Rainstorm disaster early warning method and system based on rainstorm disaster bearing coefficient

Abstract

The invention relates to the technical field of disaster forecasting and early warning, and discloses a rainstorm disaster early warning method and system based on a rainstorm disaster bearing coefficient. Comprising the following steps: acquiring multi-source data and basic geographic information of an early warning area, extracting spatial-temporal characteristics based on the multi-source data and the basic geographic information, and calculating grid point forecast rainfall within a predetermined time according to the spatial-temporal characteristics; obtaining historical rainstorm disaster data of the early warning area, and calculating a grid point rainstorm disaster bearing coefficient based on the historical rainstorm disaster data and the basic geographic information; establishing an empirical model according to historical rainstorm disaster data in combination with basic geographic information to calculate a grid point rainfall threshold value; and according to the grid point forecast rainfall, the grid point rainstorm disaster bearing coefficient, the grid point rainfall threshold value and the field measurement rainfall, a rainstorm disaster prediction result of the early warning area is obtained in combination with the algorithm model, and the problem of weak monitoring capability of existing disaster forecast and early warning is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of weather prediction, and particularly to a heavy rain disaster warning method based on a heavy rain disaster-bearing coefficient. Background Art

[0002] Rainfall is a very normal meteorological condition. During the rainfall flood season, due to its strong extremity and high short-term intensity, it is easy to trigger heavy rain floods, and heavy rain floods are often accompanied by geological disasters such as mountain torrents, landslides, and debris flows, causing huge economic losses.

[0003] The existing disaster warning technologies mainly achieve protection warnings through meteorological forecasts. The current meteorological forecasts mainly predict common weather conditions. In the face of extreme weather events, since the meteorological forecasts are only based on meteorological data for forecasting, they cannot accurately target the areas where disasters may occur for disaster warnings, with weak monitoring and forecasting capabilities and low refinement. It is impossible to effectively make an objective evaluation of whether a flood disaster occurs in a region and cannot effectively warn of flood disasters. Summary of the Invention

[0004] The present invention provides a heavy rain disaster warning method and system based on a heavy rain disaster-bearing coefficient to solve the problem of weak monitoring capabilities of existing disaster forecasts and warnings.

[0005] To achieve the above object, the present invention is realized through the following technical solutions: In a first aspect, the present invention provides a heavy rain disaster warning method based on a heavy rain disaster-bearing coefficient, including the following steps: Step 1: Obtain multi-source data and basic geographic information of the warning area, extract spatio-temporal features based on the multi-source data and basic geographic information, and calculate the grid-point predicted rainfall within a predetermined time according to the spatio-temporal features; Step 2: Obtain historical heavy rain disaster data of the warning area, calculate the grid-point heavy rain disaster-bearing coefficient based on the historical heavy rain disaster data and basic geographic information, and establish an empirical model according to the historical heavy rain disaster data combined with the basic geographic information to calculate the grid-point rainfall threshold; Step 3: Obtain the heavy rain disaster prediction result of the warning area according to the grid-point predicted rainfall, grid-point heavy rain disaster-bearing coefficient, grid-point rainfall threshold, and on-site measured rainfall in combination with an algorithm model.

[0006] Further, the multi-source data includes corrected data and measurement data; The measurement data includes satellite data, ground observation data, radar data, and sounding data.

[0007] Further, the extraction of spatio-temporal features based on multi-source data and basic geographic information includes: extracting the spatial sequence, time sequence, and image data of rainfall based on multi-source data combined with spatial sequence features, encoding the spatial sequence, time sequence, and image data, and gridifying the basic geographic information to obtain grid geographic information; Multi-source vector data is obtained through encoding based on the grid geographic information, the encoded spatial sequence, time sequence, image data, and forecast correction data, and spatio-temporal features are extracted based on the multi-source vector data.

[0008] Further, the extraction of spatio-temporal features based on multi-source vector data includes: obtaining corresponding tensor data through algorithm encoding based on multi-source vector data, and then combining with a temporal attention mechanism to optimize the tensor data to obtain spatio-temporal features.

[0009] Further, multi-modal feature extraction is adopted for the multi-source vector data, and the multi-source vector data includes a vector matrix, a physical constraint matrix, and an image feature matrix.

[0010] Further, calculating the grid forecast rainfall amount within a predetermined time according to the spatio-temporal features includes: performing feature decoding through a deep temporal neural network based on the spatio-temporal features to calculate the grid forecast rainfall amount within a predetermined time.

[0011] Further, in step 2, calculating the grid rainstorm disaster-bearing coefficient based on historical rainstorm disaster data and basic geographic information includes: analyzing historical rainstorm disaster data, obtaining rainstorm disaster factors through machine learning, gridifying the basic geographic information to obtain grid geographic information, fusing the rainstorm disaster factors and grid geographic information, and calculating the grid rainstorm disaster-bearing coefficient in combination with the certainty coefficient method.

[0012] Further, establishing an empirical model based on historical rainstorm disaster data combined with basic geographic information to calculate the grid rainfall threshold includes: obtaining rainfall data of rainstorm meteorological disasters according to the historical disaster data, gridifying the basic geographic information to obtain grid geographic information, establishing an empirical model according to the rainfall data combined with grid geographic information, and calculating the grid rainfall threshold.

[0013] Further, in step 3, the algorithm model calculates the disaster-causing probability of the grid points in the early warning area according to the grid forecast rainfall amount, the grid rainstorm disaster-bearing coefficient, the grid rainfall threshold, and the on-site measured rainfall amount, and issues an early warning signal.

[0014] In a second aspect, the present invention provides a rainstorm disaster early warning system based on a rainstorm disaster-bearing coefficient, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0015] Beneficial effects: A rainstorm disaster warning method and system based on the rainstorm disaster-bearing coefficient provided by the present invention calculates the grid forecast rainfall based on basic geographic information and multi-source data, combines the grid disaster-bearing coefficient calculated from historical rainstorm disaster data, the grid rainfall threshold, and the rainfall measured on-site, and uses an empirical model from multiple data perspectives to conduct disaster warning, improving the refinement and accuracy of disaster warning. Brief description of the drawings

[0016] Figure 1 It is a flowchart of a rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient of the present invention; Figure 2 It is a schematic structural diagram of the AutoEncoder network used in the present invention. Specific embodiments

[0017] The technical solutions of the present invention will be described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the art to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, terms such as "a" or "one" do not indicate a quantity limitation, but indicate the existence of at least one. The "connection" or "connected" and similar terms are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0019] Please refer to Figure 1 , the embodiments of the present invention provide a rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient, including the following steps: Step 1: Obtain multi-source data and basic geographic information of the warning area, extract spatio-temporal features based on the multi-source data and basic geographic information, and calculate the grid forecast rainfall within a predetermined time according to the spatio-temporal features; Among them, the multi-source data includes corrected data and measurement data; The measurement data includes satellite data, ground observation data, radar data, and sounding data; Among the corrected data of multi-source data, it is mainly through the grid forecast correction products, which mainly include: precipitation, phase state, air temperature, UV wind, relative humidity, thunderstorm, short-term heavy precipitation, hail, and thunderstorm gale. It is mainly through the precipitation meteorological element products per hour in 24 hours output from the high-resolution rapid cycling assimilation numerical forecast system, the actual situation of multi-source observations, and multi-source data at the kilometer level such as radar and satellite. In this embodiment, the 14-channel radar data and geographical information at four positions A, B, C, and D in the early warning area at four time nodes 1, 2, 3, and 4 are mainly collected. For details, please refer to Table 1 and Table 2.

[0020] Table 1: Collected radar data

[0021] Table 2: Collected geographical information

[0022] The measurement data conducts resource integration based on the meteorological geographical information data element file, including: vector format meteorological map base map, topographic data at different scales, 30-meter surface cover data, township boundary data, global topographic data, global 90-meter DEM data, comprehensive thematic data, etc. By mining the corresponding relationship between meteorological data and geographical location, please refer to Figure 2 , use the AutoEncoder network to encode the absolute geographical location, take the specific geographical location as an important feature factor; obtain an efficient encoding and semantic extraction model, map the high-precision topographic information into a high-dimensional feature vector, construct a high-dimensional topographic feature vector dataset, and form a reusable feature vector dataset.

[0023] Specifically, based on multi-source data, combined with spatial sequence feature extraction, the spatial sequence, time sequence, and image data of rainfall are obtained, the spatial sequence, time sequence, and image data are encoded, the basic geographical information is gridded, and the gridded geographical information is obtained; Based on the gridded geographical information, the encoded spatial sequence, time sequence, and image data, the forecast correction data is encoded to obtain multi-source vector data, and based on the multi-source vector data, the corresponding tensor data is obtained through algorithm encoding, and then combined with the time series attention mechanism, the tensor data is optimized to obtain spatio-temporal features.

[0024] The multi-source vector data adopts a multi-modal feature extraction method, which includes a vector matrix, an image feature matrix, and a physical constraint representation matrix. The vector matrix is obtained through encoding of measurement data, the image feature matrix mainly comes from encoding of weather radar mosaic data, and the physical constraint representation matrix is formed by encoding the physical constraints between surface runoff, reservoir capacity, water level height, precipitation, and river runoff into a probability mask matrix between 0 and 1.

[0025] Spatio-temporal feature extraction is to simultaneously obtain the feature information of spatial sequence data in the time dimension and the spatial dimension. This information refers to the tensor data that describes spatio-temporal features obtained by encoding each group of multi-source fusion data through an algorithm. The hybrid encoder is divided into a spatial transformer block and a temporal transformer block. The spatial transformer block focuses on calculating the self-attention between each data, aiming to learn the relationship between each data at each time node and promote the association between geographical information and meteorological data. The temporal transformer block calculates the self-attention between each time point, focusing on learning the correlation information in time of the fusion data, making the obtained data more accurate and effectively used for disaster warning.

[0026] Based on the spatio-temporal features, feature decoding is performed through a deep time series neural network to calculate the grid point predicted rainfall within a predetermined time.

[0027] A feature extractor is constructed using hybrid spatio-temporal coding, paying attention to the internal connections of various data at each unit time node and also the connections of data in time. Based on this spatio-temporal feature extractor, the prediction accuracy of rainfall is improved.

[0028] Spatial correlation learning of hybrid spatio-temporal coding: First, use a linear embedding layer to project various data at each time node onto a high-dimensional feature. Then, feed the spatial representation of this node into the spatial self-attention mechanism to simulate the correlation between all data and output a high-dimensional representation.

[0029] Temporal correlation learning of hybrid spatio-temporal coding: In order to inject effective motion trajectories into the learned representation, this embodiment considers the temporal correspondence of the data to model the correlation of the same joint on a long time series. Separate different categories of data in the time dimension, so that the evolution of each node data is a separate token, and different category data of the fusion data are modeled in parallel. From the perspective of the time dimension, the different evolution situations of various data of the fusion data are modeled separately to better represent the temporal correlation. In addition, regarding each category of data of the fusion data as a separate token can reduce the dimension of the model from N×dim to dim, and can also process longer sequences in the model.

[0030] The attention calculation of the query Q, key K, and value matrix V in each head of the transformer block in the hybrid spatio-temporal coding MixSTE is represented by the following formula; ; where , represents the attention mechanism, represents the normalization exponential function, N represents the types of fusion data, is the dimension of each data, and T represents matrix transpose. The multi-head attention is defined as follows: ; ; where the linear projection weight is , MSA represents the multi-head self-attention mechanism, Concat represents concatenation, and h represents the number of independent groups. In the Transformer encoder of this embodiment, each joint token from a low dimension joint projection. The joint token p embeds position information through the matrix : ; The above performs matrix dimension transformation according to different task requirements of modeling spatial correlation and temporal correlation. represents layer normalization, represents the linear embedding layer.

[0031] The feature decoding structure uses the deep temporal model LSTNet, adding a traditional autoregressive linear model on the basis of the non-linear neural network part, making the non-linear deep learning model more robust to time series with scale violation changes.

[0032] This structure consists of five major modules: a convolutional component, a recurrent component, a recurrent-skip layer, a temporal attention layer, and an autoregressive component.

[0033] Convolutional component: This component is a convolutional network without pooling, which aims to extract short-term patterns in the time dimension and local dependencies between variables.

[0034] Recurrent component: The recurrent component is a gated recurrent unit (GRU), using the RELU function as the update activation function for the hidden state. The purpose is to extract the temporal features of the features with time arrangement order extracted by the convolutional layer through the GRU.

[0035] Recurrent-skip layer: Traditional GRUs are difficult to capture long-term patterns, so a skip connection layer can be used. That is, by means of interval sampling, longer time can be looked back while the sampling sequence length remains unchanged, so as to capture long-term features.

[0036] Temporal attention layer: The recurrent-skip layer requires a predefined hyperparameter, which is disadvantageous in time series with non-periodic or dynamically changing period lengths over time. To alleviate this problem, this embodiment adopts another method, namely the attention mechanism. It learns the weighted combination of the hidden representations at each window position of the input matrix.

[0037] Autoregressive Component: Due to the non-linear nature of convolutional and recurrent components, a major drawback of neural network models is that the scale of the output is insensitive to the scale of the input. In a specific real dataset, the scale of the input signal changes in an aperiodic manner, greatly reducing the prediction accuracy of neural network models. The final prediction of LSTNet can be decomposed into a linear part and a non-linear part, where the linear part mainly focuses on local scale issues, and the non-linear part focuses on recurrent patterns. In the LSTNet architecture, a classic autoregressive (AR) model is adopted as the linear component.

[0038] Essentially, it is a kind of linear layer. Finally, the output of the neural network part is added to the AR component to obtain the final prediction of LSTNet.

[0039] In this embodiment, based on the above, the grid-point predicted rainfall at 30 minutes, 60 minutes, 90 minutes, and 120 minutes is calculated according to spatio-temporal characteristics. For details, please refer to Table 3; Table 3: Grid-point predicted rainfall within a predetermined time

[0040] Step 2: Obtain the historical heavy rain disaster data of the warning area, calculate the grid-point heavy rain disaster bearing capacity coefficient based on the historical heavy rain disaster data and basic geographic information, and establish an empirical model according to the historical heavy rain disaster data combined with basic geographic information to calculate the grid-point rainfall threshold; Specifically, obtain the rainfall data of the heavy rain meteorological disaster according to the historical heavy rain disaster data, gridify the basic geographic information to obtain the grid-point geographic information, establish an empirical model according to the rainfall data combined with the grid-point geographic information, and calculate the grid-point rainfall threshold. For details, please refer to Table 4.

[0041] Table 4: Calculated grid-point rainfall threshold

[0042] Meteorological disasters such as flood disasters, landslides, and debris flows have a strong internal correlation with precipitation. Conducting the forecast and early warning of large-scale heavy rain meteorological disasters through regional rainfall characteristics is an important way to prevent rainfall-induced meteorological disasters. Among them, the most crucial issue is to determine the critical rainfall threshold that induces rainfall-induced meteorological disasters. The research on the critical rainfall threshold of rainfall-induced meteorological disasters uses the historical statistical method to establish an empirical model.

[0043] Classify and confirm the disaster critical rainfall according to the terrain and landform conditions. Therefore, the gridified geographic information is incorporated into the model calculation. Based on the precipitation conditions in all short-term historical heavy rain meteorological disaster data, analyze the correlation between disasters and precipitation, and assign the disaster critical rainfall threshold of the grid-point geographic information according to the precipitation and correlation at the disaster occurrence location.

[0044] Calculating the grid rainstorm disaster-bearing coefficient based on historical rainstorm disaster data and basic geographic information includes: analyzing historical rainstorm disaster data, obtaining rainstorm disaster factors through machine learning, gridifying the basic geographic information to obtain grid geographic information, fusing the rainstorm disaster factors and grid geographic information, and calculating the grid rainstorm disaster-bearing coefficient by combining the certainty factor method.

[0045] Specifically, by analyzing historical rainstorm disaster data, the rainstorm disaster factors are obtained through the Apriori association analysis algorithm of machine learning; The certainty factor method (CF) is a probability function, generally used to judge the contribution rate of each index when a certain event occurs. The calculation formula is as follows: ; In the formula: represents the conditional probability of a certain disaster occurring in the influencing factor a, usually the ratio of the number of disasters occurring in a certain grid in the influencing factor a to the area of the grid. represents the probability of a certain disaster occurring in the region, usually the quotient of the number of disasters occurring in the entire region divided by the area of the entire region. The value range is in [-1, 1]. A positive value indicates a positive correlation between the factor and the disaster, and a negative value indicates the opposite.

[0046] After determining the values of each factor, the values of all factors are stacked step by step to calculate the total contribution value .

[0047] Note: Stacking step by step means adding the value of each factor to the total contribution in turn, reflecting the joint influence between factors.

[0048] ; Among them, represents the total contribution, represents the number of influencing factors, represents the CF value of different factors; Then, the independent contribution of each factor is quantified through the "exclusion method". First, for each factor, calculate the total contribution value .

[0049] ; Then the relative contribution value of this factor is: ; Finally, the relative contribution value is normalized to the weight ; Among them, represents the weights after normalization, represents the weights of different factors, represents the CF values of different factors.

[0050] Multiply the CF value of each factor by its weight and sum them up to obtain the regional disaster occurrence coefficient H: ; The range of this coefficient is [−1, 1], and the larger the value, the higher the probability of disaster occurrence.

[0051] Step 3: Obtain the prediction results of rainstorm disasters in the warning area according to the grid forecast rainfall, grid rainstorm disaster-bearing coefficient, grid rainfall threshold, and on-site measured rainfall combined with the algorithm model.

[0052] Specifically, please refer to Table 5 to obtain the prediction results of rainstorm disasters at four locations A, B, C, and D in the warning area in combination with the algorithm model; Table 5: Prediction Results of Rainstorm Disasters

[0053] The algorithm model calculates the disaster-causing probability of the grid points in the warning area according to the grid forecast rainfall, grid rainstorm disaster-bearing coefficient, grid rainfall threshold, and on-site measured rainfall, and issues a warning signal.

[0054] In this embodiment, the model calculates by combining the on-site measured rainfall of a certain grid point with the rolling output value of the precipitation forecast model to judge the probability that the predicted total precipitation in the grid point reaches the grid rainfall threshold.

[0055] ; ; Among them, the indicator function is defined as: ; In the above formula, is the precipitation disaster occurrence probability, represents the sum of the measured precipitation at the measurement point and the predicted precipitation in each time period, T represents the grid rainfall threshold. is the on-site measurement point of the on-site measured rainfall, is the grid rainstorm disaster-bearing coefficient, is the grid forecast rainfall starting from the current hour by hour, and M is the total number of on-site observation points in the grid.

[0056] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should fall within the protection scope determined by the claims.

Claims

1. A rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient, characterized in that, The method includes the following steps: Step 1: Obtain multi-source data and basic geographic information of the early warning area, extract spatio-temporal features based on the multi-source data and basic geographic information, and calculate the grid forecast rainfall within a predetermined time according to the spatio-temporal features; Step 2: Obtain historical rainstorm disaster data of the early warning area, calculate the grid rainstorm disaster-bearing coefficient based on the historical rainstorm disaster data and basic geographic information, and establish an empirical model according to the historical rainstorm disaster data combined with the basic geographic information to calculate the grid rainfall threshold; Step 3: Obtain the rainstorm disaster prediction result of the early warning area according to the grid forecast rainfall, grid rainstorm disaster-bearing coefficient, grid rainfall threshold and on-site measured rainfall by combining with an algorithm model.

2. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to claim 1, characterized in that, The multi-source data includes corrected data and measured data; The measured data includes satellite data, ground observation data, radar data and sounding data.

3. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to claim 2, wherein The extraction of spatio-temporal features based on the multi-source data and basic geographic information includes: extracting the spatial sequence, time sequence and image data of rainfall based on the multi-source data combined with the spatial sequence features, encoding the spatial sequence, time sequence and image data, and gridifying the basic geographic information to obtain grid geographic information; The multi-source vector data is obtained by encoding based on the grid geographic information and the encoded spatial sequence, time sequence, image data and forecast correction data, and spatio-temporal features are extracted based on the multi-source vector data.

4. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to claim 3, wherein, The extraction of spatio-temporal features based on the multi-source vector data includes: obtaining the corresponding tensor data by algorithm encoding based on the multi-source vector data, and then combining with the temporal attention mechanism to optimize the tensor data to obtain spatio-temporal features.

5. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to claim 4, characterized in that, The multi-source vector data adopts multi-modal feature extraction, and the multi-source vector data includes a vector matrix, a physical constraint matrix and an image feature matrix.

6. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to any one of claims 1-5, characterized in that, The calculation of the grid forecast rainfall within a predetermined time according to the spatio-temporal features includes: performing feature decoding through a deep temporal neural network based on the spatio-temporal features to calculate the grid forecast rainfall within a predetermined time.

7. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to any one of claims 1-5, characterized in that, In step 2, the calculation of the grid rainstorm disaster-bearing coefficient based on the historical rainstorm disaster data and basic geographic information includes: analyzing the historical rainstorm disaster data, obtaining the rainstorm disaster factors through machine learning, gridifying the basic geographic information to obtain grid geographic information, fusing the rainstorm disaster factors and grid geographic information, and calculating the grid rainstorm disaster-bearing coefficient by combining the certainty coefficient method.

8. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to any one of claims 1-5, characterized in that, The establishment of an empirical model according to the historical rainstorm disaster data combined with the basic geographic information to calculate the grid rainfall threshold includes: obtaining the rainfall data of the rainstorm meteorological disaster according to the historical rainstorm disaster data, gridifying the basic geographic information to obtain grid geographic information, establishing an empirical model according to the rainfall data combined with the grid geographic information, and calculating the grid rainfall threshold.

9. The rainstorm disaster warning method based on the rainstorm disaster-bearing coefficient according to any one of claims 1-5, characterized in that In step 3, the algorithm model calculates the disaster-causing probability of the grid in the early warning area according to the grid forecast rainfall, grid rainstorm disaster-bearing coefficient, grid rainfall threshold and on-site measured rainfall, and issues a warning signal.

10. A rainstorm disaster warning system based on a rainstorm disaster-bearing coefficient, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 9 above.

Images