Methods and systems for dynamic monitoring of meteorological disasters using real-time meteorological data

Abstract

This invention provides a method and system for dynamic monitoring of meteorological disasters using real-time meteorological data. First, a set of real-time meteorological data for the target area is acquired. Then, spatiotemporal feature extraction and processing are used to generate temporal variation characteristics and spatial distribution characteristics for each meteorological data unit. Based on the correlation between these temporal variation characteristics and spatial distribution characteristics, regional correlation characteristics are generated. Next, the regional correlation characteristics are input into a pre-trained meteorological disaster prediction model for disaster risk assessment, generating disaster prediction results. Finally, based on the disaster prediction results, a dynamic monitoring strategy including a monitoring equipment deployment plan and early warning information push rules is generated and fed back to the meteorological monitoring platform to trigger monitoring resource allocation operations. This allows for full utilization of real-time meteorological data, dynamic monitoring of meteorological disasters, and improved accuracy and timeliness of disaster early warnings.

Description

Technical Field

[0001] This invention relates to the field of meteorological disaster monitoring technology, and more specifically, to a method and system for dynamic monitoring of meteorological disasters using real-time meteorological data. Background Technology

[0002] Meteorological disasters such as torrential rains, typhoons, and droughts have a significant impact on human production and lives. Accurate and timely monitoring and early warning of meteorological disasters are crucial for reducing disaster losses and protecting people's lives and property. Traditional meteorological disaster monitoring methods mainly rely on observation data from fixed meteorological stations. While this data is accurate, it suffers from limited spatial coverage and insufficient temporal resolution, making it difficult to comprehensively and dynamically reflect the evolution of meteorological disasters. Furthermore, traditional methods often focus on the analysis of single meteorological elements, lacking in-depth exploration of the spatiotemporal correlations of these elements, resulting in deficiencies in the accuracy and timeliness of disaster warnings. With the continuous development of meteorological observation technology, the acquisition of real-time meteorological data has become more convenient and abundant. However, how to efficiently process and analyze this data to extract valuable information for meteorological disaster monitoring remains a major challenge facing the meteorological field. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for dynamic monitoring of meteorological disasters applied to real-time meteorological data, the method comprising:

[0004] Acquire a real-time meteorological data set of a target area containing meteorological data units with meteorological status information at multiple different timestamps;

[0005] The real-time meteorological data set is subjected to spatiotemporal feature extraction processing to generate the temporal variation characteristics and spatial distribution characteristics of each meteorological data unit. The temporal variation characteristics reflect the evolution law of basic meteorological parameters at adjacent timestamps, and the spatial distribution characteristics reflect the distribution pattern of basic meteorological parameters at different locations at the same timestamp.

[0006] Based on the correlation between the temporal change features and the spatial distribution features, regional correlation features are generated. The regional correlation features include a description of the influence of the temporal evolution pattern on the spatial distribution pattern. Specifically, by extracting key temporal features from the temporal change features and key spatial regions from the spatial distribution features, a timestamp mapping is established, and the Pearson correlation coefficient is calculated to construct a correlation description vector. This vector is then fused with the key temporal features and key spatial regions to generate regional correlation features.

[0007] The pre-trained meteorological disaster prediction model is invoked, and the regional correlation features are input into the meteorological disaster prediction model for disaster risk assessment processing, generating disaster prediction results that indicate the potential occurrence probability and impact range of meteorological disasters within the target area;

[0008] Based on the disaster prediction results, a dynamic monitoring strategy is generated that includes a deployment plan for monitoring equipment and rules for pushing early warning information. The dynamic monitoring strategy is then fed back to the meteorological monitoring platform to trigger the allocation of monitoring resources.

[0009] In another aspect, embodiments of the present invention also provide a dynamic monitoring system for meteorological disasters applied to real-time meteorological data, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the machine-readable storage medium to implement the above-described method.

[0010] Based on the above, this invention, through comprehensive analysis of meteorological data units containing meteorological state information at multiple different timestamps in the target area, achieves comprehensive extraction and in-depth mining of the spatiotemporal characteristics of meteorological elements. It not only focuses on the temporal variation characteristics of meteorological elements but also deeply analyzes the spatial distribution characteristics of meteorological elements at different locations at the same timestamp, further revealing the influence of temporal evolution patterns on spatial distribution patterns. This constructs more accurate and comprehensive regional correlation features. Inputting these regional correlation features into a pre-trained meteorological disaster prediction model significantly improves the accuracy and timeliness of disaster risk assessment, generating more precise disaster prediction results. Based on these disaster prediction results, a monitoring strategy including monitoring equipment deployment plans and early warning information push rules can be dynamically generated, achieving optimized allocation of monitoring resources and thus more effectively responding to meteorological disasters and reducing disaster losses. Therefore, by integrating multiple links such as real-time meteorological data, spatiotemporal feature extraction, correlation feature generation, and disaster prediction models, the ability to monitor and warn of meteorological disasters is significantly improved. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the execution flow of the dynamic monitoring method for meteorological disasters applied to real-time meteorological data provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of exemplary hardware and software components of a meteorological disaster dynamic monitoring system applied to real-time meteorological data, provided in an embodiment of the present invention. Detailed Implementation

[0013] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating a method for dynamic monitoring of meteorological disasters based on real-time meteorological data, provided by an embodiment of the present invention. The following is a detailed description of this method for dynamic monitoring of meteorological disasters based on real-time meteorological data.

[0014] Step S110: Obtain a real-time meteorological data set of the target area containing meteorological data units with multiple different timestamps of meteorological status information.

[0015] In this embodiment, the real-time meteorological dataset consists of multiple meteorological data units, each containing meteorological status information at a specific timestamp. The meteorological data is collected using various types of monitoring equipment distributed across different locations within the target area.

[0016] Ground-based weather stations are a crucial source of data collection, equipped with various sensors that measure basic meteorological parameters in real time. For example, temperature sensors detect the ambient air heat and record corresponding values, humidity sensors detect the water vapor content in the air, barometric pressure sensors acquire atmospheric pressure data, and wind speed and direction sensors measure the speed and direction of airflow, respectively. These sensors can record data at different times, and each recorded meteorological data point constitutes a meteorological data unit.

[0017] Satellite remote sensing equipment monitors target areas from high altitudes, capturing information such as cloud distribution and atmospheric water vapor content. The satellite scans and collects data at set time intervals, with each collected data point forming a meteorological data unit.

[0018] Radar monitoring equipment focuses on monitoring the dynamic changes of meteorological phenomena such as precipitation and storms. It analyzes information such as the location, intensity, and movement speed of meteorological targets by transmitting and receiving electromagnetic waves. Radar equipment continuously monitors the surrounding area and records relevant data at different times, forming meteorological data units.

[0019] Because different monitoring devices may collect data at different times and frequencies, the collected data needs to be integrated. During integration, the accuracy and consistency of the data must be considered. For data collection times that are inconsistent, time synchronization is required to ensure that each meteorological data unit accurately corresponds to its respective timestamp. Simultaneously, data quality control is necessary to check for missing or outlier values. If missing values ​​are found, interpolation methods can be used to supplement them; if outliers are found, they need to be corrected or removed.

[0020] Each monitoring device will send the collected data to the data center through a wired or wireless communication network. To protect the security and privacy of the data, encryption technology is adopted during the data transmission process to prevent the data from being stolen or tampered with. The data center stores and manages the received meteorological data for subsequent analysis and processing.

[0021] Step S120: Extract the spatio-temporal characteristics of the real-time meteorological data set to generate the time variation characteristics and spatial distribution characteristics of each meteorological data unit. The time variation characteristics reflect the evolution law of the basic meteorological parameters at adjacent timestamps, and the spatial distribution characteristics reflect the distribution pattern of the basic meteorological parameters at different positions at the same timestamp.

[0022] Step S121: Align the real-time meteorological data set in the time dimension to obtain a meteorological data sequence after time alignment.

[0023] Since the data collection times of different monitoring devices may vary, in order to facilitate subsequent analysis and processing, it is necessary to align the real-time meteorological data set in the time dimension. First, determine a unified time reference and time interval. The time reference can be a specific starting time point, and the time interval is set according to actual needs. For example, a shorter time interval can be selected to capture meteorological changes more precisely.

[0024] After determining the time reference and time interval, each meteorological data unit in the real-time meteorological data set needs to be arranged in chronological order. For data with timestamps not within the unified time interval, corresponding processing is required. If the timestamp of a meteorological data unit is between two unified time intervals, an interpolation method can be used to estimate the meteorological parameter value at that time point. There are various interpolation methods, such as linear interpolation, which assumes that the change of meteorological parameters between two adjacent known timestamps is linear. Specifically, if the meteorological parameter values corresponding to timestamps t1 and t2 are A1 and A2 respectively, and we want to estimate the meteorological parameter value A corresponding to timestamp t (t1 < t < t2), we can calculate the value of A according to the proportional relationship based on the relative positions of t, t1, and t2.

[0025] Through the above time dimension alignment processing, a meteorological data sequence after time alignment can be obtained. Each timestamp in this meteorological data sequence corresponds to a complete set of meteorological parameter values.

[0026] Step S122: Perform time series analysis on the meteorological data sequence after time alignment, calculate the change rate parameter and change direction parameter of each basic meteorological parameter at consecutive timestamps, and construct a time evolution curve based on the change rate parameter and change direction parameter as the time variation characteristic.

[0027] Step S1221: Extract the basic meteorological parameter values ​​of adjacent timestamps from the time-aligned meteorological data sequence, and calculate the difference between adjacent timestamp parameter values ​​as the absolute change.

[0028] In this embodiment, after obtaining the time-aligned meteorological data sequence, in order to analyze the temporal evolution of each basic meteorological parameter, it is necessary to calculate the absolute change between adjacent timestamps. Taking the temperature parameter as an example, assuming that in the time-aligned meteorological data sequence, the temperature value corresponding to timestamp t1 is T1, and the temperature value corresponding to timestamp t2 (which immediately follows t1) is T2, then the absolute change between adjacent timestamps is the difference between T2 and T1. If T2 is greater than T1, the difference is positive, indicating that the temperature is rising; if T2 is less than T1, the difference is negative, indicating that the temperature is falling.

[0029] For other basic meteorological parameters, such as humidity, air pressure, and wind speed, the same method is used to calculate the absolute changes between adjacent time stamps. These absolute changes reflect the magnitude of change of each basic meteorological parameter between adjacent time stamps.

[0030] Step S1222: Calculate the rate of change parameter per unit time based on the absolute change and the time interval length.

[0031] After calculating the absolute change between adjacent timestamps, the length of the time interval also needs to be considered to obtain the rate of change per unit time. The time interval length is the time difference between two adjacent timestamps. For example, if the time interval between timestamps t1 and t2 is Δt, and the absolute temperature change corresponding to adjacent timestamps has already been calculated as ΔT, then the rate of change of temperature per unit time is ΔT divided by Δt.

[0032] For other basic meteorological parameters, the rate of change per unit time is calculated using the same method. The rate of change per unit time parameter can more accurately reflect how quickly the basic meteorological parameters change per unit time.

[0033] Step S1223: Determine the direction of change parameter by comparing the magnitude of adjacent timestamp parameter values. The direction of change parameter includes an upward trend indicator and a downward trend indicator.

[0034] Determining the direction of change parameters allows for further analysis of the changing trends of basic meteorological parameters. For each basic meteorological parameter, the parameter values ​​corresponding to adjacent time stamps are compared. If the parameter value at a later time stamp is greater than that at a previous time stamp, it can be determined that the basic meteorological parameter shows an upward trend within that time interval, indicated by an upward trend indicator; if the parameter value at a later time stamp is less than that at a previous time stamp, the basic meteorological parameter shows a downward trend, indicated by a downward trend indicator.

[0035] For example, for temperature parameters, if T2 is greater than T1, the direction of temperature change indicates an upward trend; if T2 is less than T1, the direction of temperature change indicates a downward trend. The same method applies to other basic meteorological parameters such as humidity, air pressure, and wind speed.

[0036] Step S1224: Perform sliding window smoothing on the rate of change parameter and direction of change parameter of continuous timestamps, arrange the smoothed rate of change parameter and direction of change parameter in the order of timestamps, and generate a time evolution curve that reflects the evolution law of basic meteorological parameters over time. The horizontal axis of the time evolution curve is the timestamp, the vertical axis is the rate of change parameter, and the slope sign of the curve is determined by the direction of change parameter.

[0037] To ensure that the rate of change and direction of change parameters better reflect the true evolution of fundamental meteorological parameters, a sliding window smoothing process is needed for these parameters across consecutive time stamps. The sliding window is a fixed-size time interval that slides across the time series. Within each window, the rate of change parameter is smoothed using methods such as averaging.

[0038] Specifically, suppose there is a sliding window of size n. For a timestamp ti, the time interval covered by the window is [ti-n / 2, ti+n / 2] (assuming n is even). The rate of change parameters within the sliding window are V1, V2, ..., Vn. Then the smoothed rate of change parameter Vi' is the average of these n rate of change parameters.

[0039] For the direction parameter, a majority vote can be used within the sliding window to determine the smoothed direction parameter. Specifically, the number of upward and downward trend indicators within the window is counted, and the indicator with the most counts is used as the smoothed direction parameter.

[0040] By arranging the smoothed rate of change and direction of change parameters in timestamp order, a time evolution curve can be generated. The horizontal axis of the time evolution curve represents the timestamp, and the vertical axis represents the rate of change parameter. The sign of the curve's slope is determined by the direction of change parameter; if the direction of change parameter indicates an upward trend, the curve slope is positive; if the direction of change parameter indicates a downward trend, the curve slope is negative. This time evolution curve can intuitively reflect the evolution of basic meteorological parameters over time.

[0041] Step S123: Perform spatial location gridding processing on the real-time meteorological data set, divide the target area into uniformly distributed spatial grid units, and extract the mean and variance statistics of the basic meteorological parameters at the same time stamp in each spatial grid unit.

[0042] Step S1231: Determine the number of horizontal and vertical divisions of the spatial grid cells based on the geographical boundary of the target area and the preset grid division accuracy.

[0043] In this embodiment, the first step is to determine the number of horizontal and vertical divisions of spatial grid cells based on the geographical boundary of the target area and the preset grid division precision. The geographical boundary of the target area can be obtained using tools such as Geographic Information Systems (GIS), which clearly defines the spatial extent of the target area.

[0044] The preset grid division precision determines the size of the grid cells. If the grid division precision is high, the grid cells will be smaller, which can more precisely reflect the changes in meteorological parameters within the target area; if the grid division precision is low, the grid cells will be larger, and the description of meteorological parameters will be relatively more macroscopic.

[0045] The number of horizontal divisions can be calculated based on the horizontal length of the target area and the preset horizontal grid cell size; similarly, the number of vertical divisions can be calculated based on the vertical length of the target area and the preset vertical grid cell size.

[0046] Step S1232: Based on the number of horizontal divisions and the number of vertical divisions, perform equidistant grid division processing on the target area to generate a set of spatial grid cells covering the target area.

[0047] After determining the number of horizontal and vertical divisions of the spatial grid cells, the target area can be divided into equidistant grids. Starting from one boundary of the target area, the target area is divided into multiple spatial grid cells of equal size according to the number of horizontal and vertical divisions. These spatial grid cells are closely arranged, collectively covering the entire target area, forming a set of spatial grid cells.

[0048] Each spatial grid cell has its unique location coordinates, which allow for accurate positioning of each grid cell within the target area.

[0049] Step S1233: Collect the basic meteorological parameter values ​​at the same time stamp within each spatial grid cell, and calculate the arithmetic mean of all basic meteorological parameter values ​​within that grid cell as the mean statistic.

[0050] For each spatial grid cell, at the same timestamp, the basic meteorological parameter values ​​recorded by all meteorological monitoring equipment within that grid cell are collected. For example, at a certain timestamp t, the temperature, humidity, air pressure, and other parameter values ​​recorded by each meteorological station within that grid cell are collected.

[0051] Then, calculate the arithmetic mean of all basic meteorological parameter values ​​within the grid cell. Taking temperature as an example, if there are m meteorological stations in the grid cell that record temperature values ​​T1, T2, ..., Tm, then the mean temperature statistic is the sum of these m temperature values ​​divided by m.

[0052] For other basic meteorological parameters, such as humidity and air pressure, the same method is used to calculate the mean statistics. These mean statistics reflect the average level of the basic meteorological parameters within the grid cell.

[0053] Step S1234: Calculate the average of the squared differences between the basic meteorological parameter values ​​and the mean statistic within the grid cell as the variance statistic, which reflects the dispersion of the basic meteorological parameters within the grid cell.

[0054] After calculating the mean statistic, it is also necessary to calculate the variance statistic to reflect the dispersion of the basic meteorological parameters within the grid cell. For each basic meteorological parameter, the difference between each parameter value within the grid cell and the mean statistic is calculated, and then these differences are squared.

[0055] Taking temperature as an example, assuming the temperature mean statistic is T_mean, and the temperature values ​​recorded by each weather station in this grid cell are T1, T2, ..., Tm, then the differences between each temperature value and the mean statistic are T1-T_mean, T2-T_mean, ..., Tm-T_mean, respectively. Squaring these differences yields (T1-T_mean)^2, (T2-T_mean)^2, ..., (Tm-T_mean)^2.

[0056] Next, the average of these squared differences is calculated, which is the variance statistic. The larger the variance statistic, the greater the dispersion of the basic meteorological parameters within the grid cell, and the more uneven the distribution of the meteorological parameters; the smaller the variance statistic, the smaller the dispersion of the basic meteorological parameters within the grid cell, and the more uniform the distribution of the meteorological parameters.

[0057] Step S1235: Arrange the mean statistic and variance statistic according to the position coordinates of the spatial grid cells to generate a statistic matrix with spatial position index.

[0058] Finally, the calculated mean and variance statistics are arranged according to the position coordinates of the spatial grid cells. Each spatial grid cell has its corresponding mean and variance statistics. These statistics are arranged according to the positional order of the grid cells to form a statistics matrix.

[0059] This statistical matrix has a spatial location index, allowing precise mapping between rows and columns to each spatial grid cell. The aforementioned statistical matrix clearly demonstrates the spatial distribution of the mean and dispersion of basic meteorological parameters across all spatial grid cells within the target area.

[0060] Step S124: Construct a spatial distribution heatmap based on the mean statistic and variance statistic as a spatial distribution feature. The color gradient of the spatial distribution heatmap corresponds to the distribution density of the basic meteorological parameters.

[0061] Using the previously obtained statistical matrix with spatial location indexes, a spatial distribution heatmap is constructed. A spatial distribution heatmap is an intuitive visualization tool that can show the spatial distribution patterns of basic meteorological parameters within a target area.

[0062] When constructing a spatial distribution heatmap, the color corresponding to each spatial grid cell is determined based on the mean statistic and the variance statistic. The mean statistic reflects the average level of the basic meteorological parameters within that grid cell, while the variance statistic reflects the dispersion of the basic meteorological parameters within that grid cell. By considering both statistics, a correspondence can be established between the density of the distribution of basic meteorological parameters and the color gradient.

[0063] For example, a color mapping rule can be set up so that when the mean statistic is in a certain range and the variance statistic is small, the corresponding color is darker, indicating that the basic meteorological parameters in the area are densely distributed and the values ​​are relatively concentrated; when the mean statistic is in another range and the variance statistic is large, the corresponding color is lighter, indicating that the basic meteorological parameters in the area are more dispersed.

[0064] Each spatial grid cell is filled with a corresponding color to form a spatial distribution heatmap. By observing the spatial distribution heatmap, one can intuitively understand the distribution pattern of basic meteorological parameters within the target area, identifying which areas have densely distributed parameters and which areas have sparsely distributed parameters.

[0065] Step S125: Standardize the time variation features and the spatial distribution features to obtain a spatiotemporal feature set.

[0066] To ensure comparability and consistency between temporal variation characteristics and spatial distribution characteristics, standardization is necessary. Standardization eliminates dimensional differences between different characteristics, placing them within the same scale range.

[0067] For the time-varying characteristics, namely the previously generated time evolution curve, normalization can be used for standardization. The purpose of normalization is to map the numerical range of the time evolution curve to a set interval, such as [0, 1]. Normalization can be achieved by calculating the ratio of the difference between each data point in the time evolution curve and the maximum and minimum values ​​of the curve.

[0068] For spatial distribution characteristics, i.e., spatial distribution heatmaps, the color values ​​in the heatmap can be standardized. First, the range of color values ​​can be determined, and then the color value corresponding to each grid cell can be linearly transformed to bring it into a uniform numerical range.

[0069] After standardization, the temporal variation characteristics and spatial distribution characteristics are integrated to obtain a spatiotemporal feature set. This spatiotemporal feature set contains information on the temporal evolution patterns and spatial distribution patterns of meteorological data in the target area.

[0070] Step S130: Generate regional correlation features based on the correlation between the temporal change features and the spatial distribution features. The regional correlation features include a description of the influence of the temporal evolution pattern on the spatial distribution pattern. Specifically, by extracting key temporal features from the temporal change features and key spatial regions from the spatial distribution features, a timestamp mapping is established, and the Pearson correlation coefficient is calculated to construct a correlation description vector. This vector is then fused with the key temporal features and key spatial regions to generate regional correlation features.

[0071] Step S131: Extract the time evolution curve from the time change features and the spatial distribution heatmap from the spatial distribution features.

[0072] When generating regional correlation features, the first step is to extract key information from the previously obtained temporal variation features and spatial distribution features. The temporal evolution curve is extracted from the temporal variation features. This curve reflects the evolution of basic meteorological parameters over time, including information such as the rate of change and direction of change per unit time.

[0073] Spatial distribution heatmaps are extracted from spatial distribution features. These heatmaps visually demonstrate the spatial distribution patterns of basic meteorological parameters within the target area, and the color gradient reflects the density of the distribution of these parameters.

[0074] Step S132: Perform feature sampling processing on the time evolution curve, extract the rate of change parameters and direction of change parameters of key time nodes as key time features, and perform region segmentation processing on the spatial distribution heatmap to identify the boundary coordinates of concentrated distribution areas and discrete distribution areas as key spatial areas.

[0075] Feature sampling of the time evolution curve aims to extract information from key time points. Key time points are those points in the time evolution curve where changes are significant. A rate of change threshold can be set; when the rate of change parameter at a certain time point in the time evolution curve exceeds this threshold, that time point is identified as a key time point.

[0076] For each key time point, the corresponding rate of change and direction of change parameters are extracted. These parameters constitute the key time features. Key time features can more prominently reflect the changes in basic meteorological parameters at key time points.

[0077] Regional segmentation is performed on the spatial distribution heatmap to identify the boundary coordinates between concentrated and discrete distribution areas. Image segmentation algorithms can be used to distinguish different regions based on the color gradient of the heatmap. Darker colored areas typically indicate dense distribution of basic meteorological parameters, representing concentrated distribution areas (e.g., areas with a color gradient greater than a first set value); lighter colored areas indicate dispersed distribution of basic meteorological parameters, representing discrete distribution areas (e.g., areas with a color gradient less than a second set value). The boundary coordinates of these regions are determined through image segmentation algorithms, and these boundary coordinates constitute the key spatial regions.

[0078] Step S133: Establish a mapping relationship between the timestamps of key time features and the collection timestamps of key spatial regions, so that the time reference of the time dimension and the time reference of the spatial dimension are consistent.

[0079] To accurately analyze the impact of temporal evolution patterns on spatial distribution patterns, it is necessary to establish a mapping relationship between the timestamps of key temporal features and the collection timestamps of key spatial regions. Since key temporal features are extracted from temporal evolution curves, their timestamps reflect the changes in basic meteorological parameters over time; key spatial regions are identified from spatial distribution heatmaps, and their collection timestamps record the collection time of spatial distribution information.

[0080] By establishing a mapping relationship, the timestamps of key temporal features are matched with the collection timestamps of key spatial regions, ensuring consistency between the temporal and spatial time bases. For example, if a key temporal feature has a timestamp of t1, the corresponding region in the key spatial regions whose collection timestamp is closest to t1 is found, and a mapping relationship is established between the two. This ensures that temporal and spatial information correspond in the temporal dimension when analyzing the impact of temporal evolution patterns on spatial distribution patterns, avoiding erroneous analyses caused by temporal inconsistencies.

[0081] Step S134: For each key time feature, the rate of change parameter and direction of change parameter are statistically analyzed for the mean statistical change of the key spatial region at the corresponding timestamp, and the Pearson correlation coefficient between the rate of change parameter and the mean statistical change is calculated. The absolute value of the Pearson correlation coefficient reflects the intensity of the influence, and the sign of the correlation coefficient reflects the direction of the influence.

[0082] For each key time feature, there are parameters for the rate of change and the direction of change. Based on the established mapping relationship between the timestamps of the key time features and the collection timestamps of the key spatial regions, the change in the mean statistic of the key spatial region at the corresponding timestamp is calculated. The change in the mean statistic refers to the difference between the mean statistic of the key spatial region at that timestamp and the mean statistic at the previous relevant timestamp.

[0083] Taking temperature parameters as an example, if the timestamp of the key time feature is t, the mean temperature statistic of the corresponding key spatial region at timestamp t is M1, and the mean temperature statistic at the previous relevant timestamp t-1 is M0, then the change in the mean temperature statistic ΔM = M1 - M0.

[0084] For each key temporal feature, the rate of change parameter and the corresponding change in the mean statistic of the key spatial region are used to calculate the Pearson correlation coefficient. The Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables. By calculating the Pearson correlation coefficient between the rate of change parameter and the change in the mean statistic, the impact of the rate of change of basic meteorological parameters in the temporal evolution pattern on the change of the mean statistic in the spatial distribution pattern can be analyzed.

[0085] The larger the absolute value of the correlation coefficient, the stronger the linear relationship between the rate of change parameter and the change in the mean statistic, indicating a greater influence of the time evolution pattern on the spatial distribution pattern. A positive correlation coefficient indicates a positive correlation between the rate of change parameter and the change in the mean statistic, meaning that as the rate of change increases, the mean statistic also tends to increase, reflecting a positive influence. A negative correlation coefficient indicates a negative correlation, meaning that as the rate of change increases, the mean statistic tends to decrease, reflecting a negative influence.

[0086] Step S135: Perform trend analysis on the Pearson correlation coefficients of key features over continuous time, identify stable and fluctuating influence relationships, extract the correlation coefficients of stable influence relationships as the basic influence intensity, extract the range of changes in the correlation coefficients of fluctuating influence relationships as the influence intensity fluctuation range, and comprehensively generate a description of the direction and intensity of the influence of the time evolution law on the spatial distribution pattern.

[0087] After obtaining the Pearson correlation coefficient for each key time feature, a trend analysis was performed on the correlation coefficients of key features over continuous time. By observing the changes in the correlation coefficients over time, stable and fluctuating influence relationships were identified.

[0088] A stable influence relationship refers to a situation where the correlation coefficient remains relatively stable over a period of time, with minimal fluctuations. In this case, the correlation coefficient within this stable period is extracted as the baseline influence strength. The baseline influence strength reflects the degree to which the temporal evolution pattern affects the relative stability of the spatial distribution pattern during this period.

[0089] Fluctuation-based influence relationships refer to situations where the correlation coefficient fluctuates significantly over a period of time. For fluctuation-based influence relationships, the range of variation in the correlation coefficient, i.e., the interval between its maximum and minimum values, is extracted as the influence intensity fluctuation range. The influence intensity fluctuation range reflects the range of changes in the intensity of the influence of temporal evolution patterns on spatial distribution patterns within this period.

[0090] By combining the basic influence intensity and the fluctuation range of the influence intensity, as well as the positive and negative signs of the correlation coefficient, a description of the direction and intensity of the influence of the time evolution law on the spatial distribution pattern can be generated. It details how the time evolution law affects the spatial distribution pattern, including whether the direction of the influence is positive or negative, and whether the influence intensity is stable or fluctuates within a set range.

[0091] Step S136: Construct a correlation description vector based on the influence direction and influence intensity, and fuse the correlation description vector with the key temporal features and key spatial regions to generate regional correlation features containing spatiotemporal influence relationships.

[0092] Based on the previously obtained description of the influence direction and intensity of the temporal evolution pattern on the spatial distribution pattern, a correlation description vector is constructed. This correlation description vector contains information about the direction and intensity of influence, which can be represented by multiple dimensions. For example, one dimension represents the direction of influence (e.g., 1 for positive, -1 for negative), while other dimensions represent information such as the basic influence intensity and the fluctuation range of influence intensity.

[0093] The correlation description vector is fused with key temporal features and key spatial regions. Key temporal features include the rate and direction of change of basic meteorological parameters at key time points, while key spatial regions include the boundary coordinates of the spatially concentrated and discrete distribution areas of basic meteorological parameters. The fusion process can be performed by concatenating the correlation description vector, key temporal features, and key spatial regions in a predetermined order to form regional correlation features that incorporate spatiotemporal influences. These regional correlation features integrate information on temporal evolution patterns and spatial distribution patterns, as well as their mutual influences.

[0094] Step S140: Call the pre-trained meteorological disaster prediction model, input the regional correlation features into the meteorological disaster prediction model for disaster risk assessment, and generate disaster prediction results indicating the potential occurrence probability and impact range of meteorological disasters in the target area.

[0095] Step S141: Standardize the regional correlation features, input the standardized regional correlation features into the input layer of the meteorological disaster prediction model, and perform feature recombination processing on the regional correlation features through a fully connected network to generate an intermediate feature vector with model adaptation dimensions.

[0096] To make the regional correlation features more suitable for the input requirements of the meteorological disaster prediction model, the regional correlation features are first standardized. The standardization process can be similar to the standardization of the temporal variation features and spatial distribution features, eliminating the dimensional differences between different dimensions of the regional correlation features and mapping their numerical range to a unified interval, such as [0, 1].

[0097] The standardized regional correlation features are input into the input layer of the meteorological disaster prediction model. After receiving the regional correlation features, the input layer of the meteorological disaster prediction model performs feature recombination processing through a fully connected network. Each neuron in the fully connected network is connected to all neurons in the input layer, and the regional correlation features are recombined and transformed through a series of linear transformations and nonlinear activation functions.

[0098] During feature recombination, the fully connected network adjusts the dimensions of regional correlation features to meet the requirements of subsequent processing within the model, generating an intermediate feature vector with model-adapted dimensions. This intermediate feature vector contains information about the regional correlation features after transformation and recombination, which is more beneficial for the model's subsequent assessment of meteorological disaster risks.

[0099] Step S142: The intermediate feature vector is subjected to nonlinear transformation through the hidden layer of the meteorological disaster prediction model to extract deep semantic features containing spatiotemporal influence relationships.

[0100] The intermediate feature vectors are fed into the hidden layers of the meteorological disaster prediction model. The hidden layers are the core of the model, containing multiple neurons and multi-layered structures. Within the hidden layers, the intermediate feature vectors undergo nonlinear transformation. This nonlinear transformation is achieved through activation functions, such as the ReLU function.

[0101] Through nonlinear transformations, hidden layers can uncover more complex and deeper information from intermediate feature vectors, extracting deep semantic features that contain spatiotemporal relationships. These deep semantic features are not merely simple transformations of regional correlation features, but rather capture the more subtle interrelationships between temporal evolution patterns and spatial distribution patterns, as well as their potential connection to the occurrence of meteorological disasters. The multi-layered structure of hidden layers can progressively extract higher-level features, enabling meteorological disaster prediction models to better understand the complex patterns in meteorological data.

[0102] Step S143: Call the probability prediction layer of the meteorological disaster prediction model to perform probability distribution calculation on the deep semantic features and generate potential occurrence probability values ​​for different types of meteorological disasters in the target area.

[0103] Deep semantic features are incorporated into the probabilistic prediction layer of the meteorological disaster prediction model. The main function of the probabilistic prediction layer is to calculate the potential occurrence probability of different meteorological disaster types within the target area based on the deep semantic features. This layer maps the deep semantic features to the probability space of different meteorological disaster types through a series of calculations and processing.

[0104] The probability prediction layer may use classification algorithms or probability distribution models for calculation. For example, it can determine the correlation between each feature and different meteorological disaster types based on the feature values ​​of deep semantic features, and then calculate the potential occurrence probability of each meteorological disaster type through statistical analysis and other methods. Finally, it outputs the potential occurrence probability values ​​of different meteorological disaster types, such as rainstorms, strong winds, and lightning, within the target area.

[0105] Step S144: The spatial range inference process of the deep semantic features is performed through the range prediction layer of the meteorological disaster prediction model to generate the boundary coordinates of the impact range corresponding to different meteorological disaster types.

[0106] Step S1441: Input the deep semantic features into the spatial mapping sub-layer of the range prediction layer, and perform spatial location feature enhancement processing on the deep semantic features through a convolutional neural network to generate a feature mapping map with spatial location information.

[0107] Deep semantic features are incorporated into the spatial mapping sublayer of the range prediction layer in the meteorological disaster prediction model. This spatial mapping sublayer employs a convolutional neural network to process the deep semantic features. Convolutional neural networks possess powerful feature extraction and spatial information capture capabilities.

[0108] In the spatial mapping sublayer, the convolutional neural network performs sliding convolution operations on deep semantic features through convolutional kernels to extract spatial location features from the deep semantic features. The convolutional kernels can learn local patterns of features at different spatial locations, and through multiple convolution and pooling operations, they enhance the spatial location information in the deep semantic features.

[0109] After processing by the convolutional neural network, a feature map with spatial location information is generated. Each element in the feature map corresponds to a specific spatial location in the target region and contains relevant feature information about that location.

[0110] Step S1442: Perform threshold segmentation on the feature map and extract regions with feature values ​​exceeding a preset threshold as candidate regions for disaster impact.

[0111] After obtaining the feature map with spatial location information, threshold segmentation is performed on it. The preset threshold is a feature value limit determined based on a large amount of historical data and experience. By comparing the value of each element in the feature map with the preset threshold, areas with feature values ​​exceeding the preset threshold are extracted as candidate areas for disaster impact.

[0112] The purpose of threshold segmentation is to filter out areas with high eigenvalues ​​that may be associated with the impact of meteorological disasters. In a feature map, areas with higher eigenvalues ​​generally indicate a greater likelihood of being affected by meteorological disasters. Through threshold segmentation, the approximate range that may be affected by meteorological disasters can be preliminarily determined.

[0113] Step S1443: Perform morphological processing on the candidate disaster impact regions to eliminate isolated small regions and connect adjacent regions to generate continuous disaster impact connected regions.

[0114] Morphological processing is applied to candidate regions of disaster impact, including operations such as dilation and erosion. Dilation can expand the scope of candidate regions and connect adjacent small regions; erosion can eliminate isolated small regions within the candidate regions.

[0115] By combining expansion followed by corrosion or other suitable morphological operations, isolated small areas within the candidate disaster impact region are eliminated, while adjacent regions are connected, creating a continuous, interconnected disaster impact region. This continuous, interconnected disaster impact region better reflects the actual impact range characteristics of meteorological disasters, avoiding interference from discontinuous regions caused by noise or local characteristic fluctuations.

[0116] Step S1444: Extract the coordinates of the outer contour boundary points of the disaster-affected connected area, perform polygon fitting processing on the boundary point coordinates, and generate a polygon boundary coordinate sequence that approximately represents the affected area.

[0117] After obtaining the continuous connected regions affected by the disaster, the coordinates of their outer contour boundary points are extracted. Edge detection algorithms can be used to find the outer contour of these regions. These algorithms identify the boundary points between the connected regions affected by the disaster and their surrounding areas by calculating the gradient changes of pixel values ​​in the feature map.

[0118] After obtaining the coordinates of the outer contour boundary points, polygon fitting is performed on these coordinates. The purpose of polygon fitting is to approximate the disaster impact range with a polygon. Methods such as least squares can be used to find a polygon whose boundary is as close as possible to the coordinates of the outer contour boundary points, ultimately generating a polygon boundary coordinate sequence that approximates the impact range. This polygon boundary coordinate sequence clearly defines the approximate boundary of the meteorological disaster impact range.

[0119] Step S1445: Calculate the area parameter and center coordinate parameter of the affected area based on the polygon boundary coordinate sequence, and use the polygon boundary coordinate sequence, area parameter and center coordinate parameter as descriptive information of the boundary coordinates of the affected area.

[0120] Based on the coordinate sequence of the polygon boundary, calculate the area parameter and center coordinate parameter of the meteorological disaster impact range. The area parameter can be obtained by calculating the area of ​​the region enclosed by the polygon. Alternatively, the total area can be calculated by dividing the polygon into multiple triangles, calculating the area of ​​each triangle, and summing the results.

[0121] The center coordinate parameter can be obtained by calculating the average of the coordinates of all vertices of the polygon, which represents the approximate center location of the area affected by the meteorological disaster.

[0122] The polygon boundary coordinate sequence, area parameter, and center coordinate parameter are combined to form descriptive information for the boundary coordinates of the affected area. This descriptive information details the affected area corresponding to different meteorological disaster types, including the boundaries, size, and center location of the affected area.

[0123] Step S145: Perform correlation matching processing between the potential occurrence probability value and the boundary coordinates of the impact range to generate a disaster prediction result containing the disaster type, potential occurrence probability and boundary coordinates of the impact range.

[0124] The potential occurrence probability values ​​of different meteorological disaster types obtained from the probability prediction layer are correlated and matched with the boundary coordinates of the impact range obtained from the range prediction layer. For each meteorological disaster type, its corresponding potential occurrence probability value and the boundary coordinates of the impact range are combined accordingly.

[0125] For example, for rainstorm disasters, the potential occurrence probability value is combined with the boundary coordinates of the rainstorm disaster's impact area (including the polygon boundary coordinate sequence, area parameters, and center coordinate parameters). This correlation and matching process is performed on all meteorological disaster types, ultimately generating a disaster prediction result that includes the disaster type, potential occurrence probability, and boundary coordinates of the impact area. This disaster prediction result comprehensively reflects the potential occurrence and possible impact range of different meteorological disaster types within the target area.

[0126] Step S150: Generate a dynamic monitoring strategy that includes a monitoring equipment deployment plan and early warning information push rules based on the disaster prediction results, and feed the dynamic monitoring strategy back to the meteorological monitoring platform to trigger the monitoring resource allocation operation.

[0127] Step S151: Analyze the potential occurrence probability value and the boundary coordinates of the impact range in the disaster prediction results, and extract the target disaster type and its target impact range corresponding to the potential occurrence probability being greater than the preset probability threshold.

[0128] After receiving the disaster prediction results, they are analyzed. The disaster prediction results include the potential occurrence probability values ​​and the coordinates of the affected area boundaries for different types of meteorological disasters. The preset probability threshold is a probability limit determined based on actual needs and historical experience.

[0129] By comparing the potential occurrence probability of each meteorological disaster type with a preset probability threshold, meteorological disaster types with a potential occurrence probability greater than the preset probability threshold are extracted as target disaster types. Simultaneously, the boundary coordinates of the impact range corresponding to these target disaster types are extracted as the target impact range. Target disaster types and target impact ranges are the key focus for subsequent development of monitoring equipment deployment plans and early warning information dissemination rules.

[0130] Step S152: Determine the target area to be monitored based on the boundary coordinates of the target's influence range, and calculate the number of monitoring devices to be deployed based on the area size and geographical distribution density of the target area.

[0131] Based on the boundary coordinates of the target's impact range, the scope of the target area to be monitored is determined. The target area is the region that the target disaster type may affect. The size and geographical distribution density of the target area are analyzed. The size reflects the scale of the target area, while the geographical distribution density reflects the complexity of the meteorological conditions within the target area.

[0132] For target areas with large areas and high geographical density, more monitoring equipment may be needed to comprehensively monitor meteorological conditions; for target areas with smaller areas and lower geographical density, the number of monitoring devices required is relatively smaller. The number of monitoring devices to be deployed can be calculated based on pre-established empirical models or algorithms, combined with the size and geographical density of the target area. This calculation process takes into account the actual conditions of the target area to ensure that the monitoring equipment can effectively cover the target area.

[0133] Step S153: Combining the location distribution information of existing monitoring equipment in the target area, determine the deployment location of the new monitoring equipment through an optimization algorithm, so that the coverage of the new monitoring equipment overlaps with the target influence range to the maximum extent.

[0134] For example, step S1531: Obtain the location distribution information of existing monitoring devices within the target area, wherein the location distribution information includes the geographical coordinates of each existing monitoring device.

[0135] To determine the deployment location of new monitoring equipment, the first step is to obtain the location distribution information of existing monitoring equipment within the target area. This information is typically stored in the database of the meteorological monitoring platform. The location distribution information includes the geographic coordinates of each existing monitoring device, which allows for accurate determination of the specific location of the existing monitoring equipment within the target area.

[0136] Step S1532: Divide the target's influence range into regional sensitivity areas to generate a multi-level sensitive area set that includes core sensitive areas and extended sensitive areas.

[0137] The target's impact area is divided into regional sensitivity zones. Based on the characteristics of meteorological disasters and historical data, the degree of impact of meteorological disasters on different areas within the target's impact area is analyzed. The target's impact area is divided into core sensitive areas and extended sensitive areas, forming a multi-level sensitive area set.

[0138] The core sensitive area refers to the region within the target's influence range that is most severely affected by meteorological disasters and most sensitive to weather changes. The extended sensitive area encompasses the surrounding areas of the core sensitive area; these areas are relatively less affected by meteorological disasters but still require monitoring. This multi-level classification of sensitive areas facilitates more targeted deployment of monitoring equipment, prioritizing the monitoring needs of the core sensitive area.

[0139] Step S1533: Calculate the overlap between the coverage area of ​​each existing monitoring device and the set of multi-level sensitive areas, and generate the existing coverage-sensitive overlapping area.

[0140] Based on the location and monitoring range of existing monitoring equipment, the coverage area of ​​each existing monitoring device is calculated. Then, the coverage area of ​​each existing monitoring device is compared with a multi-level sensitive area set to calculate the overlap between them.

[0141] For example, for an existing monitoring device, calculate the overlap area between its coverage area and the core sensitive area, and the overlap area with the extended sensitive area. Summarize the overlap between the coverage areas of all existing monitoring devices and the multi-level sensitive area set to generate the existing coverage-sensitive overlapping area. The existing coverage-sensitive overlapping area reflects the sensitive areas currently covered by existing monitoring devices within the target's influence range.

[0142] Step S1534: Determine the uncovered sensitive regions in the multi-level sensitive region set that are not covered by the existing covered sensitive overlapping regions, and generate an initial set of candidate locations within the uncovered sensitive regions.

[0143] By comparing the multi-level sensitive area set with the existing coverage sensitive overlapping areas, the portion of the multi-level sensitive area set not covered by the existing coverage sensitive overlapping areas is identified as the uncovered sensitive area. The uncovered sensitive area is the area that needs to be covered by new monitoring equipment.

[0144] Within the uncovered sensitive areas, an initial set of candidate locations is generated according to predefined rules. For example, the uncovered sensitive areas can be divided into grids, and the nodes of the grids can be used as candidate locations to form the initial set of candidate locations. This initial set of candidate locations contains locations that may be suitable for deploying new monitoring equipment.

[0145] Step S1535: For each candidate location in the initial set of candidate locations, calculate the overlap ratio between its coverage area and the core sensitive area in the uncovered sensitive area, as well as the overlap ratio with the extended sensitive area. At the same time, calculate the overlap ratio between its coverage area and the coverage areas of all existing monitoring devices.

[0146] For each candidate location in the initial set of candidate locations, its coverage area is calculated based on its location and the monitoring range of the monitoring equipment. Then, the overlap ratio of this coverage area with the core sensitive area in the uncovered sensitive area, the overlap ratio with the extended sensitive area, and the overlap ratio with the coverage area of ​​all existing monitoring equipment are calculated respectively.

[0147] The overlap ratio of the core sensitive area reflects the extent to which the monitoring equipment at the candidate location can cover the core sensitive area; the overlap ratio of the extended sensitive area reflects the extent to which the extended sensitive area is covered; and the overlap ratio with the coverage area of ​​existing monitoring equipment reflects the degree of overlap between the coverage areas of the monitoring equipment at the candidate location and those of existing monitoring equipment. By calculating these ratios for each candidate location in the initial set of candidate locations, the performance of each candidate location in covering sensitive areas and avoiding overlap with existing equipment can be quantified.

[0148] Step S1536: Apply an optimization algorithm to iteratively adjust the candidate locations, with the goal of maximizing the overlap ratio of the core sensitive area, minimizing the overlap ratio of the secondary extended sensitive area, and minimizing the overlap ratio with the existing coverage. Select the location with the highest comprehensive overlap ratio after optimization as the deployment location of the new monitoring equipment.

[0149] After obtaining the overlap ratios for each candidate location, an optimization algorithm is applied to iteratively adjust the candidate locations. The goal of the optimization algorithm is to find an optimal set of deployment locations for new monitoring equipment that maximizes the overlap ratio of the core sensitive area, minimizes the overlap ratio of the extended sensitive area (i.e., covers as much of the extended sensitive area as possible), and minimizes the overlap ratio with the existing coverage.

[0150] During the iteration process, candidate locations are continuously adjusted, and a comprehensive evaluation index is calculated based on the overlap ratios of the current candidate locations. The comprehensive evaluation index can be a weighted combination of the overlap ratio of core sensitive areas, the overlap ratio of extended sensitive areas, and the overlap ratio with existing coverage. For example, a higher weight can be assigned to the overlap ratio of core sensitive areas because ensuring monitoring of core sensitive areas is more important; a moderate weight can be assigned to the overlap ratio of extended sensitive areas; and a negative weight can be assigned to the overlap ratio with existing coverage to encourage reducing overlap with existing equipment.

[0151] After each iteration, candidate locations with better performance are selected based on comprehensive evaluation metrics for further adjustments until preset stopping conditions are met, such as reaching the maximum number of iterations or the comprehensive evaluation metrics no longer showing significant improvement. Finally, the location with the highest optimized comprehensive overlap ratio is selected as the deployment location for new monitoring equipment.

[0152] Step S1537: Based on the newly added deployment locations and the existing monitoring equipment location distribution information, generate a monitoring equipment deployment plan that includes the geographical coordinates of all monitoring equipment.

[0153] After determining the deployment locations of the new monitoring equipment, the geographical coordinates of the new deployment locations are integrated with the location distribution information of existing monitoring equipment. The geographical coordinates of all monitoring equipment (including existing and new equipment) are organized and recorded according to a defined format to generate a monitoring equipment deployment plan containing the geographical coordinates of all monitoring equipment. This deployment plan clearly defines the specific locations of all monitoring equipment within the target area.

[0154] Step S154: Divide the warning level according to the magnitude of the potential occurrence probability value. The warning level includes a regular warning level and an emergency warning level.

[0155] Warning levels are classified based on the potential probability of occurrence in disaster prediction results. A probability threshold is set as the boundary between regular and emergency warning levels. When the potential probability of occurrence of a certain meteorological disaster is less than the threshold, it is classified as a regular warning level; when the potential probability is greater than or equal to the threshold, it is classified as an emergency warning level.

[0156] A standard warning level indicates that there is a certain possibility of a meteorological disaster occurring, but the risk is relatively low; an emergency warning level indicates that the possibility of a meteorological disaster is high, requiring more urgent and effective response measures. By classifying warning levels, corresponding rules for disseminating warning information can be formulated based on different levels of risk.

[0157] Step S155: Set corresponding push rules for different warning levels. The push rules include a set of push objects, a push time interval, and a push information content template.

[0158] Set corresponding push rules for different warning levels. For regular warning levels, the target audience can include general weather-related groups, such as ordinary residents and agricultural workers. The push interval can be relatively long, such as once a day or once every few days, to avoid frequent information interruptions. The push message content template can be concise and clear, mainly including the type of meteorological disaster, the potential probability of occurrence, and basic prevention suggestions.

[0159] For emergency warning levels, the target audience needs to be expanded to include not only general weather-related groups but also relevant emergency management departments and rescue teams. The notification interval should be shortened, for example, once per hour or less, to ensure timely delivery of the latest meteorological disaster information. The notification content template should be more detailed and urgent, including not only the type of meteorological disaster and its potential probability of occurrence but also specific response measures and emergency contact information.

[0160] Step S156: Integrate the number of monitoring devices, their deployment locations, the early warning levels, and the push rules to generate a dynamic monitoring strategy that includes device deployment parameters and early warning rule parameters.

[0161] The previously determined number of monitoring devices, deployment locations, early warning levels, and push rules are integrated and processed. The number of monitoring devices and deployment locations are used as device deployment parameters, and the early warning levels and push rules are used as early warning rule parameters. These early warning rule parameters are combined according to the set structure and format to generate a dynamic monitoring strategy that includes both device deployment parameters and early warning rule parameters.

[0162] The dynamic monitoring strategy comprehensively considers the potential risks and countermeasures of meteorological disasters. It includes both the rational deployment of monitoring equipment to ensure effective monitoring of meteorological conditions in the target area and push notification rules based on different warning levels to promptly convey meteorological disaster information to relevant personnel. This dynamic monitoring strategy is fed back to the meteorological monitoring platform, which can then allocate monitoring resources accordingly, such as arranging the installation and commissioning of new monitoring equipment and promptly pushing out warning information according to the push notification rules, thereby achieving dynamic monitoring and effective response to meteorological disasters.

[0163] Furthermore, the meteorological disaster prediction model is trained through the following steps.

[0164] Step S211: Obtain the regional association feature samples corresponding to each historical meteorological data unit and the disaster real label data of each historical meteorological data unit. The disaster real label data includes the actual meteorological disaster type, occurrence probability and boundary coordinates of the affected area.

[0165] To train a meteorological disaster prediction model, historical meteorological data is first required. This involves collecting meteorological data for the target area over a past period. This data consists of multiple historical meteorological data units, each containing meteorological status information at a specific timestamp.

[0166] These historical meteorological data units undergo similar processing to the real-time meteorological data, including spatiotemporal feature extraction and regional correlation feature generation, resulting in regional correlation feature samples corresponding to each historical meteorological data unit. These regional correlation feature samples contain information about the interrelationships between the temporal evolution patterns and spatial distribution patterns within the historical meteorological data.

[0167] Simultaneously, real disaster label data corresponding to each historical meteorological data unit is collected. By reviewing historical meteorological records and disaster reports, the actual meteorological disaster type, probability of occurrence, and boundary coordinates of the affected area corresponding to each historical meteorological data unit are determined. Real disaster label data serves as supervisory information for model training, guiding the meteorological disaster prediction model to learn the relationship between meteorological data and meteorological disasters.

[0168] Step S212: Combine the regional associated feature samples with the corresponding real disaster label data to form training sample pairs, and construct a model training dataset containing multiple training sample pairs.

[0169] Each region-related feature sample is paired with its corresponding real disaster label data to form a training sample pair. Each training sample pair contains a region-related feature sample and its corresponding real disaster label data, reflecting the correspondence between features in historical meteorological data and actual meteorological disasters.

[0170] Multiple training sample pairs are integrated to construct the model training dataset. The model training dataset is the foundation of model training, containing rich historical meteorological data and corresponding disaster information, providing sufficient learning materials for the model, enabling the meteorological disaster prediction model to learn the complex patterns and laws between meteorological data and meteorological disasters.

[0171] Step S213: Initialize the network parameters of the meteorological disaster prediction model. The network parameters include the weight matrix of the input layer, the nonlinear transformation parameters of the hidden layer, and the probability calculation bias of the prediction layer. The prediction layer includes a probability prediction layer and a range prediction layer.

[0172] Before training the meteorological disaster prediction model, the network parameters of the model need to be initialized. The weight matrix of the input layer is used to perform a linear transformation on the input regional correlation feature samples, mapping them to an intermediate feature vector. The initial values ​​of the elements of the weight matrix can be initialized randomly, such as by using a Gaussian distribution to randomly generate weight values.

[0173] The nonlinear transformation parameters of the hidden layers include the parameters of the activation function. The activation function is used to introduce nonlinear factors and enhance the expressive power of the model. The initial values ​​of the nonlinear transformation parameters can also be randomly initialized.

[0174] The probability calculation bias of the prediction layer is used to adjust the calculation results of the probability prediction layer and the range prediction layer. The initial value of the probability calculation bias can be set to zero or a small random value. By initializing these network parameters, an initial state is provided for the model training, enabling the meteorological disaster prediction model to gradually learn and optimize these parameters during subsequent training.

[0175] Step S214: Input the regional correlation feature samples from the model training dataset into the initialized meteorological disaster prediction model, and perform dimensional adaptation processing on the regional correlation feature samples through the input layer to generate intermediate feature vector samples.

[0176] The regional correlation feature samples from the model training dataset are sequentially input into the input layer of the initialized meteorological disaster prediction model. After receiving the regional correlation feature samples, the input layer performs a linear transformation on them using the weight matrix. The linear transformation process involves performing matrix multiplication between the regional correlation feature samples and the weight matrix to obtain intermediate feature vector samples with model adaptation dimensions.

[0177] Step S215: Perform nonlinear transformation processing on the intermediate feature vector samples through the hidden layer to extract deep semantic feature samples containing spatiotemporal influence relationships.

[0178] The intermediate feature vector samples are fed into the hidden layer of the meteorological disaster prediction model. In the hidden layer, an activation function performs a nonlinear transformation on the intermediate feature vector samples. This activation function nonlinearly maps the elements of the intermediate feature vector samples, enabling the model to learn more complex and deeper features.

[0179] The multi-layered structure of hidden layers allows for the progressive extraction of higher-level features. Each layer of neurons processes and transforms the input features, capturing the spatiotemporal relationships contained in intermediate feature vector samples, and extracting deep semantic feature samples that contain these spatiotemporal relationships. These deep semantic feature samples contain more subtle and complex patterns in meteorological data, which is of great significance for accurately predicting meteorological disasters.

[0180] Step S216: Call the prediction layer to perform probability distribution calculation on the deep semantic feature samples to generate a predicted value of the potential occurrence probability of the disaster type. At the same time, perform spatial range inference on the deep semantic feature samples to generate a predicted value of the coordinates of the boundary of the influence range.

[0181] Deep semantic feature samples enter the prediction layer, which includes a probability prediction layer and a range prediction layer. The probability prediction layer performs probability distribution calculations on the deep semantic feature samples. Through calculation and mapping, the deep semantic feature samples are transformed into predicted potential occurrence probabilities for different types of meteorological disasters. The probability prediction layer can use common classification algorithms or probability distribution models from related technologies to implement this calculation process.

[0182] The range prediction layer performs spatial range inference processing on deep semantic feature samples. Similar to the processing of the range prediction layer in the previous real-time application, it generates coordinate prediction values ​​of the influence range boundaries through steps such as spatial mapping sublayer, threshold segmentation, morphological processing, and polygon fitting. These prediction values ​​represent the range boundaries that the model predicts may be affected by different types of meteorological disasters.

[0183] Step S217: Calculate the difference between the predicted potential occurrence probability of the disaster type and the actual occurrence probability of the meteorological disaster type in the corresponding disaster real label data, and generate a probability prediction loss value.

[0184] The predicted potential occurrence probabilities of disaster types generated by the probability prediction layer are compared with the actual occurrence probabilities of meteorological disaster types in the disaster real-label data. A loss function is used to calculate the difference between the two; commonly used loss functions include the cross-entropy loss function.

[0185] The result calculated using the loss function is the probabilistic prediction loss value. The probabilistic prediction loss value reflects the degree of error of the model in predicting the probability of meteorological disasters. The smaller the loss value, the closer the model's prediction results are to the actual situation.

[0186] Step S218: Calculate the spatial location difference between the predicted coordinates of the boundary of the affected area and the corresponding actual disaster label data of the affected area boundary, and generate the predicted loss value of the affected area.

[0187] The spatial location difference between the predicted coordinates of the impact range boundary generated by the scope prediction layer and the actual impact range boundary coordinates in the disaster's true label data is calculated. For example, distance metrics such as Euclidean distance and Hausdorff distance can be used to calculate the difference between the predicted coordinates and the actual boundary coordinates.

[0188] By calculating the differences in the boundary coordinates of the impact ranges for all meteorological disaster types, and then summarizing and averaging the results, the range prediction loss value is obtained. The range prediction loss value reflects the degree of error of the meteorological disaster prediction model in predicting the impact range of meteorological disasters. The smaller the loss value, the closer the model's predicted range is to the actual impact range.

[0189] Step S219: After standardizing the probability prediction loss value and the range prediction loss value, perform a weighted summation to generate the total model loss value.

[0190] To comprehensively account for the errors of probabilistic prediction and range prediction, the probabilistic prediction loss value and the range prediction loss value are standardized. The purpose of standardization is to eliminate the dimensional differences between the two loss values ​​and bring them to the same scale.

[0191] The standardized probability prediction loss and range prediction loss are weighted and summed. Different weights are assigned to the probability prediction loss and range prediction loss, with the weights determined based on actual needs and the model's focus. For example, if the accuracy of predicting the probability of meteorological disasters is more important, a higher weight can be assigned to the probability prediction loss; if the accuracy of predicting the impact range of meteorological disasters is more important, a higher weight can be assigned to the range prediction loss. The weighted probability prediction loss and range prediction loss are then summed to obtain the total model loss. The total model loss comprehensively reflects the overall error level of the model in both probability and range prediction.

[0192] Step S220: Based on the total loss value of the model, the gradient descent optimization algorithm is used to adjust the weight matrix of the input layer, the nonlinear transformation parameters of the hidden layer, and the prediction bias of the prediction layer, and the above steps are repeated until the total loss value of the model meets the preset convergence condition.

[0193] Based on the total loss value of the model, the network parameters of the model are adjusted using the gradient descent optimization algorithm. The core idea of ​​the gradient descent optimization algorithm is to update the network parameters along the negative gradient direction of the loss function to gradually reduce the total loss value of the model.

[0194] The gradient of the model's total loss value is calculated relative to the input layer's weight matrix, the hidden layer's nonlinear transformation parameters, and the prediction bias of the prediction layer. The gradient represents the rate of change of the loss function with respect to the current parameter values; updating the parameters along the negative gradient direction reduces the value of the loss function.

[0195] Based on the calculated gradient, the network parameters are updated according to the set learning rate. The learning rate controls the step size of each parameter update; a learning rate that is too large may cause the model to fail to converge, while a learning rate that is too small may cause the convergence speed to be too slow.

[0196] Repeat steps S214-S219, that is, continuously input regional correlation feature samples into the meteorological disaster prediction model for feature extraction, prediction, loss calculation, and parameter adjustment, until the total loss value of the model meets the preset convergence condition. The preset convergence condition may be that the total loss value of the model is less than a certain threshold or that the loss value no longer decreases significantly within a set number of iterations.

[0197] Step S221: Use a model validation dataset containing historical meteorological data units that were not used in training to validate the performance of the trained meteorological disaster prediction model and determine whether the prediction accuracy of the meteorological disaster prediction model reaches a preset threshold.

[0198] After training, the performance of the meteorological disaster prediction model needs to be validated. Prepare a model validation dataset containing historical meteorological data units not used in the training. Input the regional correlation feature samples from the model validation dataset into the trained meteorological disaster prediction model, and perform predictions following the same steps as the training process to obtain the predicted potential occurrence probability of meteorological disaster types and the predicted coordinates of the affected area boundaries.

[0199] The predicted results are compared with the corresponding real-label disaster data in the model validation dataset to calculate prediction accuracy metrics, such as precision, recall, and F1 score. These prediction accuracy metrics comprehensively reflect the accuracy of the meteorological disaster prediction model in predicting the probability of meteorological disasters and their impact range.

[0200] The calculated prediction accuracy index is compared with a preset threshold to determine whether the model's prediction accuracy meets the requirements. If the prediction accuracy reaches the preset threshold, it indicates that the model has good performance and can be used for actual meteorological disaster prediction; if it does not reach the preset threshold, further adjustments and optimizations to the model are needed.

[0201] Step S222: If the prediction accuracy reaches a preset threshold, the trained meteorological disaster prediction model is used as the pre-trained meteorological disaster prediction model.

[0202] When the model's prediction accuracy reaches a preset threshold, the trained meteorological disaster prediction model is designated as the pre-trained meteorological disaster prediction model. This model has been trained and validated using a large amount of historical data, demonstrating good performance and reliability, and can be used for dynamic monitoring of meteorological disasters based on real-time meteorological data. In subsequent meteorological disaster monitoring, the regional correlation features generated from real-time meteorological data are input into this meteorological disaster prediction model to obtain accurate disaster prediction results.

[0203] Figure 2 The illustration shows exemplary hardware and software components of a meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data and to perform the functions in this application.

[0204] The meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data can be a general-purpose server or a special-purpose server; both can be used to implement the meteorological disaster dynamic monitoring method applied to real-time meteorological data of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0205] For example, a meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data also includes an I / O interface 150 between the computer and other input / output devices.

[0206] For ease of explanation, only one processor is described in the meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data. However, it should be noted that the meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data in this application may also include multiple processors. Therefore, the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the meteorological disaster dynamic monitoring system 100 applied to real-time meteorological data performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0207] Furthermore, this embodiment of the invention also provides a readable storage medium, which has computer-executable instructions pre-set in it. When the processor executes the computer-executable instructions, the above-mentioned method for dynamic monitoring of meteorological disasters applied to real-time meteorological data is implemented.

[0208] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A meteorological disaster dynamic monitoring method applied to real-time meteorological data, characterized in that, The method comprises: acquiring a real-time weather data set of a target area containing multiple weather data units of different time-stamped weather state information; performing spatio-temporal feature extraction processing on the real-time weather data set to generate time variation features and spatial distribution features of each weather data unit, the time variation features reflecting the evolution law of adjacent time-stamped basic weather parameters, and the spatial distribution features reflecting the distribution pattern of different location basic weather parameters at the same time stamp; generating a regional correlation feature according to the correlation of the time variation features and the spatial distribution features, the regional correlation feature containing a description of the influence relationship of the time evolution law on the spatial distribution pattern, wherein, by extracting time key features in the time variation features and spatial key regions in the spatial distribution features, a Pearson correlation coefficient is calculated after time stamp mapping to construct a correlation relationship description vector, and the correlation relationship description vector is fused with the time key features and the spatial key regions to generate the regional correlation feature; calling a pre-trained weather disaster prediction model, inputting the regional correlation feature into the weather disaster prediction model for disaster risk assessment processing to generate a disaster prediction result indicating the potential occurrence probability and influence range of a weather disaster in the target area; generating a dynamic monitoring strategy containing a monitoring device deployment scheme and a warning information push rule according to the disaster prediction result, and feeding back the dynamic monitoring strategy to a weather monitoring platform to trigger a monitoring resource allocation operation; the regional correlation feature is generated according to the correlation of the time variation features and the spatial distribution features, comprising: extracting the time evolution curve in the time variation features and the spatial distribution heat map in the spatial distribution features; performing feature sampling processing on the time evolution curve to extract the change rate parameter and the change direction parameter of the key time node as the time key features, and performing region segmentation processing on the spatial distribution heat map to identify the boundary coordinates of the concentrated distribution region and the discrete distribution region as the spatial key region; establishing a mapping relationship between the time stamp of the time key feature and the collection time stamp of the spatial key region, so that the time reference of the time dimension and the space dimension is consistent; for each time key feature, the mean statistical quantity change of the spatial key region corresponding to the change rate parameter and the change direction parameter at the corresponding time stamp is calculated, and the Pearson correlation coefficient of the change rate parameter and the mean statistical quantity change is calculated, the absolute value of the Pearson correlation coefficient reflecting the influence intensity, and the sign of the correlation coefficient reflecting the influence direction; performing trend analysis on the Pearson correlation coefficients of the continuous time key features to identify stable influence relationships and fluctuating influence relationships, extracting the correlation coefficients of the stable influence relationships as the basic influence intensity, and extracting the correlation coefficient change range of the fluctuating influence relationships as the influence intensity fluctuation interval, and comprehensively generating the influence direction and influence intensity description of the time evolution law on the spatial distribution pattern; based on the influence direction and the influence intensity, an association relationship description vector is constructed, the association relationship description vector is fused with the time key features and the spatial key regions, and a regional correlation feature containing a spatio-temporal influence relationship is generated. 2.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 1, characterized in that, The spatio-temporal feature extraction processing is performed on the real-time meteorological data set to generate time variation features and spatial distribution features of each meteorological data unit, including: Performing time dimension alignment processing on the real-time meteorological data set to obtain a time-aligned meteorological data sequence; Performing time series analysis processing on the time-aligned meteorological data sequence to calculate a change rate parameter and a change direction parameter of each basic meteorological parameter at consecutive time stamps, and constructing a time evolution curve as a time variation feature based on the change rate parameter and the change direction parameter; Performing spatial position gridding processing on the real-time meteorological data set to divide the target region into uniformly distributed spatial grid units, and extracting mean value statistics and variance statistics of the basic meteorological parameter at the same time stamp in each spatial grid unit; Constructing a spatial distribution heat map as a spatial distribution feature based on the mean value statistics and the variance statistics, and the color gradient of the spatial distribution heat map corresponds to the distribution density of the basic meteorological parameter; Performing standardization processing on the time variation features and the spatial distribution features to obtain a spatio-temporal feature set. 3.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 2, characterized in that, The time series analysis processing on the time-aligned meteorological data sequence to calculate a change rate parameter and a change direction parameter of each basic meteorological parameter at consecutive time stamps, and constructing a time evolution curve as a time variation feature based on the change rate parameter and the change direction parameter, includes: Extracting the basic meteorological parameter values of adjacent time stamps in the time-aligned meteorological data sequence, and calculating the difference value of the adjacent time stamp parameter values as an absolute change amount; Calculating the change rate parameter in unit time according to the absolute change amount and the time interval length; Determining the change direction parameter by comparing the size relationship of the adjacent time stamp parameter values, and the change direction parameter includes an upward trend identifier and a downward trend identifier; Performing sliding window smoothing processing on the change rate parameter and the change direction parameter of consecutive time stamps, arranging the smoothed change rate parameter and the change direction parameter in time stamp order to generate a time evolution curve reflecting the time evolution law of the basic meteorological parameter, and the horizontal coordinate of the time evolution curve is the time stamp, the vertical coordinate is the change rate parameter, and the slope sign of the curve is determined by the change direction parameter. 4.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 2, characterized in that, The spatial position gridding processing on the real-time meteorological data set to divide the target region into uniformly distributed spatial grid units, and extracting mean value statistics and variance statistics of the basic meteorological parameter at the same time stamp in each spatial grid unit, includes: Determining the horizontal division number and the vertical division number of the spatial grid unit according to the geographical boundary range of the target region and the preset grid division precision; Performing equidistant grid division processing on the target region based on the horizontal division number and the vertical division number to generate a spatial grid unit set covering the target region; Collecting the basic meteorological parameter values at the same time stamp in each spatial grid unit, and calculating the arithmetic mean of all basic meteorological parameter values in the grid unit as the mean value statistics; computing the average value of the square of the difference between the basic meteorological parameter value in the grid cell and the mean value statistics as the variance statistics, which reflects the dispersion degree of the basic meteorological parameter in the grid cell; arranging the mean value statistics and the variance statistics according to the position coordinates of the spatial grid cell to generate a statistics matrix with spatial position index. 5.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 1, characterized in that, the calling of the pre-trained meteorological disaster prediction model, the input of the regional correlation feature into the meteorological disaster prediction model for disaster risk assessment processing, and the generation of a disaster prediction result indicating the potential occurrence probability and the influence range of meteorological disasters in the target region, including: standardizing the regional correlation feature, inputting the standardized regional correlation feature into the input layer of the meteorological disaster prediction model, and performing feature reorganization processing on the regional correlation feature through a fully connected network to generate an intermediate feature vector with model adaptation dimensions; performing nonlinear transformation processing on the intermediate feature vector through the hidden layer of the meteorological disaster prediction model to extract deep semantic features containing spatio-temporal influence relationships; calling the probability prediction layer of the meteorological disaster prediction model to perform probability distribution calculation processing on the deep semantic features to generate potential occurrence probability values of different meteorological disaster types in the target region; performing spatial range inference processing on the deep semantic features through the range prediction layer of the meteorological disaster prediction model to generate influence range boundary coordinates corresponding to different meteorological disaster types; associating and matching the potential occurrence probability values and the influence range boundary coordinates to generate a disaster prediction result containing disaster types, potential occurrence probabilities, and influence range boundary coordinates. 6.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 5, characterized in that, the calling of the pre-trained meteorological disaster prediction model, the input of the regional correlation feature into the meteorological disaster prediction model for disaster risk assessment processing, and the generation of a disaster prediction result indicating the potential occurrence probability and the influence range of meteorological disasters in the target region, including: inputting the deep semantic features into the spatial mapping sub-layer of the range prediction layer, performing spatial position feature enhancement processing on the deep semantic features through a convolutional neural network to generate a feature mapping graph with spatial position information; performing threshold segmentation processing on the feature mapping graph to extract regions with feature values exceeding a preset threshold as disaster impact candidate regions; performing morphological processing on the disaster impact candidate regions to eliminate isolated small regions and connect adjacent regions to generate continuous disaster impact connected regions; extracting the outer contour boundary point coordinates of the disaster impact connected regions, performing polygon fitting processing on the boundary point coordinates to generate a polygon boundary coordinate sequence approximately representing the influence range; computing the area parameter and the center coordinate parameter of the influence range according to the polygon boundary coordinate sequence, and taking the polygon boundary coordinate sequence, the area parameter, and the center coordinate parameter as the description information of the influence range boundary coordinates. 7.The method for dynamically monitoring meteorological disasters using real-time meteorological data according to claim 1, wherein, the calling of the pre-trained meteorological disaster prediction model, the input of the regional correlation feature into the meteorological disaster prediction model for disaster risk assessment processing, and the generation of a disaster prediction result indicating the potential occurrence probability and the influence range of meteorological disasters in the target region, including: analyzing the potential occurrence probability values and the influence range boundary coordinates in the disaster prediction result, extracting target disaster types and their target influence ranges corresponding to potential occurrence probabilities greater than a preset probability threshold; Determine a target area range to be monitored according to boundary coordinates of the target influence range, and calculate a number of monitoring devices to be deployed based on an area size and a geographical distribution density of the target area range; Determine a deployment position of the added monitoring device through an optimization algorithm based on position distribution information of existing monitoring devices in the target area, so that a coverage range of the added monitoring device is maximally overlapped with the target influence range; Divide early warning levels according to sizes of the potential occurrence probability values, wherein the early warning levels include a regular early warning level and an emergency early warning level; Set corresponding push rules for different early warning levels, wherein the push rules include a push object set, a push time interval, and a push information content template; Integrate and process the number of monitoring devices, the deployment position, the early warning levels, and the push rules to generate a dynamic monitoring strategy including device deployment parameters and early warning rule parameters. 8.The method for dynamically monitoring meteorological disasters applied to real-time meteorological data according to claim 1, characterized in that, The weather disaster prediction model is trained through the following steps: Obtain regional association feature samples corresponding to each historical weather data unit and disaster real label data of each historical weather data unit, wherein the disaster real label data includes a type of actually occurred weather disaster, an occurrence probability, and boundary coordinates of an influence range; Combine the regional association feature samples and the corresponding disaster real label data to form a training sample pair, and construct a model training data set including a plurality of training sample pairs; Initialize network parameters of the weather disaster prediction model, wherein the network parameters include a weight matrix of an input layer, nonlinear transformation parameters of a hidden layer, and a probability calculation bias of a prediction layer, and the prediction layer includes a probability prediction layer and a range prediction layer; Input the regional association feature samples in the model training data set into the initialized weather disaster prediction model, perform dimension adaptation processing on the regional association feature samples through the input layer, and generate intermediate feature vector samples; Perform nonlinear transformation processing on the intermediate feature vector samples through the hidden layer to extract deep semantic feature samples including spatiotemporal influence relationships; Call the prediction layer to perform probability distribution calculation processing on the deep semantic feature samples to generate a potential occurrence probability prediction value of the disaster type, and simultaneously perform spatial range inference processing on the deep semantic feature samples to generate a coordinate prediction value of the influence range boundary; Determine a target area range to be monitored according to boundary coordinates of the target influence range, and calculate a number of monitoring devices to be deployed based on an area size and a geographical distribution density of the target area range; Determine a deployment position of the added monitoring device through an optimization algorithm based on position distribution information of existing monitoring devices in the target area, so that a coverage range of the added monitoring device is maximally overlapped with the target influence range; Divide early warning levels according to sizes of the potential occurrence probability values, wherein the early warning levels include a regular early warning level and an emergency early warning level; Set corresponding push rules for different early warning levels, wherein the push rules include a push object set, a push time interval, and a push information content template; Integrate and process the number of monitoring devices, the deployment position, the early warning levels, and the push rules to generate a dynamic monitoring strategy including device deployment parameters and early warning rule parameters. The trained meteorological disaster prediction model is verified for performance using a model verification data set comprising a historical meteorological data unit that does not participate in training, to determine whether the prediction accuracy of the meteorological disaster prediction model reaches a preset threshold value; If the prediction accuracy reaches the preset threshold value, the trained meteorological disaster prediction model is used as the pre-trained meteorological disaster prediction model.

9. A meteorological disaster dynamic monitoring system applied to real-time meteorological data, characterized in that, The application relates to a meteorological disaster dynamic monitoring method applied to real-time meteorological data, comprising a processor and a memory, wherein the memory is connected to the processor, the memory is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the memory to realize the meteorological disaster dynamic monitoring method according to any one of claims 1-8.

Images