Sea surface wind speed correction method and device based on machine learning and medium
Through the combined method of deep neural network, random forest and linear regression model, the problems of scarce sea surface wind speed observation data and unstable correction effect of existing methods under extreme weather conditions were solved, and higher-precision sea surface wind speed correction was achieved.
Patent Information
- Application Number
- CN202510672773.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-23
- Publication Date
- 2025-09-05
AI Technical Summary
Sea surface wind speed observation data is scarce, and existing methods are unable to meet the needs of machine learning models for large-scale, high-quality training data. In addition, the correction effects of existing correction methods are unstable under extreme weather conditions and fail to accurately describe the complex variation characteristics and spatial correlation of wind speed.
A sea surface wind speed correction method that combines deep neural networks, random forests, and linear regression models is adopted. By obtaining the actual values of wind speed-related variables, a preliminary prediction is made using deep neural networks and random forests, and then correction is performed in combination with a linear regression model to improve the accuracy of wind speed prediction.
The accuracy of sea surface wind speed correction has been improved, and the forecast error has been reduced. In particular, the performance is more stable under extreme wind speed conditions, which improves the accuracy of numerical forecasts.
Smart Images
Figure CN120596793A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind speed correction, and in particular to a sea surface wind speed correction method, device and medium based on machine learning. Background Art
[0002] Machine learning is a data-driven approach that improves correction capabilities by learning patterns from historical data. It has demonstrated promising results in correcting terrestrial meteorological elements such as temperature, precipitation, and wind speed, identifying complex nonlinear relationships and improving the accuracy of numerical model corrections.
[0003] However, the relative scarcity of sea surface wind speed observations has limited the application of machine learning in this area. Compared to land-based observations, offshore weather stations are fewer in number, and observational data primarily relies on buoys, ship observations, and satellite remote sensing. Due to the uneven spatial distribution of buoys and ships, the acquired wind speed data has limited coverage, making it difficult to meet the large-scale, high-quality training data requirements of machine learning models. Furthermore, while satellite wind speed data offers wide coverage, it suffers from certain errors, and its measurement frequency is limited by its orbit, resulting in low spatiotemporal resolution. Therefore, in the absence of high-quality observational data, directly using machine learning methods to correct sea surface wind speeds faces significant challenges.
[0004] Currently, methods for correcting maximum sea surface wind speeds primarily rely on physical models, statistical methods, or empirical formulas. Wind speed correction methods based on numerical weather prediction (NWP) models typically use reanalysis data to perform error correction. However, due to the uncertainty in the model's parameterization scheme for describing boundary layer wind speeds, their correction results are unstable under extreme weather conditions. Methods based on empirical formulas primarily utilize statistical relationships, such as adjusting model outputs through simple wind speed correction coefficients. These methods typically fail to account for the complex variations in wind fields and struggle to accurately describe wind speed variations under severe weather systems.
[0005] In addition, changes in sea surface wind speed are jointly influenced by weather systems, topography, and ocean dynamic processes, and exhibit significant temporal and spatial correlations. For example, when a typhoon or frontal system passes through a certain area, its peripheral wind speeds will have a significant impact on the wind speeds in the downstream sea area. However, most existing correction methods ignore this spatial correlation and usually only perform wind speed corrections based on local data, which limits their applicability over larger scales. Existing methods lack effective adaptability in the case of abnormal changes in wind speed, making it difficult to accurately correct extreme wind speeds. Summary of the Invention
[0006] The purpose of this application is to provide a sea surface wind speed correction method, device and medium based on machine learning to solve the problem of low sea surface wind speed correction accuracy.
[0007] To achieve the above objectives, this application provides the following solutions:
[0008] In a first aspect, the present application provides a sea surface wind speed correction method based on machine learning, comprising:
[0009] Obtaining actual values of wind speed-related variables in the offshore area to be corrected;
[0010] Inputting actual values of wind speed-related variables in the offshore area to be corrected into a first sea surface wind speed prediction model to obtain a first predicted value of the sea surface wind speed in the offshore area to be corrected; the first sea surface wind speed prediction model is obtained by training a deep neural network model;
[0011] Inputting the actual value of the wind speed-related variable in the offshore area to be corrected into the second sea surface wind speed prediction model to obtain a second predicted value of the sea surface wind speed in the offshore area to be corrected; the second sea surface wind speed prediction model is obtained by training a random forest model;
[0012] The first predicted value of the sea surface wind speed in the offshore area to be corrected and the second predicted value of the sea surface wind speed in the offshore area to be corrected are input into a sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed in the offshore area to be corrected; the sea surface wind speed correction model is obtained by training a linear regression model.
[0013] In one embodiment, the wind speed related variables include: an easterly wind speed component at a height of 10 meters, a northerly wind speed component at a height of 10 meters, sea level pressure, sea surface temperature, and relative humidity.
[0014] In one embodiment, the process of determining the first sea surface wind speed prediction model includes:
[0015] Obtain deep neural network model training dataset;
[0016] Initialize the deep neural network model;
[0017] The deep neural network model is trained using the deep neural network model training data set to obtain the first sea surface wind speed prediction model.
[0018] In one embodiment, obtaining a deep neural network model training dataset includes:
[0019] Acquire a first initial data set; the first initial data set includes: actual values of wind speed-related variables and actual values of sea surface wind speed in a plurality of first sample offshore areas;
[0020] Eliminating outliers from the first initial data set to obtain a first data set after the outliers have been eliminated;
[0021] Oversampling the first data set after outliers are removed to obtain an oversampled first data set;
[0022] Performing data enhancement on the oversampled first data set to obtain a data-enhanced first data set;
[0023] The first data set after data enhancement is normalized to obtain the deep neural network model training data set.
[0024] In one embodiment, the process of determining the second sea surface wind speed prediction model includes:
[0025] Get the random forest model training dataset;
[0026] Initialize the random forest model;
[0027] The random forest model is trained using the random forest model training data set to obtain the second sea surface wind speed prediction model.
[0028] In one embodiment, obtaining a random forest model training dataset includes:
[0029] Acquire a second initial data set; the second initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of second sample offshore areas;
[0030] performing outlier removal on the second initial data set to obtain a second data set after outlier removal;
[0031] Oversampling the second data set after outliers are removed to obtain an oversampled second data set;
[0032] Performing data enhancement on the oversampled second data set to obtain a data-enhanced second data set;
[0033] The second data set after data enhancement is normalized to obtain the random forest model training data set.
[0034] In one embodiment, the process of determining the sea surface wind speed correction model includes:
[0035] Get the linear regression model training data set;
[0036] Initialize the linear regression model;
[0037] The linear regression model is trained using a linear regression model training data set to obtain the sea surface wind speed correction model.
[0038] In one embodiment, obtaining a linear regression model training dataset includes:
[0039] Acquire a third initial data set; the third initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of third sample offshore areas;
[0040] performing outlier removal on the third initial data set to obtain a third data set after outlier removal;
[0041] Oversampling the third data set after outliers are removed to obtain an oversampled third data set;
[0042] Performing data enhancement on the oversampled third data set to obtain a data-enhanced third data set;
[0043] Normalizing the third data set after data enhancement to obtain a normalized third data set;
[0044] Inputting the normalized third data set into the first sea surface wind speed prediction model to obtain a normalized first prediction set; the normalized first prediction set includes a first normalized prediction value of the sea surface wind speed of each third sample offshore area;
[0045] Inputting the normalized third data set into the second sea surface wind speed prediction model to obtain a normalized second prediction set; the normalized second prediction set includes the second normalized prediction value of the sea surface wind speed of each third sample offshore area;
[0046] The linear regression model training data set is determined based on the normalized values of the sea surface wind speeds in the normalized first prediction set, the normalized second prediction set, and the normalized third data set.
[0047] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned methods for correcting sea surface wind speed based on machine learning.
[0048] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned sea surface wind speed correction methods based on machine learning.
[0049] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0050] The present application discloses a sea surface wind speed correction method, device and medium based on machine learning. First, the actual values of wind speed-related variables in the offshore area to be corrected are obtained; then, the actual values of the wind speed-related variables in the offshore area to be corrected are input into a first sea surface wind speed prediction model to obtain a first predicted value of the sea surface wind speed in the offshore area to be corrected; secondly, the actual values of the wind speed-related variables in the offshore area to be corrected are input into a second sea surface wind speed prediction model to obtain a second predicted value of the sea surface wind speed in the offshore area to be corrected; finally, the first predicted value of the sea surface wind speed in the offshore area to be corrected and the second predicted value of the sea surface wind speed in the offshore area to be corrected are input into the sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed in the offshore area to be corrected. In this application, the first sea surface wind speed prediction model is obtained by training a deep neural network model, the second sea surface wind speed prediction model is obtained by training a random forest model, and the sea surface wind speed correction model is obtained by training a linear regression model. A preliminary prediction of the sea surface wind speed is made based on the first sea surface wind speed prediction model and the second sea surface wind speed prediction model, and the first prediction value and the second prediction value are combined using the sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed, thereby improving the correction accuracy of the sea surface wind speed. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0052] Figure 1 This is an application environment diagram of a sea surface wind speed correction method based on machine learning in one embodiment of the present application;
[0053] Figure 2 A flow chart of a sea surface wind speed correction method based on machine learning provided in one embodiment of the present application;
[0054] Figure 3 Schematic diagram of the sea surface wind speed correction architecture based on machine learning;
[0055] Figure 4 This is a schematic diagram of the sea surface wind speed forecast results obtained using the global forecast system;
[0056] Figure 5 This is a schematic diagram of the sea surface wind speed prediction results obtained using the sea surface wind speed correction method based on machine learning in this application;
[0057] Figure 6 Schematic diagram of sea surface wind speed prediction results obtained using the IFS model;
[0058] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0059] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.
[0060] The purpose of this application is to provide a sea surface wind speed correction method, device and medium based on machine learning, aiming to improve the accuracy of sea surface wind speed correction.
[0061] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0062] The sea surface wind speed correction method based on machine learning provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the actual values of the wind speed-related variables in the offshore area to be corrected to the server 104. After the server 104 receives the actual values of the wind speed-related variables in the offshore area to be corrected, the server 104 inputs the actual values of the wind speed-related variables in the offshore area to be corrected into the first sea surface wind speed prediction model to obtain a first predicted value of the sea surface wind speed in the offshore area to be corrected; the actual values of the wind speed-related variables in the offshore area to be corrected are input into the second sea surface wind speed prediction model to obtain a second predicted value of the sea surface wind speed in the offshore area to be corrected;
[0063] The first predicted value of the sea surface wind speed in the offshore area to be corrected and the second predicted value of the sea surface wind speed in the offshore area to be corrected are input into the sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed in the offshore area to be corrected. The server 104 can feed back the obtained corrected value of the sea surface wind speed in the offshore area to be corrected to the terminal 102. In addition, in some embodiments, the sea surface wind speed correction method based on machine learning can also be implemented separately by the server 104 or the terminal 102. For example, the terminal 102 can directly perform a sea surface wind speed correction based on machine learning for the actual value of the wind speed-related variable in the offshore area to be corrected, or the server 104 can obtain the actual value of the wind speed-related variable in the offshore area to be corrected from the data storage system and perform a sea surface wind speed correction based on machine learning for the actual value of the wind speed-related variable in the offshore area to be corrected.
[0064] In an exemplary embodiment, Figure 2 and Figure 3 As shown in FIG, a sea surface wind speed correction method based on machine learning is provided, including:
[0065] Step 1: Obtain the actual values of wind speed-related variables in the offshore area to be corrected.
[0066] As an optional implementation, wind speed related variables include: the easterly wind speed component at a height of 10 meters (U 10 ), north wind speed component at 10 meters height (V 10 ), sea level pressure (SLP), sea surface temperature (ST) and relative humidity (RH).
[0067] Step 2: Input the actual values of the wind speed related variables in the offshore area to be corrected into the first sea surface wind speed prediction model to obtain a first predicted value of the sea surface wind speed in the offshore area to be corrected.
[0068] Among them, the first sea surface wind speed prediction model is obtained by training the deep neural network model.
[0069] As an optional implementation manner, the process of determining the first sea surface wind speed prediction model includes:
[0070] Step 21: Obtain the deep neural network model training dataset.
[0071] As an optional implementation, step 21 includes:
[0072] Step 211: Acquire a first initial data set; the first initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of first sample offshore areas.
[0073] Specifically, the actual value of the sea surface wind speed is the buoy data of the buoy on the sea surface (the wind speed collected by the buoy).
[0074] Step 212: performing outlier removal on the first initial data set to obtain a first data set after outlier removal.
[0075] Specifically, outliers are removed from the first initial data set, such as situations where the wind speed changes suddenly but there is no weather system to support it, so as to improve the data quality and ensure that each set of data in the first data set after outliers are removed contains the actual values of wind speed-related variables and the actual values of sea surface wind speed.
[0076] Step 213: oversampling the first data set after outliers are removed to obtain an oversampled first data set.
[0077] Specifically, an oversampling technique is used to repeatedly sample the first sample offshore area where the actual value of the sea surface wind speed is relatively large, so as to increase the attention of the first sea surface wind speed prediction model to extreme wind speeds.
[0078] Step 214: performing data enhancement on the oversampled first data set to obtain a data-enhanced first data set.
[0079] Specifically, the data enhancement adopts the Gaussian noise perturbation method to enhance a small amount of extremely large wind speed data, thereby improving the robustness of the first sea surface wind speed prediction model.
[0080] Step 215: Normalize the first data set after data enhancement to obtain a deep neural network model training data set.
[0081] Specifically, normalization makes the values of all data in the dataset uniformly distributed between [0, 1]. The normalization formula for any current data is as follows:
[0082]
[0083] Among them, X' is the normalized value of the current data; X is the actual value of the current data; X min is the minimum value of the current data in the data set; X max The maximum value of the current data in the dataset.
[0084] Step 22: Initialize the deep neural network model.
[0085] Specifically, the Deep Neural Network (DNN) model uses a multi-layer neural network architecture, including an input layer, multiple hidden layers, and an output layer. The hidden layers use the ReLU (Rectified Linear Unit) activation function to enhance nonlinear expression capabilities.
[0086] Step 23: Use the deep neural network model training data set to train the deep neural network model to obtain the first sea surface wind speed prediction model.
[0087] Specifically, when training the deep neural network model, the input elements of the input layer are the normalized actual values of the wind speed-related variables of each first sample offshore area in the deep neural network model training data set; the label data are the normalized actual values of the sea surface wind speed of each first sample offshore area in the deep neural network model training data set; and the output elements of the output layer are the first normalized predicted values of the sea surface wind speed of each first sample offshore area.
[0088] The mean square error (MSE) is used as the loss function during DNN model training. The calculation formula is as follows:
[0089]
[0090] Where MSE1 is the mean square error of the DNN model; N is the number of the first sample sea area in the deep neural network model training data set; y 1,i is the normalized actual value of the sea surface wind speed in the i-th first sample offshore area; is the first normalized predicted value of the sea surface wind speed in the i-th first sample offshore area.
[0091] The Adam optimization algorithm is used to adjust the network parameters of the DNN model to minimize the loss function and improve the model convergence speed. The network parameters of the DNN model are updated through repeated training until the MSE1 no longer decreases significantly or the maximum number of training times is reached.
[0092] Step 3: Input the actual values of the wind speed related variables in the offshore area to be corrected into the second sea surface wind speed prediction model to obtain a second predicted value of the sea surface wind speed in the offshore area to be corrected.
[0093] Among them, the second sea surface wind speed prediction model is obtained by training the random forest model.
[0094] As an optional implementation manner, the process of determining the second sea surface wind speed prediction model includes:
[0095] Step 31: Get the random forest model training dataset.
[0096] As an optional implementation, step 31 includes:
[0097] Step 311: Acquire a second initial data set; the second initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of second sample offshore areas.
[0098] Step 312: Remove outliers from the second initial data set to obtain a second data set after the outliers have been removed.
[0099] Specifically, outliers are removed from the second initial data set, such as situations where the wind speed changes suddenly but there is no weather system to support it, so as to improve the data quality and ensure that each set of data in the second data set after outliers are removed contains the actual values of wind speed-related variables and the actual values of sea surface wind speed.
[0100] Step 313: oversampling the second data set after outliers are removed to obtain an oversampled second data set.
[0101] Specifically, an oversampling technique is used to repeatedly sample the second sample offshore area where the actual value of the sea surface wind speed is larger, so as to increase the attention of the second sea surface wind speed prediction model to extreme wind speeds.
[0102] Step 314: performing data enhancement on the oversampled second data set to obtain a data-enhanced second data set.
[0103] Specifically, the data enhancement adopts the Gaussian noise perturbation method to enhance a small amount of extremely large wind speed data, thereby improving the robustness of the second sea surface wind speed prediction model.
[0104] Step 315: Normalize the second data set after data enhancement to obtain a random forest model training data set.
[0105] Step 32: Initialize the random forest model.
[0106] Specifically, the Random Forest (RF) model uses several decision trees for training.
[0107] Step 33: Use the random forest model training data set to train the random forest model to obtain a second sea surface wind speed prediction model.
[0108] Specifically, when training the random forest model, the input elements are the normalized actual values of the wind speed-related variables of each second sample offshore area in the random forest model training dataset; the label data are the normalized actual values of the sea surface wind speed of each second sample offshore area in the random forest model training dataset. Each decision tree randomly samples a portion of samples from the random forest model training dataset and learns the rules of wind speed correction. The output is the second normalized predicted value of the sea surface wind speed of each second sample offshore area. The second normalized predicted value of the sea surface wind speed under any training number is the weighted average of the prediction results of all decision trees:
[0109]
[0110] in, is the second normalized predicted value of sea surface wind speed; M is the number of decision trees; is the prediction result of the j-th decision tree.
[0111] The RF model training process also uses MSE as the loss function. The network parameters of the RF model are iteratively updated through repeated training until the MSE of the RF model no longer decreases significantly or the maximum number of training times is reached.
[0112] Step 4: Input the first predicted value of the sea surface wind speed in the offshore area to be corrected and the second predicted value of the sea surface wind speed in the offshore area to be corrected into the sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed in the offshore area to be corrected.
[0113] Among them, the sea surface wind speed correction model is obtained by training the linear regression model.
[0114] As an optional implementation, the process of determining the sea surface wind speed correction model includes:
[0115] Step 41: Obtain a linear regression model training dataset.
[0116] As an optional implementation, step 41 includes:
[0117] Step 411: Acquire a third initial data set; the third initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of third sample offshore areas.
[0118] Step 412: performing outlier removal on the third initial data set to obtain a third data set after outlier removal.
[0119] Specifically, by removing outliers from the third initial data set, such as the situation where the wind speed changes suddenly but there is no weather system to support it, the data quality is improved to ensure that each set of data in the third data set after the outliers are removed contains the actual values of the wind speed-related variables and the actual values of the sea surface wind speed.
[0120] Step 413: oversampling the third data set after outliers are removed to obtain an oversampled third data set.
[0121] Step 414: performing data enhancement on the oversampled third data set to obtain a data-enhanced third data set.
[0122] Step 415: normalize the data-enhanced third dataset to obtain a normalized third dataset.
[0123] Step 416: Input the normalized third data set into the first sea surface wind speed prediction model to obtain a normalized first prediction set; the normalized first prediction set includes the first normalized prediction value of the sea surface wind speed of each third sample offshore area.
[0124] Step 417: Input the normalized third data set into the second sea surface wind speed prediction model to obtain a normalized second prediction set; the normalized second prediction set includes the second normalized prediction value of the sea surface wind speed of each third sample offshore area.
[0125] Step 418: Determine a linear regression model training data set based on the normalized values of each sea surface wind speed in the normalized first prediction set, the normalized second prediction set, and the normalized third data set.
[0126] Step 42: Initialize the linear regression model.
[0127] Step 43: Use the linear regression model training data set to train the linear regression model to obtain a sea surface wind speed correction model.
[0128] Specifically, when training the linear regression model, the input elements are the first normalized predicted value of the sea surface wind speed of each third sample offshore area in the linear regression model training data set and the second normalized predicted value of the sea surface wind speed of each third sample offshore area in the linear regression model training data set; the label data is the normalized actual value of the sea surface wind speed of each third sample offshore area in the linear regression model training data set.
[0129] The MSE loss function is also used in the training process of the linear regression model. The network parameters of the linear regression model are updated through repeated training until the MSE of the linear regression model no longer decreases significantly or the maximum number of training times is reached.
[0130] After training the first sea surface wind speed prediction model, the second sea surface wind speed prediction model, and the sea surface wind speed correction model, they were tested using the test set. The testing process included:
[0131] S1: Obtain a test set; the test set includes: normalized actual values of wind speed-related variables in multiple test offshore areas and normalized actual values of sea surface wind speed.
[0132] S2: The normalized actual values of the wind speed-related variables of each test offshore area in the test set are respectively input into the first sea surface wind speed prediction model and the second sea surface wind speed prediction model to obtain the first normalized prediction value and the second normalized prediction value of the sea surface wind speed of each test offshore area.
[0133] S3: Input the first normalized predicted value and the second normalized predicted value of the sea surface wind speed of each test offshore area into the sea surface wind speed correction model to obtain the normalized comprehensive value of the sea surface wind speed of each test offshore area.
[0134] S4: Denormalize the normalized comprehensive value of the sea surface wind speed in each test offshore area to obtain the corrected value of the sea surface wind speed in each test offshore area. The denormalization formula is as follows:
[0135] Y new =Y'·(Y max -Y min )+Y min .
[0136] Among them, Y new is the corrected value of sea surface wind speed; Y' is the normalized integrated value of sea surface wind speed; Y max Y is the maximum value of the normalized integrated value of the sea surface wind speed in all the test areas in the test set; min It is the minimum value of the normalized integrated value of the sea surface wind speed in all the test offshore areas in the test set.
[0137] Furthermore, after testing, to evaluate the effectiveness of the model correction, the root mean square error (RMSE) indicator was used to calculate the error between the corrected sea surface wind speed and the observed sea surface wind speed. The formula is as follows:
[0138]
[0139] Where n is the number of offshore areas used for testing in the test set; Y k is the corrected value of the sea surface wind speed in the kth test sea area; k The observed value of the sea surface wind speed in the offshore area used in the kth test.
[0140] Furthermore, the global forecast system (GFS), IFS and the sea surface wind speed correction method based on machine learning (i.e., Model 3) of this application were used to correct the sea surface wind speed. The results are as follows: Figure 4-Figure 6 shown.
[0141] Using various meteorological elements of GFS as input elements as a training set, and using the sea surface wind speed correction method based on machine learning of this application to correct the sea surface wind speed of GFS, its forecast performance can be effectively improved. The machine learning model can effectively reduce the negative systematic deviation of GFS. Through statistical testing of one year's data, before correction, the root mean square error of the sea surface wind speed forecast of GFS with a forecast time of 120 hours was 3.376m / s; after correction, the root mean square error was 2.273m / s, a decrease of 1.103m / s. In the highest wind speed range, the root mean square error of the sea surface wind speed correction method based on machine learning of this application is 0.5m / s less than that of GFS when the forecast time is 120 hours. Figure 4 The bias (BIAS) is -1.4289, the relative error (RE) is 29.799%, and the scatter index (SI) is 39.449%. Figure 5 , BIAS is -0.2647, RE is 22.523%, and SI is 30.656%.
[0142] In an exemplary embodiment, a computer device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a sea surface wind speed correction method based on machine learning.
[0143] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, a sea surface wind speed correction method based on machine learning is implemented.
[0144] In an exemplary embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements a sea surface wind speed correction method based on machine learning.
[0145] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a sea surface wind speed correction method based on machine learning is implemented.
[0146] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0148] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.
[0149] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0150] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0151] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A sea surface wind speed correction method based on machine learning, characterized in that: The sea surface wind speed correction method based on machine learning includes: Obtaining actual values of wind speed-related variables in the offshore area to be corrected; Inputting actual values of wind speed-related variables in the offshore area to be corrected into a first sea surface wind speed prediction model to obtain a first predicted value of the sea surface wind speed in the offshore area to be corrected; the first sea surface wind speed prediction model is obtained by training a deep neural network model; Inputting the actual value of the wind speed-related variable in the offshore area to be corrected into the second sea surface wind speed prediction model to obtain a second predicted value of the sea surface wind speed in the offshore area to be corrected; the second sea surface wind speed prediction model is obtained by training a random forest model; The first predicted value of the sea surface wind speed in the offshore area to be corrected and the second predicted value of the sea surface wind speed in the offshore area to be corrected are input into a sea surface wind speed correction model to obtain a corrected value of the sea surface wind speed in the offshore area to be corrected; the sea surface wind speed correction model is obtained by training a linear regression model.
2. The sea surface wind speed correction method based on machine learning according to claim 1, characterized in that: The wind speed related variables include: the easterly wind speed component at a height of 10 meters, the northerly wind speed component at a height of 10 meters, sea level pressure, sea surface temperature and relative humidity.
3. The sea surface wind speed correction method based on machine learning according to claim 2, characterized in that: The process of determining the first sea surface wind speed prediction model includes: Obtain deep neural network model training dataset; Initialize the deep neural network model; The deep neural network model is trained using the deep neural network model training data set to obtain the first sea surface wind speed prediction model.
4. The sea surface wind speed correction method based on machine learning according to claim 3, characterized in that: Obtain a deep neural network model training dataset, including: Acquire a first initial data set; the first initial data set includes: actual values of wind speed-related variables and actual values of sea surface wind speed in a plurality of first sample offshore areas; Eliminating outliers from the first initial data set to obtain a first data set after the outliers have been eliminated; Oversampling the first data set after outliers are removed to obtain an oversampled first data set; Performing data enhancement on the oversampled first data set to obtain a data-enhanced first data set; The first data set after data enhancement is normalized to obtain the deep neural network model training data set.
5. The sea surface wind speed correction method based on machine learning according to claim 2, characterized in that: The process of determining the second sea surface wind speed prediction model includes: Get the random forest model training dataset; Initialize the random forest model; The random forest model is trained using the random forest model training data set to obtain the second sea surface wind speed prediction model.
6. The sea surface wind speed correction method based on machine learning according to claim 5, characterized in that: Get the random forest model training dataset, including: Acquire a second initial data set; the second initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of second sample offshore areas; performing outlier removal on the second initial data set to obtain a second data set after outlier removal; Oversampling the second data set after outliers are removed to obtain an oversampled second data set; Performing data enhancement on the oversampled second data set to obtain a data-enhanced second data set; The second data set after data enhancement is normalized to obtain the random forest model training data set.
7. The sea surface wind speed correction method based on machine learning according to claim 2, characterized in that: The process of determining the sea surface wind speed correction model includes: Get the linear regression model training data set; Initialize the linear regression model; The linear regression model is trained using a linear regression model training data set to obtain the sea surface wind speed correction model.
8. The sea surface wind speed correction method based on machine learning according to claim 7, characterized in that: Get the linear regression model training dataset, including: Acquire a third initial data set; the third initial data set includes: actual values of wind speed related variables and actual values of sea surface wind speed in a plurality of third sample offshore areas; performing outlier elimination on the third initial data set to obtain a third data set after outlier elimination; Oversampling the third data set after outliers are removed to obtain an oversampled third data set; Performing data enhancement on the oversampled third data set to obtain a data-enhanced third data set; Normalizing the third data set after data enhancement to obtain a normalized third data set; Inputting the normalized third data set into the first sea surface wind speed prediction model to obtain a normalized first prediction set; the normalized first prediction set includes a first normalized prediction value of the sea surface wind speed of each third sample offshore area; Inputting the normalized third data set into the second sea surface wind speed prediction model to obtain a normalized second prediction set; the normalized second prediction set includes the second normalized prediction value of the sea surface wind speed of each third sample offshore area; The linear regression model training data set is determined based on the normalized values of the sea surface wind speeds in the normalized first prediction set, the normalized second prediction set, and the normalized third data set.
9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the sea surface wind speed correction method based on machine learning as described in any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the sea surface wind speed correction method based on machine learning as described in any one of claims 1 to 8 is implemented.