Wind profile radar boundary layer height inversion method and system based on deep learning
Through a deep learning-based convolutional neural network combined with multiple atmospheric parameter profiles, the problem of cumbersome and uncertainty of traditional algorithm steps in wind profile radar measurement is solved, and the boundary layer height inversion with higher accuracy and stability is achieved , able to accurately capture daily changes in the boundary layer under complex atmospheric conditions.
Patent Information
- Application Number
- CN202410981776.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-22
AI Technical Summary
The traditional algorithms used in the prior art stroke profile radar measurement have problems such as cumbersome steps and high uncertainty in the measurement results, especially in the case of multiple local peaks, it is difficult to accurately invert the height of the atmosphere boundary layer.
A convolutional neural network based on deep learning is adopted, combining multiple atmospheric parameter profiles for boundary layer height inversion. By preprocessing and training data sets, a convolutional neural network model is constructed, and a loss function is used to converge to output boundary layer height. Inversion result.
It significantly improves the accuracy and stability of the boundary layer's high inversion, overcomes the disadvantages of the influence of multiple local peaks, has higher robustness and performance, and can accurately capture the daily changes of the boundary layer.
Smart Images

Figure CN118915006A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of atmospheric remote sensing technology, and in particular to a wind profile radar boundary layer height inversion method and system based on deep learning. Background Art
[0002] The atmospheric boundary layer is the lowest layer in the troposphere closest to the surface. It is in direct contact with the ground and is strongly affected by the surface. Surface friction drag, evaporation and transpiration, heat transfer and other forced processes lead to frequent turbulent motion in this layer. Turbulent transport provides an important channel for heat, momentum and water vapor exchange. In addition, the emission, transmission and transformation of pollutants caused by human activities will also be affected by the boundary layer. Therefore, environmental problems in the atmospheric boundary layer directly affect human activities and health. The atmospheric boundary layer height (BLH) is an important parameter to characterize the characteristics of the atmospheric boundary layer and an important physical parameter in numerical weather forecast models. However, the boundary layer is easily affected by ground thermal and dynamic forces, and the response time is less than 1 hour, so the boundary layer height has a significant daily variation. After sunrise, the heating of solar radiation strengthens turbulence, and the meteorological elements in the boundary layer are evenly mixed to form a convective boundary layer. After sunset, the ground radiation cools, and the buoyancy contribution is almost negligible, forming a neutral boundary layer that maintains neutral stratification. When the nighttime radiation cooling is strong enough, the turbulent transport weakens, the inversion layer appears, and a stable boundary layer is formed. Therefore, it is very challenging to continuously and accurately invert the boundary layer height.
[0003] Wind profiler radar can not only provide continuous boundary layer height estimation, but also operate for a long time in all weather conditions without supervision. Therefore, wind profiler radar has great potential in the field of boundary layer height inversion. Based on wind profiler radar measurements, a series of boundary layer height estimation algorithms have been developed, including threshold method, peak detection method, wavelet transform method and wind profile method. However, all of the above methods only use one atmospheric profile to invert the boundary layer height, such as signal-to-noise ratio profile or wind profile. Other atmospheric profiles are only used as auxiliary parameters to deal with the situation where there are multiple possible boundary layer heights at the same time. When there are multiple local peaks in the profile, these methods are prone to large uncertainties. Although the fuzzy logic method can improve the algorithm by considering the signal-to-noise ratio, vertical velocity variance and Doppler spectrum width at the same time, it requires the construction of complex logic functions, which is very cumbersome in practical applications.
[0004] Therefore, it is necessary to propose a more convenient and accurate adaptive boundary layer height inversion algorithm. Summary of the invention
[0005] The present invention provides a deep learning-based wind profiler radar boundary layer height inversion method and system, which are used to solve the defects of the prior art in which traditional algorithms are used for wind profiler radar measurement, such as complicated steps and high uncertainty of measurement results.
[0006] In a first aspect, the present invention provides a wind profile radar boundary layer height inversion method based on deep learning, comprising: Acquire wind profiler radar product data, preprocess and time-match the wind profiler radar product data to obtain a training data set; Filter multiple parameter profiles in wind profiler radar products to obtain parameter profiles for inverting boundary layer height; Based on the training data set and the parameter profile of the inverted boundary layer height, a convolutional neural network is trained, the convolutional neural network is converged using a loss function, and a wind profile radar boundary layer height inversion model is output; The wind profiler radar boundary layer height inversion model is used to obtain the boundary layer height inversion result.
[0007] According to a deep learning-based wind profiler radar boundary layer height inversion method provided by the present invention, wind profiler radar product data is obtained, and the wind profiler radar product data is preprocessed and time-matched to obtain a training data set, including: Screening the wind profiler radar product data to obtain wind profiler data with a preset completeness and a preset high quality, and sorting the wind profiler data based on a preset time interval to obtain screened wind profiler data; The boundary layer height inverted by sounding is time-matched with the screened wind profile data to form the training data set.
[0008] According to a deep learning-based wind profiler radar boundary layer height inversion method provided by the present invention, multiple parameter profiles in a wind profiler radar product are screened to obtain a parameter profile of the inverted boundary layer height, including: The atmospheric profile parameters in the wind profile radar product are screened out according to the principle of boundary layer physical characteristics, and the candidate atmospheric profile parameters are obtained from the correlation cases between the atmospheric profile parameters and the boundary layer height obtained by sounding inversion at the same time; The candidate atmospheric profile parameters are input into the inversion model, and the model accuracy is evaluated using the boundary layer height obtained by the sounding inversion as a reference value; Randomly adjusting the order of any atmospheric parameter profile, inputting the obtained new atmospheric parameter profile into the inversion model, and obtaining a new model accuracy; The sequential adjustment and model inversion are repeated for all atmospheric parameter profiles to obtain multiple model accuracies. All model accuracies are normalized and sorted, and the parameter profiles involved in the inversion of boundary layer height are selected according to the order of importance.
[0009] According to a deep learning-based wind profile radar boundary layer height inversion method provided by the present invention, a convolutional neural network is trained based on the training data set and the parameter profile of the inverted boundary layer height, comprising: Constructing the convolutional neural network, wherein the convolutional neural network includes a plurality of one-dimensional convolution kernels, the activation function adopts the ReLU function, and the loss function is constructed by the root mean square error between the boundary layer height calculated by the model and the actual boundary height; Dividing the training data set into a training set, a validation set, and a test set according to a preset ratio; The convolutional neural network is trained using the training set.
[0010] According to a deep learning-based wind profile radar boundary layer height inversion method provided by the present invention, the convolutional neural network is converged by using a loss function to output a wind profile radar boundary layer height inversion model, including: Judging the training result according to the change of the loss function to avoid under-fitting and over-fitting of the training result; Determine the hyperparameters in the training process including the size of several convolution kernels and the number of convolution kernels; The model with the minimum loss function is selected as the wind profile radar boundary layer height inversion model.
[0011] According to a deep learning-based wind profile radar boundary layer height inversion method provided by the present invention, after converging the convolutional neural network using a loss function and outputting a wind profile radar boundary layer height inversion model, the method further includes: Using the test set to test the wind profiler radar boundary layer height inversion model; The correlation coefficient, mean absolute error and root mean square error were used to quantitatively evaluate the accuracy of the model.
[0012] In a second aspect, the present invention further provides a wind profile radar boundary layer height inversion system based on deep learning, comprising: A preprocessing module is used to obtain wind profiler radar product data, preprocess and time match the wind profiler radar product data to obtain a training data set; A screening module is used to screen multiple parameter profiles in the wind profile radar product to obtain the parameter profile of the inverted boundary layer height; A training module, used for training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converging the convolutional neural network using a loss function, and outputting a wind profile radar boundary layer height inversion model; The inversion module is used to obtain the boundary layer height inversion result by using the wind profile radar boundary layer height inversion model.
[0013] In a third aspect, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the wind profile radar boundary layer height inversion method based on deep learning as described above is implemented.
[0014] In a fourth aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a wind profile radar boundary layer height inversion method based on deep learning as described in any one of the above.
[0015] In a fifth aspect, the present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the deep learning-based wind profile radar boundary layer height inversion methods described above.
[0016] The wind profiler radar boundary layer height inversion method and system based on deep learning provided by the present invention, by combining the advantage of deep learning in discovering complex structures in high-dimensional data, while adding multiple atmospheric parameter profiles, greatly improves the accuracy and stability of boundary layer height inversion, and provides strong support for boundary layer research and global climate change; the proposed convolutional neural network inversion algorithm significantly improves the accuracy of boundary layer inversion, can overcome the shortcomings of multiple local peak effects, has higher robustness, and has better performance than traditional algorithms. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0018] Figure 1 It is one of the flow charts of the wind profile radar boundary layer height inversion method based on deep learning provided by the present invention; Figure 2 This is the second flow chart of the wind profile radar boundary layer height inversion method based on deep learning provided by the present invention; Figure 3 It is a structural diagram of the boundary layer height convolutional neural network provided by the present invention; Figure 4 It is a structural schematic diagram of a wind profile radar boundary layer height inversion system based on deep learning provided by the present invention; Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] In view of the defects of the prior art in wind profiler radar measurement, the boundary layer height obtained by traditional algorithms at different local peaks has large errors, and the boundary layer height inverted by single profile parameters is highly sensitive to the number of local peaks under different atmospheric conditions. The present invention adopts a convolutional neural network and combines multiple atmospheric parameter profiles for inversion calculation to solve the problem of poor accuracy of traditional algorithms under complex atmospheric conditions. It can accurately capture the daily changes in boundary layer height, providing important support for boundary layer and climate change research.
[0021] Figure 1 is one of the flow charts of the wind profile radar boundary layer height inversion method based on deep learning provided by an embodiment of the present invention, such as Figure 1 As shown, including: Step 100: Acquire wind profiler radar product data, preprocess and time-match the wind profiler radar product data to obtain a training data set; Step 200: screening multiple parameter profiles in the wind profile radar product to obtain a parameter profile of the inverted boundary layer height; Step 300: Based on the training data set and the parameter profile of the inverted boundary layer height, a convolutional neural network is trained, the convolutional neural network is converged by using a loss function, and a wind profile radar boundary layer height inversion model is output; Step 400: using the wind profiler radar boundary layer height inversion model to obtain a boundary layer height inversion result.
[0022] Specifically, the algorithm steps of the embodiment of the present invention are as follows: Figure 2 As shown in the figure, the uncertainty analysis of inversion is performed based on the number of local peaks and different atmospheric conditions, and then the inversion parameters are selected. The convolutional neural network model is obtained through parameter importance analysis and sensitivity analysis, and finally the obtained model is used for high-precision boundary layer height monitoring. Specifically, it includes: The data in the wind profiler radar products are preprocessed to screen out complete and high-quality profile data, and the data set is organized in hours. In addition, the boundary layer height inverted by the sounding is time-matched with the wind profiler radar to form a complete training data set.
[0023] The numerous parameter profiles in wind profiler radar products are screened and ranked in order of importance based on physical characteristics and arrangement importance, and the parameter profiles used to invert the boundary layer height are determined.
[0024] The convolutional neural network model is trained using the filtered data set as the model input and the boundary layer height inverted by sounding as the true value. Figure 3 As shown in the figure, the convolution kernel uses an N×1 one-dimensional convolution kernel, the activation function uses the ReLU function, and the loss function is the root mean square error between the boundary layer height calculated by the model and the actual boundary layer height. The training, validation, and test data sets are divided into 7:1:2, and the training data set is used to train the boundary layer height inversion convolutional neural network.
[0025] The change of loss function during training is used to judge whether the training results are good or not, and to avoid underfitting and overfitting. The hyperparameters during training include convolution kernel size (3, 5, 7, 9, 11, 13) and number of convolution kernels (8, 16, 32, 64, 128, 256). The size and number of convolution kernels are set to the above values respectively, and the model with the smallest loss function is taken as the optimal model.
[0026] The optimal model trained in the previous step is tested using the test data set, and the accuracy of the model is quantitatively evaluated using the correlation coefficient, mean absolute error, and root mean square error. Finally, this boundary layer height convolutional neural network model can be used to obtain high-precision boundary layer height inversion results and accurately capture the daily changes in the boundary layer.
[0027] The convolutional neural network model (CNN) adopted by the present invention has two main advantages compared with the high uncertainty of the traditional algorithm when inverting the boundary layer height. First, the CNN algorithm adds more parameters that can characterize the boundary layer height, such as spectrum width, vertical velocity standard deviation, etc. When the traditional algorithm is difficult to determine the boundary layer height due to the appearance of multiple local peaks in the signal-to-noise ratio profile, the CNN algorithm can be combined with other parameter profiles to jointly determine the boundary layer height, thereby reducing the influence of multiple local peaks. The sensitivity analysis of the CNN algorithm and two traditional algorithms to the number of local peaks in the signal-to-noise ratio profile shows that the error of the traditional algorithm increases with the increase of the number of local peaks. In contrast, the CNN algorithm can maintain a low error even in the case of multiple local peaks and has good robustness. In particular, in a stable boundary layer, when the number of local peaks is 1, 2, 3 and greater than 3, the error of the CNN algorithm is reduced by 35%, 50%, 69% and 72% compared with the traditional algorithm.
[0028] In addition, the CNN algorithm learns the complex high-order features of all input parameter profiles through a convolution process. The traditional algorithm only considers one type of feature, such as maximum or gradient, while the CNN algorithm extracts different features through different convolution kernels and uses all these features to estimate the boundary layer height. Statistical analysis shows that the CNN algorithm has a high consistency with the boundary layer height inverted by soundings, with correlation coefficients, MAE and RMSE of 0.81, 0.24 km and 0.34 km, respectively. Compared with the traditional algorithm, the MAE of the CNN algorithm decreased from 0.57±0.60 km to 0.24±0.25 km, and the RMSE decreased from 0.83 km to 0.34 km.
[0029] In some embodiments, obtaining wind profiler radar product data, preprocessing and time matching the wind profiler radar product data to obtain a training data set includes: Screening the wind profiler radar product data to obtain wind profiler data with a preset completeness and a preset high quality, and sorting the wind profiler data based on a preset time interval to obtain screened wind profiler data; The boundary layer height inverted by sounding is time-matched with the screened wind profile data to form the training data set.
[0030] The uncertainty in the traditional algorithm is analyzed, the number of local peaks under different signal-to-noise ratio profiles is counted, and the mean absolute error (MAE) between the boundary layer height inverted by the traditional algorithm and the boundary layer height inverted by the sounding under different local peak numbers is calculated, and it is used as an indicator to measure the sensitivity of the algorithm to the number of local peaks:
[0031] in represents the number of samples, represents the boundary layer height inverted by the traditional method, Here, a high MAE indicates that the traditional method is sensitive to the number of local peaks when inverting the boundary layer height and is easily affected by it.
[0032] The present invention adopts a method combining physical mechanism with arrangement importance to select numerous atmospheric profile parameters in wind profiler radar products, determines the atmospheric profile highly correlated with the boundary layer height inversion, and uses it as the input of the subsequent inversion model.
[0033] In some embodiments, screening multiple parameter profiles in a wind profiler radar product to obtain a parameter profile for inverting the boundary layer height includes: The atmospheric profile parameters in the wind profile radar product are screened out according to the principle of boundary layer physical characteristics, and the candidate atmospheric profile parameters are obtained from the correlation cases between the atmospheric profile parameters and the boundary layer height obtained by sounding inversion at the same time; The candidate atmospheric profile parameters are input into the inversion model, and the model accuracy is evaluated using the boundary layer height obtained by the sounding inversion as a reference value; Randomly adjusting the order of any atmospheric parameter profile, inputting the obtained new atmospheric parameter profile into the inversion model, and obtaining a new model accuracy; The sequential adjustment and model inversion are repeated for all atmospheric parameter profiles to obtain multiple model accuracies. All model accuracies are normalized and sorted, and the parameter profiles involved in the inversion of boundary layer height are selected according to the order of importance.
[0034] Specifically, in the embodiment of the present invention, the correlation between various atmospheric parameter profiles and boundary layer height in the wind profile radar product is preliminarily analyzed and explained from the perspective of physical characteristics, and the importance of each parameter is quantitatively evaluated by using the arrangement importance, which is achieved by the following steps: (1) Preliminary screening of atmospheric profile parameters in wind profiler radar products is performed based on the physical properties of the boundary layer. Cases of these atmospheric profile parameters and boundary layer heights obtained by sounding inversion at the same time are plotted to complete the preliminary screening.
[0035] (2) The atmospheric parameter profiles obtained through preliminary screening are input into the inversion model, and the boundary layer height obtained by sounding is used as a reference value to evaluate the accuracy of the model.
[0036] (3) Randomly disrupt a parameter profile, obtain a new profile and input it into the model, and calculate the change in model accuracy as an evaluation indicator of importance.
[0037] (4) Repeat the previous step for all candidate atmospheric parameter profiles, normalize and sort the changes in the accuracy of each parameter profile after scrambling, select the parameter profiles according to their importance, and complete the selection of the parameter profiles.
[0038] The present invention adds more parameter profiles on the basis of the signal-to-noise ratio profile, including spectrum width, vertical velocity standard deviation, horizontal wind speed and direction, etc. When the traditional algorithm is difficult to determine the boundary layer height due to multiple local peaks in the signal-to-noise ratio profile, the CNN algorithm can combine other parameter profiles to jointly determine the boundary layer height, thereby reducing the impact of multiple local peaks.
[0039] In some embodiments, training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height includes: Constructing the convolutional neural network, wherein the convolutional neural network includes a plurality of one-dimensional convolution kernels, the activation function adopts the ReLU function, and the loss function is constructed by the root mean square error between the boundary layer height calculated by the model and the actual boundary height; Dividing the training data set into a training set, a validation set, and a test set according to a preset ratio; The convolutional neural network is trained using the training set.
[0040] The convolutional neural network is converged by using a loss function to output a wind profile radar boundary layer height inversion model, including: Judging the training result according to the change of the loss function to avoid under-fitting and over-fitting of the training result; Determine the hyperparameters in the training process including the size of several convolution kernels and the number of convolution kernels; The model with the minimum loss function is selected as the wind profile radar boundary layer height inversion model.
[0041] Specifically, the embodiment of the present invention studies a boundary layer height inversion algorithm based on a convolutional neural network, incorporates multiple atmospheric parameter profiles into the inversion model, and solves the problem of poor accuracy of traditional algorithms under complex atmospheric conditions.
[0042] The input of the network model consists of multiple atmospheric parameter profiles, the specific structure is as follows: Figure 3 As shown in the figure. The CNN algorithm includes two convolution processes with ReLU as the activation function, and each convolution process is followed by a maximum pooling process. After two convolution layers, two fully connected layers are used to output the boundary layer height estimated by the CNN algorithm. Considering the special vertical profile structure of the input variable, a one-dimensional convolution kernel (N×1) is used in the convolution layer. This convolution kernel can extract the complex vertical features of the five input profiles, such as gradients and maximum values, through convolution operations. Multiple convolution kernels extract a large number of different features, and superimposing two convolution and pooling layers can gradually extract high-order features. These extracted extensive complex features are conducive to the CNN algorithm to invert the boundary layer height. In addition, the network also adds two batch normalization layers to accelerate convergence.
[0043] The present invention learns the complex high-order features of all input parameter profiles through the convolution process. Different complex features are extracted through different convolution kernels, and high-order features are extracted through two convolution superpositions. These complex features are used for the inversion of the boundary layer height, improving the accuracy and robustness of the inversion.
[0044] In some embodiments, after the convolutional neural network is converged using a loss function and a wind profile radar boundary layer height inversion model is output, the method further includes: Using the test set to test the wind profiler radar boundary layer height inversion model; The correlation coefficient, mean absolute error and root mean square error were used to quantitatively evaluate the accuracy of the model.
[0045] Specifically, the embodiment of the present invention uses a test data set to test the trained optimal model, and uses the correlation coefficient, mean absolute error and root mean square error to quantitatively evaluate the accuracy of the model. Finally, using this boundary layer height convolutional neural network model, a high-precision boundary layer height inversion result can be obtained, accurately capturing the daily variation of the boundary layer.
[0046] The present invention can monitor the daily variation of boundary layer height over a long period of time. Through long-term observation, the seasonal variation and interannual variation of boundary layer height at the monitoring location can be analyzed, providing powerful data support and analysis means for climate change research.
[0047] The wind profiler radar boundary layer height inversion system based on deep learning provided by the present invention is described below. The wind profiler radar boundary layer height inversion system based on deep learning described below and the wind profiler radar boundary layer height inversion method based on deep learning described above can be referenced to each other.
[0048] Figure 4 is a schematic diagram of the structure of a wind profile radar boundary layer height inversion system based on deep learning provided by an embodiment of the present invention, such as Figure 4 As shown, it includes: a preprocessing module 41, a screening module 42, a training module 43 and an inversion module 44, wherein: The preprocessing module 41 is used to obtain wind profiler radar product data, preprocess and time-match the wind profiler radar product data, and obtain a training data set; the screening module 42 is used to screen multiple parameter profiles in the wind profiler radar product to obtain a parameter profile of the inverted boundary layer height; the training module 43 is used to train a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converge the convolutional neural network using a loss function, and output a wind profiler radar boundary layer height inversion model; the inversion module 44 is used to use the wind profiler radar boundary layer height inversion model to obtain a boundary layer height inversion result.
[0049] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute a wind profiler radar boundary layer height inversion method based on deep learning, the method comprising: obtaining wind profiler radar product data, preprocessing and time matching the wind profiler radar product data to obtain a training data set; screening multiple parameter profiles in the wind profiler radar product to obtain a parameter profile of the inverted boundary layer height; training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converging the convolutional neural network using a loss function, and outputting a wind profiler radar boundary layer height inversion model; and using the wind profiler radar boundary layer height inversion model to obtain a boundary layer height inversion result.
[0050] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0051] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the deep learning-based wind profiler radar boundary layer height inversion method provided by the above methods. The method includes: obtaining wind profiler radar product data, preprocessing and time matching the wind profiler radar product data to obtain a training data set; screening multiple parameter profiles in the wind profiler radar product to obtain a parameter profile of the inverted boundary layer height; training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converging the convolutional neural network using a loss function, and outputting a wind profiler radar boundary layer height inversion model; and using the wind profiler radar boundary layer height inversion model to obtain a boundary layer height inversion result.
[0052] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the deep learning-based wind profiler radar boundary layer height inversion method provided by the above-mentioned methods, the method comprising: obtaining wind profiler radar product data, preprocessing and time matching the wind profiler radar product data to obtain a training data set; screening multiple parameter profiles in the wind profiler radar product to obtain a parameter profile of the inverted boundary layer height; training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converging the convolutional neural network using a loss function, and outputting a wind profiler radar boundary layer height inversion model; and using the wind profiler radar boundary layer height inversion model to obtain a boundary layer height inversion result.
[0053] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0054] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0055] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A wind profile radar boundary layer height inversion method based on deep learning, characterized in that: include: Acquire wind profiler radar product data, preprocess and time-match the wind profiler radar product data to obtain a training data set; Filter multiple parameter profiles in wind profiler radar products to obtain parameter profiles for inverting boundary layer height; Based on the training data set and the parameter profile of the inverted boundary layer height, a convolutional neural network is trained, the convolutional neural network is converged using a loss function, and a wind profile radar boundary layer height inversion model is output; The wind profiler radar boundary layer height inversion model is used to obtain the boundary layer height inversion result.
2. The method for inverting the wind profile radar boundary layer height based on deep learning according to claim 1, characterized in that: Obtain wind profiler radar product data, preprocess and time match the wind profiler radar product data to obtain a training data set, including: Screening the wind profiler radar product data to obtain wind profiler data with a preset completeness and a preset high quality, and sorting the wind profiler data based on a preset time interval to obtain screened wind profiler data; The boundary layer height inverted by sounding is time-matched with the screened wind profile data to form the training data set.
3. The method for inverting the wind profile radar boundary layer height based on deep learning according to claim 1, characterized in that: Filter multiple parameter profiles in the wind profiler radar product to obtain the parameter profile of the inverted boundary layer height, including: The atmospheric profile parameters in the wind profiler radar product are screened out according to the principle of boundary layer physical characteristics, and the candidate atmospheric profile parameters are obtained from the correlation cases between the atmospheric profile parameters and the boundary layer height obtained by sounding inversion at the same time; The candidate atmospheric profile parameters are input into the inversion model, and the model accuracy is evaluated using the boundary layer height obtained by the sounding inversion as a reference value; Randomly adjusting the order of any atmospheric parameter profile, inputting the obtained new atmospheric parameter profile into the inversion model, and obtaining a new model accuracy; The sequential adjustment and model inversion are repeated for all atmospheric parameter profiles to obtain multiple model accuracies. All model accuracies are normalized and sorted, and the parameter profiles involved in the inversion of boundary layer height are selected according to the order of importance.
4. The method for inverting the wind profile radar boundary layer height based on deep learning according to claim 1, characterized in that: Based on the training data set and the parameter profile of the inverted boundary layer height, a convolutional neural network is trained, comprising: Constructing the convolutional neural network, wherein the convolutional neural network includes a plurality of one-dimensional convolution kernels, the activation function adopts the ReLU function, and the loss function is constructed by the root mean square error between the boundary layer height calculated by the model and the actual boundary height; Dividing the training data set into a training set, a validation set, and a test set according to a preset ratio; The convolutional neural network is trained using the training set.
5. The method for inverting the wind profile radar boundary layer height based on deep learning according to claim 4 is characterized in that: The convolutional neural network is converged by using a loss function to output a wind profile radar boundary layer height inversion model, including: Judging the training result according to the change of the loss function to avoid under-fitting and over-fitting of the training result; Determine the hyperparameters in the training process including the size of several convolution kernels and the number of convolution kernels; The model with the minimum loss function is selected as the wind profile radar boundary layer height inversion model.
6. The method for inverting the wind profile radar boundary layer height based on deep learning according to claim 5, characterized in that: After the convolutional neural network is converged by using the loss function and the wind profile radar boundary layer height inversion model is output, the method further includes: Using the test set to test the wind profiler radar boundary layer height inversion model; The correlation coefficient, mean absolute error and root mean square error were used to quantitatively evaluate the accuracy of the model.
7. A wind profile radar boundary layer height inversion system based on deep learning, characterized in that: include: A preprocessing module is used to obtain wind profiler radar product data, preprocess and time match the wind profiler radar product data to obtain a training data set; A screening module is used to screen multiple parameter profiles in the wind profile radar product to obtain the parameter profile of the inverted boundary layer height; A training module, used for training a convolutional neural network based on the training data set and the parameter profile of the inverted boundary layer height, converging the convolutional neural network using a loss function, and outputting a wind profile radar boundary layer height inversion model; The inversion module is used to obtain the boundary layer height inversion result by using the wind profile radar boundary layer height inversion model.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the wind profile radar boundary layer height inversion method based on deep learning as described in any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the wind profile radar boundary layer height inversion method based on deep learning as described in any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the wind profile radar boundary layer height inversion method based on deep learning as described in any one of claims 1 to 6 is implemented.
Citation Information
Patent Citations
Atmospheric boundary layer height detection method under non-precipitation condition and atmospheric boundary layer height detection system under non-precipitation condition
CN108256546A
Wind field vertical fine structure acquisition method and device based on wind profile radar data
CN116774182A
Aerosol laser radar boundary layer height inversion method based on machine learning
CN117130008A
Cited By
Wind speed three-dimensional profile intelligent reconstruction method and system based on deep learning
CN121582471A
A deep learning-based wind speed three-dimensional profile intelligent reconstruction method and system
CN121582471B
