A typhoon real-time monitoring and early warning system

Abstract

The present invention discloses a typhoon real-time monitoring and early warning system, which belongs to the technical field of disaster monitoring. The system includes: a collection system, a monitoring system and an early warning system; the collection system is used to collect real-time typhoon data, and pre-process the real-time data to obtain processed data; the monitoring system is connected to the collection system, and is used to analyze the processed data to obtain analysis results; the early warning system is connected to the monitoring system, and is used to issue typhoon risk warnings based on the analysis results. The present invention can enhance the impact of historical data on output and retain complete historical information by establishing a direct connection between the input layer and the output layer in the prediction model, that is, adding a direct connection between the input at each moment and the final output.

Description

Technical Field

[0001] The invention belongs to the technical field of disaster monitoring, and in particular relates to a typhoon real-time monitoring and early warning system. Background Art

[0002] Typhoon is one of the most severe meteorological disasters in nature. Its strong wind and huge rainfall can cause serious casualties and property losses. Therefore, real-time monitoring and prediction of typhoons can effectively help people take appropriate protective measures in time, thereby reducing casualties and property losses.

[0003] Common methods of monitoring typhoons include radiosondes, using aircraft to carry related instruments, weather radars, etc. Radiosondes are instruments carried on borehole balloons to directly measure parameters such as atmospheric pressure, temperature and relative humidity. They can scan high altitudes layer by layer from low altitudes to obtain data from each layer. However, sounding balloons are usually one-time and have a low observation frequency. The way to obtain typhoon-related data through aircraft is similar to that of sounding balloons. Aircraft carry various necessary instruments to detect whether typhoons occur over areas where typhoons may occur. After observing a typhoon, it can also pass through the typhoon coverage area in different directions and at different altitudes to detect various phenomena in the typhoon. Due to the limited temporal and spatial resolution of the data, and the fact that sounding balloons are usually one-time, and the high cost of using aircraft for observation, the two methods of sounding balloons and aircraft can no longer meet the current needs for typhoon monitoring. Meteorological radar is an active microwave atmospheric remote sensing device that measures by reflecting radar waves from the target. It has high sensitivity. my country has built a certain scale of meteorological radars and formed a network. However, for current needs, there are still certain shortcomings. For example, the coverage of a single meteorological radar for monitoring typhoons is limited, and the collected data cannot be obtained in real time. At the same time, there are situations where multiple radars operate out of sync. Summary of the invention

[0004] The present invention aims to solve the shortcomings of the prior art and proposes a typhoon real-time monitoring and early warning system. By collecting different types of data and establishing a direct connection between the input layer and the output layer in the prediction model, that is, adding a direct connection between the input at each moment and the final output, the influence of historical data on the output can be enhanced and complete historical information can be retained, thereby improving the prediction accuracy.

[0005] To achieve the above-mentioned purpose, the present invention provides the following scheme: a typhoon real-time monitoring and early warning system, comprising: a collection system, a monitoring system and an early warning system;

[0006] The acquisition system is used to collect real-time typhoon data and pre-process the real-time data to obtain processed data;

[0007] The monitoring system is connected to the acquisition system and is used to analyze the processed data to obtain analysis results;

[0008] The early warning system is connected to the monitoring system and is used to issue a typhoon risk early warning based on the analysis results.

[0009] Further preferably, the acquisition system comprises: an acquisition module, a pre-processing module and a composition module;

[0010] The acquisition module is used to acquire the real-time data and map data;

[0011] The preprocessing module is connected to the acquisition module and is used to preprocess the real-time data to obtain the processed data;

[0012] The composition module is connected to the acquisition module and is used to construct a 3D prediction map based on the map data; the 3D prediction map displays the analysis result.

[0013] Further preferably, the composition module comprises: a construction unit and a display unit;

[0014] The construction unit is used to construct the 3D prediction map based on the map data; and display the real-time data and the analysis result in the 3D prediction map;

[0015] The display unit is used to display the real-time data and the analysis results according to the operation on the 3D prediction map, thereby completing the display function.

[0016] Further preferably, the monitoring system comprises: a historical database and an analysis module;

[0017] The historical database is used to store historical typhoon data;

[0018] The analysis module is connected to the historical database and is used to analyze the processed data to obtain the analysis result.

[0019] Further preferably, the analysis module includes: a path prediction unit, a wind speed prediction unit, a rainfall prediction unit and a comprehensive unit;

[0020] The path prediction unit is used to predict the typhoon path based on the processed data to obtain a predicted path;

[0021] The wind speed prediction unit is used to calculate the wind speed at different locations on the predicted path based on the processed data;

[0022] The rainfall prediction unit is used to calculate the rainfall at different locations on the predicted path based on the processed data;

[0023] The integration unit is connected to the path prediction unit, the wind speed prediction unit and the rainfall prediction unit, and is used to obtain the analysis result based on the predicted path, the wind speed and the rainfall; the analysis result is the typhoon risk intensity.

[0024] Further preferably, the process of the path prediction unit obtaining the predicted path includes:

[0025] Building a prediction model based on the historical typhoon data and the processed data;

[0026] The predicted path is obtained based on the prediction model.

[0027] Further preferably, the prediction model comprises a three-layer neural network: an input layer, a hidden layer and an output layer;

[0028] The input layer is directly connected to the output layer;

[0029] The internal information propagation process of the prediction model includes:

[0030] y t+1 =Vs t+1 +Z+b,

[0031] In the formula, y t+1 represents the output layer output; V represents the connection weight from the hidden layer to the output layer; s t+1 represents the hidden layer output; b is the bias term;

[0032] Where Z = β1·x t-1 +β2·x t +β3·x t+1 +β4·x 2 t-1 +β5·x 2 t +β6·x 2 t+1 ,

[0033] Where β1, β2, β3, β4, β5, and β6 represent the connection weights from the input layer to the output layer at time t-1, t, and t+1, respectively; x t-1 、x t 、x t+1 They are the input layer inputs at time t-1, t and t+1 respectively.

[0034] Further preferably, the wind speed calculation method includes:

[0035] Collecting satellite microwave data and processing the satellite microwave data to obtain calculation data;

[0036] A wind speed prediction model is constructed based on the calculation data, and wind speed prediction is performed based on the wind speed prediction model.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The display method adopted by the present invention can display labels with real-time data and analysis results and can adapt to the changes in form following the operation of the 3D prediction map. At the same time, the present invention can enhance the influence of historical data on output and retain complete historical information by establishing a direct connection between the input layer and the output layer in the prediction model, that is, adding a direct connection between the input at each moment and the final output. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 This is a schematic structural diagram of a typhoon real-time monitoring and early warning system according to an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of a 3D prediction map according to an embodiment of the present invention;

[0042] Figure 3 Schematic diagram of the prediction model structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Embodiment 1:

[0046] like Figure 1As shown, this embodiment provides a typhoon real-time monitoring and early warning system, including: a collection system, a monitoring system and an early warning system; wherein the collection system is used to collect real-time typhoon data and pre-process the real-time data to obtain processed data; the monitoring system is connected to the collection system to analyze the processed data to obtain analysis results; the early warning system is connected to the monitoring system to issue a typhoon risk warning based on the analysis results. The early warning information includes: wind speed and rainfall information within the predicted path range.

[0047] In this embodiment, the acquisition system includes: an acquisition module, a preprocessing module and a composition module;

[0048] The acquisition module is used to collect real-time data and map data; the acquisition module is also used to collect historical typhoon data; in this embodiment, the acquisition module includes different types of acquisition equipment, such as satellite data acquisition equipment, observation station data acquisition equipment, device acquisition equipment equipped with detection equipment, etc. to collect real-time data as comprehensively as possible.

[0049] The preprocessing module is connected to the acquisition module and is used to preprocess the real-time data to obtain processed data.

[0050] Preprocessing methods include classification and normalization. Classification refers to classifying the same type of data collected by different devices into the same type and removing redundant data. Normalization is to scale the data to a smaller range. Normalization can remove the unit restrictions of the data and convert it into a dimensionless value.

[0051] The composition module is connected to the acquisition module and is used to construct a 3D prediction map based on the map data; the 3D prediction map displays the analysis results.

[0052] In this embodiment, the composition module includes: a construction unit and a display unit; the construction unit is used to construct a 3D prediction map based on the map data; and display the real-time data and analysis results in the 3D prediction map; Figure 2 In this embodiment, the real-time data and analysis results are displayed at the corresponding position of the 3D prediction map by matching the longitude and latitude. The display unit is used to display the real-time data and analysis results according to the operation of the 3D prediction map, thereby completing the display function.

[0053] Specifically, the working process of the display unit includes:

[0054] 1. Obtain the user's operation actions on the 3D prediction map, such as: zoom, rotate, move, etc.

[0055] 2. Based on the center point of the camera in the 3D prediction map, the distance vector difference calculation is performed on the panel position where the label displaying the real-time data and analysis results is located to obtain the difference direction.

[0056] 3. According to the difference direction, calculate the distance between the panel position where the above label is located and the position of the camera in the 3D prediction map to obtain the first distance.

[0057] 4. Redraw the size and direction of the above-mentioned labels according to the first distance to complete the display function of the follow-up operation. The redrawing operation includes: based on all the labels displayed on the 3D prediction map, obtain the label position closest to the camera and the label position farthest from the camera, and calculate the average value to obtain the intermediate vector; based on the intermediate vector, compare all labels, reduce the labels smaller than the intermediate vector, and shorten the direction line; enlarge the labels larger than the intermediate vector, and extend the direction line.

[0058] When the above method is used for display, the labels showing the real-time data and analysis results can adaptively change their shapes in accordance with the operations used for the 3D prediction map.

[0059] In this embodiment, the monitoring system includes: a historical database and an analysis module; the historical database is used to store historical typhoon data; the analysis module is connected to the historical database and is used to analyze the processed data to obtain analysis results.

[0060] Among them, historical typhoon data is collected through the China Meteorological Administration's typhoon network. The collected data include: the location of the typhoon center, moving direction, maximum ground wind speed at the typhoon center, typhoon intensity, radius of each level of wind circle, central air pressure, and moving speed of the typhoon center every few hours during the time of typhoon formation, landing, and extinction.

[0061] In this embodiment, the analysis module includes: a path prediction unit, a wind speed prediction unit, a rainfall prediction unit and a comprehensive unit; the path prediction unit is used to predict the typhoon path based on the processed data to obtain the predicted path; the wind speed prediction unit is used to calculate the wind speed at different positions on the predicted path based on the processed data; the rainfall prediction unit is used to calculate the rainfall at different positions on the predicted path based on the processed data; the comprehensive unit is connected to the path prediction unit, the wind speed prediction unit and the rainfall prediction unit, and is used to obtain the analysis result based on the predicted path, wind speed and rainfall; the analysis result is the typhoon risk intensity. Among them, the method for determining the risk intensity includes:

[0062] A data set of wind speed, wind circle radius and rainfall in historical typhoon data was selected, and a fitting function of wind force, rainfall and typhoon impact range was constructed using a neural network algorithm. Then, based on the fitting function, the coordinates of the impact range involved in the typhoon path were determined and divided into appropriate grid sizes.

[0063] Then, the typhoon impact attenuation function is calculated, and the strong wind impact branch and rainfall impact score of a certain typhoon-affected grid are determined based on the typhoon impact attenuation function. The weighted average algorithm is used to calculate the comprehensive impact score. The weight coefficients of this embodiment are 0.5 and 0.5 respectively. The comprehensive impact score of each grid is calculated, and the average is taken to obtain the final comprehensive impact score.

[0064] The final comprehensive impact score is normalized, and the risk level is divided into five levels: "extremely high", "high", "low", "medium-low" and "extremely low" according to the score results.

[0065] The process of obtaining the predicted path by the path prediction unit includes:

[0066] 1. Build a prediction model based on historical typhoon data and processed data.

[0067] 2. Obtain the prediction path based on the prediction model.

[0068] In this embodiment, an improved recurrent neural network structure is used as a prediction model. The ordinary recurrent neural network structure has a full connection from the input layer to the hidden layer and then to the output layer. The nodes between each layer are not connected. This structure makes the ordinary recurrent neural network structure powerless when facing sequence problems. However, the typhoon path is ever-changing. The ordinary recurrent neural network structure cannot accurately capture the sequence rules when the typhoon path fluctuates greatly, resulting in poor prediction results. Therefore, the improvement scheme adopted in this embodiment is to build a direct connection between the input layer and the output layer, that is, to increase the direct connection between the input and the final output at each moment. Using the improved method of this embodiment, it is possible to enhance the impact of historical data on the output and retain complete historical information.

[0069] The prediction model includes a three-layer neural network: an input layer, a hidden layer, and an output layer. Since the complexity of the model is related to the hidden layer nodes, the number of neurons in the hidden layer of the improved prediction model in this embodiment is set to 2, and the time step is set to 3. The structure of the prediction model is as follows: Figure 3 shown.

[0070] The path information of the typhoon historical data and the current path information in the processed data are used as the input of the prediction model to predict the future typhoon path and establish the functional relationship between the historical typhoon path and the future typhoon path. After the typhoon historical data is input into the prediction model in chronological order, the internal information propagation process of the prediction model includes:

[0071] y t+1 =Vs t+1 +Z+b,

[0072] In the formula, y t+1 represents the output layer output; V represents the connection weight from the hidden layer to the output layer; st+1 represents the hidden layer output; b is the bias term.

[0073] Where Z = β1·x t-1 +β2·x t +β3·x t+1 +β4·x 2 t-1 +β5·x 2 t +β6·x 2 t+1 ,

[0074] Where β1, β2, β3, β4, β5, and β6 represent the connection weights from the input layer to the output layer at time t-1, t, and t+1, respectively; x t-1 、x t 、x t+1 They are the input layer inputs at time t-1, t and t+1 respectively.

[0075] s t+1 =f(Ux t+1 +Ws t +b1),

[0076] Wherein, f() represents an activation function; this embodiment adopts a hyperbolic tangent function; U represents the connection weight from the input layer to the hidden layer, b1 is the bias number; and W represents the weight matrix between the hidden layers.

[0077] Among them, typhoon historical data is used as a training set to train the prediction model. The training method includes:

[0078] 1. Calculate the similarity between historical typhoon data and processed data, select the top 10% of historical typhoon data as the training set, and use it as the input of the prediction model.

[0079] 2. Use the legacy algorithm to initialize the weight parameters and calculate the partial derivative of the loss function with respect to each weight parameter; then use the gradient descent method for training until the convergence condition is reached and the training ends.

[0080] In this embodiment, the wind speed calculation method includes:

[0081] 1. Collect satellite microwave data and process them to obtain calculation data.

[0082] The satellite microwave data in this embodiment uses the brightness temperature data of the microwave sounder ATMS, which uses advanced technology and is carried on the polar-orbiting satellite SNPP of the United States. First, the satellite microwave data is classified according to the typhoon date, and then the classified satellite microwave data is screened according to the longitude and latitude of the typhoon center, and the range of distance is set according to actual needs. The screened data is further removed of redundant and useless data to obtain calculation data.

[0083] 2. Construct a wind speed prediction model based on the calculated data, and perform wind speed prediction based on the wind speed prediction model.

[0084] In this embodiment, the wind speed prediction model adopts an improved VGG multi-scale target detection network; on the basis of the VGG multi-scale target detection network, the branches of candidate box generation and prediction in the backbone network are removed, and regression tasks are added to replace the tasks of different target classifications with the tasks of wind speed value distribution learning. The L2 regularization in the backbone network structure is removed and replaced with batch regularization before Relu after each convolution. The network outputs 85 wind speed value distribution probabilities and 1 regression wind speed value. The expected distribution probability and the mean of the regression wind speed value are used as the final estimated wind speed. The distribution probability of each wind speed value is calculated by the last FC layer through the softmax function.

[0085] In this embodiment, the rainfall prediction method includes:

[0086] The landing direction, landing center, landing wind speed and distance from the predicted location of the typhoon historical data and the above information in the processed data are used as the input of the prediction model to predict the rainfall on the future typhoon path.

[0087] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A typhoon real-time monitoring and early warning system, characterized in that: include: Collection system, monitoring system and early warning system; The acquisition system is used to collect real-time typhoon data and pre-process the real-time data to obtain processed data; The acquisition system includes: an acquisition module, a preprocessing module and a composition module; The acquisition module is used to acquire the real-time data and map data; The preprocessing module is connected to the acquisition module and is used to preprocess the real-time data to obtain the processed data; The composition module is connected to the acquisition module and is used to construct a 3D prediction map based on the map data; the 3D prediction map displays the analysis results; the composition module includes: a construction unit and a display unit; The construction unit is used to construct the 3D prediction map based on the map data; and display the real-time data and the analysis result in the 3D prediction map; The display unit is used to display the real-time data and the analysis results according to the operation on the 3D prediction map, thereby completing the display function; The working process of the display unit includes: Obtain the user's operation actions on the 3D prediction map, such as scaling, rotating, and moving; Based on the center point of the camera in the 3D prediction map, the distance vector difference calculation is performed on the panel position where the label showing the real-time data and analysis results is located to obtain the difference direction; According to the difference direction, the distance between the panel position where the label is located and the position of the camera in the 3D prediction map is calculated to obtain a first distance; The size and direction of the label are redrawn according to the first distance to complete the display function of the follow-up operation; wherein the redrawing operation includes: based on all the labels displayed on the 3D prediction map, the label position closest to the camera and the label position farthest from the camera are obtained, and the average value is calculated to obtain the intermediate vector; based on the intermediate vector, all labels are compared, and the labels smaller than the intermediate vector are reduced, and the direction lines are shortened; the labels larger than the intermediate vector are enlarged, and the direction lines are extended; the monitoring system is connected to the acquisition system to analyze the processed data and obtain the analysis results; the monitoring system includes: a historical database and an analysis module; The historical database is used to store historical typhoon data; The analysis module is connected to the historical database and is used to analyze the processed data to obtain the analysis result; The analysis module includes: a path prediction unit, a wind speed prediction unit, a rainfall prediction unit and a comprehensive unit; The path prediction unit is used to predict the typhoon path based on the processed data to obtain a predicted path; The wind speed prediction unit is used to calculate the wind speed at different positions on the predicted path based on the processed data; the wind speed calculation method includes: Collect satellite microwave data and process the satellite microwave data to obtain calculation data; First, the satellite microwave data are classified according to the typhoon date, and then the classified satellite microwave data are screened according to the longitude and latitude of the typhoon center, and the distance range is set according to actual needs; the redundant and useless data are further removed from the screened data to obtain the calculated data; Building a wind speed prediction model based on the calculated data, and performing wind speed prediction based on the wind speed prediction model; The wind speed prediction model uses an improved VGG multi-scale target detection network. Based on the VGG multi-scale target detection network, the branches of candidate box generation and prediction in the backbone network are removed, and regression tasks are added to replace the tasks of different target classification with the tasks of wind speed value distribution learning. The L2 regularization in the backbone network structure is removed and replaced with batch regularization before Relu after each convolution. The network outputs 85 wind speed value distribution probabilities and 1 regression wind speed value. The expected distribution probability and the mean of the regression wind speed value are used as the final estimated wind speed. The distribution probability of each wind speed value is calculated by the last FC layer through the softmax function. The rainfall prediction unit is used to calculate the rainfall at different locations on the predicted path based on the processed data; The integration unit is connected to the path prediction unit, the wind speed prediction unit and the rainfall prediction unit, and is used to obtain the analysis result based on the predicted path, the wind speed and the rainfall; the analysis result is the typhoon risk intensity; The early warning system is connected to the monitoring system and is used to issue a typhoon risk early warning based on the analysis results.

2. A typhoon real-time monitoring and early warning system according to claim 1, characterized in that: The process of obtaining the predicted path by the path prediction unit includes: Building a prediction model based on the historical typhoon data and the processed data; The predicted path is obtained based on the prediction model.

3. A typhoon real-time monitoring and early warning system according to claim 2, characterized in that: The prediction model includes a three-layer neural network: an input layer, a hidden layer, and an output layer; The input layer is directly connected to the output layer; The internal information propagation process of the prediction model includes: y t+1 =Vs t+1 +Z+b, In the formula, y t+1 represents the output layer output; V represents the connection weight from the hidden layer to the output layer; s t+1 represents the hidden layer output; b is the bias term; where Z = β1·x t-1 + β2·x t + β3·x t+1 + β4·x 2 t-1 + β5·x 2 t + β6·x 2 t+1 , Where β1, β2, β3, β4, β5, and β6 represent the connection weights from the input layer to the output layer at time t-1, t, and t+1, respectively; x t-1 、x t 、x t+1 They are the input layer inputs at time t-1, t and t+1 respectively.

4. A typhoon real-time monitoring and early warning system according to claim 2, characterized in that: The wind speed calculation method includes: Satellite microwave data is collected and processed to obtain calculation data; a wind speed prediction model is constructed based on the calculation data, and wind speed prediction is performed based on the wind speed prediction model.

Images