Multi-source meteorological data fusion method and system based on beidou space-time reference

Abstract

The application relates to the field of meteorological data processing and satellite navigation application, and provides a multi-source meteorological data fusion method and system based on a Beidou space-time reference. The method comprises the following steps: acquiring Beidou data and multi-source meteorological data to be fused, taking the coordinated universal time timestamp output by a Beidou ground-based enhancement station as a global time reference, and setting a sliding window; taking the world geodetic coordinate system corresponding to the Beidou positioning data as a space reference, resampling the multi-source meteorological data to a preset standard grid, and combining terrain elevation data to construct a unified space coordinate system; calculating the confidence degrees of fused static terrain features and dynamic change features based on Beidou satellite orbit residual and clock difference parameters, adjusting the feature weights of the fused static terrain features and dynamic change features, and generating an enhanced fusion feature matrix.

Description

Technical Field

[0001] This application relates to the fields of meteorological data processing and satellite navigation application technology, and in particular to a method and system for multi-source meteorological data fusion based on the BeiDou spatiotemporal reference. Background Technology

[0002] With the increasing demands for accuracy and timeliness in short-term and nowcasting weather forecasts, river basin flood control, and urban flooding early warning systems, the fusion analysis of multi-source heterogeneous meteorological data has become a core technological foundation in the field of meteorological forecasting. Currently, the data sources used for meteorological forecasting encompass various heterogeneous data, including ground monitoring stations, weather radars, satellite remote sensing, and numerical weather prediction models. While these data types are complementary in representing meteorological elements, they also suffer from core problems such as mismatched spatiotemporal scales, significant differences in feature dimensions, and inconsistent data quality, severely restricting the accuracy and reliability of meteorological forecasts. Summary of the Invention

[0003] This application provides a method and system for fusing multi-source meteorological data based on the BeiDou spatiotemporal reference. It aims to address the problem that in the existing technology, the application of the BeiDou satellite navigation system is mostly limited to using its retrieved atmospheric parameters as an auxiliary supplementary data source. It has not fully explored and utilized the high-precision timing and positioning capabilities of the BeiDou system, nor has it constructed a full-process spatiotemporal reference system with BeiDou as the core. As a result, it cannot fundamentally solve the core pain point of inconsistent spatiotemporal references for multi-source heterogeneous meteorological data, and it is difficult to meet the application needs of current refined meteorological services.

[0004] In a first aspect, embodiments of this application provide a method for fusing multi-source meteorological data based on the BeiDou spatiotemporal reference, the method comprising: The process involves acquiring BeiDou data and multi-source meteorological data to be merged, using the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference, and setting a sliding window. Based on the sliding window and setting the BeiDou data to be merged as the time anchor point, the process involves time synchronization of multi-source meteorological data with different sampling frequencies using multiple spline interpolation methods. Using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, multi-source meteorological data is resampled to a preset standard grid and combined with topographic elevation data to construct a unified spatial coordinate system. In the unified spatial coordinate system, static topographic features derived from topographic elevation data are integrated with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from the multi-source meteorological data. The confidence levels of the fused static terrain features and dynamic change features are calculated based on the orbital residuals and clock error parameters of BeiDou satellites. The feature weights of the fused static terrain features and dynamic change features are then adjusted to generate an enhanced fused feature matrix.

[0005] In some embodiments, setting a sliding window with the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference includes: acquiring nanosecond-level timing data output by the BeiDou ground-based augmentation station, generating a unified UTC timestamp sequence as the globally unique time reference; setting a sliding window with a duration of 15 minutes, using the timestamp of the BeiDou data as the anchor point for window matching, and completing the time interval division of the sliding window.

[0006] In some embodiments, the step of using a sliding window and setting the BeiDou data to be fused as a time anchor point to perform time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation includes: using the timestamp of the BeiDou data as a time anchor point, identifying the original sampling frequencies of different multi-source meteorological data within the time interval of the sliding window; and performing interpolation processing on the multi-source meteorological data with different sampling frequencies through cubic spline interpolation to achieve minute-level time synchronization of multi-source meteorological data.

[0007] In some embodiments, the step of using the World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling multi-source meteorological data to a preset standard grid, and constructing a unified spatial coordinate system in combination with terrain elevation data includes: using the 1984 World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling all multi-source meteorological data to be fused to a preset 500-meter resolution standard grid; and using terrain elevation data as a static background field to complete the construction of a unified spatial coordinate system covering the target area.

[0008] In some embodiments, the process of fusing static terrain features derived from terrain elevation data with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from multi-source meteorological data in the unified spatial coordinate system includes: calculating static terrain features of slope, aspect, and terrain roughness based on terrain elevation data; extracting atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from BeiDou data, and calculating the time variation rate of each parameter to obtain dynamic variation features; and splicing and fusing the static terrain features with the dynamic variation features to generate a standardized multi-dimensional spatiotemporal feature dataset covering the target area.

[0009] In some embodiments, the step of calculating the confidence level of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters includes: obtaining the BeiDou satellite orbit residuals and clock error parameters; calculating the accuracy of the BeiDou positioning data within each standard grid under a unified spatial coordinate system; and calculating the data quality confidence level of the fused static terrain features and dynamic change features within the corresponding grid based on the BeiDou positioning accuracy, thereby generating a confidence score factor for each grid.

[0010] In some embodiments, adjusting the feature weights corresponding to the fused static terrain features and dynamic change features to generate an enhanced fused feature matrix includes: dynamically adjusting the feature weights of the static terrain features and dynamic change features within the corresponding grid based on the confidence score factor corresponding to each grid; calculating the spatiotemporal correlation between each feature and the measured data from the ground rain gauge through mutual information entropy; repairing low consistency features with correlation below a preset threshold using spatial neighborhood interpolation; and finally generating a fused feature matrix with enhanced reliability.

[0011] In some embodiments, the method further includes: performing fast Fourier transform analysis on the generated fusion feature matrix for the temporal feature sequence of each standard grid to identify the dominant cycle of the corresponding meteorological element; and optimizing and adjusting the feature weights of the corresponding cycle in combination with the characteristics of rainstorm extreme weather events to generate a cycle-optimized fusion feature set adapted to the time series prediction model.

[0012] In some embodiments, the method further includes: acquiring measured ground rainfall data transmitted via BeiDou short message in real time, calculating the residual between the predicted value and the measured data corresponding to the fused feature matrix; when the residual exceeds a preset threshold, initiating incremental learning to fine-tune the feature weights online, completing the real-time correction of the fused feature matrix, and broadcasting the corrected fused feature data with confidence labels through the BeiDou system.

[0013] Secondly, this application provides a multi-source meteorological data fusion system based on the BeiDou spatiotemporal reference, the system comprising: The data acquisition unit is used to acquire the BeiDou data and multi-source meteorological data to be fused. It uses the Coordinated Universal Time timestamp output by the BeiDou ground-based augmentation station as the global time reference and sets a sliding window. Based on the sliding window and setting the BeiDou data to be fused as the time anchor point, it performs time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation. The feature fusion unit is used to resample multi-source meteorological data to a preset standard grid using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, and to construct a unified spatial coordinate system by combining it with terrain elevation data. In the unified spatial coordinate system, the static terrain features derived from the terrain elevation data are fused with the dynamic variation features of atmospheric precipitable water, tropospheric delay and total ionospheric electron content retrieved from the multi-source meteorological data. The matrix generation unit is used to calculate the confidence levels of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters, adjust the feature weights of the fused static terrain features and dynamic change features, and generate an enhanced fused feature matrix.

[0014] This application fundamentally solves the problems of asynchronous time and misaligned time series features of multi-source data in the prior art. Compared with traditional time synchronization methods, the error is reduced by more than 60%, and the consistency of time series features is greatly improved.

[0015] This invention constructs a globally unified Coordinated Universal Time (UTC) reference using high-precision timing data from BeiDou ground-based augmentation stations. Using BeiDou data as the time anchor, it achieves minute-level time synchronization of multi-source meteorological data with different sampling frequencies through cubic spline interpolation, strictly constraining the interpolation error to within ±15 seconds. This fundamentally solves the problems of asynchronous time and misaligned time series characteristics of multi-source data in existing technologies. Compared with traditional time synchronization methods, the error is reduced, and the consistency of time series characteristics is greatly improved.

[0016] This invention uses the 1984 Coordinate System corresponding to BeiDou positioning data as a rigid spatial reference, resamples all multi-source meteorological data to a 500-meter standard grid, and constructs a unified spatial coordinate system by combining it with terrain elevation data. This completely eliminates the spatial heterogeneity of multi-source data, solves the problems of insufficient spatial matching accuracy and poor consistency of features in complex terrain areas in existing technologies, improves spatial feature consistency by more than 35%, and achieves the unification of spatial features in fine grids.

[0017] This invention uses the BeiDou spatiotemporal reference as the core anchor point for the fusion of multi-source meteorological data, rather than using BeiDou data as an auxiliary supplementary data source. It integrates the high-precision spatiotemporal capabilities of BeiDou throughout the entire data fusion process, fundamentally solving the industry pain point of inconsistent spatiotemporal references for multi-source heterogeneous meteorological data, and providing standardized input with strict spatiotemporal alignment for subsequent meteorological forecasting models.

[0018] This invention quantifies the confidence level of fused features based on the orbital residuals and clock error parameters of BeiDou satellites, dynamically adjusts the feature weights, and performs spatial neighborhood interpolation repair on features with low consistency. This achieves dynamic quality control and reliability enhancement of fused features, effectively shielding the interference of low-quality data on the fusion results, and providing high-quality data support for high-precision weather forecasting.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1This is a schematic flowchart illustrating the steps of a multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference provided in an embodiment of this application; Figure 2 This is a schematic block diagram of a multi-source meteorological data fusion system based on the BeiDou spatiotemporal reference provided in one embodiment of this application; Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0025] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0026] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] With the increasing demands for accuracy and timeliness in short-term and nowcasting weather forecasts, river basin flood control, and urban flooding early warning systems, the fusion analysis of multi-source heterogeneous meteorological data has become a core technological foundation in the field of meteorological forecasting. Currently, the data sources used for meteorological forecasting encompass various heterogeneous data, including ground monitoring stations, weather radars, satellite remote sensing, and numerical weather prediction models. While these data types are complementary in representing meteorological elements, they also suffer from core problems such as mismatched spatiotemporal scales, significant differences in feature dimensions, and inconsistent data quality, severely restricting the accuracy and reliability of meteorological forecasts.

[0029] At the time synchronization level, the sampling frequencies of existing multi-source meteorological data vary significantly, with sampling intervals ranging from minutes to hours. Existing technologies lack a unified global high-precision time reference and often use the timestamps provided by each data source for rough matching. The interpolation error in the time synchronization process lacks a rigid constraint mechanism, which can easily lead to misalignment of time series characteristics and fail to meet the high-precision time series analysis requirements for rapidly evolving extreme weather processes such as short-term strong convection.

[0030] At the spatial unification level, existing multi-source meteorological data exhibit significant differences in spatial coordinate systems and grid resolutions. Different data sources employ varying geographic projections and spatial scales, making spatial matching between discrete station data and gridded remote sensing and numerical weather prediction data challenging. Current technologies lack a rigid and unified spatial benchmark system, resulting in insufficient matching accuracy during spatial resampling. Particularly in areas with complex terrain, the influence of topographical elements on meteorological elements is not fully coupled, leading to poor spatial feature consistency and hindering the deep fusion of meteorological features from fine-grid systems.

[0031] In terms of feature fusion and quality control, existing technologies mostly adopt fixed-weight fusion strategies, which cannot dynamically adjust feature weights according to the real-time accuracy of the data source. They also fail to quantitatively evaluate the reliability of fused features and perform anomaly repair. Low-quality and low-consistency feature data can easily interfere with the fusion results, making it impossible for the fused dataset to provide standardized and highly reliable input for subsequent high-precision prediction models.

[0032] Furthermore, in existing technologies, the application of the BeiDou Navigation Satellite System is mostly limited to using its retrieved atmospheric parameters as an auxiliary supplementary data source. It has not fully explored and utilized the high-precision timing and positioning capabilities of the BeiDou system, nor has it constructed a full-process spatiotemporal reference system with BeiDou as the core. As a result, it cannot fundamentally solve the core pain point of inconsistent spatiotemporal references for multi-source heterogeneous meteorological data, and it is difficult to meet the current application needs of refined meteorological services.

[0033] To solve the above problem, please refer to Figure 1This application provides a method for fusing multi-source meteorological data based on the BeiDou spatiotemporal reference, applied to computer equipment. The computer equipment can be deployed on a single server or server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc. It should be noted that each piece of information involved in the method provided in this application is extracted with the authorization of the relevant user and in accordance with relevant regulations, and will not infringe on user privacy.

[0034] The provided method for multi-source meteorological data fusion based on the BeiDou spatiotemporal reference includes steps S101 to S103. Details are as follows: Step S101. Obtain the BeiDou data and multi-source meteorological data to be fused. Use the Coordinated Universal Time timestamp output by the BeiDou ground-based augmentation station as the global time reference and set a sliding window. Based on the sliding window and setting the BeiDou data to be fused as the time anchor point, use multiple spline interpolation to synchronize the time of multi-source meteorological data with different sampling frequencies.

[0035] Specifically, this step forms the temporal foundation of the entire fusion method. It addresses the industry pain points of inconsistent sampling frequencies, inconsistent time bases, and misaligned temporal features in existing technologies for multi-source heterogeneous meteorological data. By leveraging the high-precision time synchronization capability of the BeiDou system, a globally unique rigid time base is constructed to achieve high-precision time synchronization of all data to be fused, eliminating the time asynchrony problem of different data sources and providing a unified and accurate temporal dimension scale for subsequent spatial fusion and feature fusion.

[0036] The first step involves the computer equipment completing the compliant collection and access of the data to be integrated, specifically including two types of core data: the first is BeiDou-related data, covering positioning data, nanosecond-level timing data, satellite orbit parameters, clock bias data output by BeiDou ground-based augmentation stations, as well as atmospheric meteorological parameters retrieved by BeiDou remote sensing payloads; the second is multi-source meteorological data, including measured rainfall data from ground rain gauges, meteorological radar reflectivity data, infrared cloud image data from Fengyun-4 satellite, ERA5 reanalysis meteorological data, GFS numerical forecast data, and other heterogeneous meteorological data sources.

[0037] The second step involves using the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the core of the computer equipment to construct a globally unique time reference. This time reference is then set as the unique time alignment scale for all data to be fused, thus shielding the deviation and asynchronous issues caused by different data sources having timestamps.

[0038] The third step involves the computer equipment setting up a sliding window adapted to the dynamic changes of meteorological elements based on the established global time reference, defining a standardized processing range for time synchronization, and explicitly using BeiDou data as the sole time anchor point for the entire time synchronization process. The time alignment of all multi-source meteorological data is performed based on the time node of BeiDou data.

[0039] The fourth step involves the computer equipment performing interpolation using cubic spline interpolation on multi-source meteorological data with different sampling frequencies. This process unifies heterogeneous data with different sampling intervals, ranging from minutes to hours, to the same minute-level time sampling granularity, completing the time synchronization of all data to be fused and ensuring that all data are strictly aligned in the time dimension.

[0040] Step S102. Using the world geodetic coordinate system corresponding to the BeiDou positioning data as the spatial reference, resample the multi-source meteorological data to a preset standard grid, and construct a unified spatial coordinate system by combining it with the terrain elevation data; in the unified spatial coordinate system, integrate the static terrain features derived from the terrain elevation data with the dynamic change features of atmospheric precipitable water, tropospheric delay and total ionospheric electron content retrieved from the multi-source meteorological data.

[0041] Specifically, this step is the core of the spatial and feature fusion of the entire fusion method. It addresses the problems in existing technologies such as inconsistent spatial coordinate systems of multi-source meteorological data, large differences in grid resolution, strong heterogeneity of spatial features, and insufficient coupling between terrain and meteorological features. Using the geographic coordinate system corresponding to BeiDou positioning data as a rigid spatial reference, a unified gridded spatial coordinate system is constructed. At the same time, the deep fusion of static terrain features and dynamic BeiDou atmospheric features is completed to generate a standardized spatiotemporal feature dataset, providing a unified spatial carrier and feature foundation for subsequent quality control and feature enhancement.

[0042] The first step involves the computer equipment using the BeiDou positioning data that has been synchronized in step S101 as a basis to extract the 1984 World Geodetic Coordinate System corresponding to the BeiDou positioning data as the globally unique spatial reference, thereby eliminating the spatial misalignment problem caused by the differences in geographic coordinate systems and projection methods used by different data sources.

[0043] The second step involves the computer equipment pre-setting the resolution parameters of the standard grid, and then using spatial resampling to uniformly map all multi-source meteorological data that have completed time synchronization to the standard grid with the preset resolution, thereby achieving uniform grid resolution across all data sources in the spatial dimension.

[0044] The third step involves the computer equipment accessing the terrain elevation data of the target area, using the terrain elevation data as the static background field of the spatial coordinate system, and matching and overlaying it grid by grid with the multi-source meteorological data grid that has undergone spatial resampling, thereby completing the construction of a unified spatial coordinate system covering the entire target monitoring area.

[0045] The fourth step involves the computer equipment extracting static terrain features derived from terrain elevation data within the constructed unified spatial coordinate system, and then combining these with the dynamic features corresponding to atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from BeiDou data. The two types of features are then fused together in a grid-to-grid manner to generate a standardized multi-dimensional spatiotemporal feature dataset covering the entire grid of the target area, thus completing the spatial dimension feature fusion processing.

[0046] Step S103. Calculate the confidence levels of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters, adjust the feature weights of the fused static terrain features and dynamic change features, and generate an enhanced fused feature matrix.

[0047] Specifically, this step is the core of quality control and output of the entire fusion method. It addresses the problems in existing technologies where multi-source data fusion uses fixed weights, cannot dynamically adjust based on real-time data quality, and low-quality features interfere with the reliability of the fusion results. Based on the real-time operating parameters of BeiDou satellites, it quantifies feature confidence, dynamically adjusts feature weights, and simultaneously verifies and repairs low-consistency features. Finally, it generates a fusion feature matrix with enhanced reliability, providing standardized and highly reliable input data for subsequent meteorological forecasting models.

[0048] The first step is to use computer equipment to obtain the orbital residual and clock error parameters transmitted in real time by Beidou satellites. Based on these parameters, the quality of the feature data corresponding to each grid in the spatiotemporal feature dataset generated in step S102 is quantitatively evaluated, and the confidence level of the corresponding feature data is calculated.

[0049] The second step involves the computer equipment adjusting the weight coefficients of various fusion features within a grid based on the confidence level of each grid feature calculated. This achieves dynamic weight allocation based on real-time data quality, strengthens the contribution of high-quality features, and reduces the interference of low-quality features on the fusion results.

[0050] The third step involves the computer equipment performing consistency checks on the weighted feature data, identifying anomalous feature data with low consistency, and repairing the anomalous features using spatial neighborhood interpolation. Finally, the optimization processing of the full grid features is completed, generating a fusion feature matrix with enhanced reliability, thus completing the entire fusion processing flow of multi-source meteorological data.

[0051] In some embodiments, setting a sliding window with the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference includes: acquiring nanosecond-level timing data output by the BeiDou ground-based augmentation station, generating a unified UTC timestamp sequence as the globally unique time reference; setting a sliding window with a duration of 15 minutes, using the timestamp of the BeiDou data as the anchor point for window matching, and completing the time interval division of the sliding window.

[0052] This embodiment is a detailed implementation of the global time reference construction and sliding window setting step S101. The core purpose is to further clarify the reference role of BeiDou high-precision time synchronization, standardize the setting rules of the sliding window, ensure the uniqueness of the time reference and the meteorological adaptability of the sliding window, and provide accurate interval constraints for subsequent time synchronization.

[0053] The first step is for the computer equipment to acquire timing data with nanosecond-level precision from the BeiDou ground-based augmentation station. Based on this timing data, a continuous and equally spaced Coordinated Universal Time (UTC) timestamp sequence is generated. This timestamp sequence is set as the globally unique time reference in the entire multi-source data fusion process. The time alignment of all data to be fused must strictly follow this reference, and independent time references from other data sources are not accepted.

[0054] The second step involves setting the duration of a sliding window to 15 minutes based on the dynamic evolution characteristics of meteorological elements. This window duration is suitable for monitoring rapidly evolving meteorological processes such as short-term severe convection and rainstorms, and it also matches the sampling frequency of BeiDou data.

[0055] The third step involves the computer equipment using the timestamp of the BeiDou data as the unique anchor point for matching the sliding window. According to the preset sliding step size, the time interval of the sliding window is divided within the entire monitoring period, ensuring that each sliding window contains a complete BeiDou reference time node, thus providing a standardized time processing interval for subsequent multi-source data time synchronization.

[0056] In some embodiments, the step of using a sliding window and setting the BeiDou data to be fused as a time anchor point to perform time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation includes: using the timestamp of the BeiDou data as a time anchor point, identifying the original sampling frequencies of different multi-source meteorological data within the time interval of the sliding window; and performing interpolation processing on the multi-source meteorological data with different sampling frequencies through cubic spline interpolation to achieve minute-level time synchronization of multi-source meteorological data.

[0057] This embodiment is a detailed implementation of the multi-source data time synchronization step based on cubic spline interpolation in step S101. The core purpose is to clarify the interpolation processing logic with BeiDou as the anchor point, standardize the synchronization process of data with different sampling frequencies, and ensure that all multi-source meteorological data achieve accurate minute-level time alignment.

[0058] The first step involves the computer equipment identifying the original sampling frequencies of all multi-source meteorological data to be fused within the 15-minute sliding window time interval, based on the BeiDou global time reference constructed in step S101. This includes heterogeneous data with different sampling intervals, such as minute-level ground rain gauge data, 30-minute-level GFS numerical forecast data, and 1-hour-level ERA5 reanalysis data.

[0059] The second step involves using the timestamp of BeiDou data as the sole anchor point for time synchronization, and determining the minute-level target time node that needs to be aligned within the sliding window. All multi-source meteorological data must be aligned to this target time node.

[0060] The third step involves using cubic spline interpolation to interpolate multi-source meteorological data with different sampling frequencies. Based on the original sampling data from each data source, a continuous time-series variation curve is fitted. Then, based on the target time node, the corresponding interpolated data is extracted, and the sampling granularity of all heterogeneous data is unified to the minute level, achieving strict time synchronization of all multi-source meteorological data under the BeiDou time reference.

[0061] In some embodiments, the step of using the World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling multi-source meteorological data to a preset standard grid, and constructing a unified spatial coordinate system in combination with terrain elevation data includes: using the 1984 World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling all multi-source meteorological data to be fused to a preset 500-meter resolution standard grid; and using terrain elevation data as a static background field to complete the construction of a unified spatial coordinate system covering the target area.

[0062] This embodiment is a detailed implementation of the unified spatial coordinate system construction step S102. The core purpose is to clarify the rigid constraints of the BeiDou spatial reference, standardize the spatial resampling process, refine the application of terrain elevation data, completely eliminate the spatial heterogeneity of multi-source data, and construct a high-precision unified gridded spatial coordinate system.

[0063] The first step involves using the time-synchronized BeiDou positioning data as a basis to extract the 1984 Coordinate System corresponding to the BeiDou positioning data as the globally unique spatial reference. This coordinate system is defined as the sole benchmark for spatial mapping of all multi-source meteorological data, and all data sources must be converted to this coordinate system for processing.

[0064] The second step involves setting the standard grid resolution of the computer equipment to 500 meters, which is sufficient to meet the spatial accuracy requirements of refined meteorological operations such as urban flooding early warning and regional flood control.

[0065] The third step involves the computer equipment performing spatial resampling on all multi-source meteorological data that have completed time synchronization, including discrete ground station data, satellite remote sensing data of different resolutions, and gridded numerical forecast data. This process maps the data to a standard grid with a resolution of 500 meters, achieving complete spatial resolution uniformity across all data sources.

[0066] The fourth step involves the computer equipment accessing the SRTM terrain elevation data of the target area. The terrain elevation data is used as a static background field for a unified spatial coordinate system and is matched and superimposed with a 500-meter standard grid grid by grid. Finally, a unified spatial coordinate system covering the entire target monitoring area is constructed, providing a unified spatial carrier for subsequent feature fusion.

[0067] In some embodiments, the process of fusing static terrain features derived from terrain elevation data with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from multi-source meteorological data in the unified spatial coordinate system includes: calculating static terrain features of slope, aspect, and terrain roughness based on terrain elevation data; extracting atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from BeiDou data, and calculating the time variation rate of each parameter to obtain dynamic variation features; and splicing and fusing the static terrain features with the dynamic variation features to generate a standardized multi-dimensional spatiotemporal feature dataset covering the target area.

[0068] This embodiment is a detailed implementation of the fusion of static terrain features and dynamic BeiDou atmospheric features in step S102. The core purpose is to clarify the specific process of feature extraction and fusion, refine the calculation dimensions of terrain features and BeiDou atmospheric features, generate a standardized multi-dimensional spatiotemporal feature dataset, and strengthen the coupling between terrain and meteorological elements.

[0069] The first step involves the computer equipment calculating, within the established unified spatial coordinate system and based on the accessed terrain elevation data, the three core static terrain features of each 500-meter standard grid: slope, aspect, and terrain roughness, thus fully characterizing the topographic features of the target area.

[0070] The second step involves the computer equipment extracting the spatiotemporally aligned BeiDou data and analyzing it to obtain three core atmospheric parameters retrieved by the remote sensing payload: precipitable water, tropospheric delay, and total electron content of the ionosphere.

[0071] The third step involves the computer equipment calculating the rate of change of each of the three types of BeiDou atmospheric parameters within a preset time period, thereby obtaining the dynamic change characteristics that characterize the dynamic evolution trend of atmospheric elements.

[0072] The fourth step involves the computer equipment stitching and fusing the static terrain features corresponding to each grid with the BeiDou atmospheric dynamic change features corresponding to the same grid in a one-to-one grid correspondence manner, generating a standardized spatiotemporal feature dataset that covers the entire grid of the target area and contains multi-dimensional features, thus completing the deep coupling and fusion of terrain and meteorological features.

[0073] In some embodiments, the step of calculating the confidence level of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters includes: obtaining the BeiDou satellite orbit residuals and clock error parameters; calculating the accuracy of the BeiDou positioning data within each standard grid under a unified spatial coordinate system; and calculating the data quality confidence level of the fused static terrain features and dynamic change features within the corresponding grid based on the BeiDou positioning accuracy, thereby generating a confidence score factor for each grid.

[0074] This embodiment is a detailed implementation of the step S103, which calculates the confidence level of features based on BeiDou parameters. The core purpose is to clarify the quantitative calculation logic of confidence level, refine the grid-level quality assessment process based on BeiDou satellite operation parameters, and provide accurate quantitative basis for subsequent feature weight adjustment.

[0075] The first step is for computer equipment to acquire the orbital residual and clock error parameters issued by the BeiDou Navigation Satellite System in real time. These parameters can directly characterize the real-time operating status of the BeiDou satellite and the accuracy level of the corresponding positioning data.

[0076] The second step involves the computer equipment calculating the real-time accuracy of the BeiDou positioning data within each 500-meter standard grid under a unified spatial coordinate system, based on the corresponding BeiDou positioning data, satellite orbit residuals, and clock error parameters.

[0077] The third step involves the computer equipment calculating the data quality confidence level of the fused static terrain features and dynamic change features within each grid, based on the BeiDou positioning accuracy corresponding to each grid and the acquisition quality of the feature data within that grid.

[0078] The fourth step involves the computer equipment generating a confidence score factor that corresponds one-to-one with the standard grid based on the confidence calculation results of the entire grid. The value range of this score factor can be dynamically adjusted between 0.3 and 0.7 according to the BeiDou positioning accuracy, providing a quantitative basis for the subsequent dynamic adjustment of feature weights.

[0079] In some embodiments, adjusting the feature weights corresponding to the fused static terrain features and dynamic change features to generate an enhanced fused feature matrix includes: dynamically adjusting the feature weights of the static terrain features and dynamic change features within the corresponding grid based on the confidence score factor corresponding to each grid; calculating the spatiotemporal correlation between each feature and the measured data from the ground rain gauge through mutual information entropy; repairing low consistency features with correlation below a preset threshold using spatial neighborhood interpolation; and finally generating a fused feature matrix with enhanced reliability.

[0080] This embodiment is a detailed implementation of the feature weight adjustment and fusion feature matrix generation step S103. The core purpose is to clarify the rules for dynamic weight adjustment, standardize the verification and repair process of low consistency features, and finally generate a highly reliable enhanced fusion feature matrix.

[0081] The first step is to use computer equipment to obtain the confidence score factor corresponding to each generated standard grid. Based on this score factor, the feature weights of static terrain features and dynamic features within the corresponding grid are dynamically adjusted. The higher the confidence score, the larger the corresponding weight coefficient, and vice versa, thus realizing adaptive weight allocation based on real-time data quality.

[0082] The second step involves the computer equipment using the mutual information entropy calculation method to calculate the spatiotemporal correlation between the weighted feature data and the measured data from ground rain gauges in the corresponding area, thereby quantifying the degree of correlation between each fused feature and the measured rainfall data.

[0083] The third step involves the computer equipment to preset a correlation threshold. Features with spatiotemporal correlation below the preset threshold are marked as low-consistency features. For these low-consistency features, spatial neighborhood interpolation is used to perform interpolation repair based on the surrounding grid feature data with high correlation and high confidence, thereby eliminating the interference of abnormal features on the fusion results.

[0084] The fourth step involves the computer equipment integrating the full-grid feature data that has undergone weight adjustment and anomaly repair, ultimately generating a fusion feature matrix that covers the entire target area and has enhanced reliability, thus completing the fusion processing of the entire multi-source meteorological data.

[0085] In some embodiments, the method further includes: performing fast Fourier transform analysis on the generated fusion feature matrix for the temporal feature sequence of each standard grid to identify the dominant cycle of the corresponding meteorological element; and optimizing and adjusting the feature weights of the corresponding cycle in combination with the characteristics of rainstorm extreme weather events to generate a cycle-optimized fusion feature set adapted to the time series prediction model.

[0086] This embodiment is an extended and optimized implementation based on the core method. The core purpose is to perform time-series periodic optimization processing on the fusion feature matrix, enhance the adaptability of the fusion features to the characteristics of meteorological time-series periods, improve the support capability of the fusion features for subsequent meteorological prediction models, and especially improve the predictive adaptability of extreme rainstorm weather.

[0087] The first step involves the computer equipment acquiring the generated reliability-enhanced fusion feature matrix. For the time-series feature sequence corresponding to each standard grid in the matrix, a fast Fourier transform analysis is performed to convert the time-domain time-series features to the frequency domain for processing.

[0088] The second step involves the computer equipment identifying the dominant cycles of each meteorological element within the corresponding grid based on the frequency domain analysis results. These cycles include short cycles on an hourly scale, medium cycles on a weekly scale, and long cycles on a yearly scale, thus fully capturing the temporal and periodic variation patterns of meteorological elements.

[0089] The third step involves using computer equipment to combine the temporal characteristics of extreme weather events such as rainstorms, and to optimize and adjust the weight coefficients of the corresponding periodic features for high-frequency, short-cycle events, thereby enhancing the contribution of temporal features related to extreme weather.

[0090] The fourth step involves the computer equipment generating a periodically optimized fusion feature set adapted to the time series prediction model based on the results of periodic identification and weight optimization. This further enhances the ability of the fusion features to represent the evolution of meteorological time series, providing higher-quality input data for subsequent high-precision rainfall prediction.

[0091] In some embodiments, the method further includes: acquiring measured ground rainfall data transmitted via BeiDou short message in real time, calculating the residual between the predicted value and the measured data corresponding to the fused feature matrix; when the residual exceeds a preset threshold, initiating incremental learning to fine-tune the feature weights online, completing the real-time correction of the fused feature matrix, and broadcasting the corrected fused feature data with confidence labels through the BeiDou system.

[0092] This embodiment is a closed-loop optimization implementation based on the core method. The core objective is to construct a real-time feedback closed loop based on BeiDou short messages, realize online dynamic correction of the fused feature matrix, and simultaneously complete the real-time broadcast distribution of fused data, thereby improving the real-time performance, reliability, and engineering application value of the method.

[0093] The first step involves the computer equipment acquiring real-time ground rainfall data uploaded by ground monitoring terminals within the target area through the short message communication channel of the BeiDou satellite navigation system. This measured data is also spatiotemporally aligned based on the BeiDou spatiotemporal reference.

[0094] The second step involves the computer equipment comparing the acquired ground rainfall data with the rainfall-related predicted values ​​corresponding to the fused feature matrix, calculating the residual between the two, and quantifying the prediction bias of the fused features.

[0095] The third step involves the computer equipment to preset a residual threshold. When the calculated residual exceeds the preset threshold, the incremental learning process is automatically triggered, the parameters of the basic feature extraction module are frozen, and online fine-tuning is performed only on modules related to feature weights to complete the real-time correction of the fused feature matrix. The response time of the entire online fine-tuning is controlled within 180 seconds.

[0096] The fourth step involves the computer equipment transmitting the corrected fusion feature data, along with the confidence level label for each grid, to various receiving terminals within the target area via the broadcast channel of the BeiDou satellite system. This provides real-time, highly reliable fusion meteorological data support for emergency operations such as flood warnings and flood control scheduling.

[0097] In some embodiments, multi-source data acquisition and spatiotemporal alignment include: BeiDou data acquisition: acquiring millimeter-level positioning data (longitude / latitude / elevation / timestamp) from ground-based augmentation stations; analyzing meteorological elements retrieved from remote sensing payloads: atmospheric precipitable water (PWV), tropospheric delay (ZTD), and ionospheric TEC value; and synchronizing BeiDou satellite orbital parameters (clock error, orbital residual) for subsequent model calibration.

[0098] The fusion of multi-source observation data includes: ground monitoring: minute-level rainfall from rain gauges and reflectivity from meteorological radar; satellite remote sensing: Fengyun-4 infrared cloud image (cloud top temperature); numerical forecasting: ERA5 reanalysis data (temperature / humidity / wind field vertical profiles) and GFS forecast data (precipitation probability / vertical velocity).

[0099] Spatiotemporal alignment processing includes: Time alignment: unifying the UTC timestamp benchmark; 15-minute sliding window matching (using BeiDou data as the time anchor); achieving minute-level synchronization of ERA5 / GFS data through cubic spline interpolation; Spatial gridding: resampling all data to a 500-meter standard grid (WGS-84 coordinate system); using terrain elevation data as a static background field; Feature engineering: generating the BeiDou PWV time variation rate; calculating ERA5 water vapor flux divergence and GFS convective available potential energy. Dataset construction includes: ; The feature channels include: BeiDou positioning data (3D), BeiDou meteorological inversion (2D), ground rainfall (1D), radar reflectivity (1D), satellite cloud image (1D), ERA5 variables (5D: temperature, humidity, air pressure, wind speed, wind direction), and GFS forecast variables (2D: precipitation forecast and temperature forecast for the next 6 hours).

[0100] TimesNet spatiotemporal feature extraction includes: Temporal periodicity decomposition: performing FFT on the time series of each grid point. ; In the formula, For frequency domain coefficients, This is time-domain data, where T is the time series length. The periodic frequency.

[0101] Dominant period for detection: Short period: 1 hour scale (daily variation); Medium period: 1 week scale (weather system evolution); Long period: 1 year scale (seasonal cycle); Dominant period is analyzed by Fast Fourier Transform (FFT): BeiDou PWV spectrum characteristics optimization period selection: Rainstorm events are given an additional 30-minute period.

[0102] Spatiotemporal joint embedding includes: spatial location encoding: three-dimensional geographic coordinates (latitude / longitude / elevation) are generated into a 128-dimensional vector through spherical convolution, and terrain complexity features are used as static enhancements.

[0103] ; In the formula, Geohash is for spatial coding, while MLP is for latitude and longitude coding. This is element-wise addition.

[0104] BeiDou Feature Enhancement: PWV change rate is used as a temporal gradient feature, and ZTD and ionospheric TEC values ​​are mapped to high-dimensional meteorological embeddings.

[0105] ; In the formula, To enhance the characteristics of BeiDou, ∥ represents a one-dimensional convolution operation, and ∥ represents a concatenation operation.

[0106] Temporal 2D reshaping includes: converting 1D temporal series into 2D tensors: ; In the formula, Describes the Reshape it into a new tensor shape, where: It is a dimension. It is based on and parameters The calculated dimensions It is the channel dimension.

[0107] TimesNet core processing includes: Multi-periodic spatiotemporal convolution: Inception-style convolution processing uses parallel multi-scale convolution kernels (3×3, 5×5, 7×7), with diagonal convolution kernels capturing frontal movement features. Period-specific modeling is shown in the table below. BeiDou dynamic calibration injects BeiDou orbital parameters (clock error / orbit residual) as calibration factors into the feature fusion layer: ; In the formula, α: calibration factor, W: weight matrix, Track residuals Clock difference.

[0108] Features after calibration: ; In the formula, Periodic observation data : BeiDou feature enhancement, α: weighting factor.

[0109] Terrain-meteorological adaptive (spatial attention enhancement) includes: dynamic weight adjustment matrix, incorporating meteorological condition weights, and constructing a terrain attention mask. ; In the formula, Dynamic weight matrix Multilayer perceptron, Topographic slope map ERA5 stability information :GFS Precipitation Probability.

[0110] Final characteristics: ; In the formula, This indicates a normalization operation. Element-wise multiplication (Hadamard product).

[0111] The prediction head architecture corresponding to the multi-scale prediction output includes: The probability prediction output corresponding to uncertainty quantification includes: In the formula, Predicted value, μ: mean :variance.

[0112] Variance prediction input: In the formula, Multilayer perceptron, Final characteristic.

[0113] Dynamic confidence assessment includes: The multi-stage training strategy corresponding to training optimization and real-time correction includes: basic pre-training: data: 5 years of ERA5 reanalysis data + BeiDou historical PWV.

[0114] Loss function: ; In the formula, the loss function It contains multiple items, each representing a different goal or constraint. Indicates the target value. This represents the model's predicted value. It refers to the number of time steps or the number of samples. Represents the model's predicted values The gradient in space (i.e., the rate of change of the predicted value at each location). This is the L2 norm of the spatial gradient, which measures the degree of variation of the predicted value in space. λ is a regularization parameter that controls the weight of the spatial smoothing term. It is a continuous ranking probability score used to assess the differences between probability distributions. It is a hyperparameter that controls the weight of the CRPS term in the total loss.

[0115] Incremental fine-tuning: Extreme weather events automatically trigger the freezing of the TimesBlock underlying layer, only optimizing the sample weight of the first typhoon area in the forecast, increasing it by 5 times.

[0116] Four-dimensional assimilation correction includes: ; In the formula, This is the transformation coefficient matrix.

[0117] Kalman filter updates include: ; in, Kalman gain for calculating BeiDou positioning accuracy; Ground station / BeiDou PWV real-time observation.

[0118] The BeiDou feedback loop includes: 1. Terminal feedback: Users send real-time rainfall observations via BeiDou short message service; 2. Error detection: Calculate the residuals ; 3. Decision Trigger: Initiate online fine-tuning at any time; 4. Model update: Incremental learning updates the prediction head parameters (response time < 180 seconds); 5. Result delivery: Predictive grid labeled with confidence level broadcast by the BeiDou system; This technical process is based on the TimesNet model and combines multi-source data such as BeiDou satellite, ERA5, and GFS. Through spatiotemporal coding, feature fusion, and dynamic adjustment technologies, it provides high-precision and high-reliability rainfall forecasts to meet the forecasting needs of different meteorological conditions and regions. Users can use these forecast data to carry out emergency response or preventive measures.

[0119] The technical advantages of this embodiment include: 1. Periodic pattern capture: TimesNet's periodic decomposition capability significantly improves the prediction accuracy of daily / seasonal variations; 2. Multi-source data fusion: ERA5 provides the climate background field, and GFS provides prior knowledge for numerical forecasting; 3. BeiDou Dynamic Calibration: Real-time correction feature representation of satellite orbit parameters; 4. Optimized computational efficiency: 2D convolution reduces the computational cost of self-attention by 40% compared to Transformer (for the same input size); 5. Terrain Adaptation: Enhances the prediction capability for complex terrain areas through the spatial gradient loss function.

[0120] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of a multi-source meteorological data fusion system 200 based on the BeiDou spatiotemporal reference provided in this application embodiment. This multi-source meteorological data fusion system 200 based on the BeiDou spatiotemporal reference is used to execute the steps of the multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference shown in the above embodiments. The multi-source meteorological data fusion system 200 based on the BeiDou spatiotemporal reference can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0121] like Figure 2 As shown, the multi-source meteorological data fusion system 200 based on the BeiDou spatiotemporal reference includes: The data acquisition unit 201 is used to acquire the BeiDou data to be fused and multi-source meteorological data. It sets a sliding window with the Coordinated Universal Time timestamp output by the BeiDou ground-based augmentation station as the global time reference. Based on the sliding window and setting the BeiDou data to be fused as the time anchor point, it performs time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation. The feature fusion unit 202 is used to resample multi-source meteorological data to a preset standard grid using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, and to construct a unified spatial coordinate system by combining it with terrain elevation data. In the unified spatial coordinate system, the static terrain features derived from the terrain elevation data are fused with the dynamic variation features of atmospheric precipitable water, tropospheric delay and total ionospheric electron content retrieved from the multi-source meteorological data. The matrix generation unit 203 is used to calculate the confidence levels of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters, adjust the feature weights of the fused static terrain features and dynamic change features, and generate an enhanced fused feature matrix.

[0122] In some embodiments, setting a sliding window with the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference includes: acquiring nanosecond-level timing data output by the BeiDou ground-based augmentation station, generating a unified UTC timestamp sequence as the globally unique time reference; setting a sliding window with a duration of 15 minutes, using the timestamp of the BeiDou data as the anchor point for window matching, and completing the time interval division of the sliding window.

[0123] In some embodiments, the step of using a sliding window and setting the BeiDou data to be fused as a time anchor point to perform time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation includes: using the timestamp of the BeiDou data as a time anchor point, identifying the original sampling frequencies of different multi-source meteorological data within the time interval of the sliding window; and performing interpolation processing on the multi-source meteorological data with different sampling frequencies through cubic spline interpolation to achieve minute-level time synchronization of multi-source meteorological data.

[0124] In some embodiments, the step of using the World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling multi-source meteorological data to a preset standard grid, and constructing a unified spatial coordinate system in combination with terrain elevation data includes: using the 1984 World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling all multi-source meteorological data to be fused to a preset 500-meter resolution standard grid; and using terrain elevation data as a static background field to complete the construction of a unified spatial coordinate system covering the target area.

[0125] In some embodiments, the process of fusing static terrain features derived from terrain elevation data with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from multi-source meteorological data in the unified spatial coordinate system includes: calculating static terrain features of slope, aspect, and terrain roughness based on terrain elevation data; extracting atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from BeiDou data, and calculating the time variation rate of each parameter to obtain dynamic variation features; and splicing and fusing the static terrain features with the dynamic variation features to generate a standardized multi-dimensional spatiotemporal feature dataset covering the target area.

[0126] In some embodiments, the step of calculating the confidence level of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters includes: obtaining the BeiDou satellite orbit residuals and clock error parameters; calculating the accuracy of the BeiDou positioning data within each standard grid under a unified spatial coordinate system; and calculating the data quality confidence level of the fused static terrain features and dynamic change features within the corresponding grid based on the BeiDou positioning accuracy, thereby generating a confidence score factor for each grid.

[0127] In some embodiments, adjusting the feature weights corresponding to the fused static terrain features and dynamic change features to generate an enhanced fused feature matrix includes: dynamically adjusting the feature weights of the static terrain features and dynamic change features within the corresponding grid based on the confidence score factor corresponding to each grid; calculating the spatiotemporal correlation between each feature and the measured data from the ground rain gauge through mutual information entropy; repairing low consistency features with correlation below a preset threshold using spatial neighborhood interpolation; and finally generating a fused feature matrix with enhanced reliability.

[0128] In some embodiments, the method further includes: performing fast Fourier transform analysis on the generated fusion feature matrix for the temporal feature sequence of each standard grid to identify the dominant cycle of the corresponding meteorological element; and optimizing and adjusting the feature weights of the corresponding cycle in combination with the characteristics of rainstorm extreme weather events to generate a cycle-optimized fusion feature set adapted to the time series prediction model.

[0129] In some embodiments, the method further includes: acquiring measured ground rainfall data transmitted via BeiDou short message in real time, calculating the residual between the predicted value and the measured data corresponding to the fused feature matrix; when the residual exceeds a preset threshold, initiating incremental learning to fine-tune the feature weights online, completing the real-time correction of the fused feature matrix, and broadcasting the corrected fused feature data with confidence labels through the BeiDou system.

[0130] It should be noted that, for the sake of convenience and brevity, the specific working process of the multi-source meteorological data fusion system and its modules based on the BeiDou spatiotemporal reference described above can be found in the corresponding contents of the various embodiments of the multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference, and will not be repeated here.

[0131] The aforementioned multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference can be implemented as a computer program, which can be used in various ways, such as... Figure 2 It runs on the device shown.

[0132] Please see Figure 3 , Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0133] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference.

[0134] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0135] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference.

[0136] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0137] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0138] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The process involves acquiring BeiDou data and multi-source meteorological data to be merged, using the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference, and setting a sliding window. Based on the sliding window and setting the BeiDou data to be merged as the time anchor point, the process involves time synchronization of multi-source meteorological data with different sampling frequencies using multiple spline interpolation methods. Using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, multi-source meteorological data is resampled to a preset standard grid and combined with topographic elevation data to construct a unified spatial coordinate system. In the unified spatial coordinate system, static topographic features derived from topographic elevation data are integrated with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from the multi-source meteorological data. The confidence levels of the fused static terrain features and dynamic change features are calculated based on the orbital residuals and clock error parameters of BeiDou satellites. The feature weights of the fused static terrain features and dynamic change features are then adjusted to generate an enhanced fused feature matrix.

[0139] In some embodiments, setting a sliding window with the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference includes: acquiring nanosecond-level timing data output by the BeiDou ground-based augmentation station, generating a unified UTC timestamp sequence as the globally unique time reference; setting a sliding window with a duration of 15 minutes, using the timestamp of the BeiDou data as the anchor point for window matching, and completing the time interval division of the sliding window.

[0140] In some embodiments, the step of using a sliding window and setting the BeiDou data to be fused as a time anchor point to perform time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation includes: using the timestamp of the BeiDou data as a time anchor point, identifying the original sampling frequencies of different multi-source meteorological data within the time interval of the sliding window; and performing interpolation processing on the multi-source meteorological data with different sampling frequencies through cubic spline interpolation to achieve minute-level time synchronization of multi-source meteorological data.

[0141] In some embodiments, the step of using the World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling multi-source meteorological data to a preset standard grid, and constructing a unified spatial coordinate system in combination with terrain elevation data includes: using the 1984 World Geodetic Coordinate System corresponding to the BeiDou positioning data as a spatial reference, resampling all multi-source meteorological data to be fused to a preset 500-meter resolution standard grid; and using terrain elevation data as a static background field to complete the construction of a unified spatial coordinate system covering the target area.

[0142] In some embodiments, the process of fusing static terrain features derived from terrain elevation data with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from multi-source meteorological data in the unified spatial coordinate system includes: calculating static terrain features of slope, aspect, and terrain roughness based on terrain elevation data; extracting atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from BeiDou data, and calculating the time variation rate of each parameter to obtain dynamic variation features; and splicing and fusing the static terrain features with the dynamic variation features to generate a standardized multi-dimensional spatiotemporal feature dataset covering the target area.

[0143] In some embodiments, the step of calculating the confidence level of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters includes: obtaining the BeiDou satellite orbit residuals and clock error parameters; calculating the accuracy of the BeiDou positioning data within each standard grid under a unified spatial coordinate system; and calculating the data quality confidence level of the fused static terrain features and dynamic change features within the corresponding grid based on the BeiDou positioning accuracy, thereby generating a confidence score factor for each grid.

[0144] In some embodiments, adjusting the feature weights corresponding to the fused static terrain features and dynamic change features to generate an enhanced fused feature matrix includes: dynamically adjusting the feature weights of the static terrain features and dynamic change features within the corresponding grid based on the confidence score factor corresponding to each grid; calculating the spatiotemporal correlation between each feature and the measured data from the ground rain gauge through mutual information entropy; repairing low consistency features with correlation below a preset threshold using spatial neighborhood interpolation; and finally generating a fused feature matrix with enhanced reliability.

[0145] In some embodiments, the method further includes: performing fast Fourier transform analysis on the generated fusion feature matrix for the temporal feature sequence of each standard grid to identify the dominant cycle of the corresponding meteorological element; and optimizing and adjusting the feature weights of the corresponding cycle in combination with the characteristics of rainstorm extreme weather events to generate a cycle-optimized fusion feature set adapted to the time series prediction model.

[0146] In some embodiments, the method further includes: acquiring measured ground rainfall data transmitted via BeiDou short message in real time, calculating the residual between the predicted value and the measured data corresponding to the fused feature matrix; when the residual exceeds a preset threshold, initiating incremental learning to fine-tune the feature weights online, completing the real-time correction of the fused feature matrix, and broadcasting the corrected fused feature data with confidence labels through the BeiDou system.

[0147] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the multi-source meteorological data fusion method based on the BeiDou spatiotemporal reference provided in any embodiment of this application.

[0148] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMediaCard (SMC), SecureDigital (SD) card, or FlashCard equipped on the computer device.

[0149] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for fusing multi-source meteorological data based on the BeiDou spatiotemporal reference, characterized in that, include: The process involves acquiring BeiDou data and multi-source meteorological data to be merged, using the Coordinated Universal Time (UTC) timestamp output by the BeiDou ground-based augmentation station as the global time reference, and setting a sliding window. Based on the sliding window and setting the BeiDou data to be merged as the time anchor point, the process involves time synchronization of multi-source meteorological data with different sampling frequencies using multiple spline interpolation methods. Using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, multi-source meteorological data is resampled to a preset standard grid and combined with topographic elevation data to construct a unified spatial coordinate system. In the unified spatial coordinate system, static topographic features derived from topographic elevation data are integrated with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from the multi-source meteorological data. The confidence levels of the fused static terrain features and dynamic change features are calculated based on the orbital residuals and clock error parameters of BeiDou satellites. The feature weights of the fused static terrain features and dynamic change features are then adjusted to generate an enhanced fused feature matrix.

2. The method according to claim 1, characterized in that, The method of using the Coordinated Universal Time timestamp output by the BeiDou ground-based augmentation station as the global time reference and setting a sliding window includes: The system acquires nanosecond-level timing data output from BeiDou ground-based augmentation stations and generates a unified Coordinated Universal Time (UTC) timestamp sequence as a globally unique time reference. A sliding window with a duration of 15 minutes is set, and the time interval of the sliding window is divided using the timestamp of the BeiDou data as the anchor point for window matching.

3. The method according to claim 1, characterized in that, The process of synchronizing multi-source meteorological data with different sampling frequencies by using a sliding window and setting the BeiDou data to be merged as time anchors, and employing multiple spline interpolation methods, includes: Using the timestamp of BeiDou data as the time anchor, the original sampling frequency of different multi-source meteorological data is identified within the time interval of the sliding window; the cubic spline interpolation method is used to interpolate the multi-source meteorological data with different sampling frequencies to achieve minute-level time synchronization of multi-source meteorological data.

4. The method according to claim 1, characterized in that, The process involves using the global geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, resampling multi-source meteorological data to a preset standard grid, and constructing a unified spatial coordinate system by combining it with terrain elevation data. This includes: Using the 1984 World Geodetic Coordinate System corresponding to BeiDou positioning data as the spatial reference, all multi-source meteorological data to be fused are resampled and unified to a preset 500-meter resolution standard grid. Using terrain elevation data as a static background field, a unified spatial coordinate system covering the target area is constructed.

5. The method according to claim 1, characterized in that, The process of integrating static terrain features derived from terrain elevation data with dynamic variation features of atmospheric precipitable water, tropospheric delay, and total ionospheric electron content retrieved from multi-source meteorological data within the unified spatial coordinate system includes: Static topographic features such as slope, aspect, and topographic roughness are calculated based on topographic elevation data; atmospheric precipitable water, tropospheric delay, and total ionospheric electron content are extracted from BeiDou data and the time variation rate of each parameter is calculated to obtain dynamic variation features. Static terrain features are spliced ​​and fused with dynamic change features to generate a standardized multi-dimensional spatiotemporal feature dataset covering the target area.

6. The method according to claim 1, characterized in that, The confidence levels corresponding to the static terrain features and dynamic change features calculated and fused based on the BeiDou satellite orbit residuals and clock error parameters include: Obtain the orbital residuals and clock errors of BeiDou satellites, and calculate the accuracy of BeiDou positioning data within each standard grid under a unified spatial coordinate system. Based on the BeiDou positioning accuracy, the data quality confidence scores of the fused static terrain features and dynamic change features within the corresponding grid are calculated, and a confidence score factor is generated for each grid.

7. The method according to claim 1, characterized in that, The process of adjusting the feature weights corresponding to the fused static terrain features and dynamic change features to generate an enhanced fused feature matrix includes: Based on the confidence score factor corresponding to each grid, the feature weights of static terrain features and dynamic changes within the corresponding grid are dynamically adjusted. The spatiotemporal correlation between each feature and the measured data from ground rain gauges is calculated by mutual information entropy. For low consistency features with correlation below a preset threshold, spatial neighborhood interpolation is used for repair, and finally, a fusion feature matrix with enhanced reliability is generated.

8. The method according to claim 1, characterized in that, The method further includes: For the generated fusion feature matrix, fast Fourier transform analysis is performed on the temporal feature sequence of each standard grid to identify the dominant cycle of the corresponding meteorological element; combined with the characteristics of rainstorm extreme weather events, the feature weights of the corresponding cycles are optimized and adjusted to generate a cycle-optimized fusion feature set adapted to the time series prediction model.

9. The method according to claim 1, characterized in that, The method further includes: Real-time acquisition of ground rainfall data transmitted via BeiDou short message; calculation of the residual between the predicted value and the measured data corresponding to the fused feature matrix; when the residual exceeds a preset threshold, incremental learning is initiated to fine-tune the feature weights online, completing the real-time correction of the fused feature matrix, and broadcasting the corrected fused feature data with confidence labels through the BeiDou system.

10. A multi-source meteorological data fusion system based on the BeiDou spatiotemporal reference, characterized in that, The method applied to any one of claims 1-9 includes: The data acquisition unit is used to acquire the BeiDou data and multi-source meteorological data to be fused. It uses the Coordinated Universal Time timestamp output by the BeiDou ground-based augmentation station as the global time reference and sets a sliding window. Based on the sliding window and setting the BeiDou data to be fused as the time anchor point, it performs time synchronization of multi-source meteorological data with different sampling frequencies through multiple spline interpolation. The feature fusion unit is used to resample multi-source meteorological data to a preset standard grid using the world geodetic coordinate system corresponding to BeiDou positioning data as a spatial reference, and to construct a unified spatial coordinate system by combining it with terrain elevation data. In the unified spatial coordinate system, the static terrain features derived from the terrain elevation data are fused with the dynamic variation features of atmospheric precipitable water, tropospheric delay and total ionospheric electron content retrieved from the multi-source meteorological data. The matrix generation unit is used to calculate the confidence levels of the fused static terrain features and dynamic change features based on the BeiDou satellite orbit residuals and clock error parameters, adjust the feature weights of the fused static terrain features and dynamic change features, and generate an enhanced fused feature matrix.

Images