Grid-based intelligent fusion method and system for sea surface wind fields based on multi-source satellite payloads

Abstract

The present invention relates to a gridded intelligent fusion method and system for sea surface wind fields based on multi-source satellite payloads. The method comprises: constructing a deep neural network model and combining the advantages of the D-Matrix statistical algorithm for medium and low wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range; performing spatiotemporal matching processing between multi-source satellite data and an EC background field to obtain the data source for the fusion analysis period; combining the EC background field with the wind speed, divergence, vorticity, and second-order derivative terms of wind speed at each point in the fused wind field to obtain a cost function; obtaining the gradient terms of each cost function within the cost function based on the cost function and summing them to obtain the function gradient; and obtaining the final fused wind field based on the cost function and the function gradient. The present invention achieves efficient fusion of sea surface wind field data from multi-source satellite payloads, and can improve the accuracy and spatiotemporal resolution of sea surface wind field data.

Description

Technical Field

[0001] The present invention relates to the field of satellite technology, and in particular to a sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads. Background Art

[0002] In meteorology, the accuracy and spatial and temporal resolution of sea surface wind data are crucial for weather forecasting, marine environmental research, and maritime safety. However, due to the vastness of the ocean, the observation capabilities of a single satellite payload are limited, making it difficult to meet the demand for high-precision, high-temporal and high-resolution sea surface wind data.

[0003] To address this issue, multi-source satellite payload sea surface wind data fusion technology has emerged. By integrating observation data from different satellite payloads, this technology can significantly improve the coverage and spatiotemporal resolution of sea surface wind data. However, due to differences in observation principles, accuracy, and spatiotemporal characteristics among different satellite payloads, effectively fusing this data has become a technical challenge.

[0004] For example, due to the varying observation principles and technical specifications of different satellite payloads, the data formats, resolutions, and coverage they provide can also vary. This leads to compatibility issues during data fusion, requiring complex preprocessing to ensure that data from different sources are compatible. The observation accuracy of various satellite payloads also varies; some may provide more accurate data, while others may contain significant errors. Balancing data of varying precision to ensure the accuracy of the fusion results is a technical challenge during the fusion process. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads, thereby realizing efficient fusion of sea surface wind field data from multi-source satellite payloads and improving the accuracy and spatiotemporal resolution of sea surface wind field data.

[0006] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:

[0007] The first aspect is the grid-based intelligent fusion method of sea surface wind fields based on multi-source satellite payloads, including:

[0008] Perform star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field;

[0009] Build a deep neural network model and combine the advantages of the D-Matrix statistical algorithm for low and medium wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range;

[0010] Multi-source satellite data are processed with spatiotemporal matching of EC background fields to obtain the data source for the fusion analysis period;

[0011] The cost function is obtained by combining the wind speed, divergence, vorticity and second-order derivative of wind speed at each point of the EC background field and the fused wind field;

[0012] According to the cost function, the gradient terms of each cost function in the cost function are calculated and summed to obtain the function gradient;

[0013] According to the cost function and function gradient, the final fused wind field is obtained.

[0014] Preferably, performing star-to-star cross calibration on the wind and cloud and ocean satellite multi-source scatterometers, and performing wind speed correction on the forecast field data to obtain the sea breeze fusion background field, includes:

[0015] Using cyclone wind speed products, the multi-source scatterometers of Fengyun and Ocean satellites were calibrated by star-to-star cross-calibration to obtain coordinate data.

[0016] Perform wind speed correction on the forecast field data to obtain the corrected forecast field data;

[0017] According to the coordinate data and forecast field data, the sea breeze fusion background field is established to obtain the sea breeze fusion background field.

[0018] Preferably, a deep neural network model is constructed and combined with the advantages of the D-Matrix statistical algorithm in inversion of medium and low wind speeds to obtain MWRI sea surface wind speed products for the full wind speed range, including:

[0019] Based on the sea breeze fusion background field, a deep neural network based on physical constraints and suitable for MWRI wind inversion is constructed to obtain an inversion deep neural network;

[0020] The inversion deep neural network is combined with the medium and low wind speed inversion advantages of the D-Matrix statistical algorithm, and a two-dimensional variational assimilation multi-source sea breeze fusion technology is constructed for optimization to obtain the MWRI sea surface wind speed product.

[0021] Preferably, multi-source satellite data are processed with the EC background field in time and space to obtain the data source of the fusion analysis period;

[0022] Generate a time window with 3 hours before and after the selected time stamp for the specified date and time, and convert the data file into a time stamp to obtain the time window and file time stamp;

[0023] Based on the time window and file timestamp, the files with timestamps falling within the time window are screened, and the multi-source satellite sea surface wind field data are read respectively. The data are screened according to the orbital area and the data quality control and processing are completed to obtain the processed data;

[0024] Obtain EC background field data, and use the EC background field as an initial estimate of the fused wind field to obtain the EC background field and fused wind field data;

[0025] The cost function is obtained by combining the EC background field with the wind speed, divergence, vorticity and second-order derivative of wind speed at each point in the fused wind field.

[0026] Preferably, the formula of the cost function is:

[0027]

[0028] Where j1 represents the cost function value; N represents the number of points in the fused wind field; M represents the number of observation data points; v EC，i represents the wind speed of the EC background field at point i; v Fuse，i Indicates the wind speed at point i in the fused wind field; div EC，i represents the divergence of the EC background field at point i; div Fuse，i represents the divergence of the fused wind field at point i; vor EC，i represents the vorticity of the EC background field at point i; vor Fuse，i represents the vorticity of the fused wind field at point i; Δv EC，i represents the second-order derivative of wind speed at point i in the EC background field; Δv Fuse，i represents the second-order derivative of the wind speed at point i in the fused wind field; v Obs，j represents the observed wind speed at point j; v FuseInterp，j It represents the wind speed at point j after spatial interpolation of the fused wind field.

[0029] Preferably, the formula for the function gradient is:

[0030]

[0031] Among them, θ represents the parameter vector of the fusion wind field model, The partial derivative of θ, each element of which corresponds to the partial derivative of the cost function with respect to each parameter in θ; and j represents the index of the fused wind field and observation data; j2 represents the function gradient value.

[0032] Preferably, the final fused wind field is obtained according to the cost function and the function gradient, including:

[0033] According to the cost function and the function gradient, through θ k+1 =θ k +α k p k Calculate the minimum value of the cost function, where k = 0, 1, 2, ..., θ k+1 is the parameter vector after the k+1th iteration, θ kis the parameter vector for the kth iteration, α k is the step size determined by line search in the kth iteration, p k is the search direction in the kth iteration;

[0034] According to the minimum value of the cost function, the fused wind field is obtained to obtain the final fused wind field.

[0035] The second aspect is the sea surface wind field grid intelligent fusion system based on multi-source satellite payloads, including:

[0036] The acquisition module is used to perform star-to-star cross-calibration on the Fengyun and Ocean Satellite multi-source scatterometers, and to perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field; build a deep neural network model, and combine the advantages of the D-Matrix statistical algorithm for medium and low wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range.

[0037] The processing module performs spatiotemporal matching of multi-source satellite data and the EC background field to obtain the data source for the fusion analysis period; the wind speed, divergence, vorticity and second-order derivative of wind speed at each point of the EC background field and the fused wind field are combined to obtain the cost function; based on the cost function, the gradient terms of each cost function within the cost function are obtained and summed to obtain the function gradient; based on the cost function and the function gradient, the final fused wind field is obtained.

[0038] The third aspect is computing equipment, including:

[0039] one or more processors;

[0040] The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0041] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method described above when executed by a processor.

[0042] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0043] By performing star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and performing wind speed correction on the forecast field data, this method can eliminate systematic errors between different satellite payloads, thereby improving the accuracy and reliability of sea surface wind field data. This method integrates data from multiple satellite payloads, effectively increasing the data coverage and observation frequency, thereby improving the spatiotemporal resolution of sea surface wind field data. By defining a cost function and calculating the function gradient, this method can achieve intelligent fusion of multi-source data. This fusion method not only takes into account the spatial distribution and temporal variation of the data, but also automatically adjusts the weights of different data sources through an optimization algorithm, thereby obtaining more accurate and consistent sea surface wind field data. By providing real-time and accurate sea surface wind field information, this method helps ships avoid adverse weather conditions and reduce accident risks, while also contributing to the rational development and utilization of marine resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. Some specific embodiments of the present application will be described in detail in an illustrative and non-restrictive manner with reference to the drawings. The same reference numerals in the drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the drawings:

[0045] Figure 1 It is a flow chart of the sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads of the present invention.

[0046] Figure 2 It is a schematic diagram of the sea surface wind field gridding intelligent fusion system based on multi-source satellite payloads of the present invention. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0048] The following embodiment of the present application takes the sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads as an example to explain the solution of the present application in detail, but this embodiment does not limit the scope of protection of the present application.

[0049] like Figure 1 As shown, the present invention provides a sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads, including:

[0050] Step 11, perform star-to-star cross calibration on the Fengyun and Ocean satellite multi-source scatterometers, and perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field;

[0051] Step 12: Build a deep neural network model and combine the advantages of the D-Matrix statistical algorithm in low and medium wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range.

[0052] Step 13: Perform spatiotemporal matching of multi-source satellite data and EC background field to obtain the data source for the fusion analysis period;

[0053] Step 14: Combine the EC background field with the wind speed, divergence, vorticity, and second-order derivative of wind speed at each point in the fused wind field to obtain the cost function;

[0054] Step 15: according to the cost function, calculate the gradient terms of each cost function in the cost function and sum them to obtain the function gradient;

[0055] Step 16: Obtain the final fused wind field based on the cost function and the function gradient.

[0056] In an embodiment of the present invention, by performing star-to-star cross-calibration on the Fengyun and ocean satellite multi-source scatterometers, and performing wind speed correction on the forecast field data, the method can eliminate the systematic errors between different satellite payloads, thereby improving the accuracy and reliability of the sea surface wind field data. The method integrates data from multiple satellite payloads, effectively increasing the data coverage and observation frequency, thereby improving the spatiotemporal resolution of the sea surface wind field data. By defining a cost function and obtaining the function gradient, the method can achieve intelligent fusion of multi-source data. This fusion method not only takes into account the spatial distribution and temporal variation of the data, but also automatically adjusts the weights of different data sources through an optimization algorithm, thereby obtaining more accurate and consistent sea surface wind field data. By providing real-time and accurate sea surface wind field information, the method helps ships avoid adverse weather conditions and reduce accident risks, while also contributing to the rational development and utilization of marine resources.

[0057] The present invention provides a grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads, performs star-to-star cross-calibration on wind-cloud and ocean satellite multi-source scatterometers, and performs wind speed correction on forecast field data to obtain a sea breeze fusion background field, including:

[0058] Using cyclone wind speed products, the multi-source scatterometers of Fengyun and Ocean satellites were calibrated by star-to-star cross-calibration to obtain coordinate data.

[0059] Perform wind speed correction on the forecast field data to obtain the corrected forecast field data;

[0060] According to the coordinate data and forecast field data, the sea breeze fusion background field is established to obtain the sea breeze fusion background field.

[0061] In an embodiment of the present invention, cyclone wind speed products are selected as reference standards, multi-source scatterometer data from Fengyun satellites and ocean satellites are processed, and star-to-star cross-calibration is performed to obtain original forecast field data. Wind speed correction is performed on the forecast field data to obtain corrected forecast field data. The cross-calibrated coordinate data and the corrected forecast field data are then matched and integrated. The coordinate data and the corrected forecast field data are fused together using weighted average data fusion technology to form a sea breeze fused background field.

[0062] In the present invention, through star-to-star cross-calibration and forecast field data correction, system errors and deviations are reduced, the accuracy and reliability of sea surface wind field data are improved, and the fusion of multi-source data makes the data observed by different satellite payloads more consistent in geographic coordinates and wind speed values. By integrating multi-source data, the sea breeze fusion background field can cover a wide ocean area.

[0063] The present invention provides a gridded intelligent fusion method for sea surface wind fields based on multi-source satellite payloads, constructs a deep neural network model, and combines the advantages of the D-Matrix statistical algorithm for medium and low wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range, including:

[0064] Based on the sea breeze fusion background field, a deep neural network based on physical constraints and suitable for MWRI wind inversion is constructed to obtain an inversion deep neural network;

[0065] The inversion deep neural network is combined with the medium and low wind speed inversion advantages of the D-Matrix statistical algorithm, and a two-dimensional variational assimilation multi-source sea breeze fusion technology is constructed for optimization to obtain the MWRI sea surface wind speed product.

[0066] In an embodiment of the present invention, sea breeze fusion background field data is collected and used as training samples; the data is preprocessed, including normalization and removal of outliers, to ensure data quality; based on the characteristics of MWRI data, a deep neural network structure that can capture the complex characteristics of the sea surface wind field is designed; the deep neural network structure includes multiple hidden layers, each layer contains multiple neurons, and an activation function (such as ReLU); using the sea breeze fusion background field data as training samples, the network is trained by a backpropagation algorithm and optimization techniques (such as gradient descent). During the training process, the network parameters (such as weights and biases) are continuously adjusted to minimize the difference between the predicted wind speed and the actual wind speed, and cross-validation and other techniques are used to evaluate the performance of the model and prevent overfitting; after training and optimization, a deep neural network model suitable for MWRI high wind inversion is obtained.

[0067] For MWRI data, the traditional D-Matrix statistical algorithm was selected, which has advantages in inverting medium and low wind speeds. The trained deep neural network was combined with the D-Matrix algorithm. For medium and low wind speed segments, the D-Matrix algorithm was used for inversion; for high wind speed segments, the prediction of the deep neural network was relied upon. The two-dimensional variational (2D-Var) method was used to assimilate and fuse multi-source sea breeze data. By constructing a cost function, the wind speed data and background field information from different sources were comprehensively considered to minimize the difference between observations and model predictions. The multi-source data was used to perform quantitative accuracy testing of the fused product, and the model was tuned based on the test results, including adjusting the network parameters and the weights of the assimilation technology. After the above steps, the fused method was used to invert the MWRI data for the entire wind speed segment, ultimately obtaining a high-precision sea surface wind speed product. This product combines the prediction capabilities of the deep neural network with the inversion advantages of the D-Matrix algorithm in medium and low wind speed segments.

[0068] In this invention, by combining deep neural networks and traditional D-Matrix statistical algorithms, we can fully utilize the advantages of both, improve the inversion accuracy in the entire wind speed range, broaden the applicability of MWRI sea surface wind speed products, integrate deep learning technology and traditional statistical algorithms, and promote the cross-integration and development between different technical fields.

[0069] The present invention provides a grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads, which performs spatiotemporal matching processing of multi-source satellite data and EC background fields to obtain the data source for the fusion analysis period, including:

[0070] Based on MWRI sea surface wind speed products, a multi-source satellite data resource pool is built to obtain data files;

[0071] Generate a time window with 3 hours before and after the selected time stamp for the specified date and time, and convert the data file into a time stamp to obtain the time window and file time stamp;

[0072] Based on the time window and file timestamp, the files with timestamps falling within the time window are screened, and the multi-source satellite sea surface wind field data are read respectively. The data are screened according to the orbital area and the data quality control and processing are completed to obtain the processed data;

[0073] Obtain EC background field data, and use the EC background field as an initial estimate of the fused wind field to obtain the EC background field and fused wind field data;

[0074] The cost function is obtained by combining the EC background field with the wind speed, divergence, vorticity and second-order derivative of wind speed at each point in the fused wind field. The formula of the cost function is:

[0075]

[0076] Where j1 represents the cost function value; N represents the number of points in the fused wind field; M represents the number of observation data points; v EC,i represents the wind speed of the EC background field at point i; v Fuse，i Indicates the wind speed at point i in the fused wind field; div EC，i represents the divergence of the EC background field at point i; div Fuse，i represents the divergence of the fused wind field at point i; vor EC，i represents the vorticity of the EC background field at point i; vor Fuse，i represents the vorticity of the fused wind field at point i; Δv EC，i represents the second-order derivative of wind speed at point i in the EC background field; Δv Fuse，i represents the second-order derivative of the wind speed at point i in the fused wind field; v Obs，j represents the observed wind speed at point j; v FuseInterp，j It represents the wind speed at point j after spatial interpolation of the fused wind field.

[0077] In an embodiment of the present invention, by constructing a multi-source satellite data resource pool and generating a time window, relevant data can be retrieved and utilized more efficiently, and through a strict data screening and quality control process, inaccurate or abnormal data points can be removed, thereby improving the accuracy and reliability of the overall data; by reading the EC background field data based on a selected time, the EC background field is used as the initial estimate of the fused wind field, and the sum of the squares of the wind speed, divergence, vorticity and wind speed laplace term errors between the EC background field and each point in the fused wind field, as well as the sum of the squares of the wind speed errors between the fused wind field after spatial interpolation and the observed data are used as cost functions.

[0078] In the present invention, by minimizing the cost function, a fused wind field that is more consistent with the EC background field and observation data can be obtained, thereby improving the accuracy of the data. The cost function takes into account multiple physical quantities and data characteristics, and can integrate information from different sources to obtain a comprehensive wind field description. The form of the cost function can be adjusted according to specific needs, such as adding or subtracting different terms to adapt to different application scenarios and analysis purposes.

[0079] The present invention provides a grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads. According to the cost function, the gradient terms of each cost function in the cost function are calculated and summed to obtain the function gradient. The formula is:

[0080]

[0081] Among them, θ represents the parameter vector of the fusion wind field model, The partial derivative of θ, each element of which corresponds to the partial derivative of the cost function with respect to each parameter in θ; and j represents the index of the fused wind field and observation data; j2 represents the function gradient value.

[0082] In an embodiment of the present invention, the function gradient is obtained by determining the model parameters, calculating the gradient of the wind speed cost term, calculating the gradient of the vorticity cost term, calculating the gradient of the wind speed second derivative cost term, and calculating the gradient of the observation data cost term.

[0083] In the present invention, by calculating the gradient of the cost function, we can accurately know how to adjust the model parameters to minimize the cost function. The gradient information can help the optimization algorithm converge to the optimal solution faster, reduce computing time and resource consumption, and is applicable to a variety of different types of cost functions and models, providing great flexibility. By including cost terms of physical quantities such as divergence, vorticity, and second-order derivatives of wind speed, it can ensure that the fused wind field is more physically consistent and reasonable.

[0084] The present invention provides a grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads, which obtains the final fused wind field according to a cost function and a function gradient, including:

[0085] According to the cost function and the function gradient, through θ k+1 =θ k +α k p k Calculate the minimum value of the cost function, where k = 0, 1, 2, ..., θ k+1 is the parameter vector after the k+1th iteration, θ k is the parameter vector for the kth iteration, α k is the step size determined by line search in the kth iteration, p k is the search direction in the kth iteration.

[0086] In an embodiment of the present invention, the L-BFGS algorithm combines the fast convergence of the BFGS method and the simplicity of the gradient descent method, and can maintain efficient memory usage when processing large-scale problems. Through an iterative optimization process, the parameters that minimize the cost function can be accurately found. The L-BFGS algorithm is insensitive to the choice of initial parameters and can exhibit good convergence under different cost function shapes. It is not only suitable for wind farm fusion problems, but can also be applied to other scenarios that require cost function optimization.

[0087] The sea surface wind field gridding intelligent fusion system 20 based on multi-source satellite payloads described in an embodiment of the present invention includes:

[0088] Acquisition module 21 is used to perform star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and to perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field; construct a deep neural network model, and combine the advantages of the D-Matrix statistical algorithm in the inversion of medium and low wind speeds to obtain MWRI sea surface wind speed products for the entire wind speed range.

[0089] Processing module 22 is used for spatiotemporal matching of multi-source satellite data and EC background field to obtain the data source of the fusion analysis period; the wind speed, divergence, vorticity and second-order derivative of wind speed of each point of the EC background field and the fused wind field are combined to obtain the cost function; based on the cost function, the gradient terms of each cost function in the cost function are obtained and summed to obtain the function gradient; based on the cost function and the function gradient, the final fused wind field is obtained.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A grid-based intelligent fusion method for sea surface wind fields based on multi-source satellite payloads is characterized by: include: Perform star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field; Build a deep neural network model and combine the advantages of the D-Matrix statistical algorithm for low and medium wind speed inversion to obtain MWRI sea surface wind speed products for the entire wind speed range; Multi-source satellite data are processed with spatiotemporal matching of EC background fields to obtain the data source for the fusion analysis period; The cost function is obtained by combining the EC background field with the wind speed, divergence, vorticity and second-order derivative of wind speed at each point in the fused wind field; According to the cost function, the gradient terms of each cost function in the cost function are calculated and summed to obtain the function gradient; According to the cost function and function gradient, the final fused wind field is obtained. The formula of the cost function is: Where j1 represents the cost function value; N represents the number of points in the fused wind field; M represents the number of observation data points; v EC,i represents the wind speed of the EC background field at point i; v Fuse,i Indicates the wind speed at point i in the fused wind field; div EC,i represents the divergence of the EC background field at point i; div Fuse,i represents the divergence of the fused wind field at point i; vor EC,i represents the vorticity of the EC background field at point i; vor Fuse,i represents the vorticity of the fused wind field at point i; Δv EC,i represents the second-order derivative of wind speed at point i in the EC background field; Δv Fuse,i represents the second-order derivative of the wind speed at point i in the fused wind field; v Obs,j represents the observed wind speed at point j; v FuseInterp,j It represents the wind speed at point j after spatial interpolation of the fused wind field; the formula of the function gradient is: Among them, θ represents the parameter vector of the fusion wind field model, The partial derivative of θ, each element of which corresponds to the partial derivative of the cost function with respect to each parameter in θ; i and j represent the indexes of the fused wind field and observation data; j2 represents the function gradient value.

2. The sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads according to claim 1 is characterized in that: Perform star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field, including: Perform star-to-star cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers to obtain coordinate data; Perform wind speed correction on the forecast field data to obtain the corrected forecast field data; According to the coordinate data and forecast field data, the sea breeze fusion background field is established to obtain the sea breeze fusion background field.

3. The sea surface wind field gridding intelligent fusion method based on multi-source satellite payloads according to claim 2 is characterized in that: By building a deep neural network model and combining the advantages of the D-Matrix statistical algorithm for low and medium wind speed inversion, we can obtain MWRI sea surface wind speed products for the entire wind speed range, including: Based on the sea breeze fusion background field, a deep neural network based on physical constraints and suitable for MWRI wind inversion is constructed to obtain an inversion deep neural network; The inversion deep neural network is combined with the medium and low wind speed inversion advantages of the D-Matrix statistical algorithm, and a two-dimensional variational assimilation multi-source sea breeze fusion technology is constructed for optimization to obtain the MWRI sea surface wind speed product.

4. The method for intelligent fusion of sea surface wind fields based on multi-source satellite payloads according to claim 3 is characterized in that: Multi-source satellite data is processed by temporal and spatial matching with the EC background field to obtain the data source for the fusion analysis period, including: Based on MWRI sea surface wind speed products, a multi-source satellite data resource pool is built to obtain data files; Generate a time window with 3 hours before and after the selected time stamp for the specified date and time, and convert the data file into a time stamp to obtain the time window and file time stamp; Based on the time window and file timestamp, the files with timestamps falling within the time window are screened, and the multi-source satellite sea surface wind field data are read respectively. The data are screened according to the orbital area and the data quality control and processing are completed to obtain the processed data; The EC background field data is obtained and used as the initial estimate of the fused wind field to obtain the EC background field and the fused wind field data.

5. The method for intelligent fusion of sea surface wind fields based on multi-source satellite payloads according to claim 4 is characterized in that: According to the cost function and function gradient, the final fused wind field is obtained, including: According to the cost function and the function gradient, through θ k+1 =θ k +α k p k Calculate the minimum value of the cost function, where k = 0, 1, 2, ..., θ k+1 is the parameter vector after the k+1th iteration, θ k is the parameter vector for the kth iteration, α k is the step size determined by line search in the kth iteration, p k is the search direction in the kth iteration; According to the minimum value of the cost function, the fused wind field is obtained to obtain the final fused wind field.

6. A gridded intelligent fusion system for sea surface wind fields based on multi-source satellite payloads, including: The acquisition module is used to perform satellite-to-satellite cross-calibration on the Fengyun and Ocean satellite multi-source scatterometers, and to perform wind speed correction on the forecast field data to obtain the sea breeze fusion background field. A deep neural network model is constructed, and the advantages of the D-Matrix statistical algorithm for medium and low wind speed inversion are combined to obtain MWRI sea surface wind speed products for the entire wind speed range. The processing module is used for spatiotemporal matching of multi-source satellite data and EC background field to obtain the data source of the fusion analysis period; the wind speed, divergence, vorticity and second-order derivative of wind speed at each point of the EC background field and the fusion wind field are combined to obtain the cost function; based on the cost function, the gradient terms of each cost function in the cost function are obtained and summed to obtain the function gradient; based on the cost function and the function gradient, the final fusion wind field is obtained. The formula of the cost function is: Where j1 represents the cost function value; N represents the number of points in the fused wind field; M represents the number of observation data points; v EC,i represents the wind speed of the EC background field at point i; v Fuse,i Indicates the wind speed at point i in the fused wind field; div EC,i represents the divergence of the EC background field at point i; div Fuse,i represents the divergence of the fused wind field at point i; vor EC,i represents the vorticity of the EC background field at point i; vor Fuse,i represents the vorticity of the fused wind field at point i; Δv EC,i represents the second-order derivative of wind speed at point i in the EC background field; Δv Fuse,i represents the second-order derivative of the wind speed at point i in the fused wind field; v Obs,j represents the observed wind speed at point j; v FuseInterp,j It represents the wind speed at point j after spatial interpolation of the fused wind field; the formula of the function gradient is: Among them, θ represents the parameter vector of the fusion wind field model, The partial derivative of θ, each element of which corresponds to the partial derivative of the cost function with respect to each parameter in θ; i and j represent the indexes of the fused wind field and observation data; j2 represents the function gradient value.

7. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images