Coastal channel operation state real-time monitoring and early warning method based on digital twinning

Abstract

This invention relates to the field of intelligent waterway information technology, specifically a method for real-time monitoring and early warning of coastal waterway operation status based on digital twins. It establishes a digital twin architecture and integrates multi-source operational data to achieve real-time synchronous mapping from physical waterways to virtual waterways; it introduces a multi-dimensional mechanism model and its historical mechanism functions to drive an AI model for function mapping, constructing a dynamic siltation prediction model to obtain navigation risk indicators for future waterways; based on the speed and heading data of nearby vessels and the vessel itself, it constructs a spatiotemporal occupancy matrix and a spatiotemporal safety envelope model to predict and assess future vessel collision risks and generate navigation safety indicators; it uses a marginal dependency distribution model to couple and assess navigation risks, forming a comprehensive risk and triggering dynamic early warnings; and it disseminates early warning information and assists in decision-making through shipborne terminals and the VTS center. This invention constructs a complete ship-shore-cloud collaborative digital twin system, realizing full-domain digital and intelligent management of coastal waterways.

Description

Technical Field

[0001] This invention relates to the field of intelligent waterway information technology, and in particular to a method for real-time monitoring and early warning of the operational status of coastal waterways based on digital twins. Background Technology

[0002] Currently, coastal waterway management primarily employs decentralized and independent technical solutions, which have significant limitations. Firstly, the widespread use of decentralized monitoring systems presents challenges. Vessel Traffic Management Systems (VTS) often operate independently, focusing solely on vessel dynamics; hydrological and meteorological monitoring systems are deployed separately, mainly providing basic environmental data; and waterway facility management systems operate independently, responsible for maintaining infrastructure such as navigation aids. These systems lack effective data exchange mechanisms, resulting in severe "information silos." Secondly, traditional early warning methods are inadequate. Most existing early warning systems rely on threshold warnings from a single data source, such as issuing alerts based solely on wind speed or visibility data. This method heavily depends on human experience and judgment, requiring on-duty personnel to simultaneously monitor multiple system interfaces and make comprehensive assessments. This leads to delayed early warning responses, limited accuracy, and a complete lack of capability for coupled analysis of multiple risks. This current technological state cannot meet the safety, efficiency, and economic requirements of modern coastal waterways.

[0003] First, there is a bottleneck in data fusion. Existing systems cannot effectively integrate heterogeneous information from multiple sources, such as ship dynamics data, hydrological and meteorological data, and waterway status data, and lack a unified data governance and fusion mechanism. Second, there is a bottleneck in early warning capabilities. Existing early warning methods are mostly reactive, lacking predictive analysis capabilities for waterway operation status and thus failing to provide early warnings. Third, there is a bottleneck in collaborative decision-making. There is a lack of effective collaboration mechanisms between various aspects such as waterway management, ship navigation, and port scheduling, resulting in isolated decision-making processes. Finally, there is a bottleneck in model accuracy. Existing predictive models rely on traditional mechanistic models and typically use single data-driven models, resulting in inadequate prediction accuracy and reliability. These bottlenecks severely restrict the improvement of coastal waterway management. Summary of the Invention

[0004] This invention overcomes the shortcomings of existing technologies and provides a method for real-time monitoring and early warning of the operational status of coastal waterways based on digital twins.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] The first aspect of this invention provides a method for real-time monitoring and early warning of the operational status of coastal waterways based on digital twins, comprising the following steps:

[0007] S102: Construct a digital twin architecture, collect multi-source operational data of the coastal waterway in real time, and use the attention mechanism backbone network to fuse the multi-source operational data across sources. Then, synchronously map the data from the source data domain to the target data domain of the digital twin architecture to drive updates and generate a multi-source data fusion digital twin of the coastal waterway.

[0008] S104: Based on the past mechanism functions of the multi-dimensional mechanism model corresponding to the coastal waterway, the AI ​​model is driven to perform function mapping using high-dimensional feature input data. A dynamic prediction model for siltation is coupled and constructed to simulate the overall trend of sediment transport. The navigation risk indicators of the coastal waterway in the future period are obtained through the dynamic prediction model for siltation.

[0009] S106: Discretize the sea area plane environment space based on the speed and heading data of nearby and current vessels, and infer and update the sea area plane environment space based on navigation observation data to obtain the spatiotemporal occupancy matrix. Construct the spatiotemporal safety envelope model of the current vessel, and combine the spatiotemporal occupancy matrix and the spatiotemporal safety envelope model to predict and judge whether there is a collision risk between nearby and current vessels in the future period, and output the vessel navigation safety index.

[0010] S108: Quantitatively construct a marginal dependency distribution model of relevant risk characteristics, and based on the marginal dependency distribution model, perform superimposed coupling of waterway navigation risk indicators and ship navigation safety indicators in a multi-source data fusion digital twin to assess risk, and execute dynamic early warning actions for coastal waterways based on the comprehensive risk determination.

[0011] S110: Through the shipborne intelligent terminal, push early warning information and suggested routes to relevant vessels, and push early warning information and auxiliary decision-making solutions to the VTS center.

[0012] Preferably, step S102 specifically includes the following steps:

[0013] Acquire various historical observation data of the coastal waterway within a preset time period, and pre-construct a digital twin architecture of the coastal waterway based on the various historical observation data; wherein, the historical observation data includes survey data, remote sensing imagery, and UAV aerial photography data;

[0014] Multi-source operational data of the coastal waterway is collected in real time through a waterway IoT array and defined as the first multi-source operational data domain. At the same time, multi-source operational data that drives the dynamic updates of the digital twin architecture at the termination time frame is extracted and defined as the second multi-source operational data domain.

[0015] High-dimensional comparative features of the data are extracted. Based on the high-dimensional comparative features, the second-order statistical structure of feature correlation is calculated for the first multi-source operating data domain and the second multi-source operating data domain to obtain the covariance matrix of the first multi-source operating data domain, which is set as the source domain covariance matrix; and the covariance matrix of the second multi-source operating data domain, which is set as the target domain covariance matrix.

[0016] A redundant correlation algorithm is introduced to construct an uncorrelated whitening space. In the uncorrelated whitening space, the internal correlation redundancy of the source domain covariance matrix is ​​removed to generate whitened source domain features.

[0017] Add a structured identifier code to each multi-source operational data in the first multi-source operational data domain, and then concatenate the source domain features of all multi-source operational data into a cross-source fusion sequence based on the structured identifier code.

[0018] Construct an attention mechanism backbone network, feed the cross-source fusion sequence into the attention mechanism backbone network to learn the cross-source relationship interaction between features of each source domain, establish the cross-dependency of cross-source attention fusion, and output multi-source fusion encoding;

[0019] The target domain covariance space is established based on the statistical structure of the target domain covariance matrix. After redundancy removal, the whitened source domain features are matched and projected onto the target domain covariance space based on multi-source fusion coding. In this way, the digital twin architecture is updated through cross-source synchronization of real-time collected multi-source operation data, generating a multi-source data fusion digital twin of the coastal waterway.

[0020] Preferably, step S104 specifically includes the following steps:

[0021] Obtain the construction assessment report of the coastal waterway, and retrieve the waterway mechanism assessment parameters from the construction assessment report through big data network to obtain the multi-dimensional mechanism model and empirical mechanism case of the coastal waterway; among which, the multi-dimensional mechanism model includes sediment mechanism, hydrodynamic mechanism and wave mechanism;

[0022] By extracting past mechanism functions and applied past mechanism equations of coastal waterways in the multidimensional mechanism model simulation of sediment transport within a preset time period through empirical mechanism cases, a data-driven AI model based on LSTM architecture is obtained, and a high-dimensional convolutional layer is constructed through the data-driven AI model.

[0023] A high-dimensional feature space is constructed by high-dimensional convolutional layers. The discrete past mechanism functions are embedded into the high-dimensional feature space and nonlinearly vectorized to obtain the high-dimensional feature vector points where the past mechanism functions are represented in the data-driven AI model.

[0024] An integral operator is introduced and used in the multi-source data fusion digital twin to learn the dependency relationship between the channel mechanism input and output of each high-dimensional feature vector point, forming a global integral layer;

[0025] Repeat the learning steps of the above integral operator to traverse all high-dimensional feature vector points, output multiple global integral layers and perform layer stacking processing to obtain a function mapping deep network that reflects the overall trend of sediment transport on the coastal waterway in historical periods.

[0026] Based on past mechanistic equations, a function space for a data-driven AI model is created. Each high-dimensional feature vector point is projected back into the function space for solution through a deep network of function mapping. In this way, the high-dimensional feature vector is restored to a coupled architecture of multi-dimensional mechanistic model and data-driven AI model for observing the dynamics of sediment transport in coastal waterways, and a siltation dynamic prediction model is generated.

[0027] Obtain sediment transport parameters for future time periods, import these parameters into a dynamic prediction model for siltation, and use them for rolling predictions to obtain navigation risk indicators for the siltation location and amount in the coastal waterway during future time periods.

[0028] Preferably, step S106 specifically includes the following steps:

[0029] The speed and heading of nearby vessels relative to the current vessel in a future time period are obtained and defined as relative speed and relative heading. The navigation drive step size is set based on the relative speed, relative heading, predetermined speed and predetermined heading.

[0030] Obtain a remote sensing geographic map of the coastal waterway, construct the sea area planar environmental space of the coastal waterway based on the remote sensing geographic map, and discretize the sea area planar environmental space into M sub-sea area environmental grids based on the navigation drive step size.

[0031] The navigation observation data and observation time sequence strategy between the current ship and nearby ships are obtained by monitoring the relative position of the ship through the relative pose sensor. Based on the observation time sequence strategy, the future time period is divided into N consecutive observation time steps at equal intervals.

[0032] A Bayesian inference network is introduced. Based on navigation observation data, the Bayesian inference network is used at each consecutive observation time step to infer and update the occupancy probability of whether nearby ships enter or pass through the sub-ocean environment grid. The occupancy probability is stacked in a time-series chain according to the observation time step to obtain the spatiotemporal occupancy matrix.

[0033] The spatiotemporal navigation trajectory point set of the ship is obtained. Based on the ship's external dimensions, steering performance index, predetermined speed, predetermined course and international maritime collision avoidance rules, the spatial convex hull planning of the spatiotemporal navigation trajectory point set for future time periods is constructed to generate the spatiotemporal safety envelope model of the current ship.

[0034] The spatiotemporal occupancy matrix is ​​used to obtain the sub-sea area environmental grids occupied by nearby ships in the future time period, which are marked as occupied sea area environmental grids. If the occupied sea area environmental grid is located within the spatiotemporal safety envelope model, it is considered that the current ship has a collision risk, and the ship navigation safety index is output.

[0035] Preferably, the acquisition of the ship's spatiotemporal navigation trajectory point set involves constructing a spatial convex hull for future time periods based on the ship's external dimensional parameters, steering performance index, predetermined speed, predetermined course, and international maritime collision avoidance rules, thereby generating a spatiotemporal safety envelope model for the current ship. This process specifically includes the following steps:

[0036] Obtain the ship's model information, and retrieve the ship's external dimensions, steering performance index, and the international maritime collision avoidance rules it follows based on big data network retrieval of model information;

[0037] The ship's navigation memorandum is used to obtain the planned speed and course for the future time period. At the same time, the spatiotemporal navigation trajectory point set of the ship in the coastal waterway area in a continuous time sequence is extracted from the coastal navigation log.

[0038] Obtain a remote sensing geographic map of the coastal waterway, construct a spatial trajectory coordinate domain of the coastal waterway on the remote sensing geographic map, obtain the coordinates of the current ship in the spatial trajectory coordinate domain, and mark it as a predetermined trajectory coordinate point;

[0039] Using predetermined trajectory coordinates as the reference center of the dynamic safety planning, and with the help of relative pose measurement tools, the spatial polar angle value of each spatiotemporal trajectory point relative to the reference center is calculated based on external dimensions, predetermined speed, predetermined course, turning performance index, and international maritime collision avoidance rules.

[0040] Construct an empty convex hull stack, sort the navigation trajectory points of each sea area in descending order according to the spatial polar angle value, generate a spatiotemporal navigation trajectory point sorting table, push each spatiotemporal navigation trajectory point into the empty convex hull stack according to the spatiotemporal navigation trajectory point sorting table, and then obtain the topmost stack vertex of the empty convex hull stack and the nearest adjacent stack vertex to the stack vertex.

[0041] If the direction between the stack vertex and its adjacent stack vertex is consistent, then the boundary of the empty stack of the convex hull is planned and maintained based on the spatial polar angle value corresponding to the stack vertex and its adjacent stack vertex, and finally the dynamic trajectory convex hull is output. The spatiotemporal safety envelope model of the current ship is constructed based on the dynamic trajectory convex hull.

[0042] Preferably, step S108 specifically includes the following steps:

[0043] Based on big data network retrieval, a risk assessment system for coastal waterways and ship navigation is obtained, and a normal distribution model for marginal risk assessment is identified through the risk assessment system.

[0044] By fitting the distribution parameters of the waterway navigation risk index and the ship navigation safety index to the prior probability distribution of the normal distribution model, the marginal distribution function of the waterway navigation risk index is obtained, which is defined as the first marginal distribution function; and the marginal distribution function of the ship navigation safety index is defined as the second marginal distribution function.

[0045] By mapping the waterway navigation risk index and the ship navigation safety index to the probability integral transformation interval [0, 1] through the first marginal distribution function and the second marginal distribution function, a marginal dependency distribution model is generated.

[0046] Based on the uniform distribution pattern of the marginal dependency distribution model, the relevant risk characteristics between the waterway navigation risk indicators and the ship navigation safety indicators are determined, and the connection function of risk coupling is preset based on the relevant risk characteristics.

[0047] A maximum likelihood estimation algorithm is introduced to decouple and estimate the linkage distribution characteristics between waterway navigation risk indicators and ship navigation safety indicators based on the linkage function, resulting in a series of linkage vector parameters; wherein, the linkage vector parameters include the risk coupling strength and direction between waterway navigation risk indicators and ship navigation safety indicators;

[0048] By using a series of connection vector functions located in the big data network, multiple joint random samples of the comprehensive navigation risk assessment of waterway navigation risk indicators and ship navigation safety indicators are retrieved and constructed. The joint random samples are simulated on the original margin to generate a comprehensive risk map. Based on the comprehensive risk map, the comprehensive risk value of the coastal waterway in the future time period is determined.

[0049] If the overall risk value exceeds the preset risk threshold, a dynamic early warning instruction is generated and uploaded to the early warning terminal to execute dynamic early warning actions for the coastal waterway.

[0050] A second aspect of this invention provides a real-time monitoring and early warning system for the operational status of coastal waterways based on digital twins, applicable to any of the methods for real-time monitoring and early warning of the operational status of coastal waterways based on digital twins, wherein the system specifically includes:

[0051] The ship-side sensing module includes an intelligent navigation terminal, an AIS automatic identification module, and a shipborne navigation data recorder, used to collect ship position, speed, and heading data.

[0052] The shore-based sensing module includes a shore-based AIS base station, underwater radar, CCTV video monitoring module, hydrological and meteorological station, intelligent navigation beacon and underwater topography monitoring module, which is responsible for measuring tide level, current velocity, wind direction and wind speed, and also has a status self-diagnosis function.

[0053] Digital Twin Engine: The digital twin engine has the functions of data-driven operation, model calculation, real-time rendering and simulation, and is responsible for driving the operation of the digital twin.

[0054] Channel operation monitoring and early warning module: The channel operation monitoring and early warning module is based on a comprehensive risk map provided by a digital twin platform, which realizes real-time monitoring and multi-level early warning of channel operation status;

[0055] Channel siltation analysis module: The channel siltation analysis module is a coupled channel siltation model based on a digital twin platform, used for channel siltation prediction and siltation volume analysis and calculation;

[0056] Intelligent Navigation Assistance Module for Ships: The intelligent navigation assistance module for ships has functions such as recommended routes, speed suggestions, and hazard warnings, which are used to provide personalized navigation services for ships.

[0057] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows:

[0058] A digital twin architecture is constructed, and multi-source operational data of the coastal waterway is collected in real time. After cross-source fusion of the multi-source operational data using an attention mechanism backbone network, the data is synchronously mapped from the source data domain to the target data domain of the digital twin architecture to drive updates, generating a multi-source data fusion digital twin of the coastal waterway. The past mechanism functions of the corresponding multi-dimensional mechanism model of the coastal waterway are used as high-dimensional feature input data to drive the AI ​​model for function mapping, coupled to construct a dynamic prediction model for siltation that simulates the overall trend of sediment transport. The siltation dynamic prediction model is used to obtain navigation risk indicators for the coastal waterway in future periods. Based on the speed data of nearby and current vessels, and the navigation... This invention discretizes the marine planar environmental space into data, and updates the marine planar environmental space based on navigation observation data to obtain a spatiotemporal occupancy matrix. A spatiotemporal safety envelope model of the current vessel is constructed. Combining the spatiotemporal occupancy matrix and the spatiotemporal safety envelope model, the invention predicts and analyzes whether there is a collision risk between nearby vessels and the current vessel in the future, outputting vessel navigation safety indicators. A marginal dependency distribution model of relevant risk characteristics is quantified and constructed. Based on this marginal dependency distribution model, a risk assessment is superimposed and coupled between waterway navigation risk indicators and vessel navigation safety indicators in a multi-source data fusion digital twin. Dynamic early warning actions for coastal waterways are executed based on the comprehensive risk determination. This invention constructs a complete ship-shore-cloud collaborative digital twin system, coupling a mechanistic model and a data-driven model for waterway siltation prediction, improving the accuracy of operational risk prediction for coastal waterways. Meanwhile, this invention also utilizes the spatiotemporal safety envelope to assess ship collision risks, enabling more timely and accurate detection of collision risks. It achieves multi-risk coupling analysis and dynamic collaborative early warning, improving the overall digital and intelligent management level of coastal waterways and providing closed-loop decision support for both ship and shore ends. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0060] Figure 1 A flowchart of the first method for real-time monitoring and early warning of coastal waterway operation status based on digital twin is shown;

[0061] Figure 2 A flowchart of the second method for real-time monitoring and early warning of coastal waterway operation status based on digital twin is shown;

[0062] Figure 3 A system framework diagram of a real-time monitoring and early warning system for the operation status of coastal waterways based on digital twins is shown. Detailed Implementation

[0063] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0064] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0065] The first aspect of this invention provides a method for real-time monitoring and early warning of the operational status of coastal waterways based on digital twins, such as... Figure 1 As shown, it includes the following steps:

[0066] S102: Construct a digital twin architecture, collect multi-source operational data of the coastal waterway in real time, and use the attention mechanism backbone network to fuse the multi-source operational data across sources. Then, synchronously map the data from the source data domain to the target data domain of the digital twin architecture to drive updates and generate a multi-source data fusion digital twin of the coastal waterway.

[0067] S104: Based on the past mechanism functions of the multi-dimensional mechanism model corresponding to the coastal waterway, the AI ​​model is driven to perform function mapping using high-dimensional feature input data. A dynamic prediction model for siltation is coupled and constructed to simulate the overall trend of sediment transport. The navigation risk indicators of the coastal waterway in the future period are obtained through the dynamic prediction model for siltation.

[0068] S106: Discretize the sea area plane environment space based on the speed and heading data of nearby and current vessels, and infer and update the sea area plane environment space based on navigation observation data to obtain the spatiotemporal occupancy matrix. Construct the spatiotemporal safety envelope model of the current vessel, and combine the spatiotemporal occupancy matrix and the spatiotemporal safety envelope model to predict and judge whether there is a collision risk between nearby and current vessels in the future period, and output the vessel navigation safety index.

[0069] S108: Quantitatively construct a marginal dependency distribution model of relevant risk characteristics, and based on the marginal dependency distribution model, perform superimposed coupling of waterway navigation risk indicators and ship navigation safety indicators in a multi-source data fusion digital twin to assess risk, and execute dynamic early warning actions for coastal waterways based on the comprehensive risk determination.

[0070] S110: Through the shipborne intelligent terminal, push early warning information and suggested routes to relevant vessels, and push early warning information and auxiliary decision-making solutions to the VTS center.

[0071] Preferably, step S102 specifically includes the following steps:

[0072] Acquire various historical observation data of the coastal waterway within a preset time period, and pre-construct a digital twin architecture of the coastal waterway based on the various historical observation data; wherein, the historical observation data includes survey data, remote sensing imagery, and UAV aerial photography data;

[0073] Multi-source operational data of the coastal waterway is collected in real time through a waterway IoT array and defined as the first multi-source operational data domain. At the same time, multi-source operational data that drives the dynamic updates of the digital twin architecture at the termination time frame is extracted and defined as the second multi-source operational data domain.

[0074] High-dimensional comparative features of the data are extracted. Based on the high-dimensional comparative features, the second-order statistical structure of feature correlation is calculated for the first multi-source operating data domain and the second multi-source operating data domain to obtain the covariance matrix of the first multi-source operating data domain, which is set as the source domain covariance matrix; and the covariance matrix of the second multi-source operating data domain, which is set as the target domain covariance matrix.

[0075] A redundant correlation algorithm is introduced to construct an uncorrelated whitening space. In the uncorrelated whitening space, the internal correlation redundancy of the source domain covariance matrix is ​​removed to generate whitened source domain features.

[0076] Add a structured identifier code to each multi-source operational data in the first multi-source operational data domain, and then concatenate the source domain features of all multi-source operational data into a cross-source fusion sequence based on the structured identifier code.

[0077] Construct an attention mechanism backbone network, feed the cross-source fusion sequence into the attention mechanism backbone network to learn the cross-source relationship interaction between features of each source domain, establish the cross-dependency of cross-source attention fusion, and output multi-source fusion encoding;

[0078] The target domain covariance space is established based on the statistical structure of the target domain covariance matrix. After redundancy removal, the whitened source domain features are matched and projected onto the target domain covariance space based on multi-source fusion coding. In this way, the digital twin architecture is updated through cross-source synchronization of real-time collected multi-source operation data, generating a multi-source data fusion digital twin of the coastal waterway.

[0079] It should be noted that multi-source operational data includes ship AIS data, hydro-meteorological data, underwater topographic data, and video data. By extracting high-dimensional comparative features from the data, the feature vectors of the multi-source operational data from the real-time acquired source domain to the target domain in the digital twin architecture can maintain a high degree of consistency, ensuring the covariance alignment of the physical waterway synchronously mapped to the virtual waterway. The covariance matrix is ​​the overall alignment of the multi-source operational data from the real-time acquisition at the physical level to the rendering at the virtual level within the digital twin architecture; it represents the second-order statistical structure between the source and target domains, quantifying the correlation and offset of local data feature mappings. Subsequently, the source domain is mapped to a neutral statistical space, i.e., an uncorrelated whitening space, to blank out the feature interior of the source domain covariance matrix. This completely eliminates the correlation of the multi-source operational data source domain features, transforming the source domain covariance into an identity matrix. By eliminating the inherent correlation structure of the source domain, a clean and unbiased starting point is created for the subsequent remapping of its covariance structure to the target domain, ensuring the accuracy of the real-time multi-source operational data fusion and digital twin mapping. Next, a structured identifier code is added to each data point in the first multi-source operational data domain. This enables the attention mechanism to effectively distinguish the source and modality of cross-source data, allowing the attention mechanism backbone network to correctly receive and understand multi-source inputs and avoid subsequent multi-source fusion failures due to data incompatibility. The source domain features of all multi-source operational data are sequentially concatenated into a cross-source fusion sequence, ensuring that the operational data from all sources are immediately organized into the same attention level at the virtual input end of the digital twin, thereby establishing interactive relationships between different data sources. This sequence is then fed into the attention mechanism backbone network, where it establishes cross-dependencies among all data sources, models fine-grained cross-source feature interactions, directly learns cross-source relationships between different data sources, and outputs a multi-source fusion code that simultaneously contains a global representation of multi-source information. Finally, the whitened source domain is mapped to a space with the covariance structure of the target domain, enabling the source domain features to match the target domain statistically. At this point, the second-order statistical structure of the source domain features is actively adjusted to ensure data-driven consistency between the target domain and the target domain in terms of correlation patterns. This reduces the local distribution differences between physical and virtual waterways, effectively improving the generalization ability of the digital twin after cross-domain data fusion. This method can drive and update the digital twin of coastal waterways based on multi-source ship operation data, achieving synchronous mapping between physical and virtual waterways. This makes the real-time snapshots of coastal waterways in the digital twin more realistic and accurate, improving the reliability of waterway operation status monitoring and avoiding information silos in coastal waterways.

[0080] Preferably, step S104 specifically includes the following steps:

[0081] Obtain the construction assessment report of the coastal waterway, and retrieve the waterway mechanism assessment parameters from the construction assessment report through big data network to obtain the multi-dimensional mechanism model and empirical mechanism case of the coastal waterway; among which, the multi-dimensional mechanism model includes sediment mechanism, hydrodynamic mechanism and wave mechanism;

[0082] By extracting past mechanism functions and applied past mechanism equations of coastal waterways in the multidimensional mechanism model simulation of sediment transport within a preset time period through empirical mechanism cases, a data-driven AI model based on LSTM architecture is obtained, and a high-dimensional convolutional layer is constructed through the data-driven AI model.

[0083] A high-dimensional feature space is constructed by high-dimensional convolutional layers. The discrete past mechanism functions are embedded into the high-dimensional feature space and nonlinearly vectorized to obtain the high-dimensional feature vector points where the past mechanism functions are represented in the data-driven AI model.

[0084] An integral operator is introduced and used in the multi-source data fusion digital twin to learn the dependency relationship between the channel mechanism input and output of each high-dimensional feature vector point, forming a global integral layer;

[0085] Repeat the learning steps of the above integral operator to traverse all high-dimensional feature vector points, output multiple global integral layers and perform layer stacking processing to obtain a function mapping deep network that reflects the overall trend of sediment transport on the coastal waterway in historical periods.

[0086] Based on past mechanistic equations, a function space for a data-driven AI model is created. Each high-dimensional feature vector point is projected back into the function space for solution through a deep network of function mapping. In this way, the high-dimensional feature vector is restored to a coupled architecture of multi-dimensional mechanistic model and data-driven AI model for observing the dynamics of sediment transport in coastal waterways, and a siltation dynamic prediction model is generated.

[0087] Obtain sediment transport parameters for future time periods, import these parameters into a dynamic prediction model for siltation, and use them for rolling predictions to obtain navigation risk indicators for the siltation location and amount in the coastal waterway during future time periods.

[0088] It should be noted that by constructing a high-dimensional feature space for high-dimensional convolutional layers through a data-driven AI model, this high-dimensional feature space can be used to map mechanistic functions into high-dimensional feature vectors. This allows for the capture of the nonlinear relationships in the siltation mechanism of sediment transport in coastal waterways. By abstracting and quantifying nonlinear high-dimensional feature vector points, the clarity and expressive power of the local waterway mechanism can be significantly improved. Then, an integral operator is used to learn the nonlocal relationships between any high-dimensional feature vector point and other arbitrary high-dimensional feature vector points, thereby obtaining the global dependencies between functions and achieving a global propagation effect. This allows the data-driven AI model based on the LSTM architecture to reflect the overall siltation behavior of sediment transport in the waterway mechanism model at high resolution, rather than the local neighborhood capture of traditional CNN methods. This further improves the accuracy and reliability of dynamic prediction of waterway siltation and reduces the noise bias of dynamic prediction. By stacking multiple global integral layers, a deep network of function mappings representing the overall trend of sediment transport is formed on the data-driven AI model. This allows for the capture and extraction of function patterns of different scales and complexities through multi-layered high-dimensional feature abstraction of waterway mechanisms. This approach explores and simulates the global coupling within the physical waterway mechanism system, enhancing the interpretability and generalization ability of the data-driven AI model for waterway mechanisms and sediment backfilling. This method enables the input of function calculation results from the waterway mechanism model (e.g., flow velocity or sediment concentration function) as high-dimensional mechanism features into the data-driven model within a digital twin. This couples the waterway mechanism model with the LSTM-based data-driven AI model, allowing for dynamic and rolling prediction of backfilling locations and volumes in coastal waterways during specific future periods. This provides reliable and accurate navigation risk indicators for subsequent operational status monitoring and risk assessment, effectively overcoming the limitations of single-mechanism models relying on assumptions or single-data models relying on data, significantly improving dynamic prediction accuracy.

[0089] Preferably, S106, as Figure 2 As shown, the specific steps include:

[0090] The speed and heading of nearby vessels relative to the current vessel in a future time period are obtained and defined as relative speed and relative heading. The navigation drive step size is set based on the relative speed, relative heading, predetermined speed and predetermined heading.

[0091] Obtain a remote sensing geographic map of the coastal waterway, construct the sea area planar environmental space of the coastal waterway based on the remote sensing geographic map, and discretize the sea area planar environmental space into M sub-sea area environmental grids based on the navigation drive step size.

[0092] The navigation observation data and observation time sequence strategy between the current ship and nearby ships are obtained by monitoring the relative position of the ship through the relative pose sensor. Based on the observation time sequence strategy, the future time period is divided into N consecutive observation time steps at equal intervals.

[0093] A Bayesian inference network is introduced. Based on navigation observation data, the Bayesian inference network is used at each consecutive observation time step to infer and update the occupancy probability of whether nearby ships enter or pass through the sub-ocean environment grid. The occupancy probability is stacked in a time-series chain according to the observation time step to obtain the spatiotemporal occupancy matrix.

[0094] The spatiotemporal navigation trajectory point set of the ship is obtained. Based on the ship's external dimensions, steering performance index, predetermined speed, predetermined course and international maritime collision avoidance rules, the spatial convex hull planning of the spatiotemporal navigation trajectory point set for future time periods is constructed to generate the spatiotemporal safety envelope model of the current ship.

[0095] The spatiotemporal occupancy matrix is ​​used to obtain the sub-sea area environmental grids occupied by nearby ships in the future time period, which are marked as occupied sea area environmental grids. If the occupied sea area environmental grid is located within the spatiotemporal safety envelope model, it is considered that the current ship has a collision risk, and the ship navigation safety index is output.

[0096] It should be noted that since the current navigation of ships and nearby ships involves regional dynamic transfers, this method reveals the transfer intervals of ships reaching different local areas at different speeds and directions by setting navigation drive step lengths adapted to relative speed, relative heading, predetermined speed, and predetermined heading. Furthermore, based on the navigation drive step lengths, the entire coastal channel area is discretized into local area blocks that accept different trajectory migrations, i.e., sub-sea area environmental grids. The observation timing strategy is a monitoring response (or triggering) timing sequence for the relative movement distance and trajectory between ships, preset by the relative pose sensor or adjusted according to the navigation task. The observation time steps defined by this strategy define the calculation window in the observation timing sequence, forming a time-level occupancy grid sequence. This introduces a time dimension, improving the fine-grained inference of observation data by the Bayesian inference network. Moreover, this method stacks occupancy probabilities in a time-series chain according to the observation time steps, enabling the environmental grid to depict the spatiotemporal linearity from current occupancy to potential future occupancy, ensuring more accurate capture and dynamic description of encounter situations (encounter, overtaking, or intersection) between ships in both time and space dimensions. If the occupied marine environment grid is located within the spatiotemporal safety envelope model, it indicates that the location of a nearby vessel, after reaching a certain position in the future with relative speed and heading, is within the safe collision avoidance range of the current vessel. This suggests that a collision between the nearby vessel and the current vessel is possible in the future, thus representing a high collision risk. Conversely, if the grid is outside the current vessel's range, it represents a low collision risk. This method, based on real-time vessel AIS data, uses a pre-defined spatiotemporal safety envelope model in a digital twin to calculate the encounter situation between the current vessel and nearby vessels in real time, and predicts the collision navigation risk in the future. It fully considers the relative distance between vessels and introduces the time dimension, enabling a forward-looking assessment of collision avoidance risks and significantly improving the accuracy of dynamic early warnings of navigation anomalies and dangers in coastal waterways.

[0097] Preferably, the acquisition of the ship's spatiotemporal navigation trajectory point set involves constructing a spatial convex hull for future time periods based on the ship's external dimensional parameters, steering performance index, predetermined speed, predetermined course, and international maritime collision avoidance rules, thereby generating a spatiotemporal safety envelope model for the current ship. This process specifically includes the following steps:

[0098] Obtain the ship's model information, and retrieve the ship's external dimensions, steering performance index, and the international maritime collision avoidance rules it follows based on big data network retrieval of model information;

[0099] The ship's navigation memorandum is used to obtain the planned speed and course for the future time period. At the same time, the spatiotemporal navigation trajectory point set of the ship in the coastal waterway area in a continuous time sequence is extracted from the coastal navigation log.

[0100] Obtain a remote sensing geographic map of the coastal waterway, construct a spatial trajectory coordinate domain of the coastal waterway on the remote sensing geographic map, obtain the coordinates of the current ship in the spatial trajectory coordinate domain, and mark it as a predetermined trajectory coordinate point;

[0101] Using predetermined trajectory coordinates as the reference center of the dynamic safety planning, and with the help of relative pose measurement tools, the spatial polar angle value of each spatiotemporal trajectory point relative to the reference center is calculated based on external dimensions, predetermined speed, predetermined course, turning performance index, and international maritime collision avoidance rules.

[0102] Construct an empty convex hull stack, sort the navigation trajectory points of each sea area in descending order according to the spatial polar angle value, generate a spatiotemporal navigation trajectory point sorting table, push each spatiotemporal navigation trajectory point into the empty convex hull stack according to the spatiotemporal navigation trajectory point sorting table, and then obtain the topmost stack vertex of the empty convex hull stack and the nearest adjacent stack vertex to the stack vertex.

[0103] If the direction between the stack vertex and its adjacent stack vertex is consistent, then the boundary of the empty stack of the convex hull is planned and maintained based on the spatial polar angle value corresponding to the stack vertex and its adjacent stack vertex, and finally the dynamic trajectory convex hull is output. The spatiotemporal safety envelope model of the current ship is constructed based on the dynamic trajectory convex hull.

[0104] It should be noted that since collision avoidance perception typically covers a safe area extending outward from the ship as the center, the current spatial trajectory coordinates of the ship must be anchored as the envelope reference point, i.e., the reference envelope reference center. This provides a source point coordinate reference for the relative trajectory of the ship's dynamic navigation in the future, ensuring that the planning of the spatiotemporal safety envelope interval is adaptively updated with the ship's dynamic navigation. The spatial polar angle integrates speed, heading, and collision avoidance rules, giving the convex hull structure a spatiotemporal safety geometric meaning rather than a simple geometric shape. Next, the trajectory points in each sea area are sorted in descending order based on the spatial polar angle values, creating a sequence of trajectory points from the outside in. This descending order considers the ship's maximum safe collision avoidance limit. Furthermore, the convex hull stack is used to retain only the outermost trajectory boundary points, preventing internal trajectory points from interfering with the overall convex hull structure, thus achieving the initial extraction of the spatiotemporal trajectory's spatial envelope shape. In this method, the stack vertices and adjacent stack vertices jointly determine the rationality of the expansion of the safety envelope convex hull, ensuring that the convex hull boundary always expands adaptively based on data such as speed, heading, and collision avoidance rules, thus preventing erroneous envelopes and reducing planning deviations in the safety envelope. If the directions between the stack vertices and adjacent stack vertices are consistent, it indicates that the envelope is the smallest convex boundary enclosing all trajectory points. Therefore, the boundary of the empty stack of the convex hull is planned and maintained based on the spatial polar angle values ​​corresponding to the stack vertices and adjacent stack vertices, ultimately constructing the spatiotemporal safety envelope model of the current ship. This method can generate a dynamic safety zone for each ship that varies with time and space. Its planning comprehensively considers the ship's size, speed, heading, turning ability, and international maritime collision avoidance rules (COLREGs), making the collision risk assessment of each ship more accurate and reliable, and improving the response rate and perception accuracy of dynamic early warning of coastal waterway risks.

[0105] Preferably, step S108 specifically includes the following steps:

[0106] Based on big data network retrieval, a risk assessment system for coastal waterways and ship navigation is obtained, and a normal distribution model for marginal risk assessment is identified through the risk assessment system.

[0107] By fitting the distribution parameters of the waterway navigation risk index and the ship navigation safety index to the prior probability distribution of the normal distribution model, the marginal distribution function of the waterway navigation risk index is obtained, which is defined as the first marginal distribution function; and the marginal distribution function of the ship navigation safety index is defined as the second marginal distribution function.

[0108] By mapping the waterway navigation risk index and the ship navigation safety index to the probability integral transformation interval [0, 1] through the first marginal distribution function and the second marginal distribution function, a marginal dependency distribution model is generated.

[0109] Based on the uniform distribution pattern of the marginal dependency distribution model, the relevant risk characteristics between the waterway navigation risk indicators and the ship navigation safety indicators are determined, and the connection function of risk coupling is preset based on the relevant risk characteristics.

[0110] A maximum likelihood estimation algorithm is introduced to decouple and estimate the linkage distribution characteristics between waterway navigation risk indicators and ship navigation safety indicators based on the linkage function, resulting in a series of linkage vector parameters; wherein, the linkage vector parameters include the risk coupling strength and direction between waterway navigation risk indicators and ship navigation safety indicators;

[0111] By using a series of connection vector functions located in the big data network, multiple joint random samples of the comprehensive navigation risk assessment of waterway navigation risk indicators and ship navigation safety indicators are retrieved and constructed. The joint random samples are simulated on the original margin to generate a comprehensive risk map. Based on the comprehensive risk map, the comprehensive risk value of the coastal waterway in the future time period is determined.

[0112] If the overall risk value exceeds the preset risk threshold, a dynamic early warning instruction is generated and uploaded to the early warning terminal to execute dynamic early warning actions for the coastal waterway.

[0113] It should be noted that the normal distribution model selects and fits appropriate probability distributions for waterway navigation risk indicators and ship navigation safety indicators to obtain the cumulative distribution of each indicator, namely the first marginal distribution function and the second marginal distribution function. Since the comprehensive evaluation of different risk indicators only observes the dependency structure and does not depend on the specific marginal distribution, this method maps the two to the probability integral transformation interval [0, 1] through the marginal distribution function. At this time, the waterway navigation risk and ship safety indicators are transformed into a unified probability scale, eliminating the difference in dimensions and scale between different risk indicators. This allows the two indicators to be placed in the same statistical framework for mathematical coupling analysis, ensuring that different risk indicators have a consistent scale benchmark, quantifiability, and comparability when superimposed and coupled in modeling. Subsequently, based on the characteristics of the data distribution after marginal distribution, the dependency structure between channel risk and ship safety is analyzed, including positive correlation, negative correlation, tail correlation, and nonlinear dependency structures. This is used to determine the connection function that matches the true risk correlation. This connection function provides nonlinear risk coupling modeling capabilities that traditional linear models cannot achieve, accurately capturing the nonlinear risk assessment propagation between extreme sea states, abnormal channel events, and ship behavior, improving the reference value and credibility of subsequent comprehensive risk assessments. Next, the maximum likelihood estimation method is used to solve the connection distribution characteristics between channel navigation risk indicators and ship navigation safety indicators, ensuring the probability distribution of the coupling model is most consistent with the indicator data. This transforms the risk interaction relationship from qualitative to quantitative, further providing a concrete basis for predicting future risks. The coupling direction clearly indicates whether changes in ship navigation behavior will amplify or mitigate channel risks, serving as a core premise for intelligent dynamic early warning. The comprehensive risk value can be expressed as: R = f(R_s, R_c, R_e); where R_s is siltation risk, R_c is collision risk, and R_e is environmental risk (such as severe weather). If the overall risk value exceeds a preset risk threshold, a tiered early warning system is generated, such as red, orange, and yellow levels. This method enables the overlay and coupling analysis of predicted waterway navigation risk indicators (such as shallowness and siltation) and assessed ship navigation safety indicators (such as collisions and yaws) in a digital twin, generating comprehensive risk assessment information. This allows for precise transmission and control of the dynamic early warning system's response, providing timely alerts to dynamic navigation risks along coastal waterways.

[0114] The second aspect of this invention provides a real-time monitoring and early warning system for the operational status of coastal waterways based on digital twins, such as... Figure 3 As shown, the system, applied to the implementation of any of the digital twin-based methods for real-time monitoring and early warning of coastal waterway operation status, specifically includes:

[0115] The ship-side sensing module includes an intelligent navigation terminal, an AIS automatic identification module, and a shipborne navigation data recorder, used to collect ship position, speed, and heading data.

[0116] The shore-based sensing module includes a shore-based AIS base station, underwater radar, CCTV video monitoring module, hydrological and meteorological station, intelligent navigation beacon and underwater topography monitoring module, which is responsible for measuring tide level, current velocity, wind direction and wind speed, and also has a status self-diagnosis function.

[0117] Digital Twin Engine: The digital twin engine has the functions of data-driven operation, model calculation, real-time rendering and simulation, and is responsible for driving the operation of the digital twin.

[0118] Channel operation monitoring and early warning module: The channel operation monitoring and early warning module is based on a comprehensive risk map provided by a digital twin platform, which realizes real-time monitoring and multi-level early warning of channel operation status;

[0119] Channel siltation analysis module: The channel siltation analysis module is a coupled channel siltation model based on a digital twin platform, used for channel siltation prediction and siltation volume analysis and calculation;

[0120] Intelligent Navigation Assistance Module for Ships: The intelligent navigation assistance module for ships has functions such as recommended routes, speed suggestions, and hazard warnings, which are used to provide personalized navigation services for ships.

[0121] It should be noted that the response behavior data of ships and VTS are collected again by the system to evaluate the effectiveness of the early warning, such as whether ships sail according to the recommended route and whether VTS takes control measures. The feedback data is used to optimize the model, such as by adjusting the risk threshold or model parameters through reinforcement learning, forming a continuously self-evolving "perception-decision-control-optimization" closed loop.

[0122] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A digital-twin-based real-time monitoring and early warning method for the operating state of a coastal waterway, characterized in that, The method comprises the following steps: S102: Constructing a digital twin architecture, collecting multi-source operation data of the coastal waterway in real time, using an attention mechanism backbone network to fuse the multi-source operation data, and driving update from the source data domain to the target data domain of the digital twin architecture to generate a multi-source data fusion digital twin of the coastal waterway; S104: According to the past mechanism function of the corresponding multi-dimensional mechanism model of the coastal waterway as high-dimensional feature input data, driving an AI model to perform function mapping, coupling to construct a back silting dynamic prediction model of the overall trend of sediment transport, and obtaining the waterway navigation risk index of the coastal waterway in the future period through the back silting dynamic prediction model; S106: Based on the speed data and heading data of the nearby ship and the current ship, a discrete sea area plane environment space is obtained, the sea area plane environment space is updated based on the navigation observation data, a time-space occupation matrix is obtained, a time-space safety envelope line model of the current ship is constructed, and whether there is a collision risk between the nearby ship and the current ship in the future period is predicted and judged by combining the time-space occupation matrix and the time-space safety envelope line model, and a ship navigation safety index is output; S108: Quantitatively constructing a marginal dependence distribution model of related risk characteristics, coupling and superimposing the marginal dependence distribution model in the multi-source data fusion digital twin to evaluate the joint risk of the waterway navigation risk index and the ship navigation safety index, and performing a dynamic early warning action on the coastal waterway according to the judgment of the comprehensive risk; S110: Through a shipborne intelligent terminal, early warning information and a recommended route are pushed to the related ship, and early warning information and an auxiliary decision scheme are pushed to the VTS center; The S104 specifically comprises the following steps: Obtaining a construction evaluation report of the coastal waterway, and obtaining a multi-dimensional mechanism model and an experience mechanism case of the coastal waterway through a big data network to search for channel mechanism evaluation parameters of the construction evaluation report; wherein the multi-dimensional mechanism model comprises a sediment mechanism, a hydrodynamic mechanism and a wave mechanism; Extracting a past mechanism function and an applied past mechanism equation of the coastal waterway in a preset period under the multi-dimensional mechanism model simulating sediment transport through the experience mechanism case, and obtaining a data-driven AI model based on an LSTM architecture, and constructing a high-dimensional convolution layer through the data-driven AI model; Constructing a high-dimensional feature space through the high-dimensional convolution layer, embedding the discrete past mechanism function into the high-dimensional feature space for nonlinear vectorization to obtain a high-dimensional feature vector point represented by the past mechanism function in the data-driven AI model; Introducing an integral operator, learning the dependence relationship between the channel mechanism input and the output overall behavior of each high-dimensional feature vector point in the multi-source data fusion digital twin using the integral operator, and forming a global integral layer; Iterating the learning steps of the integral operator to traverse all high-dimensional feature vector points, outputting multiple global integral layers and performing layer stacking processing to obtain a function mapping deep network of the past mechanism function reflecting the overall trend of sediment transport on the coastal waterway in the historical period; The function space of the data-driven AI model is created based on the past mechanism equation, each high-dimensional feature vector point is projected back to the function space solution through function mapping deep network, the high-dimensional feature vector is restored to the multi-dimensional mechanism model-data-driven AI model, the coupling architecture form of observing the sediment transport dynamics of the coastal waterway is generated, and the back silting dynamic prediction model is generated; The sediment transport parameters of the future time period are obtained, the sediment transport parameters are input into the back silting dynamic prediction model for rolling prediction, and the navigation risk index of the back silting position and the back silting amount of the coastal waterway in the future time period is obtained.

2. The digital-twin-based real-time monitoring and early warning method for the operational state of a coastal waterway according to claim 1, characterized in that, The S102 specifically includes the following steps: Obtain a plurality of historical observation data of the coastal waterway in a preset time period, and pre-construct a digital twin architecture of the coastal waterway based on the plurality of historical observation data; wherein the historical observation data includes survey data, remote sensing image and unmanned aerial vehicle aerial photography data; Real-time collection of multi-source operation data of the coastal waterway through the channel Internet of Things array is defined as a first multi-source operation data domain, and multi-source operation data driving the digital twin architecture to update dynamically at the terminal time sequence frame is extracted, which is defined as a second multi-source operation data domain, Extract high-dimensional comparative features of the data, and perform second-order statistical structure calculation on the first multi-source operation data domain and the second multi-source operation data domain based on the high-dimensional comparative features, to obtain the covariance matrix of the first multi-source operation data domain, which is set as the source domain covariance matrix; and the covariance matrix of the second multi-source operation data domain, which is set as the target domain covariance matrix; Introduce a redundancy correlation algorithm to construct an uncorrelated whitening space, remove the internal correlation redundancy of the source domain covariance matrix in the uncorrelated whitening space, and generate whitened source domain features; Add a structured identification code to each multi-source operation data of the first multi-source operation data domain, and sequentially splice the source domain features of all multi-source operation data into a cross-source fusion sequence based on the structured identification code; Construct an attention mechanism backbone network, input the cross-source fusion sequence into the attention mechanism backbone network to learn the cross-source relationship interaction between the source domain features, establish cross-source attention fusion cross-dependence, and output multi-source fusion encoding; Based on the statistical structure of the target domain covariance matrix, a target domain covariance space is established, and after redundancy removal, the whitened source domain features are matched and projected to the target domain covariance space based on the multi-source fusion encoding, so as to update the digital twin architecture through the cross-source synchronous driving of the real-time collected multi-source operation data, and generate a multi-source data fusion digital twin of the coastal waterway. 3.The digital-twin-based real-time monitoring and early warning method for the operational state of a coastal waterway according to claim 1, characterized in that, The S106 specifically includes the following steps: Obtain the speed and heading of the nearby ship relative to the current ship in the future time period, which is defined as the relative speed and relative heading, and set the sailing driving step based on the relative speed, the relative heading, the predetermined speed and the predetermined heading; Obtain a remote sensing geographical sketch map of the coastal waterway, construct a sea area plane environment space of the coastal waterway according to the remote sensing geographical sketch map, and discretely segment the sea area plane environment space into M sub-sea area environment grids based on the sailing driving step; The navigation observation data of the relative position between the current ship and the nearby ship and the observation time sequence strategy are obtained through the relative position sensor, the future time period is divided into continuous N observation time steps at equal intervals based on the observation time sequence strategy; The Bayesian inference network is introduced, the occupation probability of the nearby ship entering or passing through the sub-marine environment grid is inferred and updated at each continuous observation time step based on the navigation observation data, and the time-space occupation matrix is obtained according to the time sequence chain stacking occupation probability of the observation time step; The time-space navigation trajectory point set of the ship is obtained, the time-space navigation trajectory point set is planned and constructed for the spatial convex hull in the future time period based on the size parameters, turning performance index, predetermined speed, predetermined heading and international maritime collision avoidance rules of the ship, and the time-space safety envelope line model of the current ship is generated; The sub-marine environment grid occupied by the nearby ship in the future time period is obtained through the time-space occupation matrix, and is marked as an occupied marine environment grid. If the occupied marine environment grid is located within the time-space safety envelope line model, it is considered that the current ship has a collision risk, and the ship navigation safety index is output.

4. The method according to claim 3, wherein, The time-space navigation trajectory point set of the ship is obtained, the time-space navigation trajectory point set is planned and constructed for the spatial convex hull in the future time period based on the size parameters, turning performance index, predetermined speed, predetermined heading and international maritime collision avoidance rules of the ship, and the time-space safety envelope line model of the current ship is generated, specifically including the following steps: The model information of the ship is obtained, and the size parameters, turning performance index and international maritime collision avoidance rules of the ship are obtained based on the big data network search of the model information; The planned predetermined speed and predetermined heading in the future time period are obtained through the navigation memorandum of the ship, and the time-space navigation trajectory point set of the ship located in the coastal waterway region in the continuous time sequence is extracted according to the coastal navigation log; The remote sensing geographical sketch map of the coastal waterway is obtained, the spatial trajectory coordinate field of the coastal waterway is constructed on the remote sensing geographical sketch map, the coordinates of the current ship in the spatial trajectory coordinate field are obtained, and the coordinates are calibrated as the given trajectory coordinate points; The given trajectory coordinate points are taken as the reference center of dynamic safety planning, and the space polar angle value of each time-space navigation trajectory point in the time-space navigation trajectory point set relative to the reference center of the envelope is calculated based on the size parameters, predetermined speed, predetermined heading, turning performance index and international maritime collision avoidance rules by means of the relative position measurement tool; The convex hull empty stack is constructed, the space polar angle value is used to arrange the sea area running trajectory points in descending order, the time-space navigation trajectory point sorting table is generated, and each time-space navigation trajectory point is pressed into the convex hull empty stack according to the time-space navigation trajectory point sorting table, and the top point of the convex hull empty stack and the adjacent top point adjacent to the top point are obtained; If the direction between the top point and the adjacent top point is consistent, the boundary of the convex hull empty stack is planned and maintained based on the space polar angle value corresponding to the top point and the adjacent top point, and finally the dynamic trajectory convex hull is output, and the time-space safety envelope line model of the current ship is constructed according to the dynamic trajectory convex hull.

5. The digital-twin-based real-time monitoring and early warning method for the operational state of a coastal waterway according to claim 1, characterized in that, The S108 specifically includes the following steps: Based on the big data network retrieval, a risk evaluation system for coastal waterway-ship running is obtained, and a normal distribution model of marginal risk evaluation is obtained through the risk evaluation system; The prior probability distribution of the normal distribution model is used to fit the distribution parameters of the waterway navigation risk index and the ship navigation safety index respectively, to obtain a marginal distribution function of the waterway navigation risk index, defined as a first marginal distribution function, and a marginal distribution function of the ship navigation safety index, defined as a second marginal distribution function; The waterway navigation risk index and the ship navigation safety index are mapped into the probability integral transformation interval [0, 1] through the first marginal distribution function and the second marginal distribution function, and a marginal dependent distribution model is generated; According to the uniform distribution pattern of the marginal dependent distribution model, the correlation risk characteristics between the waterway navigation risk index and the ship navigation safety index are determined, and a coupling function of the risk is preset based on the correlation risk characteristics; The maximum likelihood estimation algorithm is introduced, the coupling distribution characteristics between the waterway navigation risk index and the ship navigation safety index are decoupled based on the coupling function, and a series of coupling vector parameters are obtained; wherein the coupling vector parameters include the risk coupling strength and direction between the waterway navigation risk index and the ship navigation safety index; A series of joint random samples of the waterway navigation risk index and the ship navigation safety index are used to construct a comprehensive navigation risk evaluation of the waterway navigation risk index and the ship navigation safety index, the joint distribution of the index risk coupling is simulated on the original margin, a comprehensive risk map is generated, and the comprehensive risk value of the coastal waterway in the future time period is determined according to the comprehensive risk map; If the comprehensive risk value is greater than the preset risk threshold, a dynamic warning instruction is generated and uploaded to the warning terminal to perform a dynamic warning action on the coastal waterway.

6. A digital-twin-based real-time monitoring and early warning system for the operating state of a coastal waterway, characterized in that The system is applied to realize the real-time monitoring and early warning method of the coastal waterway operation state based on digital twinning, and specifically includes: A ship end perception module, the ship end perception module includes an intelligent navigation terminal, an AIS automatic identification module and a shipborne navigation data recorder, for collecting ship position, speed and heading data; A shore-based perception module, the shore-based perception module includes a shore-based AIS base station, an underwater radar, a CCTV video monitoring module, a hydrological and meteorological station, an intelligent navigation mark and an underwater terrain monitoring module, responsible for measuring tide level, flow rate, wind direction and wind speed, and having a state self-diagnosis function; A digital twinning engine: the digital twinning engine has the functions of data driving, model calculation, real-time rendering and simulation deduction, and is responsible for driving the operation of the digital twinning body; A waterway operation monitoring and warning module: the waterway operation monitoring and warning module is based on the comprehensive risk map provided by the digital twinning platform, and realizes real-time monitoring and multi-level warning of the waterway operation state; A waterway siltation analysis module: the waterway siltation analysis module is a coupled waterway siltation model based on the digital twinning platform, used for waterway siltation prediction and siltation amount analysis. The ship intelligent navigation service module has the function items of route recommendation, speed suggestion and danger warning, and is used for providing personalized navigation service for the ship.

Images