Marine disaster risk prevention and control early warning system and method based on big data

Abstract

The invention discloses a big data-based marine disaster risk prevention and control early warning system and method, and belongs to the technical field of marine disaster monitoring and early warning. The invention discloses a big data-based marine disaster risk prevention and control early warning system and method. The system comprises a multi-source data fusion module, an intelligent modeling module and a real-time early warning module, the method comprises the following steps: cleaning and storing multi-source data after wavelet transform downsampling through an edge computing node; a space-time Transform network is adopted to extract space-time features, a Coriolis force correction equation is combined to calibrate a predicted value, and the disaster risk probability and tide level and wind speed key parameters in the next 24 hours are generated; graded early warning is triggered based on a dynamic risk map, and second-level pushing is carried out through multiple channels such as a mobile network and emergency broadcast. The tide level prediction error is reduced to be smaller than or equal to 0.2 m, the early warning response delay is smaller than 10 seconds, and the problems that a traditional system is high in false alarm rate and updating lags are effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of marine disaster monitoring and early warning, and specifically relates to a marine disaster risk prevention and control early warning system and method based on big data. Background Art

[0002] Marine disasters such as storm surges, tsunamis, and giant waves pose a serious threat to coastal urban safety, marine economic activities, and the ecological environment. With the widespread adoption of technologies such as satellite remote sensing and sensor networks, marine monitoring data has exploded in volume. Leveraging big data to improve disaster prediction accuracy and response efficiency has become a key research focus in the international field of marine disaster prevention and mitigation. Traditional marine disaster early warning systems face the following bottlenecks: inefficient data processing. The heterogeneous formats and high-frequency sampling of multi-source monitoring data create significant transmission and storage pressures. Traditional filtering algorithms struggle to effectively remove noise, resulting in data utilization rates of less than 10%. Prediction models are limited. Purely physical models rely on idealized assumptions and fail to capture complex spatiotemporal correlations. Purely data-driven models lack physical constraints, resulting in low reliability and tide level prediction errors often exceeding 0.5 meters. Warning responses are crude, with limited risk visualization methods and fixed warning grading strategies that fail to dynamically match risk levels across regions, leading to high information latency. Model updates lag, with post-disaster data not effectively utilized for model optimization, requiring manual retraining and making them incapable of adapting to dynamic changes in the marine environment.

[0003] Therefore, there is an urgent need for a marine disaster risk prevention and control early warning system and method that integrates big data processing, intelligent modeling and dynamic response to solve the core problems of traditional systems in data efficiency, prediction accuracy, response effectiveness and model adaptability. Summary of the Invention

[0004] In response to the above pain points, the present invention provides a marine disaster risk prevention and control early warning system and method based on big data. Through multi-source data fusion, spatiotemporal Transformer and physical formula hybrid modeling, dynamic risk map and graded early warning, and post-disaster data-driven closed-loop training mechanism, it can achieve accurate prediction and efficient response to marine disaster risks.

[0005] The scheme of the present invention is as follows: The marine disaster risk prevention and control early warning system based on big data is characterized by including a multi-source data fusion module, an intelligent modeling module, and a real-time early warning module: The multi-source data fusion module is used to access multi-source real-time monitoring data such as satellite remote sensing and buoy monitoring, as well as historical disaster data, and to clean and feature-engineer the data before storing it in a distributed storage system. The intelligent modeling module builds a hybrid prediction model based on spatiotemporal sequence data. The hybrid prediction model combines a spatiotemporal Transformer network with physical correction formulas to predict disaster risk probability and key parameters. The module also has a closed-loop training mechanism that can automatically trigger incremental updates of model parameters based on post-disaster assessment data to continuously optimize prediction accuracy. The real-time warning module generates a dynamic risk map and graded warning information based on the prediction results, and pushes them to the terminal device.

[0006] Preferably, the multi-source data fusion module includes an edge computing node, which obtains the 50Hz high-frequency sampling real-time monitoring raw data output by the device end, and performs multi-resolution decomposition on the raw data through the wavelet transform algorithm: performs abnormal filtering on the high-frequency components, eliminates noise and outliers that exceed the preset physical threshold, retains the low-frequency trend characteristics, and verifies the rationality of the low-frequency components in combination with the statistical laws of historical data; and finally generates a minute-level down-sampling data sequence that meets the needs of marine disaster warning.

[0007] Preferably, the distributed storage system uses the HBase engine to store time-series monitoring data such as wave height, wind speed, and tide level, and uses the Elasticsearch engine to store disaster report texts, sensor logs, and satellite remote sensing image metadata, and achieves efficient data management through the following technologies: Unified API Gateway: Provides a standardized data access interface and automatically routes query requests to the corresponding storage engine, supporting cross-database association queries based on time and regional dimensions; Index optimization strategy: Establishes a time plus regional composite index in HBase, and builds a text inverted index and a geographic spatial index in Elasticsearch to shorten the single-database query response time; Redis cache mechanism: Implements a cache strategy for high-frequency access data in the past 7 days, reducing the repeated query response time from an average of 200ms to ≤20ms; Through the collaboration of the above technologies, the cross-database query delay of the storage system is ≤500ms, meeting the real-time data call requirements of marine disaster warning.

[0008] Preferably, the hybrid prediction model combines the spatiotemporal Transformer network with a physical correction formula, wherein the spatiotemporal Transformer network encodes marine disaster-related features such as wind speed, air pressure, and tide level into a spatiotemporal matrix according to a 10-minute time step, captures cross-feature correlations through a multi-head attention mechanism, and outputs the initial value of the disaster risk probability and a time series feature vector; the physical correction formula is based on the basic equation of tidal theory, and uses the Coriolis force correction term and the seabed friction parameter correction term to calibrate the initial value of the disaster risk probability and key parameters such as tide level output by the spatiotemporal Transformer network; forms a disaster risk probability prediction value that includes physical mechanism constraints and prediction results of key parameters such as tide level, wind speed, and wave height, thereby realizing the prediction of disaster risk probability and key parameters in the next 24 hours.

[0009] Preferably, the physical correction formula is based on the basic equation of tidal theory:

[0010] Where, Indicates the storm surge water value (unit: meter), which is a key parameter that needs to be calibrated; is the seabed friction parameter correction term, where Respectively represent the horizontal flow velocity in the east-west direction and the north-south direction, unit: m / s, is the friction coefficient related to the seabed roughness; is the Coriolis force correction term, is the Coriolis parameter, which is approximately , which is related to geographical latitude; is the seawater velocity vector, The curl of the velocity vector is used to correct the influence of the Coriolis force caused by the Earth's rotation. The tide level prediction results output by the spatiotemporal Transformer network are calibrated by the above correction term to make the tide level prediction error ≤ 0.2 meters.

[0011] Preferably, the real-time warning module develops a dynamic risk map based on Leaflet.js, supports the superposition of coastal population density and important facility location data, and automatically identifies high-risk areas; the graded warning includes four levels: blue, yellow, orange, and red. The red warning requires pre-verification and manual secondary confirmation, and the warning response delay is <10 seconds.

[0012] Preferably, the closed-loop training mechanism includes: automatically synchronizing disaster assessment data from the distributed storage system after a disaster, the assessment data including post-disaster real-time monitoring data processed by a multi-source data fusion module and data on disasters of the same type in a historical case library, wherein the real-time monitoring data includes measured values ​​of wind speed, wave height, and tide level; when more than 95% of the effective feature dimensions of the assessment data are free of missing information and the number of available samples reaches 15% of the total amount of training data, an incremental update of the parameters of the hybrid prediction model is triggered, and the update process performs gradient optimization based on the current parameters of the model.

[0013] Preferably, the warning information push channels of the real-time warning module include mobile communication networks, Internet platforms, emergency broadcast systems and smart terminal devices. Through the preset hierarchical strategy configuration table, blue warnings are pushed to monitoring personnel once a day by SMS, yellow warnings are added with graphic and text notifications on WeChat public accounts, orange warnings start emergency broadcasts on the hour, and red warnings simultaneously trigger strong reminders on smart terminal APPs and are pushed to emergency departments and coastal community leaders every 10 minutes, thereby achieving differentiated reach of warnings at different levels.

[0014] Preferably, the multi-source data fusion module, intelligent modeling module and real-time warning module are deployed using a unified technical architecture, and automatic connection is achieved through the following methods: standardized data interface: data is transmitted between modules through predefined data formats and interactive protocols; task scheduling mechanism: based on the workflow engine, task chains are automatically arranged in the order of data processing, model training, and warning generation; cross-module linkage: warning results trigger model incremental training requests to form a closed-loop feedback.

[0015] A marine disaster risk prevention and control early warning method based on big data, the implementation steps are as follows: S1. Multi-source data fusion processing: This approach integrates real-time monitoring data from multiple sources, such as satellite remote sensing and buoy monitoring, as well as historical disaster data. Edge computing nodes use a wavelet transform algorithm to downsample and filter out anomalies on the 50Hz high-frequency data at the minute level. Data cleaning is then performed to repair missing values, remove outliers, and perform feature engineering to generate time series data that meets early warning requirements. S2. Hybrid model prediction calculation: The processed data is input into the hybrid prediction model. The hybrid prediction model encodes marine disaster-related features such as wind speed, air pressure, and tide level into a spatiotemporal matrix at a 10-minute time step through a spatiotemporal Transformer network. The multi-head attention mechanism is used to extract spatiotemporal correlation features and output the initial value of the disaster risk probability and the time series feature vector. Based on the basic equations of tidal theory, the prediction results are physically calibrated using the Coriolis force correction term and the seabed friction parameter correction term to generate the predicted values ​​of the disaster risk probability and key parameters such as tide level, wind speed, and wave height for the next 24 hours. S3. Dynamic warning and information push: Based on the forecast results, a dynamic risk map developed with Leaflet.js is overlaid with coastal population density and key facility location data to automatically identify high-risk areas. Alerts are triggered based on four preset thresholds: blue, yellow, orange, and red. Red alerts require pre-verification and manual secondary confirmation. Differentiated warning information is pushed through multiple channels, including mobile communication networks and emergency broadcasts, to ensure response delays are less than 10 seconds. S4. Model closed-loop training mechanism: After a disaster, disaster assessment data is automatically synchronized from the distributed storage system. When the effective feature dimensions of the assessment data exceed 95% without missing and the number of available samples reaches 15% of the total training data, the parameters of the hybrid prediction model are incrementally updated. The update process performs gradient optimization based on the current parameters of the model.

[0016] Compared with the prior art, the advantages of the present invention are: (1) In this invention, edge computing nodes are combined with wavelet transform algorithms to downsample 50Hz high-frequency data into minute-level sequences, reducing 90% of invalid data transmission, lowering network load and storage costs. A distributed storage system is used to optimize multi-engine collaboration and indexing, making cross-database query latency ≤ 500ms. The response time for high-frequency data queries in the past 7 days is reduced from 200ms to ≤ 20ms, meeting the data call requirements for real-time early warning. (2) In this invention, a spatiotemporal Transformer network is used to extract cross-feature spatiotemporal correlations through a multi-head attention mechanism, output the initial value of the risk probability, and combine the Coriolis force and seabed friction correction terms of the tidal theory equation to control the tide level prediction error to ≤ 0.2 meters, which is 60% higher than the accuracy of the traditional pure data model. A closed-loop training mechanism is used to automatically trigger the incremental update of the model using post-disaster assessment data, without the need for full training, so that the model prediction accuracy continues to improve with the accumulation of disaster cases and adapt to the dynamic changes of the marine environment. (3) In the present invention, a dynamic risk map is used to support real-time superposition of population density and important facility data, automatically identify high-risk areas, and provide visual support for emergency decision-making; a graded warning strategy is adopted to achieve differentiated reach through multiple channels, and the red warning response delay is less than 10 seconds, which is 98% higher than the traditional system, ensuring that risk information of different levels is accurately and timely conveyed. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a structural diagram of the marine disaster risk prevention and control early warning system based on big data.

[0018] Figure 2 This is a flow chart of the marine disaster risk prevention, control and early warning method based on big data. DETAILED DESCRIPTION

[0019] The technical solutions of the embodiments of the present invention will be explained and described below, but the following embodiments are only preferred embodiments of the present invention and are not exhaustive. Based on the embodiments in the implementation manner, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.

[0020] Example 1 1. Deployment Environment and Hardware Configuration 1. Scenario positioning: The marine disaster monitoring center in a coastal province covers an area of ​​approximately 20,000 square kilometers of sea area, deploys 50 buoy monitoring stations and 3 shore-based radar stations, and connects to the remote sensing data interface of the Satellite Ocean Application Center.

[0021] 2. Hardware facilities: 2-1. Edge computing nodes: NVIDIA Jetson Xavier NX edge computing boxes are deployed at each buoy station. They are equipped with ARM architecture processors, support 50Hz real-time data processing, and have built-in wavelet transform algorithm firmware. 2-2. Distributed storage cluster: Time series data cluster: 10 HBase servers: configured with Intel Xeon Silver 4310, 256GB memory, and SSD storage; table storage is divided into "time (day) + region (1°×1° longitude and latitude grid)"; Unstructured data cluster: 5 Elasticsearch servers (with the same configuration as above), storing satellite remote sensing metadata, such as Sentinel-1 SAR image metadata, about 10GB per day; Cache layer: 3 Redis servers with 16GB of memory, caching the tide and wind speed data that has been frequently accessed over the past 7 days; 2-3. Model training server: 2 GPU servers, NVIDIA A100×4, 2TB of video memory, used for spatiotemporal Transformer network training and physical correction calculations.

[0022] 3. Software stack: 3-1. Edge computing: Implementing the wavelet transform algorithm in C++ and integrating it into the JetsonOS system; 3-2. Data storage: HBase 2.4.16 + Elasticsearch 7.17.9, through a unified API gateway, developed based on Spring Boot, providing access interfaces; 3-3. Model development: PyTorch 2.1 to implement the spatiotemporal Transformer, Python scientific computing libraries: NumPy and SciPy, to implement the physical correction formula; 3-4. Early warning system: The front-end is developed based on Leaflet.js 1.9.4, and the back-end is built with Go language to build microservices, connecting to the SMS gateway and emergency broadcast protocol.

[0023] 2. Multi-source data fusion module 1. Data access Buoy monitoring data: Data from the SBE37-SI temperature, salinity, and depth instrument is collected via an RS-485 interface at a sampling frequency of 50 Hz. The data includes temperature, salinity, current velocity (u / v components), and tide level. The data is processed by the edge computing node and transmitted to the central terminal via the 4G network. 1-2. Satellite remote sensing data: Daily ocean wave data from the HY-2B satellite of the Satellite Ocean Application Center is pulled regularly. The data has a spatial resolution of 10 km and a temporal resolution of 6 hours, and includes significant wave height, wave direction, and wind speed. 1-3. Historical data: Synchronize typhoon path data for the past 10 years from the provincial disaster recovery database in CSV format, including time, center latitude and longitude, wind speed level; and storm surge historical reports.

[0024] 2. Edge computing: Wavelet transform: db8 wavelet is applied to 50Hz flow velocity data for a two-layer decomposition. High-frequency components greater than 1Hz are filtered with a threshold. For example, flow velocity absolute values ​​greater than 12m / s are considered noise, while low-frequency components ≤ 1Hz retain trend characteristics. Combined with historical data statistics, for example, if the normal flow velocity in the sea area is ≤ 8m / s, outliers are removed, and finally one-minute interval data is generated, reducing the data volume from 288,000 / day to 3,456 / day, with a compression rate of 98.8%.

[0025] 3. Storage Verification: Perform a cross-database query: Retrieve tide level data for the area "29°-30°N, 121°-122°E, 00:00-24:00, August 1, 2024" (HBase) and correlate it with disaster reports for the same period (Elasticsearch). A JMeter stress test of 100 queries showed an average latency of 420ms, meeting the ≤500ms requirement. Redis cache hit rate: Using the Redis monitoring tool, we found that the data access hit rate over the past seven days was 92%, and the response time for repeated queries was stable at 18ms.

[0026] 3. Intelligent Modeling Module 1. Hybrid model construction: 1-1. Spatiotemporal Transformer Network Structure: Input layer: Encode wind speed (m / s), air pressure (hPa), tide level (m), and u / v flow rate (m / s) with a 10-minute step into a spatiotemporal matrix: the temporal dimension is 144 steps, and the spatial dimension is a 3×3 grid centered on the buoy station, with a total of 9 nodes; Encoder: 6 Transformer layers, 8 attention heads per layer, 512 hidden layer dimensions, capturing cross-feature correlations, such as the lagged correlation between wind speed and wave height: historical statistics show that for every 1m / s increase in wind speed, wave height increases by 0.2m after 2 hours; Output layer: The fully connected layer outputs the risk probability and time series feature vector, which is the tide level change trend in the next 24 hours.

[0027] 2. Physical correction implementation: Implement the tidal theory equation in Python for a coastal province (typical latitude is about 30° north latitude, Coriolis parameter C= ), using PyTorch’s automatic differentiation function to combine the correction term with the Transformer output to form an end-to-end trainable model; Calibration example: During a typhoon on September 10, 2024, the Transformer predicted a tide level of 3.2m, which was corrected to 3.45m. The actual measured value was 3.52m, with an error of 0.07m (<0.2m).

[0028] 3. Closed-loop training verification: Post-disaster data synchronization: After the typhoon, 320 pieces of measured data (tide level, wind speed, and wave height completeness rate of 100%) were automatically obtained. The number of samples accounted for 17% of the training set (the total training set was 1,880), triggering an incremental update. Training process: Using the AdamW optimizer, a learning rate of 1e-4, and 50 iterations, it took 1 hour and 45 minutes. After the update, the model's risk probability prediction accuracy for similar storm surges increased from 85% to 88%.

[0029] 4. Real-time warning module 1. Dynamic risk map: related overlay data: Population density: Grid data of coastal population distribution in a coastal province in 2023: resolution 100m, unit: person / hectare; Important facilities: Contains shapefile vector data of three nuclear power plants and five ports; High-risk identification: Generate a storm surge risk heat map using Leaflet's Heatmap plug-in. When the predicted flooding depth is greater than 1m, the flooded area is marked with a red polygon and the population in the area is automatically associated: an area is marked with "Affected population: 5,000 people."

[0030] 2. Graded warning response: 2-1. Blue warning: risk probability 30% Triggering conditions: The wave height forecast for the next 24 hours is 2.5-3.0m, and the blue warning threshold for the sea area is 2.5-3.5m; Push content: SMS messages were sent to 200 monitoring personnel via SMS service, with the message: "[Ocean Warning] Blue Warning: Wave height in a certain sea area will be 2.8m in the next 24 hours, please pay attention to monitoring"; 2-2. Yellow warning: risk probability 50% Trigger condition: The wave height forecast for the next 24 hours is 3.0-3.5m, with yellow corresponding to the 3.0-3.5m range; Response measures: SMS messages will be sent to monitoring personnel once a day (same as blue alerts), and WeChat official accounts will be updated with graphic notifications, including wave forecast trend charts. Push content: Published via WeChat official account: "Yellow Warning: Wave height in a certain sea area will be 3.2m in the next 24 hours"; 2-3. Orange warning: risk probability 70% Trigger conditions: The tide level forecast for the next 24 hours is 3.5m, reaching the local warning tide level of 3.5m but not exceeding the red threshold of 3.8m; Response measures: Activate hourly emergency broadcasts: The local emergency broadcast system broadcasts for three minutes every hour from 6:00 AM to 10:00 PM daily; send a JSON file containing tide level data to emergency departments every hour, and automatically push an "Orange Warning Defense Notice" to coastal community leaders, including a list of key areas for levee inspections; 2-4. Red alert: risk probability 85% Trigger condition: The predicted tide level is 3.8m, exceeding the local warning tide level of 3.5m. A pre-verification report is automatically generated, including model output and historical similar case comparison. After manual confirmation by the on-duty staff, if an alert is generated at 10:05:12, it will be pushed to the smart terminal at 10:05:20 with a response delay of 8 seconds. At the same time, a JSON file containing the latest forecast data will be sent to the emergency department every 10 minutes.

[0031] 3. Channel verification: SMS arrival rate: According to operator receipt statistics, all 200 blue warning SMS messages were successfully delivered, with an average delivery time of 2.3 seconds; Emergency broadcast announcements: Connected to the local emergency broadcast system, the hourly broadcast recordings during the orange warning period can be verified by playback through the recording system. The time interval between the first broadcast and the warning generation is 7 seconds.

[0032] V. Comparison of Marine Disaster Risk Prevention and Control Early Warning Systems

[0033] 1. Summary: This invention surpasses traditional physical models, commercial systems, and existing regional solutions in terms of data processing efficiency, prediction accuracy, and response speed. It is particularly suitable for the increasingly frequent marine disasters. It can transform the "delayed response" of disaster warnings into "real-time prevention and control", and is expected to reduce the risk of casualties by more than 30% and reduce economic losses by over 100 million yuan per year.

[0034] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. The marine disaster risk prevention and control early warning system based on big data is characterized by: Including multi-source data fusion module, intelligent modeling module, and real-time warning module: The multi-source data fusion module is used to access multi-source real-time monitoring data such as satellite remote sensing and buoy monitoring, as well as historical disaster data, and to clean and feature-engineer the data before storing it in a distributed storage system. The intelligent modeling module builds a hybrid prediction model based on spatiotemporal sequence data. The hybrid prediction model combines a spatiotemporal Transformer network with physical correction formulas to predict disaster risk probability and key parameters. The module also has a closed-loop training mechanism that can automatically trigger incremental updates of model parameters based on post-disaster assessment data to continuously optimize prediction accuracy. The real-time warning module generates a dynamic risk map and graded warning information based on the prediction results, and pushes them to the terminal device.

2. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The multi-source data fusion module includes an edge computing node, which obtains 50Hz high-frequency sampling real-time monitoring raw data output by the device end, and performs multi-resolution decomposition on the raw data through a wavelet transform algorithm: performing abnormal filtering on the high-frequency components, eliminating noise and outliers that exceed the preset physical threshold, retaining low-frequency trend characteristics, and verifying the rationality of the low-frequency components in combination with the statistical laws of historical data; finally, generating a minute-level down-sampling data sequence that meets the needs of marine disaster warning.

3. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The distributed storage system uses the HBase engine to store time-series monitoring data such as wave height, wind speed, and tide level, and uses the Elasticsearch engine to store disaster report texts, sensor logs, and satellite remote sensing image metadata, and achieves efficient data management through the following technologies: Unified API Gateway: Provides a standardized data access interface and automatically routes query requests to the corresponding storage engine, supporting cross-database association queries based on time and regional dimensions; Index optimization strategy: Establishes a time and region composite index in HBase, and builds a text inverted index and a geographic spatial index in Elasticsearch to shorten the single-database query response time; Redis cache mechanism: Implements a cache strategy for high-frequency access data in the past 7 days, reducing the response time of repeated queries from an average of 200ms to ≤20ms; Through the collaboration of the above technologies, the cross-database query delay of the storage system is ≤500ms, meeting the real-time data call requirements of marine disaster warning.

4. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The hybrid prediction model combines the spatiotemporal Transformer network with a physical correction formula, wherein the spatiotemporal Transformer network encodes marine disaster-related features such as wind speed, air pressure, and tide level into a spatiotemporal matrix according to a 10-minute time step, captures cross-feature correlations through a multi-head attention mechanism, and outputs the initial value of the disaster risk probability and a time series feature vector; the physical correction formula is based on the basic equation of tidal theory, and uses the Coriolis force correction term and the seabed friction parameter correction term to calibrate the initial value of the disaster risk probability and key parameters such as tide level output by the spatiotemporal Transformer network; forms a disaster risk probability prediction value and a prediction result of key parameters such as tide level, wind speed, and wave height that include physical mechanism constraints, and realizes the prediction of disaster risk probability and key parameters in the next 24 hours.

5. The marine disaster risk prevention and control early warning system based on big data according to claim 4 is characterized in that: The physical correction formula is based on the basic equation of tidal theory: ； Where, Indicates the storm surge water value in meters, which is a key parameter that needs to be calibrated; is the seabed friction parameter correction term, where Respectively represent the horizontal flow velocity in the east-west direction and the north-south direction, unit: m / s, is the friction coefficient related to the seabed roughness; is the Coriolis force correction term, is the Coriolis parameter, which is approximately , which is related to geographical latitude; is the seawater velocity vector, The curl of the velocity vector is used to correct the influence of the Coriolis force caused by the Earth's rotation. The tide level prediction results output by the spatiotemporal Transformer network are calibrated by the above correction term to make the tide level prediction error ≤ 0.2 meters.

6. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The real-time warning module develops a dynamic risk map based on Leaflet.js, supports the overlay of coastal population density and important facility location data, and automatically identifies high-risk areas; the graded warning includes four levels: blue, yellow, orange, and red. The red warning requires pre-verification and manual secondary confirmation, and the warning response delay is less than 10 seconds.

7. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The closed-loop training mechanism includes: automatically synchronizing disaster assessment data from the distributed storage system after a disaster, wherein the assessment data includes post-disaster real-time monitoring data processed by a multi-source data fusion module and data on disasters of the same type in a historical case library, wherein the real-time monitoring data includes measured values ​​of wind speed, wave height, and tide level; when more than 95% of the effective feature dimensions of the assessment data are free of missing information and the number of available samples reaches 15% of the total training data, an incremental update of the parameters of the hybrid prediction model is triggered, and the update process performs gradient optimization based on the current parameters of the model.

8. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The warning information push channels of the real-time warning module include mobile communication networks, Internet platforms, emergency broadcast systems and smart terminal devices. Through the preset hierarchical strategy configuration table, blue warnings are pushed to monitoring personnel once a day by SMS, yellow warnings are added with WeChat public account graphic notifications, orange warnings start emergency broadcasts on the hour, and red warnings simultaneously trigger strong reminders on smart terminal APPs and are pushed to emergency departments and coastal community leaders every 10 minutes, achieving differentiated reach of different levels of warnings.

9. The marine disaster risk prevention and control early warning system based on big data according to claim 1 is characterized in that: The multi-source data fusion module, intelligent modeling module and real-time warning module are deployed using a unified technical architecture and achieve automated connection through the following methods: Standardized data interface: data is transmitted between modules through predefined data formats and interactive protocols; Task scheduling mechanism: Based on the workflow engine, the task chain is automatically arranged in the order of data processing, model training, and warning generation; Cross-module linkage: Early warning results trigger model incremental training requests, forming a closed-loop feedback loop.

10. A marine disaster risk prevention and control early warning method based on big data, characterized in that: The following steps are involved: S1. Multi-source data fusion processing: This approach integrates real-time monitoring data from multiple sources, such as satellite remote sensing and buoy monitoring, as well as historical disaster data. Edge computing nodes use a wavelet transform algorithm to downsample and filter out anomalies on the 50Hz high-frequency data at the minute level. Data cleaning is then performed to repair missing values, remove outliers, and perform feature engineering to generate time series data that meets early warning requirements. S2. Hybrid model prediction calculation: The processed data is input into the hybrid prediction model. The hybrid prediction model encodes marine disaster-related features such as wind speed, air pressure, and tide level into a spatiotemporal matrix at a 10-minute time step through a spatiotemporal Transformer network. The multi-head attention mechanism is used to extract spatiotemporal correlation features and output the initial value of the disaster risk probability and the time series feature vector. Based on the basic equations of tidal theory, the prediction results are physically calibrated using the Coriolis force correction term and the seabed friction parameter correction term to generate the predicted values ​​of the disaster risk probability and key parameters such as tide level, wind speed, and wave height for the next 24 hours. S3. Dynamic warning and information push: Based on the forecast results, a dynamic risk map developed with Leaflet.js is overlaid with coastal population density and key facility location data to automatically identify high-risk areas. Alerts are triggered based on four preset thresholds: blue, yellow, orange, and red. Red alerts require pre-verification and manual secondary confirmation. Differentiated warning information is pushed through multiple channels, including mobile communication networks and emergency broadcasts, to ensure response delays are less than 10 seconds. S4. Model closed-loop training mechanism: After a disaster, disaster assessment data is automatically synchronized from the distributed storage system. When the effective feature dimensions of the assessment data exceed 95% without missing and the number of available samples reaches 15% of the total training data, the parameters of the hybrid prediction model are incrementally updated. The update process performs gradient optimization based on the current parameters of the model.

Images