A storm surge approach monitoring and early warning method and system

Abstract

The present invention relates to the field of storm surge early warning technology, and discloses a storm surge approach monitoring and early warning method and system, wherein the storm surge approach monitoring and early warning method comprises: when the storm surge is approaching, obtaining storm surge related information and astronomical tide data; inputting the storm surge related information into a storm surge numerical simulation model to calculate the storm surge numerical simulation calculation result of the future time period, the storm surge numerical simulation calculation result includes the storm surge water increase height; calculating the water level of the future time period according to the storm surge numerical simulation calculation result and the astronomical tide data; comparing the water level of the future time period with the warning water level to determine whether to issue a storm surge warning, and the condition for issuing a storm surge warning is: the water level of the future time period is greater than the warning water level. The present invention can warn of the tidal disaster caused by the superposition of the astronomical tide and the storm surge in the future, leaving a long time for responding to the tidal disaster, which is conducive to reducing the loss of the tidal disaster.

Description

Technical Field

[0001] The present invention relates to the technical field of storm surge early warning, and more particularly to a storm surge approach monitoring and early warning method and system. Background Art

[0002] Storm surge is a phenomenon of abnormal sea level rise caused by strong atmospheric disturbances. It can be divided into tropical storm surge caused by tropical cyclones and temperate storm surge caused by temperate cyclones. Storm surges following hurricanes and typhoons usually cause great disasters to coastal areas. If storm surges overlap with astronomical tides, they will cause extremely severe tidal disasters. When warnings are issued by measuring and observing storm surges reaching warning water levels, the time left for disaster avoidance is short. Summary of the invention

[0003] The present invention provides a storm surge approach monitoring and early warning method and system, which solves the technical problem in the related art that when early warning is given by measuring and observing the storm surge reaching the early warning water level, the time left for disaster avoidance is short.

[0004] The present invention provides a storm surge approach monitoring and early warning method, comprising the following steps:

[0005] Step 1, when a storm surge is approaching, obtain storm surge related information and astronomical tide data, the astronomical tide data includes the astronomical tide water height for each future time period; perform unstructured triangular meshing on the predicted storm surge sea area, number the grids of the processed grid system, the resolution of the grid decreases as the distance from the coast increases, the grid is divided into three layers vertically, the storm surge related information collected in the upper layer includes ocean current, temperature, salinity, water quality, and meteorology, the storm surge related information collected in the middle layer includes ocean current, temperature, salinity, and water quality, and the storm surge related information collected in the bottom layer includes seabed topography; connect the storm surge related information collected in each time period before the current moment with the grid to form a grid structured data, the grid structured data includes grid data corresponding to the grid one by one, and the grid data is divided into three channels, which respectively store the storm surge related information collected in the upper layer, middle layer, and bottom layer of the grid;

[0006] Step 2, input the grid structured data into the storm surge numerical simulation model to calculate the storm surge numerical simulation results of the future time period, and the storm surge numerical simulation results include the storm surge water height of the future N time periods; the storm surge numerical simulation model includes a grid space fusion module and a time series recognition module, wherein the grid space fusion module performs data fusion on each grid structured data on the grid system to obtain spatial fusion features, and then inputs the spatial fusion features obtained by processing the grid structured data into the time series recognition module in sequence, and the order of input is consistent with the order of the time period corresponding to the grid structured data. The time series recognition module identifies the correlation of the spatial fusion features in time, obtains the time series features to the output module, and the output module maps the time series features to the prediction space to obtain the storm surge numerical simulation results;

[0007] Step 3, calculating the water level in the future time period according to the numerical simulation results of the storm surge and the astronomical tide data;

[0008] Step 4, judging whether to issue a storm surge warning based on the comparison of the water level in the future time period with the warning water level, the condition for issuing a storm surge warning is that the water level in the future time period is greater than the warning water level.

[0009] Furthermore, the storm surge warning includes more than one level, each level corresponds to a warning water level, and a storm surge warning of the corresponding level is issued when the water level in a future time period is greater than the corresponding warning water level.

[0010] Furthermore, an aggregation network is used as a grid space fusion module. The aggregation network is a multi-layer network. The expression of the layer is as follows:

[0011] ;

[0012] : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data, hour , For Grid The spatial fusion feature is the concatenation of the representation features of all grids. is the total number of layers of the aggregation network, ranging from 2 to 5;

[0013] : Grid In the Feature representation of the layer;

[0014] : From the grid To Grid news;

[0015] ;

[0016] : message calculation function;

[0017] : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data;

[0018] : Aggregate function;

[0019] : Update function;

[0020] : Grid The collection of neighbor grids, grid The neighbor grid is the grid The number of grids of spacing is less than 3 grids.

[0021] Furthermore, LSTM, RNN, and xLSTM are applied as time series recognition modules.

[0022] Furthermore, a multilayer perceptron is applied as an output module, and the output module outputs a vector, the value of the rth component of the vector represents the per-unit value of the storm surge height in the rth cycle in the future.

[0023] The present invention provides a storm surge approach monitoring and early warning system, comprising:

[0024] A plurality of fixed data collectors, a plurality of mobile data collectors, a data acquisition control system and a data processing system, wherein the data processing system is used to execute the aforementioned storm surge approach monitoring and early warning method, and the mobile data collector can be controlled by the data acquisition control system to execute the action of collecting storm surge related information;

[0025] The data acquisition control system performs the following steps to complete the control:

[0026] Based on the acquired storm surge related information, object recognition is performed to obtain specific objects;

[0027] generating first structured data according to storm surge associated information and storm surge generation information associated with a specific object;

[0028] The first structured data includes two structural dimensions. The basic structural elements included in the first structural dimension are specific objects, including: monitoring sea area divisions, data collectors, and storm surges; the basic structural elements included in the second structural dimension are time series units.

[0029] Identifying the first structured data to obtain a control scheme for the mobile data collector for the next time period;

[0030] The method of constructing the first structured data includes:

[0031] Extract the features associated with the object, and divide the extracted features into multiple parts according to the collection period. One part is mapped to a time sequence unit. The order of the time sequence units is consistent with the time order of the feature collection period.

[0032] Establishing a mapping between the extracted objects and their associated features;

[0033] For any two objects, if there is a correlation between the associated features related to the completion control or there is a correlation between the objects related to the completion control, an object relationship is established between the two objects.

[0034] Furthermore, the characteristics associated with the zoning of the monitored sea area are: water level, seabed topography, ocean currents, temperature, salinity, water quality, and weather;

[0035] Features associated with data loggers: collected water level, seafloor topography, currents, temperature, salinity, water quality, and weather;

[0036] Characteristics of storm surge associations: Storm surge generation information;

[0037] Relevance to completion control includes:

[0038] Monitor the adjacency of the subdivisions of the sea area;

[0039] Monitoring the boundaries or coverage of sub-areas of the sea area;

[0040] The data collector is used to collect data in the sub-areas of the monitored sea area;

[0041] Two data collectors collect data in the same subarea of ​​the monitored sea area;

[0042] There is a correlation between storm surge and the zoning of all monitored sea areas;

[0043] Storm surge is relevant for all data loggers.

[0044] Furthermore, the data acquisition control system applies the following neural network model to execute the steps, and the neural network model includes a graph network layer, a timing layer and an output layer, wherein the graph network layer inputs the first structured data, outputs the first identification feature to the timing layer, the timing layer outputs the second identification feature to the output layer, and the output layer outputs the control plan of the mobile data collector for the next time period.

[0045] Furthermore, the output layer outputs two matrices. The cth component of the i-th row vector of one matrix represents the probability that the i-th mobile data collector selects the c-th target position, and the target position with the largest probability value is selected as the position that the mobile data collector needs to collect;

[0046] The cth component of the i-th row vector of another matrix represents the probability that the i-th mobile data collector selects the c-th action, and the action with the largest probability value is selected as the action that the mobile data collector needs to perform.

[0047] The present invention provides a computer-readable storage medium for storing computer-readable instructions, which can execute the aforementioned storm surge approach monitoring and early warning method when the computer-readable instructions are read by a computer.

[0048] The beneficial effect of the present invention is that the present invention can give early warning of tidal disasters caused by the superposition of astronomical tides and storm tides in the future, leaving a long time for responding to tidal disasters, which is conducive to reducing the losses caused by tidal disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of the storm surge approach monitoring and early warning method of the present invention;

[0050] Figure 2 is an example of a control scheme for a mobile data collector of the present invention;

[0051] Figure 3 It is the result of gridding the sea area studied by the present invention;

[0052] Figure 4 This is a list of marine meteorological observation data that is publicly available to the public. DETAILED DESCRIPTION

[0053] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0054] Astronomical tides are tidal fluctuations caused by the gravitational effects of the moon and the sun, and can usually be predicted very accurately. Astronomical tide data are generally obtained through tidal forecasting systems or astronomical tide models. Astronomical tides reflect the regular fluctuations of natural tides, which are usually not sudden or drastic. Therefore, the combination of astronomical tide data and storm surge numerical simulation data can predict the worst possible water level conditions in the future. For example, when a storm surge occurs during the high tide period, the tidal height of the astronomical tide will increase the base of the storm surge water level, causing the water level to reach a higher warning value.

[0055] It not only monitors the current water level, but also predicts water level changes in the future. For example, if the current water level is 80cm, and according to the simulation calculation of the storm surge, an astronomical tide is expected to occur in 2 hours, and the expected height of the astronomical tide is 120cm, then the device determines through simulation calculation that the current water level (80cm) will continue in 2 hours, and superimposes the 120cm height of the astronomical tide, and finally concludes that the water level will reach 200cm (80+120cm) in 2 hours. If this water level value exceeds the set warning standard (such as the orange storm surge warning level), the system will trigger an early warning.

[0056] In short, the system can not only predict future water levels by monitoring real-time water level data and combining it with astronomical tide and storm surge simulation calculation results, but also provide advance warning of approaching storm surges, thereby providing earlier warning information for disaster prevention.

[0057] The role of water level data:

[0058] Water level data is the basic data in the monitoring and early warning system. Real-time acquisition of current water level information is the first step in determining whether there is a storm surge threat. Water level monitoring uses sensors deployed on the coastline or at sea (such as pressure sensors, buoys, tide gauges, etc.) to obtain the current sea level in real time. The main functions of this data are:

[0059] Real-time monitoring: directly reflects the trend of sea level rise or fall and helps identify potential abnormal water level changes.

[0060] Fundamentals of imminent warning: Combining storm surge simulations with astronomical tide data helps calculate water level trends over the next few hours or days.

[0061] The role of numerical simulation of storm surges:

[0062] Storm surge numerical simulation is a comprehensive model based on meteorological, oceanographic, topographical and other factors to predict the intensity, impact range and duration of storm surge. Its functions include:

[0063] Predict the intensity and duration of storm surges: Through simulation results, it is possible to predict the water level changes caused by storm surges and calculate the time it takes for the water level to exceed the warning line.

[0064] Calculate water level changes: In the case mentioned, through numerical simulation, the device can predict the water level rise caused by the storm surge (for example, increase the current water level of 80cm by a certain height until the expected peak value calculated by the simulation).

[0065] At least one embodiment of the present invention discloses a storm surge approach monitoring and early warning method, such as Figure 1 As shown, the following steps are included:

[0066] Step 1: When a storm surge is approaching, storm surge related information and astronomical tide data are obtained, where the astronomical tide data includes the astronomical tide surge height for each future time period;

[0067] The predicted storm surge sea area is processed into unstructured triangle grids, and the grids of the grid system obtained by the processing are numbered. The resolution of the grid decreases with the distance from the coast. The grid is divided into three layers vertically. The storm surge related information collected in the upper layer includes ocean currents, temperature, salinity, water quality, and meteorology. The storm surge related information collected in the middle layer includes ocean currents, temperature, salinity, and water quality. The storm surge related information collected in the bottom layer includes seabed topography.

[0068] Connecting the storm surge correlation information collected in each time period before the current moment with the grid to form a grid structured data, the grid structured data includes grid data corresponding to the grid one by one, and the grid data is divided into three channels, which respectively store the storm surge correlation information collected from the upper layer, middle layer and bottom layer of the grid;

[0069] If some or all of the layers of the grid corresponding to the grid data have not collected storm surge associated information within the corresponding time period, the channels of the grid data corresponding to the layers for which the storm surge associated information has not been collected are filled with storm surge associated information with a value of 0;

[0070] The sources of data include a certain province's marine monitoring survey data, a certain province's coastal areas of the ocean online monitoring data, socially disclosed marine meteorological observation data, satellite observation data and reanalysis data and other channels.

[0071] The public marine meteorological observation data to be collected include: Figure 4 shown.

[0072] Data collection is carried out in accordance with the relevant requirements of the "Ocean Survey Specifications", "Ocean Monitoring Specifications", "Ocean Monitoring Technical Regulations", etc., and complies with relevant regulations. For example, water temperature and salinity are measured using the CTD method, and the normative document implemented is "GB / T12763.2-2007", and dissolved oxygen is measured using the electrochemical probe method, and the normative document implemented is "HJ 506-2009".

[0073] According to the complex terrain and shore boundary characteristics of a certain province, an unstructured triangular grid system covering the sea area of ​​a certain province is designed, with a longitude range of 119°E~126°E and a latitude range of 28°N~37.5°N. The grid system range completely includes the maximum area of ​​storm surge warning; the resolution of the grid system in the coastal area of ​​a certain province is about 2km, and the resolution decreases with the increase of the distance from the coast. The farthest resolution is about 50km. It is divided into three layers vertically, the upper layer is 5~10m below the sea surface, the middle layer is the middle position of the water level of the grid, and the bottom layer is the bottom layer of the grid.

[0074] The resolution of the nearshore area is about 2km, which means that in the nearshore area, the model can capture relatively small-scale ocean dynamic characteristics, such as local water flow changes, small-scale water temperature differences, etc. Because the nearshore ocean dynamic characteristics are relatively more complex and smaller in granularity, and the cost of deploying data collectors near the coast is lower, the density of data collectors near the coast is also greater. Therefore, through grid division, the global spatial structuring of the predicted sea area can be established, and the spatial structuring is consistent with the data collection scene, which can provide a good foundation for the next step of integrating global data.

[0075] Storm surge prediction relies on long-term observation data. Since the data collector includes fixed buoys, observation stations and mobile unmanned boats, not all sea areas can be monitored, and the monitored local areas are not relatively fixed. Therefore, the local areas mapped by the global observation data in the long-term space are not fixed. In order to capture the dynamics in time and space, the sea area space is structured through grid division. The fusion of storm surge related information based on the grid system can maintain higher correlation in the long time series.

[0076] Step 2, inputting the grid structured data into the storm surge numerical simulation model to calculate the storm surge numerical simulation results of the future time period, the storm surge numerical simulation results including the storm surge water height of the future N time periods;

[0077] The storm surge numerical simulation model includes a grid space fusion module and a time series recognition module, wherein the grid space fusion module performs data fusion on each grid structured data on the grid system to obtain spatial fusion features, and then sequentially inputs the spatial fusion features obtained by processing the grid structured data into the time series recognition module, and the order of input is consistent with the order of time periods corresponding to the grid structured data. The time series recognition module identifies the correlation of the spatial fusion features in time, obtains the time series features to the output module, and the output module maps the time series features to the prediction space to obtain the storm surge numerical simulation calculation results;

[0078] In some embodiments of the present invention, an aggregation network is applied as a grid space fusion module.

[0079] The aggregation network is a multi-layer network. The expression of the layer is as follows:

[0080]

[0081] : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data, hour , For Grid The spatial fusion feature is the concatenation of the representation features of all grids. The total number of layers of the aggregation network. The value ranges from 2 to 5. The default value is 3.

[0082] : Grid In the Feature representation of the layer;

[0083] : From the grid To Grid news;

[0084] ;

[0085] : Message calculation function, which can be implemented in many ways, for example:

[0086] Simple linear transformation: , Indicates the message calculation weight.

[0087] Simple linear transformation: , Indicates splicing.

[0088] Use non-linear transformations, such as ReLU, to enhance expressiveness.

[0089] : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data.

[0090] : An aggregate function (such as sum, average, or maximum).

[0091] : Update function, such as GRU (Gated Recurrent Unit) or a simple MLP (Multilayer Perceptron).

[0092] : Grid The collection of neighbor grids, grid The neighbor grid is the grid The number of grids of spacing is less than 3 grids.

[0093] It should be noted that the grid data needs to be processed through feature engineering before being input into the aggregation network.

[0094] In some embodiments of the present invention, LSTM (Long Short-Term Memory), RNN (Recurrent Neural Network), xLSTM (Extended Long Short-Term Memory), and 2D CNN (convolutional neural network) are used as time series recognition modules.

[0095] In some embodiments of the present invention, a multi-layer perceptron is applied as an output module.

[0096] In a specific embodiment, LSTM is applied as a time series recognition module, the output of the last time step of LSTM is input into the output module, and the output module outputs a vector, the value of the rth component of the vector represents the per-unit value of the storm surge water height in the future rth cycle.

[0097] For example, if the value of the rth component is 0.63 and the baseline value of the storm surge height is 300 cm, then the storm surge height in the rth cycle is 189 cm.

[0098] Step 3, calculating the water level in the future time period according to the numerical simulation results of the storm surge and the astronomical tide data;

[0099] Step 4, judging whether to issue a storm surge warning based on the comparison of the water level in the future time period with the warning water level, the condition for issuing a storm surge warning is that the water level in the future time period is greater than the warning water level.

[0100] In some embodiments of the present invention, the storm surge warning includes more than one level, and each level corresponds to a warning water level.

[0101] For example, storm surge alerts include orange and red levels, corresponding to orange warning water levels and red warning water levels respectively.

[0102] In some embodiments of the present invention, since high-dimensional comprehensive data are called, the storm surge numerical simulation model applies a neural network model for fitting. In some specific feasible methods, CNN (convolutional neural network) or RNN (temporal neural network) is applied as the storm surge numerical simulation model, and the storm surge numerical simulation model is trained through supervised training.

[0103] In some embodiments of the present invention, although high-dimensional comprehensive data are used, the data modality is single, and a general nonlinear regression model or decision tree model can still be tried for fitting.

[0104] Here, the present invention provides an example of calculation in implementation:

[0105] The numerical simulation results of storm surge are as follows: ,in The height of the storm surge, is the sth time period;

[0106] Astronomical tide data is ,in The height of the astronomical tide. is the sth time period;

[0107] The water level for the future time period is ,in For the water level of the future time period, is the sth time period;

[0108]

[0109] The actual application sea area is extremely vast. Obtaining storm surge related information at multiple observation points is a necessary measure to improve the robustness of the system. However, fixed observation points are generally arranged to ensure the consistency of information expressed by the data, but a small number of observation points are difficult to achieve good results. Increasing the number blindly will cause uncontrollable costs and it is difficult to balance the number of observation points. On the other hand, storm surges are affected by various factors such as weather, seabed topography, seasons, and ocean currents. The optimal layout of observation points for each storm surge observation is very different, and it is difficult to achieve the global optimal layout.

[0110] In some further embodiments of the present invention, a hardware system based on a combination of a fixed data collector and a mobile data collector is provided to realize the combination of fixed observation points and mobile observation points, wherein the fixed observation points provide basic information collection, and the mobile observation points provide information collection solutions that can be dynamically and flexibly combined according to the current storm surge conditions. In addition, during the implementation process, a variety of optional information collection tasks are configured for the mobile data collector, which can achieve global optimization at a lower cost.

[0111] The benefits of this approach include:

[0112] Improve data quality. According to the key characteristics of storm surge development, flexibly collect the most valuable data and reduce information redundancy.

[0113] Save observation costs. Avoid wasting resources in low-risk areas and maximize the effectiveness of limited observation resources.

[0114] Reduce warning uncertainty. Increase observation density near possible storm surge landing points to improve warning confidence.

[0115] Such a hardware system requires dynamic policy control, and in order to cooperate with such a hardware system, a software system is required to cooperate.

[0116] In some embodiments of the present invention, a storm surge approach monitoring and early warning system is provided, comprising:

[0117] Multiple fixed data collectors, multiple mobile data collectors, a data collection control system and a data processing system. The information collected by the data collectors is transmitted to the data processing system, wherein the data processing system is used to execute the aforementioned storm surge proximity monitoring and early warning method, and the mobile data collector can be controlled by the data collection control system to execute the action of collecting storm surge related information.

[0118] The fixed data collectors are arranged in a conventional manner, such as evenly distributed.

[0119] In some specific implementations, the data acquisition control system performs the following steps to complete the control:

[0120] Based on the acquired storm surge related information, object recognition is performed to obtain specific objects;

[0121] generating first structured data according to storm surge associated information and storm surge generation information associated with a specific object;

[0122] The first structured data includes two structural dimensions. The basic structural elements included in the first structural dimension are specific objects, including: monitoring sea area divisions, data collectors, and storm surges; the basic structural elements included in the second structural dimension are time series units.

[0123] The first structured data is identified to obtain a control scheme of the mobile data collector for the next time period.

[0124] In some specific implementations of the present invention, the method for constructing the first structured data includes:

[0125] Extract the features associated with the object, and divide the extracted features into multiple parts according to the collection period. One part is mapped to a time sequence unit. The order of the time sequence units is consistent with the time order of the feature collection period.

[0126] Establishing a mapping between the extracted objects and their associated features;

[0127] For any two objects, if there is a correlation between the associated features related to the completion of control or there is a correlation between the objects related to the completion of control, an object relationship is established between the two objects;

[0128] In some embodiments of the present invention, the characteristics of the subareas associated with the monitored sea area include: water level, seafloor topography, ocean currents, temperature, salinity, water quality, and weather;

[0129] Features associated with the data logger: collected water level, seafloor topography, ocean currents, temperature, salinity, water quality, and weather (if there are uncollected items, interpolate 0 values ​​as fill);

[0130] Characteristics of storm surge associations: Storm surge generation information (expressing the causes of storm surge formation, information about associated storms, and estimated duration of storm surges, etc.);

[0131] In some embodiments of the present invention, the associations related to completion control include:

[0132] Monitor the adjacency of the subdivisions of the sea area;

[0133] Monitoring the boundaries or coverage of sub-areas of the sea area;

[0134] The data collector is used to collect data in the sub-areas of the monitored sea area;

[0135] Two data collectors collect data in the same subarea of ​​the monitored sea area;

[0136] There is a correlation between storm surge and the zoning of all monitored sea areas;

[0137] Storm surge is relevant for all data loggers.

[0138] The aforementioned objects: the partition of the monitored sea area and the data collector refer to the type of object, not an object. For example, there are multiple objects belonging to the partition of the monitored sea area, and one object only refers to a partition of the monitored sea area.

[0139] like Figure 2 As shown, in a specific embodiment, the mobile data collector may adopt the following equipment:

[0140] Unmanned Ships / Autonomous Surface Vehicles (USV / ASV): These devices can sail autonomously on the sea surface, carrying various meteorological and oceanographic sensors to collect key data such as wind speed, air pressure, wave height, and ocean currents in real time. They are flexible and can be deployed to estimated high-risk areas before a storm arrives to perform encrypted observation tasks.

[0141] Underwater Glider / AUV: These devices can perform preset "flying" navigation tasks underwater, using sensors such as CTD and ADCP to collect profile parameters such as temperature and salinity and ocean currents. They can provide the ocean subsurface environment information required for storm surge warning.

[0142] Drifting buoys / submersible buoys: Drifting and anchored buoys of various sizes can be used to collect ocean dynamic environmental elements at the surface and at different depths. They can be deployed in key areas affected by storm surges to obtain long-term fixed-point observation data.

[0143] Fixed-wing / rotor-wing UAVs: Mobile observation systems based on UAVs are also developing rapidly. They can map coastal topography and obtain high-resolution geographic information data required for numerical simulation of storm surges and risk assessment.

[0144] Tide gauges and wave radars: Although not necessarily mobile platforms, they can be deployed flexibly by setting up temporary observation stations. The tide and wave information obtained can directly indicate the level of storm surge disaster risk.

[0145] For the control of mobile data collectors Figure 2 As shown, the data collection action column expresses the information collection task that the mobile data collector needs to perform.

[0146] The data acquisition control system applies the following neural network model to perform the steps, the neural network model includes a graph network layer, a timing layer and an output layer, wherein the graph network layer inputs the first structured data, outputs the first identification feature to the timing layer, and the timing layer outputs the second identification feature to the output layer;

[0147] The expression of the graph network layer is as follows:

[0148]

[0149] in represents the initial hidden features of the vth object in the tth time sequence unit, represents the aggregation coefficient of the vth object in the tth time series unit, represents the set of objects that have object relationships with the vth object of the tth time sequence unit, represents the activation function, represents the graph weight coefficient;

[0150]

[0151]

[0152]

[0153] in and Respectively represent the associated features of the vth and uth objects of the tth time sequence unit, and Respectively represent the linear transformation features of the vth and uth objects in the tth time sequence unit, represents the concatenation function, represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transposition, represents an exponential function with a natural constant as the base, LeakyRelu represents a modified linear unit function, Represents the set of objects that have an object relationship with the vth object.

[0154] The expression of the timing layer is as follows:

[0155]

[0156] in represents the t-th deep hidden feature, represents the t-1th deep hidden feature, , Represents the initial hidden features of the vth object in the tth time sequence unit, when t=1 , Represents the collection of all objects, and are the first and second sequence weight parameters, is the sequence bias parameter, and tanh is the hyperbolic tangent function.

[0157] In one embodiment of the present invention, the output layer outputs two matrices, wherein the cth component of the i-th row vector of one matrix represents the probability that the i-th mobile data collector selects the c-th target position, and the target position with the largest probability value is selected as the position that the mobile data collector needs to collect;

[0158] The cth component of the i-th row vector of another matrix represents the probability that the i-th mobile data collector selects the c-th action, and the action with the largest probability value is selected as the action that the mobile data collector needs to perform.

[0159] Train the first output layer through supervised learning;

[0160] The second output layer is then trained through Q-learning, and the reward function of the training is to minimize the cost and minimize the error of the numerical simulation results of the storm surge.

[0161] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A storm surge approach monitoring and early warning method, characterized in that: The following steps are involved: Step 1, when a storm surge is approaching, obtain storm surge related information and astronomical tide data, the astronomical tide data includes the astronomical tide water height for each future time period; perform unstructured triangular meshing on the predicted storm surge sea area, number the grids of the processed grid system, the resolution of the grid decreases as the distance from the coast increases, the grid is divided into three layers vertically, the storm surge related information collected in the upper layer includes ocean current, temperature, salinity, water quality, and meteorology, the storm surge related information collected in the middle layer includes ocean current, temperature, salinity, and water quality, and the storm surge related information collected in the bottom layer includes seabed topography; connect the storm surge related information collected in each time period before the current moment with the grid to form a grid structured data, the grid structured data includes grid data corresponding to the grid one by one, and the grid data is divided into three channels, which respectively store the storm surge related information collected in the upper layer, middle layer, and bottom layer of the grid; Step 2, input the grid structured data into the storm surge numerical simulation model to calculate the storm surge numerical simulation results of the future time period, and the storm surge numerical simulation results include the storm surge water height of the future N time periods; the storm surge numerical simulation model includes a grid space fusion module and a time series recognition module, wherein the grid space fusion module performs data fusion on each grid structured data on the grid system to obtain spatial fusion features, and then inputs the spatial fusion features obtained by processing the grid structured data into the time series recognition module in sequence, and the order of input is consistent with the order of the time period corresponding to the grid structured data. The time series recognition module identifies the correlation of the spatial fusion features in time, obtains the time series features to the output module, and the output module maps the time series features to the prediction space to obtain the storm surge numerical simulation results; Step 3, calculating the water level in the future time period according to the numerical simulation results of the storm surge and the astronomical tide data; Step 4, judging whether to issue a storm surge warning based on the comparison of the water level in the future time period with the warning water level, the condition for issuing a storm surge warning is that the water level in the future time period is greater than the warning water level.

2. A storm surge approach monitoring and early warning method according to claim 1, characterized in that: The storm surge warning includes more than one level, each level corresponds to a warning water level, and a storm surge warning of the corresponding level is issued when the water level in the future time period is greater than the corresponding warning water level.

3. A storm surge approach monitoring and early warning method according to claim 1, characterized in that: The aggregation network is used as the grid space fusion module. The aggregation network is a multi-layer network. The expression of the layer is as follows: ； : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data, hour , For Grid The spatial fusion feature is the concatenation of the representation features of all grids. is the total number of layers of the aggregation network, ranging from 2 to 5; : Grid In the Feature representation of the layer; : From the grid To Grid Message; ； : message calculation function; : Grid In the The feature representation of the layer, hour , For Grid Connected mesh data; : Aggregate function; : Update function; : Grid The collection of neighbor grids, grid The neighbor grid is the grid The number of grids of spacing is less than 3 grids.

4. A storm surge approach monitoring and early warning method according to claim 1, characterized in that: LSTM, RNN, and xLSTM are used as time series recognition modules.

5. A storm surge approach monitoring and early warning method according to claim 1, characterized in that: A multilayer perceptron is used as an output module, and the output module outputs a vector, the value of the rth component of the vector represents the per-unit value of the storm surge height in the rth cycle in the future.

6. A storm surge approach monitoring and early warning system, characterized in that: include: A plurality of fixed data collectors, a plurality of mobile data collectors, a data acquisition control system and a data processing system, wherein the data processing system is used to execute the storm surge approach monitoring and early warning method as described in any one of claims 1 to 5, and the mobile data collector can be controlled by the data acquisition control system to execute the action of collecting storm surge related information; The data acquisition control system performs the following steps to complete the control: Based on the acquired storm surge related information, object recognition is performed to obtain specific objects; generating first structured data according to storm surge associated information and storm surge generation information associated with a specific object; The first structured data includes two structural dimensions. The basic structural elements included in the first structural dimension are specific objects, including: monitoring sea area divisions, data collectors, and storm surges; the basic structural elements included in the second structural dimension are time series units. Identifying the first structured data to obtain a control scheme for the mobile data collector for the next time period; The method of constructing the first structured data includes: Extract the features associated with the object, and divide the extracted features into multiple parts according to the collection period. One part is mapped to a time sequence unit. The order of the time sequence units is consistent with the time order of the feature collection period. Establishing a mapping between the extracted objects and their associated features; For any two objects, if there is a correlation between the associated features related to the completion control or there is a correlation between the objects related to the completion control, an object relationship is established between the two objects.

7. A storm surge approach monitoring and early warning system according to claim 6, characterized in that: Monitor the characteristics of the sea area: water level, seabed topography, currents, temperature, salinity, water quality, and weather; Features associated with data loggers: collected water level, seafloor topography, currents, temperature, salinity, water quality, and weather; Characteristics of storm surge associations: Storm surge generation information; Relevance to completion control includes: Monitor the adjacency of the subdivisions of the sea area; Monitoring the boundaries or coverage of sub-areas of the sea area; The data collector is used to collect data in the sub-areas of the monitored sea area; Two data collectors collect data in the same subarea of ​​the monitored sea area; There is a correlation between storm surge and the zoning of all monitored sea areas; Storm surge is relevant for all data loggers.

8. A storm surge approach monitoring and early warning system according to claim 6, characterized in that: The data acquisition control system applies the following neural network model to execute the steps, and the neural network model includes a graph network layer, a timing layer and an output layer, wherein the graph network layer inputs the first structured data, outputs the first identification feature to the timing layer, the timing layer outputs the second identification feature to the output layer, and the output layer outputs the control plan of the mobile data collector for the next time period.

9. A storm surge approach monitoring and early warning system according to claim 8, characterized in that: The output layer outputs two matrices. The cth component of the i-th row vector of one matrix represents the probability that the i-th mobile data collector selects the c-th target position. The target position with the largest probability value is selected as the position that the mobile data collector needs to collect. The cth component of the i-th row vector of another matrix represents the probability that the i-th mobile data collector selects the c-th action, and the action with the largest probability value is selected as the action that the mobile data collector needs to perform.

10. A computer-readable storage medium, characterized in that: It is used to store computer-readable instructions, and when the computer-readable instructions are read by a computer, the storm surge approach monitoring and early warning method as described in any one of claims 1 to 5 can be executed.

Images