Meteorological risk identification and intelligent engine digital coupling method and system

Abstract

The invention relates to the technical field of meteorological data processing, and discloses a meteorological risk identification and intelligent engine digital coupling method and system, and the method comprises the following steps: S1, multi-source data collection and integration, S2, building of a multi-dimensional grid system, S3, implementation of a multi-scene coupling technology, S4, model operation and loop optimization, and S5, product generation and service pushing. Radar and satellite cloud picture risk features are mined through the convolutional neural network and are converted into grid data, and integrated learning is combined to fuse a multi-forecast model result, so that the meteorological risk identification precision is greatly improved, and typhoon, air temperature and other related risks can be more accurately judged. Meanwhile, graphs such as a traffic network and terrain vegetation are coupled with meteorological data, the visualization and analysis capability is enhanced, and powerful support is provided for urban emergency, forest fire prevention and the like. In addition, parallel computing accelerates model operation, and an intelligent algorithm dynamically optimizes model parameters, so that the system operation efficiency is improved, and the adaptability to complex weather and various user requirements is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological data processing, and particularly to a method and system for digital coupling of meteorological risk identification and intelligent engine. Background Art

[0002] Smart meteorology is achieved through the in-depth application of new technologies such as cloud computing, Internet of Things, mobile Internet, big data, and intelligence. Relying on the progress of meteorological science and technology, the meteorological system becomes a system with the capabilities of self-perception, judgment, analysis, selection, action, innovation, and self-adaptation, making the entire process of meteorological business, services, and management activities full of wisdom. The status and role of digitalization in building the "digital foundation" of smart meteorological services are becoming increasingly obvious.

[0003] With the rapid development of information technology, the intelligent engine technology has emerged. Based on advanced technical means such as big data, artificial intelligence, and machine learning, the intelligent engine can process, analyze, and mine various types of data.

[0004] However, in actual use, on the one hand, the sources of meteorological data are extensive and the formats are diverse, making it difficult to effectively integrate and synergistically apply different types of data, resulting in low information utilization rate. On the other hand, traditional meteorological service systems cannot accurately match the personalized needs of various users and are difficult to provide customized meteorological solutions. In addition, during the meteorological risk assessment process, due to the lack of efficient models and algorithms, the data value cannot be fully exploited, resulting in poor accuracy and timeliness of risk identification. These problems limit the quality and application scope of meteorological services. In view of this, we propose a method and system for digital coupling of meteorological risk identification and intelligent engine. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a method and system for digital coupling of meteorological risk identification and intelligent engine, which solves the problem that the sources of meteorological data are extensive and the formats are diverse, making it difficult to effectively integrate and synergistically apply different types of data, resulting in low information utilization rate.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A method and system for digital coupling of meteorological risk identification and intelligent engine, including the following steps:

[0007] S1: Multi-source data collection and integration

[0008] Collect real-time meteorological monitoring data, forecast and early warning data, risk threshold data, and user object data from multiple channels, and perform standardization processing;

[0009] S2: Construct a multi-dimensional grid system

[0010] Construct a real-time grid based on different data, generate a forecast grid for meteorological forecast and warning data, generate a risk grid for meteorological risk threshold data, and generate a service grid for user object data;

[0011] S3: Implement multi-scenario coupling technology

[0012] Use the grid to implement various coupling technologies including meteorological risk monitoring and identification, digital data and layer images;

[0013] S4: Model operation and loop optimization

[0014] Perform digital and grid-based model operations, conduct multi-source data coupling and multi-model operations based on the grid, and optimize in a loop;

[0015] S5: Product generation and service push

[0016] Obtain the actual user needs and classify them. At the same time, generate digital products according to the operation results and push them according to user needs.

[0017] Preferably, in the step S1 of multi-source data acquisition and integration, meteorological data is collected by using a sensor network, meteorological satellites, and ground monitoring stations, forecast and warning data is obtained through a network interface, risk thresholds are set by combining historical data and expert experience, user object data is obtained through questionnaires and interface docking, missing data is supplemented by using a data interpolation algorithm, and abnormal fluctuation data is processed by using a moving average filtering method.

[0018] Preferably, in the step S2 of constructing a multi-dimensional grid system, grid cells are divided based on a geographic information system, the grid size is adaptively adjusted according to meteorological element changes and user distribution, the real-time grid is constructed by calculating or matching meteorological element values, the forecast grid is generated by predicting meteorological changes, the risk grid is formed corresponding to the risk level, and the service grid is constructed by associating user information.

[0019] Preferably, in the step S3 of implementing multi-scenario coupling technology, for the coupling of meteorological risk monitoring and identification, a convolutional neural network is used to extract risk features from radar images and satellite cloud images and convert them into grid data. For the coupling of meteorological monitoring and forecast and warning models, an ensemble learning method is used to fuse the results of multiple forecast models.

[0020] Preferably, in the step S3 of implementing multi-scenario coupling technology, when coupling meteorological digital data and layer images, graphics such as traffic networks and urban layouts are gridified and then overlaid with meteorological risk level data. For the coupling of meteorological basic data and image data, a comprehensive model is constructed by combining terrain, vegetation data, and remote sensing images.

[0021] Preferably, in the S4 model operation and loop optimization step, parallel computing technology is adopted for model operation, and a graphics processing unit or cluster computing resources are utilized for acceleration. For dynamic optimization, intelligent algorithms such as genetic algorithms and particle swarm optimization algorithms are used to search for and optimize model parameters.

[0022] Preferably, in the S5 product generation and service push step, the digital products adopt templatized and parameterized designs, support user-defined configurations, and dynamically update user requirements by analyzing user behaviors and feedback.

[0023] A meteorological risk identification and intelligent engine digital coupling system includes the following modules:

[0024] Data acquisition module: Configured to collect real-time meteorological monitoring data from various data sources such as meteorological satellites, ground monitoring stations, and sensor networks, obtain meteorological forecast and early warning data from meteorological departments, set meteorological risk threshold data based on historical data and expert experience, and collect user object data;

[0025] Grid construction module: Used to divide the target area into grid cells based on geographic information system technology, construct a factual grid according to real-time meteorological monitoring data, generate a forecast grid based on meteorological forecast and early warning data, form a risk grid according to meteorological risk threshold data, and establish a service grid in combination with user object data;

[0026] Coupling processing module: Capable of using the factual grid, forecast grid, risk grid, and service grid to achieve the coupling of meteorological risk monitoring and meteorological risk identification, the coupling of meteorological digital data and layer images, the coupling of meteorological monitoring and forecast and early warning models, the coupling of meteorological basic data and image data, the coupling of meteorological element changes and image changes, and the assimilation coupling of different data materials;

[0027] Model operation module: Adopts a digital and grid-based model to perform comparison analysis and model operation on numerical grid data according to grid thresholds;

[0028] Loop optimization module: Based on the above four grids, perform multi-source data coupling and multi-model operation, loop execute relevant operations, and optimize the model according to the operation results and newly collected data;

[0029] Product generation and service module: Combine the operation results after loop optimization to generate digital products, and push the digital products to users through various channels according to the risk identification results and user needs.

[0030] Preferably, the data acquisition module further includes a data preprocessing unit, which is used to clean the collected real-time meteorological monitoring data, meteorological forecast and early warning data, meteorological risk threshold data, and user object data, remove outliers and error data, unify the data format, and perform interpolation processing on missing data.

[0031] Preferably, when implementing the coupling of meteorological element changes and image changes, the coupling processing module uses the physical quantity data changes of radar and numerical weather prediction to establish a physical quantity element change analysis model, and combines graphic image recognition technology to dynamically track and analyze weather systems.

[0032] The present invention provides a method and system for digital coupling of meteorological risk identification and intelligent engine, having the following beneficial effects:

[0033] 1. The present invention mines the risk characteristics of radar and satellite cloud images through a convolutional neural network and converts them into grid data, and combines ensemble learning to fuse the results of multiple prediction models, greatly improving the accuracy of meteorological risk identification and enabling more accurate judgment of related risks such as typhoons and temperatures. At the same time, coupling graphics such as transportation networks, terrain, and vegetation with meteorological data enhances visualization and analysis capabilities, providing strong support for urban emergencies, forest fire prevention, etc. In addition, parallel computing accelerates model operations, and intelligent algorithms dynamically optimize model parameters, not only improving the system operation efficiency but also enhancing the adaptability to complex meteorology and diverse user needs.

[0034] 2. The present invention adopts the digital coupling technology of "four grids + grid model + model drive" for meteorological risk identification and intelligent engine. The six types basically integrate the digital coupling data resources and model computing power resources in four dimensions, namely "time (real-time grid and forecast and early warning grid implementation)", "space (all four grids are implemented)", "meteorological elements (real-time grid and forecast and early warning grid implementation)", and "service objects (user grid implementation)" related to risk identification and intelligent engine. Thus, it provides a standardized development plan for the top-level design of the service system and the data integration at the bottom layer of the service model. The development plan technology is replicable and popularizable, can integrally and efficiently utilize various relevant data resources, reduces costs, and improves efficiency.

[0035] 3. Through the established missing data filling algorithm, on the one hand, the present invention can improve the integrity and accuracy of meteorological data, ensuring that subsequent analysis and model operations based on these data are more reliable. On the other hand, filling in missing data can enhance the continuity of data, making the analysis of meteorological change trends more logical and scientific, providing a solid data foundation for meteorological prediction, disaster early warning, etc., helping relevant departments make more accurate and effective decisions, and protecting social production and life from the adverse effects of meteorological disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flowchart of the method for digital coupling of this meteorological risk identification and intelligent engine;

[0037] Figure 2 is a schematic diagram of the multi-scenario coupling technology of the present invention;

[0038] Figure 3 This is the module diagram of the meteorological risk identification and intelligent engine digital coupling system;

[0039] Figure 4 This is the schematic diagram of the coupling process of meteorological risk monitoring and risk identification of the present invention;

[0040] Figure 5 This is the schematic diagram of the coupling process of digital data and graphic images of the present invention;

[0041] Figure 6 This is the coupling of the meteorological monitoring and forecasting and early warning model of the present invention;

[0042] Figure 7 This is the schematic diagram of the coupling process of basic data and image data of the present invention. Specific implementation manners

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] Embodiment 1:

[0045] Please refer to the attached Figure 1 - attached Figure 7 , the embodiment of the present invention provides a meteorological risk identification and intelligent engine digital coupling method, including the following steps:

[0046] S1: Multi-source data collection and integration

[0047] Collect real-time meteorological monitoring data, forecasting and early warning data, risk threshold data and user object data from multiple channels, and perform standardization processing;

[0048] S2: Construct a multi-dimensional grid system

[0049] Construct a live grid based on different data, generate a forecast grid for meteorological forecasting and early warning data, generate a risk grid for meteorological risk threshold data, and generate a service grid for user object data;

[0050] S3: Implement multi-scenario coupling technology

[0051] Use the grid to implement various coupling technologies including meteorological risk monitoring and identification, digital data and layer images;

[0052] S4: Model operation and loop optimization

[0053] Perform digital and grid-based model operations, conduct multi-source data coupling and multi-model operations based on grids, and perform cyclic optimization;

[0054] S5: Product generation and service push

[0055] Obtain actual user requirements and classify them. At the same time, generate digital products according to the operation results and push them according to user requirements.

[0056] In the step of multi-source data collection and integration of S1, use sensor networks, meteorological satellites, and ground monitoring stations to collect meteorological data, obtain forecast and warning data through network interfaces, set risk thresholds in combination with historical data and expert experience, obtain user object data through questionnaires and interface docking, use data interpolation algorithms to supplement missing data, and use moving average filtering methods to process abnormally fluctuating data.

[0057] In the step of constructing a multi-dimensional grid system of S2, divide grid cells based on a geographic information system, adaptively adjust the grid size according to meteorological element changes and user distribution, calculate or match meteorological element values to construct real-time grids, predict meteorological changes to generate forecast grids, form risk grids corresponding to risk levels, and associate user information to construct service grids. The following algorithms are established during the process:

[0058] Algorithm for dividing grid cells based on a geographic information system (GIS):

[0059] Commonly, a grid division method based on longitude and latitude is used. For example, the earth's surface is divided into regular grids at certain longitude and latitude intervals (such as Δλ longitude interval and Δφ latitude interval). Assuming that the longitude and latitude coordinates of a point on the earth's surface are (λ, φ), the grid row and column numbers (i, j) to which the point belongs can be calculated by the following formula:

[0060]

[0061] where λ min and φ min are the minimum longitude and minimum latitude of the study area, represents the floor operation;

[0062] Algorithm for adaptively adjusting the grid size according to meteorological element changes and user distribution:

[0063] Adaptive adjustment based on meteorological element changes: Methods such as analysis of variance can be used to measure the degree of change of meteorological elements. For example, for the temperature element in a certain area, calculate the variance σ 2 :

[0064]

[0065] where n is the number of temperature measurement points in the area, Tk is the temperature value at the k-th measurement point, is the average temperature of this area;

[0066] When the variance is large, it indicates that the meteorological elements change violently. At this time, the grid size can be appropriately reduced to capture the changes of meteorological elements more precisely; when the variance is small, the grid size can be increased to reduce the computational amount;

[0067] Adaptive adjustment based on user distribution: The density estimation method, such as kernel density estimation, can be used to estimate the density of user distribution. For the user location x i , the kernel density estimation formula is:

[0068]

[0069] where n is the number of users, h is the bandwidth parameter, and K(·) is the kernel function (such as Gaussian kernel function, etc.). When the user distribution density is high, reduce the grid size to better serve users; when the user distribution density is low, increase the grid size;

[0070] Algorithm for calculating or matching meteorological element values to construct the real-time grid:

[0071] If calculating meteorological element values, for some continuous meteorological elements (such as temperature, humidity, etc.), interpolation algorithms can be used. For example, for the linear interpolation algorithm, for two known points (x1, y1) and (x2, y2), to calculate the meteorological element value y at x (x1 < x < x2), the formula is:

[0072]

[0073] If matching meteorological element values, according to the position information of the grid, the data of the station closest to the grid center can be directly found from the data of meteorological monitoring stations as the meteorological element value of this grid;

[0074] Algorithm for predicting meteorological changes to generate the forecast grid:

[0075] Common meteorological prediction models such as numerical prediction models involve complex hydrodynamic and thermodynamic equations. For example, the basic equations of atmospheric motion, and its component form in the Cartesian coordinate system is:

[0076] Momentum equation (taking the x direction as an example):

[0077]

[0078] where u, v, and w are the wind speed components in the x, y, and z directions respectively, t is the time, ρ is the air density, p is the air pressure, and F x is the external force term such as friction in the x direction;

[0079] These equations are solved on the grid through numerical discretization methods (such as finite difference method, finite element method, etc.), so as to obtain the forecast values ​​of meteorological elements at each grid point at future times and generate a forecast grid.

[0080] In the S3 step of implementing the multi-scenario coupling technology, the meteorological risk monitoring and identification coupling uses a convolutional neural network to extract risk features from radar images and satellite cloud images and convert them into grid data, and the meteorological monitoring and forecast and warning model coupling uses integrated learning to fuse multiple forecast model results.

[0081] In the S3 implementation of the multi-scenario coupling technology step, when the meteorological digital data is coupled with the layer image, the graphics such as the traffic network and the urban layout are gridded and superimposed with the meteorological risk level data. The meteorological basic data and the image data are coupled to build a comprehensive model by combining terrain, vegetation data and remote sensing images. For the data coupling part, the following algorithm is proposed:

[0082] Convolutional neural network in the coupling of meteorological risk monitoring and identification: When coupling meteorological risk monitoring and identification, convolutional neural network is used to extract risk features from radar images and satellite cloud images and convert them into grid data;

[0083] Convolution layer calculation: For the input radar image or satellite cloud image (which can be regarded as a two-dimensional input feature map X), assume that the convolution kernel is W and the output feature map is Y. Taking two-dimensional convolution as an example, its calculation formula is:

[0084]

[0085] Among them, (i, j) is the coordinate of the output feature map, (m, n) is the coordinate of the convolution kernel, and M and N are the sizes of the convolution kernel. Through the convolution operation, local features in the image can be extracted to capture risk-related patterns in meteorological images;

[0086] Pooling layer calculation (taking maximum pooling as an example): Assuming the input feature map is A and the output feature map is B, taking the common 2×2 pooling window as an example, the maximum pooling operation takes the maximum value in the window as the output, and the formula is:

[0087]

[0088] Pooling operations can reduce data dimensions and computational complexity while retaining key features. After multiple layers of convolution and pooling operations, the extracted features are converted into grid data for subsequent meteorological risk analysis.

[0089] Integrated learning in the coupling of meteorological monitoring and forecasting and warning models: When the meteorological monitoring and forecasting and warning models are coupled, integrated learning is used to fuse the results of multiple forecast models. Assume that there are T different meteorological forecast models h1(x),h2(x),…,hT (x), for the input meteorological monitoring data x (such as multi-dimensional data like temperature, humidity, wind speed, etc.), the final forecast result H(x) can be obtained in the following way;

[0090] For regression problems, such as air temperature forecasting, etc.:

[0091]

[0092] That is, the arithmetic mean of the output results of T forecasting models is taken to obtain the final forecast value.

[0093] For classification problems, such as weather condition classification, sunny, cloudy, rainy, etc.:

[0094]

[0095] Among them, is the category set, I(·) is the indicator function, and the I value is 1 when the condition in the parentheses is true, otherwise it is 0. That is, the number of votes for different categories predicted by each model is counted, and the category with the most votes is used as the final classification result

[0096] Coupling of meteorological digital data and layer images: After gridifying graphics such as traffic networks and urban layouts and overlaying them with meteorological risk level data, essentially based on the corresponding relationship of geographical coordinates, the graphic data and meteorological data are associated under the same grid system;

[0097] For example, for each grid point in the traffic network graphic, find the corresponding grid point of the meteorological risk level data, and fuse and display or analyze the information of the two.

[0098] Coupling of meteorological basic data and image data: By combining terrain, vegetation data and remote sensing images to build a comprehensive model, according to specific data types and analysis purposes, using relevant technologies and algorithms of geographic information system (GIS), spatial matching and association of different types of data are carried out, so as to build a model that can reflect the comprehensive relationship between meteorology and geographical environment.

[0099] In the S4 model operation and loop optimization step, the model operation adopts parallel computing technology, accelerates using graphics processors or cluster computing resources, and the dynamic optimization uses intelligent algorithms such as genetic algorithms and particle swarm optimization algorithms to search and optimize model parameters.

[0100] In the S5 product generation and service push step, the digital products adopt templatized and parameterized designs, support user-defined configurations, and dynamically update user requirements by analyzing user behaviors and feedback.

[0101] A meteorological risk identification and intelligent engine digital coupling system includes the following modules:

[0102] Data acquisition module: Configured to collect real-time meteorological monitoring data from various data sources such as meteorological satellites, ground monitoring stations, and sensor networks, obtain meteorological forecast and early warning data from meteorological departments, set meteorological risk threshold data based on historical data and expert experience, and collect user object data;

[0103] Grid construction module: Used to divide the target area into grid cells based on geographic information system technology, construct a real-time grid according to real-time meteorological monitoring data, generate a forecast grid based on meteorological forecast and early warning data, form a risk grid according to meteorological risk threshold data, and establish a service grid in combination with user object data;

[0104] Coupling processing module: Can utilize the real-time grid, forecast grid, risk grid, and service grid to achieve coupling of meteorological risk monitoring and meteorological risk identification, coupling of meteorological digital data and layer images, coupling of meteorological monitoring and forecast and early warning models, coupling of meteorological basic data and image data, coupling of meteorological element changes and image changes, and assimilation coupling of different data materials;

[0105] Model operation module: Adopts a digital and grid-based model to perform comparison analysis and model operations on numerical grid data according to grid thresholds;

[0106] Circular optimization module: Based on the four grids, perform multi-source data coupling and multi-model operation, circularly execute relevant operations, and optimize the model according to the operation results and newly collected data;

[0107] Product generation and service module: Combine the operation results after circular optimization to generate digital products, and push the digital products to users through various channels according to the risk identification results and user needs.

[0108] The data acquisition module further includes a data preprocessing unit, which is used to clean the collected real-time meteorological monitoring data, meteorological forecast and early warning data, meteorological risk threshold data, and user object data, remove outliers and error data, unify the data format, and perform interpolation processing on missing data. When data errors are caused by external abnormal environments during the process, the following algorithm is established to repair the data:

[0109] Suppose the interval [a, b] is divided into n subintervals [x i , x i+1 , i = 0, 1, …, n - 1, where a = x0 < x1 < … < x n = b. On each subinterval [x i , x i+1 , construct a cubic polynomial S i (x):

[0110] Si S(x) = a i + b i (x - x i ) + c i (x - x i ) 2 + d i (x - x i ) 3

[0111] To determine the coefficients a i , b i , c i , d i of these polynomials, the following conditions need to be satisfied:

[0112] 1. Function value continuity condition: At the node x i , S i (x i ) = f(x i ), S i-1 (x i ) = f(x i ), that is:

[0113] a i = f(x i )

[0114] a i-1 + b i-1 (x i - x i-1 ) + c i-1 (x i - x i-1 ) 2 + d i-1 (x i - x i-1 ) 3 = f(x i )

[0115] 2. First derivative continuity condition: At the node x i , S′ i (x i ) = S′ i-1 (x i ). Differentiating S i (x) gives S′ i (x) = b i + 2c i (x - x i ) + 3d i (x - x i ) 2 , then there is:

[0116] b i = bi-1 +2c i-1 (x i -x i-1 )+3d i-1 (x i -x i-1 ) 2

[0117] Second - derivative continuity condition: At the node \(x\) i , \(S''\) i (x i ) = \(S''\) i-1 (x i ). Differentiating \(S'\) i (x) gives \(S'\) i (x)=2c i +6d i (x - x i ). Then we have:

[0118] 2c i = 2c i-1 +6d i-1 (x i -x i-1 )

[0119] Let \(h\) i = x i+1 -x i . Through rearrangement and derivation (the process is relatively complex), a tridiagonal linear system of equations about \(c\) i can be obtained:

[0120]

[0121] Among them

[0122] After solving the above tridiagonal linear system of equations to obtain \(c\) i , then according to the previous conditions, \(d\) i , \(b\) i can be obtained in turn.

[0123] 1. Determine the interval where the missing data point is located: First, it is necessary to clarify which interval \([x\) i , \(x\) i+1 the missing data point \(x\) lies within the known data points. For example, the known data points are \(x_0,x_1,\cdots,x\) n . Find the \(i\) value that satisfies \(x\) i < x < \(x\) i+1 ;

[0124] 2. Substitute the cubic spline polynomial of the corresponding interval: For the interval \([x\) i , \(x\) i+1 , its corresponding cubic spline polynomial is \(S\)i S(x) = a i + b i (x - x i ) + c i (x - x i ) 2 + d i (x - x i ) 3 . Where a i , b i , c i , d i are coefficients obtained by solving the system of equations for cubic spline interpolation;

[0125] 3. Calculate the estimated value of the missing point: Substitute the missing data point x into the above cubic spline polynomial S i (x), and calculate the value of S i (x). This value is the estimated value of the missing data point, that is, the result of data missing processing;

[0126] For example, given the temperature values T0, T1, T2, T3 of a meteorological data sequence at time points t0, t1, t2, t3, the coefficients a i , b i , c i , d i (i = 0, 1, 2) of each interval are obtained by cubic spline interpolation. Now, to estimate the temperature value at time point t (t1 < t < t2), then use the cubic spline polynomial S1(t) = a1 + b1(t - t1) + c1(t - t1) 2 + d1(t - t1) 3 corresponding to the interval [t1, t2]. Substitute t into this formula, and the calculated S1(t) is the estimated value of the temperature at time point t, so as to fill the missing temperature data at this time point.

[0127] When implementing the coupling of meteorological element changes and image changes, the coupling processing module uses the changes in physical quantity data of radar and numerical weather prediction to establish an analysis model for physical quantity element changes, and combines graphic image recognition technology to dynamically track and analyze weather systems.

[0128] Example 2:

[0129] Urban traffic meteorological service scenario

[0130] Data collection and integration: Utilize the sensor networks, meteorological satellites, and ground monitoring stations distributed in the city to collect real-time meteorological monitoring data. For example, the sensor network collects data every 10 minutes, and the collected real-time temperature is 25°C, humidity is 60%, wind speed is 3 m / s, and precipitation is 0 mm; obtain the forecast and early warning data for the next 24 hours released by the meteorological department through the network interface. It is predicted that there will be moderate rain at night, with precipitation of 10 - 20 mm and the temperature dropping to 20°C; combine the 50-year historical meteorological data of the city and the experience of meteorological experts to set the meteorological risk thresholds for urban transportation. When the rainfall reaches 30 mm, the risk level of road waterlogging increases, and when the visibility is lower than 500 meters, the risk level of fog affecting traffic increases; obtain user object data through the interface docking with the traffic management department and the questionnaire survey of citizens' travel habits. It is known that during the morning rush hour from 7 to 9 on weekdays, the traffic flow on a certain main road is 2,000 vehicles per hour. Use data interpolation algorithms to supplement missing data, use the moving average filtering method to process abnormally fluctuating data, and then perform standardization processing.

[0131] Construct a multi-dimensional grid system: Based on the geographic information system, divide the urban area into grid units of 1 km × 1 km according to the urban road layout, functional area division, etc. Adaptively adjust the grid size according to the changes in meteorological elements and user distribution. Near transportation hubs and main roads, due to the density of people and vehicles, the grid size is reduced to 0.5 km × 0.5 km. Calculate or match meteorological element values to construct real-time grids; predict meteorological changes based on meteorological forecast models to generate forecast grids; corresponding to different meteorological risk levels, such as yellow rainstorm warning (rainfall of 50 - 100 mm), orange fog warning (visibility of 200 - 500 meters), etc., form risk grids; associate user information, such as traffic flow and travel demand in different regions, to construct service grids.

[0132] Implement multi-scenario coupling technology: In terms of the coupling of meteorological risk monitoring and identification, use convolutional neural networks to process radar images and satellite cloud images, extract risk features such as rainfall cloud clusters and fog areas and convert them into grid data to timely identify possible meteorological risk areas in the city. When coupling meteorological digital data with layer images, grid the urban traffic network graph and overlay it with meteorological risk level data to intuitively display the risk status of different road sections under different meteorological conditions. The coupling of meteorological monitoring and forecast and early warning models uses ensemble learning to fuse the results of 3 forecast models to improve the forecast accuracy, and grid the forecast results and fuse them with risk level data.

[0133] Model operation and loop optimization: Use digital and grid models to compare and analyze numerical grid data and perform model operations according to grid thresholds. Use graphics processors to accelerate the operation process, and dynamically optimize model parameters through genetic algorithms, such as adjusting the weights of meteorological elements and optimizing the forecast model structure, to improve the model's ability to predict urban traffic meteorological risks. Based on the "four grids", multivariate data coupling and multi-model operation are carried out, and operations are performed cyclically, and the model is continuously optimized based on newly collected data and operation results.

[0134] Product generation and service push: Categorize the actual user needs obtained, such as travel planning needs, traffic management needs, etc. Generate digital products based on the calculation results, such as personalized travel weather recommendations for citizens, road risk warning maps for traffic management departments, etc. Use templates and parameterized design products to support user-defined configurations. By analyzing user behavior and feedback, such as users' adoption of travel recommendations and the effect of traffic management departments' use of warning information, dynamically update user needs and accurately push products.

[0135] Embodiment three:

[0136] Agricultural production meteorological service scenario

[0137] Data collection and integration: With the help of sensor networks, meteorological satellites and ground-based agricultural meteorological monitoring stations distributed in farmland, real-time meteorological monitoring data is collected. For example, the sensor network collects data every 30 minutes, and collects real-time temperature of 28°C, light intensity of 5000 lux, and soil moisture of 65%; obtains forecast and warning data for agricultural production in the next 7 days from the meteorological department, and it is expected that there will be high temperature weather in the next 3 days, with a maximum temperature of 35°C and no precipitation; combined with the local 30-year agricultural meteorological historical data and the experience of agricultural meteorological experts, the agricultural meteorological risk threshold is set. When the temperature is higher than 33°C for 3 consecutive days, there is a risk of heat damage to crop growth, and when the soil moisture is lower than 40%, the risk of drought increases; through the interface connection with the agricultural department and the questionnaire survey of farmers, the user object data is obtained, and it is known that 1,000 acres of wheat are planted in a certain area, and the irrigation facilities cover 80% of the area. The data is processed by data interpolation algorithm and sliding average filtering method and then standardized.

[0138] Construct a multi-dimensional grid system: Based on the geographic information system, divide grid units according to farmland distribution, topography, etc. The initial grid size is 2 km × 2 km. Adaptively adjust the grid size according to meteorological element changes and crop planting distribution, and encrypt the grid to 1 km × 1 km in areas with intensive crop planting. Calculate or match meteorological element values to construct a real-time grid; predict meteorological changes to generate a forecast grid; corresponding to different agricultural meteorological risk levels, such as frost risk level (temperature below 0°C), drought risk level (soil humidity below 30%), etc., form a risk grid; associate user information, such as farmer information, crop planting information, etc., to construct a service grid.

[0139] Implement multi-scenario coupling technology: When coupling meteorological risk monitoring and identification, use a convolutional neural network to process radar images and satellite cloud images to identify meteorological disaster areas that may affect crop growth, such as the path of hail cloud clusters, the expansion of drought areas, etc., and convert them into grid data. In terms of coupling meteorological basic data and image data, combine terrain data, vegetation cover data (crop planting situation) and remote sensing images to construct a comprehensive model to analyze the impact of meteorological conditions on crop growth. The coupling of meteorological monitoring and forecast and early warning models uses ensemble learning to fuse the results of 4 forecast models, grid the forecast results and fuse them with agricultural meteorological risk level data to provide accurate meteorological services for farmers and agricultural departments.

[0140] Model operation and loop optimization: Use digital and grid models to compare and analyze numerical grid data and perform model operations, and use cluster computing resources to accelerate the operations. Use the particle swarm optimization algorithm to optimize model parameters, such as optimizing the model parameters of the relationship between meteorological elements and crop growth. Based on the "four grids", perform multi-source data coupling and multi-model operation, loop optimize the model, and continuously improve the model according to new meteorological data and crop growth status data.

[0141] Product generation and service push: Classify and obtain actual user needs, such as crop planting decision-making needs, farmland irrigation needs, etc. Generate digital products according to the operation results, such as crop planting risk assessment reports, suggestions on the timing of farmland irrigation, etc. The products adopt template-based and parameterized designs, support user-defined configuration, and dynamically update user needs by analyzing user behavior and feedback, such as the implementation of irrigation suggestions by farmers, changes in crop yields, etc., and accurately push products to farmers and agricultural departments.

[0142] Example 4

[0143] Meteorological service scenario for tourist attractions

[0144] Data collection and integration: Collect real-time meteorological monitoring data through the sensor network, meteorological satellites, and surrounding ground monitoring stations within the scenic area. For example, the sensor network collects data every 15 minutes, obtaining real-time temperature of 26°C, wind force of level 3, and air quality index of 80. Obtain the 48-hour forecast and early warning data for the area where the scenic area is located from the meteorological department. It is predicted that there will be thunderstorms tomorrow afternoon, with precipitation of 5 - 10 mm and wind force increasing to level 4 - 5. Combine the 20-year historical meteorological data of the scenic area and the research experience of meteorological experts on the scenic area to set the tourism meteorological risk threshold. When the wind force reaches level 6, it poses a risk to the safety of tourists climbing the mountain. When the precipitation exceeds 20 mm, some roads in the scenic area may be waterlogged. Through the interface connection with the scenic area management department and the questionnaire survey of tourists, obtain user object data and learn that the average daily tourist volume of a popular scenic spot is 5,000 people, and the average stay time of tourists is 3 hours. Use data interpolation algorithms and moving average filtering methods to process the data and then standardize it.

[0145] Construct a multi-dimensional grid system: Based on the geographic information system, divide grid units according to the distribution of scenic spots and terrain in the scenic area. The initial grid size is 0.8 km × 0.8 km. Adaptively adjust the grid size according to meteorological element changes and tourist distribution, and reduce the grid size to 0.4 km × 0.4 km in popular scenic spots and tourist-dense areas. Calculate or match meteorological element values to construct a real-time grid; predict meteorological changes to generate a forecast grid; corresponding to different tourism meteorological risk levels, such as a blue gale warning (wind force 6 - 7 levels), a yellow rainstorm warning (precipitation 50 - 100 mm), etc., form a risk grid; associate user information, such as tourist information, scenic spot information, etc., to construct a service grid.

[0146] Implement multi-scenario coupling technology: For meteorological risk monitoring and identification coupling, use convolutional neural networks to process radar images and satellite cloud images, extract meteorological risk features affecting tourist safety and scenic area operation, and convert them into grid data. When coupling meteorological digital data with layer images, grid the scenic area map and scenic spot distribution graphics, and overlay them with meteorological risk level data to visually display the risk status of different scenic spots under different meteorological conditions. For meteorological monitoring and forecast warning model coupling, adopt ensemble learning to fuse the results of 5 forecast models, grid the forecast results and fuse them with tourism meteorological risk level data to provide accurate meteorological information for the scenic area management department and tourists.

[0147] Model operation and cyclic optimization: Use digital and grid models to conduct comparative analysis and model operations on numerical grid data, and use a graphics processing unit to accelerate the operation. Optimize model parameters through genetic algorithms, such as adjusting the weights of meteorological elements on tourists. Based on the "four grids", conduct multi-source data coupling and multi-model operation, cyclically optimize the model, and continuously improve the model according to newly collected meteorological data and tourist flow data.

[0148] Product generation and service push: Classify and obtain actual user needs, such as tourist travel planning needs, scenic area operation management needs, etc. Generate digital products according to the operation results, such as tourist play weather guides, scenic area risk warning information, etc. The products adopt template-based and parameterized designs, support user-defined configurations, and dynamically update user needs by analyzing user behaviors and feedback, such as the usage of play guides by tourists and the response effects of warning information by scenic area management departments, and accurately push products to tourists and scenic area management departments.

[0149] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for digital coupling of meteorological risk identification and intelligent engine, characterized in that: The following steps are involved: S1: Multi-source data collection and integration Collect real-time meteorological monitoring data, forecast and warning data, risk threshold data and user object data from multiple channels and perform standardized processing; S2: Building a multi-dimensional grid system Build a real-time grid based on different data, build a forecast grid for weather forecast and warning data, build a risk grid for weather risk threshold data, and build a service grid for user object data. S3: Implementing multi-scenario coupling technology Using grid to realize multiple coupling technologies including meteorological risk monitoring and identification, digital data and layer images; S4: Model operation and loop optimization Use digital and grid model operations, grid-based multivariate data coupling and multi-model operation, and cycle optimization; S5: Product generation and service delivery Acquire actual user needs and classify them, generate digital products based on the calculation results, and push them according to user needs.

2. According to claim 1, a method for digital coupling of meteorological risk identification and intelligent engine is characterized in that: In the S1 multi-source data collection and integration step, meteorological data are collected using sensor networks, meteorological satellites, and ground monitoring stations, forecast and warning data are obtained through network interfaces, risk thresholds are set in combination with historical data and expert experience, user object data are obtained through questionnaires and interface docking, missing data are supplemented by data interpolation algorithms, and abnormal fluctuation data are processed using sliding average filtering.

3. According to claim 1, a method for digital coupling of meteorological risk identification and intelligent engine is characterized in that: In the step S2 of constructing a multidimensional grid system, grid units are divided based on the geographic information system, the grid size is adaptively adjusted according to the changes in meteorological elements and user distribution, the meteorological element values ​​are calculated or matched to construct a real-time grid, meteorological changes are predicted to generate a forecast grid, corresponding risk levels form a risk grid, and user information is associated to construct a service grid.

4. According to claim 1, a method for digital coupling of meteorological risk identification and intelligent engine is characterized in that: In the S3 step of implementing the multi-scenario coupling technology, the meteorological risk monitoring and identification coupling uses a convolutional neural network to extract risk features from radar images and satellite cloud images and convert them into grid data, and the meteorological monitoring and forecast and warning model coupling uses integrated learning to fuse multiple forecast model results.

5. According to claim 1, a method for digital coupling of meteorological risk identification and intelligent engine is characterized in that: In the S3 step of implementing the multi-scenario coupling technology, when the meteorological digital data is coupled with the layer image, the graphics such as the transportation network and urban layout are gridded and superimposed with the meteorological risk level data. The meteorological basic data and image data are coupled to build a comprehensive model by combining terrain, vegetation data and remote sensing images.

6. According to claim 1, a method for digital coupling of meteorological risk identification and intelligent engine is characterized in that: In the S4 model calculation and loop optimization step, the model calculation adopts parallel computing technology and is accelerated by using graphics processors or cluster computing resources. Dynamic optimization uses intelligent algorithms such as genetic algorithms and particle swarm optimization algorithms to search and optimize model parameters.

7. The method for digital coupling of meteorological risk identification and intelligent engine according to claim 1 is characterized in that: In the S5 product generation and service push step, digital products adopt template and parameterized design, support user-defined configuration, and dynamically update user needs by analyzing user behavior and feedback.

8. According to claim 1, a digital coupling system for meteorological risk identification and intelligent engine is characterized in that: Includes the following modules: Data collection module: It is configured to collect real-time meteorological monitoring data from multiple data sources such as meteorological satellites, ground monitoring stations, sensor networks, etc., obtain meteorological forecast and warning data from meteorological departments, set meteorological risk threshold data based on historical data and expert experience, and collect user object data; Grid construction module: used to divide the target area into grid units based on geographic information system technology, build a live grid according to real-time meteorological monitoring data, generate a forecast grid according to meteorological forecast and warning data, form a risk grid according to meteorological risk threshold data, and establish a service grid in combination with user object data; Coupling processing module: capable of using the live grid, forecast grid, risk grid and service grid to realize the coupling of meteorological risk monitoring and meteorological risk identification, the coupling of meteorological digital data and layer images, the coupling of meteorological monitoring and forecast warning models, the coupling of meteorological basic data and image data, the coupling of meteorological element changes and image changes, and the coupling of different data assimilation; Model operation module: using digital and grid models, the numerical grid data is compared and analyzed and modeled according to the grid threshold; Loop optimization module: performs multivariate data coupling and multi-model operation based on the four grids, executes related operations cyclically, and optimizes the model according to the operation results and newly collected data; Product generation and service module: Generate digital products based on the calculation results after cycle optimization, and push digital products to users through various channels according to risk identification results and user needs.

9. A meteorological risk identification and intelligent engine digital coupling system according to claim 8, characterized in that: The data acquisition module also includes a data preprocessing unit, which is used to clean the collected real-time meteorological monitoring data, meteorological forecast and warning data, meteorological risk threshold data and user object data, remove abnormal values ​​and erroneous data, unify the data format, and interpolate the missing data.

10. A meteorological risk identification and intelligent engine digital coupling system according to claim 8, characterized in that: When coupling the changes in meteorological elements and images, the coupling processing module uses the changes in physical quantity data from radar and numerical weather forecasts to establish a physical quantity element change analysis model, and combines graphic image recognition technology to dynamically track and analyze the weather system.

Images