measurement data interpolation device based on machine learning according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0039] Specifically, as an emerging wind monitoring device, LiDAR can accurately capture real-time changes in wind conditions, providing multi-faceted support for the prediction of future wind condition trends for wind turbines. Nacelle-type LiDAR is mainly used for performance optimization and overall control of wind farms. It can accurately and remotely measure wind conditions in multiple ranges in front of wind turbines, automatically and accurately detect flow field conditions, and has significant advantages over traditional mechanical wind measuring instruments.

[0040] However, since the laser radar uses the Doppler shift principle to measure wind and relies on aerosol particles in the air to obtain wind speed and direction data, its wind measurement results are not only easily affected by environmental factors (topography, weather conditions, etc.), but the effectiveness of the measurement data usually decreases with the increase of the measurement distance. Although the data of the close-range layer is efficient, it is greatly different from the free-flow wind conditions due to the influence of the wind wheel blocking effect; the long-distance wind measurement data is closer to the free-flow wind conditions, but the data efficiency is poor and it is difficult to meet the actual requirements of the project. Therefore, it is of great engineering value to interpolate the long-range layer data of the cabin wind radar to obtain sufficient and reliable free-flow wind data.

[0041] In order to ensure the availability of laser radar long-distance layer wind measurement data, meet the actual needs of engineering applications, and provide technical support and guarantee for the development and intelligent operation and maintenance of the wind power industry, the embodiment of the present disclosure proposes a laser radar wind measurement data interpolation method based on machine learning. That is to say, in view of the problem that the laser radar farthest distance layer wind measurement data is close to the free flow wind condition data but has low efficiency, a laser radar wind measurement data interpolation method based on machine learning is proposed. First, data screening and elimination are performed based on the laser radar measured data, and the data rules of different distance layers are analyzed. Then, the correlation of data of different distance layers is calculated based on the correlation algorithm and the target correlation coefficient is constructed in combination with the data distribution law. The distance layer with the highest correlation coefficient with the farthest distance layer target is selected as the target distance layer. Finally, based on the machine learning algorithm, the selected target distance layer data is used as input and the farthest distance layer data is used as output to establish a laser radar data interpolation model, so as to realize the interpolation of the laser radar farthest distance layer wind measurement data to improve the efficiency and availability of long-distance data.

[0042] Figure 1 This is a flowchart of a method for interpolating wind data from a laser radar based on machine learning provided in an embodiment of the present disclosure. Figure 1 As shown in FIG. 1 , the lidar wind data interpolation method based on machine learning includes:

[0043] Step 101: Obtain all range layer lidar data measured by the lidar, and pre-process all range layer lidar data to obtain data regularities of each range layer.

[0044] In the disclosed embodiment, laser radar is used as a wind monitoring device, such as a nacelle-type laser radar, which can measure wind speed and direction data in multiple ranges in front of the wind turbine. The principle of laser radar wind measurement mainly relies on the Doppler frequency shift of the laser backscatter echo of particles in the air to measure parameters such as wind speed and wind direction. That is, when the emitted laser pulse encounters particles in the air, part of the laser will be reflected back along the original path, resulting in a Doppler frequency shift. By analyzing this frequency shift, the wind speed and direction at a certain height in the air can be calculated.

[0045] In the embodiment of the present disclosure, the laser radar measurement includes multiple distance layers, for example, a maximum of ten distance layers, that is, the farthest distance layer is the tenth distance layer; for another example, a maximum of twelve distance layers, that is, the farthest distance layer is the twelfth distance layer; wherein, multiple distance layers refer to different measurement distances.

[0046] In some embodiments, the lidar data of all distance layers are preprocessed to obtain the data patterns of each distance layer. For example, after reducing the time resolution, screening and eliminating the data of all distance layer lidar data, the data patterns of different distance layers of the lidar are analyzed, including but not limited to numerical distribution patterns (such as wind speed mean, wind speed variance, absolute angle coordinates of wind direction, etc.), data validity patterns (such as different validity of different distance layers, for example, in the first distance layer, 9 out of 10 measured data are valid, and the data validity pattern is 90%), etc.

[0047] Step 102: Calculate the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer.

[0048] In the disclosed embodiment, the correlation algorithm can be selected and set according to the application needs, such as the Pearson correlation coefficient algorithm, the Spearman correlation coefficient algorithm, etc. Among them, the farthest distance layer refers to the farthest measurement distance of the laser radar, and the measurement data of the long-distance layer of the laser radar is closer to the actual wind speed and wind direction data.

[0049] In an embodiment of the present disclosure, there are many ways to calculate the target correlation coefficient between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer. As an example, the wind speed or wind direction data of other distance layers and the wind speed or wind direction data of the farthest distance layer are obtained, and calculation is performed based on a preset first correlation algorithm, the wind speed data or wind direction data of other distance layers, and the wind speed data or wind direction data of the farthest distance layer to obtain the first correlation coefficient between other distance layers and the farthest distance layer as the target correlation coefficient.

[0050] As another example, based on a preset first correlation algorithm, wind speed data or wind direction data of other distance layers, and wind speed data or wind direction data of the farthest distance layer, calculations are performed to obtain a first correlation coefficient between other distance layers and the farthest distance layer; based on a preset second correlation algorithm, wind speed data or wind direction data of other distance layers, and wind speed data or wind direction data of the farthest distance layer, calculations are performed to obtain a second correlation coefficient between other distance layers and the farthest distance layer; and based on the data rules of each distance layer, the first correlation coefficient and the second correlation coefficient, the target correlation coefficient between other distance layers and the farthest distance layer is determined.

[0051] As another example, based on obtaining wind speed or wind direction data of other distance layers and the wind speed or wind direction data of the farthest distance layer, the wind speed data or wind direction data of other distance layers and the wind speed or wind direction data of the farthest distance layer are respectively calculated based on multiple correlation algorithms to obtain multiple correlation coefficients between other distance layers and the farthest distance layer, and the target correlation coefficients between other distance layers and the farthest distance layer are determined based on the data rules of each distance layer and the multiple correlation coefficients.

[0052] Step 103: determine the target distance layer from other distance layers based on the target correlation coefficient, and perform model training based on the lidar data of the target distance layer, the lidar data of the farthest distance layer and a preset machine learning model to obtain a lidar data interpolation model.

[0053] In the embodiment of the present disclosure, the target correlation coefficients can be sorted, and the distance layer corresponding to the highest target correlation coefficient can be taken as the target distance layer, that is, the target distance layer, and model training can be performed based on the lidar data of the target distance layer, the lidar data of the farthest distance layer and the preset machine learning model to obtain the lidar data interpolation model.

[0054] Specifically, an appropriate machine learning model is selected as the lidar data interpolation model, including but not limited to support vector regression (SVR), long short-term memory network (LSTM), multilayer perceptron (MLP), stacking generalization (Stacking) and other machine learning models. The selected target distance layer wind speed (or wind direction) data is used as input and the farthest distance layer wind speed (or wind direction) data is used as output to establish a lidar data interpolation model based on machine learning.

[0055] Step 104: Acquire real-time lidar data of the target distance layer, input the real-time lidar data into the lidar data interpolation model, and obtain lidar data of the farthest distance layer.

[0056] Specifically, after training and acquiring the lidar data interpolation model, the real-time lidar data of the target distance layer can be acquired and input into the lidar data interpolation model to obtain more accurate lidar data of the farthest distance layer.

[0057] In summary, the laser radar wind measurement data interpolation method based on machine learning of the embodiment of the present disclosure includes: obtaining laser radar data of all distance layers measured by the laser radar, and preprocessing the laser radar data of all distance layers to obtain the data rules of each distance layer, calculating the target correlation coefficient between other distance layers and the farthest distance layer based on the preset correlation algorithm and the data rules of each distance layer, determining the target distance layer from other distance layers based on the target correlation coefficient, and training the model based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and the preset machine learning model to obtain the laser radar data interpolation model, obtaining the real-time laser radar data of the target distance layer, inputting the real-time laser radar data into the laser radar data interpolation model, and obtaining the laser radar data of the farthest distance layer. Therefore, considering a variety of correlation coefficients and data rules, analyzing the target distance layer with the strongest comprehensive correlation with the farthest distance layer, and establishing a data interpolation model based on the machine learning algorithm with the target distance layer data as input and the farthest distance layer as output can interpolate the farthest layer of wind measurement data in the wind turbine cabin laser radar monitoring data, effectively improving the availability of laser radar wind measurement data.

[0058] As an example, the above embodiment is described by taking two related algorithms as examples. Figure 2 A flow chart of another method for interpolating wind measurement data using a laser radar based on machine learning provided in an embodiment of the present disclosure, the method for interpolating wind measurement data using a laser radar based on machine learning comprising:

[0059] Step 201: Obtain all range layer lidar data measured by the lidar, reduce the time resolution of all range layer lidar data, and obtain the lidar data of the target measurement period.

[0060] Step 202: perform outlier analysis on the laser radar data of the target measurement period, and screen and eliminate the laser radar data of the target measurement period based on the analysis results to obtain candidate laser radar data, and perform data regularity analysis on the candidate laser radar data based on different distance layers to obtain the numerical distribution law and data validity law of each distance layer as the data law.

[0061] In the disclosed embodiment, after acquiring all distance layer lidar data of lidar measurement, time resolution reduction processing is performed on all distance layer lidar data. For example, during measurement, one measurement data is acquired every 5 seconds. In order to improve subsequent data processing, time resolution reduction processing can be performed to acquire one measurement data every 10 minutes, thereby obtaining lidar data of the target measurement period.

[0062] Furthermore, the laser radar data is analyzed for outliers, and based on the analysis results, the laser radar data is screened and eliminated to obtain candidate laser radar data. For example, wind speeds that are significantly higher than a certain value can be screened and eliminated. Finally, data regularity analysis of different distance layers is performed on the candidate laser radar data, including but not limited to numerical distribution regularities (such as wind speed mean, wind speed variance, absolute angle coordinates of wind direction, etc.), data validity regularities (such as different validity at different distance layers, for example, at the first distance layer, 9 out of 10 measured data are valid, and the data validity regularity is 90%), etc.

[0063] Step 203, obtain the wind speed or wind direction data of other distance layers and the wind speed or wind direction data of the farthest distance layer, and calculate based on a preset first correlation algorithm, the wind speed or wind direction data of other distance layers and the wind speed or wind direction data of the farthest distance layer to obtain a first correlation coefficient between the other distance layers and the farthest distance layer.

[0064] Step 204: Calculate based on a preset second correlation algorithm, the wind speed data or wind direction data of other distance layers, and the wind speed data or wind direction data of the farthest distance layer to obtain a second correlation coefficient between the other distance layers and the farthest distance layer.

[0065] Step 205: Determine the target correlation coefficients between other distance layers and the farthest distance layer based on the data regularity of each distance layer and the multiple correlation coefficients.

[0066] In some embodiments, a calculation is performed based on a preset first correlation algorithm, wind speed data or wind direction data of other distance layers, and wind speed data or wind direction data of the farthest distance layer to obtain a first correlation coefficient between other distance layers and the farthest distance layer, including: obtaining wind speed data or wind direction data of each distance layer at the same time in other distance layers, and the wind speed average or wind direction average of all wind speed data or all wind direction data of each distance layer; obtaining wind speed data or wind direction data of the farthest distance layer at the same time, and the wind speed average or wind direction average of the farthest distance layer; calculating a first difference between the wind speed data or wind direction data of each distance layer and the wind speed average or wind direction average of each distance layer, and calculating the wind speed data or wind direction data of the farthest distance layer. The method comprises the following steps: calculating a first difference between the wind speed data or wind direction data of each distance layer and the mean wind speed or the mean wind direction of the farthest distance layer, calculating a plurality of products between the first difference and the second difference, and obtaining a first result based on the sum of the plurality of products; calculating a first difference square between the wind speed data or wind direction data of each distance layer and the mean wind speed or the mean wind direction of each distance layer, and obtaining a first square sum based on the sum of the plurality of first difference squares, and calculating a second difference square between the wind speed data or wind direction data of the farthest distance layer and the mean wind speed or the mean wind direction of the farthest distance layer, determining a second square sum based on the second difference square, and multiplying the first square and the second square sum and averaging them to obtain a second result; determining a first correlation coefficient based on the ratio of the first result to the second result.

[0067] In some embodiments, a calculation is performed based on a preset second correlation algorithm, wind speed data or wind direction data of other distance layers, and wind speed data or wind direction data of the farthest distance layer to obtain a second correlation coefficient between other distance layers and the farthest distance layer, including: determining the difference values ​​of multiple ranks between each distance layer and the farthest distance layer based on the wind speed data or wind direction data of other distance layers and the wind speed data or wind direction data of the farthest distance layer; and calculating based on the difference values ​​of multiple ranks, the total number of distance layers and the target value to obtain the second correlation coefficient.

[0068] Step 206: determine the target distance layer from other distance layers based on the target correlation coefficient, pre-process the lidar data of the target distance layer and the lidar data of the farthest distance layer, and divide them into a training set and a test set according to a certain ratio.

[0069] Step 207: Based on the machine learning model, the lidar data of the target distance layer is used as input and the lidar data of the farthest distance layer is used as output to construct a lidar data interpolation model; wherein the model is trained by the training set and tested by the test set.

[0070] Step 208: Acquire real-time lidar data of the target distance layer, input the real-time lidar data into the lidar data interpolation model, and obtain lidar data of the farthest distance layer.

[0071] As an example, Figure 3 As shown, the lidar data is preprocessed, including but not limited to reducing the time resolution, data screening and elimination, etc., and the data patterns of different distance layers of the lidar are analyzed, including but not limited to the numerical distribution patterns (wind speed mean and variance, wind direction), data validity patterns, etc.

[0072] Furthermore, a target correlation coefficient is constructed based on data correlation and distribution law, and the distance layer with the highest correlation coefficient with the farthest distance layer target is selected as the target distance layer, laying the foundation for subsequent model establishment, including: calculating the correlation between wind speed data and wind direction data of different distance layers based on the correlation algorithm, wherein the correlation algorithm includes but is not limited to the Pearson correlation coefficient (such as formula 1), the Spearman correlation coefficient (such as formula 2), etc.

[0073]

[0074] Among them, x i is the data of a certain distance layer at the same time, y i It is the same type of data (absolute angle coordinates of wind speed or wind direction) at the farthest distance layer at the same time. is the mean of all data in a certain distance layer, is the mean of the same type of data (wind speed or wind direction) in the tenth distance layer; d i is the difference in the ranks of the two variables, n is the number of samples, that is, the total number of distance layers, and the target value is 1.

[0075] Then, a weighted correlation coefficient is created by combining multiple correlation coefficients (Pearson correlation coefficient, Stolman correlation coefficient, etc.) with data rules (such as data efficiency, etc.). The weighted scores of each distance layer are compared to obtain the target distance layer with the strongest correlation with the farthest distance layer and strong data validity.

[0076] Finally, with the selected target distance layer wind speed (or wind direction) data as input and the farthest distance layer wind speed (or wind direction) data as output, a lidar data interpolation model based on machine learning is established, including: preprocessing the target distance layer and the farthest distance layer lidar data, including but not limited to removing vacant values ​​and normalization, such as dividing the training set and the test set according to a certain ratio (9:1 or 8:2); selecting an appropriate machine learning model as a data interpolation model, including but not limited to support vector regression (SVR), long short-term memory network (LSTM), multi-layer perceptron (MLP), stacking generalization (Stacking) and other machine learning models, based on the selected machine learning model, with the target distance layer wind speed (or wind direction) data as input and the farthest distance layer wind speed (or wind direction) data as output, to build a lidar data interpolation model. Among them, the training set is used for model training, and the test set is used to check the regression accuracy of the model.

[0077] Therefore, the laser radar wind measurement data is limited by its wind measurement principle. The data of the close-range layer is highly efficient, but it cannot be used as free-flow wind conditions due to the influence of the wind wheel blocking effect; the long-range wind measurement data is closer to the free-flow wind conditions, but the data efficiency is poor. The present invention comprehensively considers a variety of correlation coefficients and data rules, analyzes the target distance layer with the strongest comprehensive correlation with the farthest distance layer, and establishes a data interpolation model based on a machine learning algorithm with the target distance layer data as input and the farthest distance layer as output to interpolate the farthest layer of wind measurement data in the wind turbine cabin laser radar monitoring data, effectively improving the availability of laser radar wind measurement data.

[0078] In the above embodiment, a cabin lidar long-distance wind measurement data interpolation method based on a machine learning algorithm is proposed, and a wind speed and wind direction data interpolation model is established. The proposed method has high accuracy and low error. Through the application of the machine learning algorithm, the availability of lidar wind measurement data is significantly improved, providing an important reference for wind energy resource assessment and optimal design of wind farms.

[0079] Figure 4 A schematic diagram of the structure of a laser radar wind measurement data interpolation device based on machine learning provided in an embodiment of the present disclosure, the laser radar wind measurement data interpolation device based on machine learning includes:

[0080] An acquisition module 401 is used to acquire all range layer lidar data measured by the lidar;

[0081] A pre-model application module 402 is used to pre-process the laser radar data of all distance layers to obtain the data regularity of each distance layer;

[0082] A calculation module 403 is used to calculate the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer;

[0083] A determination module 404, configured to determine a target distance layer from the other distance layers based on the target correlation coefficient;

[0084] A training module 405 is used to perform model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model;

[0085] The model application module 406 is used to obtain real-time lidar data of the target distance layer, input the real-time lidar data into the lidar data interpolation model, and obtain the lidar data of the farthest distance layer.

[0086] In an optional embodiment of the present disclosure, the preprocessing module 402 is specifically configured to:

[0087] Performing time resolution reduction processing on all the distance layer lidar data to obtain lidar data of the target measurement period;

[0088] Performing outlier analysis on the laser radar data of the target measurement period, and screening and eliminating the laser radar data of the target measurement period based on the analysis result to obtain candidate laser radar data;

[0089] The candidate laser radar data is subjected to data regularity analysis based on different distance layers, and the numerical distribution law and data validity law of each distance layer are obtained as the data regularity.

[0090] In an optional embodiment of the present disclosure, the computing module includes:

[0091] A determination unit, configured to obtain the wind speed or wind direction data of the other distance layers and the wind speed or wind direction data of the farthest distance layer;

[0092] a calculation unit, configured to calculate based on a plurality of preset correlation algorithms, the wind speed data or wind direction data of the other distance layers, and the wind speed data or wind direction data of the farthest distance layer, to obtain a plurality of correlation coefficients between the other distance layers and the farthest distance layer;

[0093] A processing unit is used to determine the target correlation coefficients between the other distance layers and the farthest distance layer based on the data regularity of the respective distance layers and the multiple correlation coefficients.

[0094] In an optional embodiment of the present disclosure, the determining module 404 is specifically configured to:

[0095] The target correlation coefficients between the other distance layers and the farthest distance layer are sorted, and the distance layer corresponding to the maximum target correlation coefficient is determined as the target distance layer.

[0096] In an optional embodiment of the present disclosure, the training module is specifically used to:

[0097] Preprocessing the laser radar data of the target distance layer and the laser radar data of the farthest distance layer, and dividing them into a training set and a test set according to a certain ratio;

[0098] Based on the machine learning model, the lidar data interpolation model is constructed with the lidar data of the target distance layer as input and the lidar data of the farthest distance layer as output; wherein the model is trained by the training set and tested by the test set.

[0099] The laser radar wind measurement data interpolation device based on machine learning provided in the embodiments of the present disclosure can execute the laser radar wind measurement data interpolation method based on machine learning provided in any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0100] The embodiments of the present disclosure also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the laser radar wind measurement data interpolation method based on machine learning provided by any embodiment of the present disclosure.

[0101] According to one or more embodiments of the present disclosure, the present disclosure provides an electronic device, including:

[0102] processor;

[0103] a memory for storing instructions executable by the processor;

[0104] The processor is used to read the executable instructions from the memory and execute the instructions to implement any of the machine learning-based lidar wind measurement data interpolation methods provided in the present disclosure.

[0105] According to one or more embodiments of the present disclosure, the present disclosure provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program is used to execute any of the machine learning-based lidar wind measurement data interpolation methods provided by the present disclosure.

[0106] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0107] The above description is only a specific embodiment of the present disclosure, so that those skilled in the art can understand or implement the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A laser radar wind measurement data interpolation method based on machine learning, characterized in that: include: Acquire all range layer lidar data measured by the lidar, and pre-process all range layer lidar data to obtain data rules of each range layer; Calculate the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer; Determine a target distance layer from the other distance layers based on the target correlation coefficient, and perform model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model; The real-time laser radar data of the target distance layer is acquired, and the real-time laser radar data is input into the laser radar data interpolation model to obtain the laser radar data of the farthest distance layer.

2. The laser radar wind measurement data interpolation method based on machine learning according to claim 1 is characterized in that: The preprocessing of the laser radar data of all distance layers to obtain the data regularity of each distance layer includes: Performing time resolution reduction processing on all the distance layer lidar data to obtain lidar data of the target measurement period; Performing outlier analysis on the laser radar data of the target measurement period, and screening and eliminating the laser radar data of the target measurement period based on the analysis result to obtain candidate laser radar data; The candidate laser radar data is subjected to data regularity analysis based on different distance layers, and the numerical distribution law and data validity law of each distance layer are obtained as the data regularity.

3. The laser radar wind measurement data interpolation method based on machine learning according to claim 1 is characterized in that: The calculation of the target correlation coefficients between other distance layers and the farthest distance layer based on the preset correlation algorithm and the data rules of each distance layer includes: Obtaining wind speed or wind direction data of the other distance layers and wind speed or wind direction data of the farthest distance layer; Calculating based on a plurality of preset correlation algorithms, the wind speed data or wind direction data of the other distance layers, and the wind speed data or wind direction data of the farthest distance layer, to obtain a plurality of correlation coefficients between the other distance layers and the farthest distance layer; A target correlation coefficient between the other distance layers and the farthest distance layer is determined based on the data regularity of the respective distance layers and the plurality of correlation coefficients.

4. The laser radar wind measurement data interpolation method based on machine learning according to claim 1 is characterized in that: The determining the target distance layer from the other distance layers based on the target correlation coefficient includes: The target correlation coefficients between the other distance layers and the farthest distance layer are sorted, and the distance layer corresponding to the maximum target correlation coefficient is determined as the target distance layer.

5. The laser radar wind measurement data interpolation method based on machine learning according to claim 1 is characterized in that: Model training is performed based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model, including: Preprocessing the laser radar data of the target distance layer and the laser radar data of the farthest distance layer, and dividing them into a training set and a test set according to a certain ratio; Based on the machine learning model, the lidar data interpolation model is constructed with the lidar data of the target distance layer as input and the lidar data of the farthest distance layer as output; wherein the model is trained by the training set and tested by the test set.

6. A laser radar wind measurement data interpolation device based on machine learning, characterized in that: include: An acquisition module is used to acquire lidar data of all range layers measured by the lidar; A preprocessing module, used to preprocess the laser radar data of all distance layers to obtain the data regularity of each distance layer; A calculation module, used for calculating the target correlation coefficients between other distance layers and the farthest distance layer based on a preset correlation algorithm and the data rules of each distance layer; A determination module, configured to determine a target distance layer from the other distance layers based on the target correlation coefficient; A training module, used for performing model training based on the laser radar data of the target distance layer, the laser radar data of the farthest distance layer and a preset machine learning model to obtain a laser radar data interpolation model; The model application module is used to obtain real-time laser radar data of the target distance layer, input the real-time laser radar data into the laser radar data interpolation model, and obtain the laser radar data of the farthest distance layer.

7. The laser radar wind measurement data interpolation method based on machine learning according to claim 6 is characterized in that: The pre-model application module is specifically used for: Performing time resolution reduction processing on all the distance layer lidar data to obtain lidar data of the target measurement period; Performing outlier analysis on the laser radar data of the target measurement period, and screening and eliminating the laser radar data of the target measurement period based on the analysis result to obtain candidate laser radar data; The candidate laser radar data is subjected to data regularity analysis based on different distance layers, and the numerical distribution law and data validity law of each distance layer are obtained as the data regularity.

8. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing instructions executable by the processor; The processor is used to read the executable instructions from the memory and execute the instructions to implement the lidar wind measurement data interpolation method based on machine learning as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the lidar wind measurement data interpolation method based on machine learning as described in any one of claims 1 to 5.

10. A computer program product, characterized in that It comprises a computer program, wherein the computer program is executed by a processor as a method for interpolating lidar wind measurement data based on machine learning as described in any one of claims 1 to 5 above.