A ship emergency path planning method based on offshore multi-source data fusion

Abstract

The application relates to the technical field of marine emergency rescue, and discloses a ship emergency path planning method based on marine multi-source data fusion, which comprises the following steps: collecting multi-source core original data of a marine new energy station emergency scene and completing preprocessing to obtain a standardized data matrix; generating comprehensive cost coefficients of each grid in a sea area through data level, feature level and decision level hierarchical fusion; constructing a multi-level rigid constraint model, matching constraint elastic adjustment rules and safety cost compensation coefficient calculation rules; dividing differentiated grids based on constraint rigidity level, and generating an initial emergency path through an improved A algorithm of constraint-path bidirectional iteration; and monitoring data changes in real time, performing local re-planning when a threshold is triggered, and closing loop optimization algorithm parameters. The application has good practicability and expansibility by deep cooperation of the constraint model and the path planning, improves emergency rescue efficiency while keeping the safety red line, and adapts to the exclusive emergency demand of the marine new energy station.

Description

Technical Field

[0001] This application relates to the technical field of maritime emergency rescue, and in particular to a ship emergency route planning method based on the fusion of multi-source maritime data. Background Technology

[0002] With the rapid development of offshore wind farms, photovoltaic power plants, and other new energy facilities, the sea area for offshore new energy development is constantly expanding. The facilities within these farms, including wind turbines, photovoltaic arrays, and underwater cables, are densely distributed, making the constraints on ship navigation and emergency rescue in the corresponding sea areas more complex. Furthermore, meteorological and hydrological data, ship dynamic data, rescue resource distribution data, and farm facility data for the sea areas where offshore new energy farms are located are mostly scattered and heterogeneous, making data integration and utilization difficult.

[0003] Existing ship emergency route planning methods are mostly designed for general near-shore waters, failing to fully consider the dense facility characteristics of offshore new energy power stations. They also do not use data such as the location and operational status of these facilities as rigid constraints on route planning, easily leading to risks of emergency routes encroaching on station restricted areas or colliding with station facilities. These methods are ill-suited to the specific emergency needs of offshore new energy power stations. Furthermore, the unidirectional static logic commonly used in existing methods presents a core contradiction: strictly enforcing station constraints leads to detours and insufficient timeliness in emergency routes, while relaxing constraints introduces uncontrollable safety risks. The constraint model and route planning algorithm are completely decoupled, making collaborative optimization for real-time emergency scenarios impossible. Moreover, there is insufficient integration of multi-source heterogeneous data at sea, resulting in mostly static routes that cannot respond in real-time to sudden weather changes or dynamic ship deviations, compromising the safety and timeliness of emergency rescue. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a ship emergency route planning method based on the fusion of multi-source marine data.

[0005] Firstly, this application provides a ship emergency route planning method based on multi-source marine data fusion, employing the following technical solution: A ship emergency route planning method based on multi-source maritime data fusion includes the following steps: Step S1: Perform multi-source data acquisition and preprocessing to obtain a standardized data matrix corresponding to the multi-source core raw data of the emergency scenario of offshore new energy power stations; Step S2: Perform data-level fusion, feature-level fusion, and decision-level fusion on the standardized data matrix in sequence to generate the comprehensive cost coefficients corresponding to each grid in the sea area; Step S3: Construct a multi-level rigid constraint model for emergency scenarios of offshore new energy power stations, classify constraint types of different rigidity levels, set constraint elastic adjustment mapping rules and safety cost compensation coefficient calculation rules, and form a path planning constraint model that can be dynamically interacted. Step S4: Divide the rigidity level based on the constraint model into differentiated meshes, and improve A through constraint-path bidirectional iteration. The algorithm searches for a low-cost path that meets the constraints, starting from the location of the rescue ship and ending at the accident point, and outputs an initial emergency path that includes waypoints and speed prompts. Step S5: Collect new field data in real time and compare it with historical data. When the data change triggers a preset threshold, start local path replanning. At the same time, record the deviation between the actual path execution data and the planned value, optimize the algorithm parameters, and form a closed-loop iteration of the whole process.

[0006] Optionally, in step S3, the constraints are divided into four levels: absolutely rigid constraints, high rigid constraints, medium rigid constraints, and soft constraints. The absolute rigid constraint is an inviolable safety red line that cannot be adjusted under any circumstances; the high rigid constraint and medium rigid constraint can be relaxed to a limited extent according to the timeliness requirements of emergency scenarios and the mapping rules. At the same time, the corresponding safety cost compensation coefficient is calculated and superimposed for the relaxed area.

[0007] Optionally, in the constraint elastic adjustment mapping rule of step S3, the relaxation of high / medium rigid constraints is only allowed when the time gap of path rescue exceeds the preset threshold of the emergency scenario. The relaxation range is proportional to the rescue time benefit, and the maximum relaxation ratio boundary of each constraint is preset.

[0008] Optionally, in step S4, the division of the differentiated grid is coupled with the constraint rigidity level: the region corresponding to the absolute rigid constraint adopts an ultra-fine grid, the region corresponding to the high rigidity constraint adopts a fine grid, and the open sea area corresponding to the medium rigidity constraint and the soft constraint adopts a coarse grid.

[0009] Optionally, the constraint-path bidirectional iterative process in step S4 is as follows: First, the initial baseline path is searched using the default highest rigidity constraint; If the baseline path does not meet the emergency response time requirements, the time gap and information on the optimizable path segments will be fed back to the constraint model. The constraint model dynamically adjusts the constraint boundaries and calculates the safety cost compensation coefficient for the corresponding region. The optimization path is re-searched based on the adjusted constraint model and safety cost compensation coefficient.

[0010] Optionally, in step S4, A is improved. The algorithm's heuristic function incorporates a safety cost compensation coefficient for constraint adjustment, and the expression for the heuristic function is: ; in, The heuristic cost of the current node n; The comprehensive cost coefficient of the sea area grid corresponding to node n is derived from the fusion result of step S2; Let n be the Euclidean distance from node n to the target accident node; The compensation coefficient for the safety cost resulting from the relaxation of constraints corresponding to node n; , , These are the comprehensive cost weight, distance weight, and safety cost weight, respectively, and satisfy the following conditions: .

[0011] Optionally, the formula for calculating the security cost compensation coefficient S(n) is as follows: ; in, The basic risk coefficient corresponding to the constraint type, and the high rigidity constraint. Rigid constraints Areas where restrictions have not been relaxed ; To constrain the relaxation of proportions, , The range of values ​​is , This is the preset maximum relaxation ratio.

[0012] Optionally, in step S1, the multi-source core raw data includes four categories: ship dynamic data, meteorological environment data, rescue resource data, and station facility data. The station facility data is obtained based on the BIM+GIS model of the offshore new energy station and includes the location coordinates, physical boundaries, and real-time operation status data of the station's wind turbines, photovoltaic arrays, and underwater cables.

[0013] Optionally, in the decision-level fusion of step S2, the weight of each dimension of data is calculated using the analytic hierarchy process (AHP). The weight is dynamically adjusted according to the type of emergency: for accidents involving people falling into water, the weight of the rescue efficiency dimension is increased; for accidents involving facility collisions, the weight of the facility protection dimension is increased.

[0014] In summary, this application includes at least one of the following beneficial technical effects: 1. This application adopts a coupled and coordinated mechanism of multi-level rigid constraint model and bidirectional iterative path search. The constraint level of S3 directly determines the mesh generation accuracy of S4, and the path timeliness gap of S4 directly drives the dynamic adjustment of the constraint of S3. The two are mutually dependent and form a bidirectional closed loop, which solves the core pain point that the existing technology cannot balance security and timeliness.

[0015] 2. This application uses a four-level rigid constraint system to accurately match the facility risk level of offshore new energy power stations. It not only safeguards the absolute safety red line of core facilities such as wind turbine tower foundations and underwater cables, but also allows for limited relaxation of non-core constraints to improve rescue efficiency in extreme emergency scenarios such as personnel falling into the water. At the same time, through a quantifiable safety cost compensation coefficient, the risk of constraint relaxation is directly embedded into the algorithm logic of path search, avoiding the safety hazards caused by unlimited relaxation of constraints. It has good scalability and practicality. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the ship emergency route planning method described in this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0018] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0019] This application discloses a ship emergency route planning method based on multi-source marine data fusion, referring to... Figure 1 This includes the following steps: Step S1: Perform multi-source data acquisition and preprocessing to obtain a standardized data matrix corresponding to the multi-source core raw data of the emergency scenario of offshore new energy power stations; Step S2: Perform data-level fusion, feature-level fusion, and decision-level fusion on the standardized data matrix in sequence to generate the comprehensive cost coefficients corresponding to each grid in the sea area; Step S3: Construct a multi-level rigid constraint model for emergency scenarios of offshore new energy power stations, classify constraint types of different rigidity levels, set constraint elastic adjustment mapping rules and safety cost compensation coefficient calculation rules, and form a path planning constraint model that can be dynamically interacted. Step S4: Divide the rigidity level based on the constraint model into differentiated meshes, and improve A through constraint-path bidirectional iteration. The algorithm searches for a low-cost path that meets the constraints, starting from the location of the rescue ship and ending at the accident point, and outputs an initial emergency path that includes waypoints and speed prompts. Step S5: Collect new field data in real time and compare it with historical data. When the data change triggers a preset threshold, start local path replanning. At the same time, record the deviation between the actual path execution data and the planned value, optimize the algorithm parameters, and form a closed-loop iteration of the whole process.

[0020] Specifically, this embodiment discloses a ship emergency path planning method based on multi-source data fusion at sea. The specific application scenario is an emergency rescue operation for personnel falling into the water at an offshore wind farm. The core area of ​​the wind farm has 8 wind turbines per square kilometer. The preset golden rescue time threshold for personnel falling into the water is 15 minutes. The rescue vessel is located in open water outside the wind farm, 5 nautical miles away from the point of fall. The default safe avoidance distance for wind turbines within the wind farm is 50 meters. The underwater cable duct area is an absolute no-navigation zone. The specific implementation steps are as follows: Step S1: Multi-source data acquisition and preprocessing. First, collect four types of core raw data: The first category is ship dynamic data, including AIS message data of rescue ships and ships operating in wind farms, and real-time ship position, speed and heading data obtained by drone inspections; The second category is meteorological and environmental data, including wind speed, wave height, current direction, and current velocity data obtained from marine buoys and meteorological satellite monitoring. The third category is rescue resource data, including the location of shore-based rescue stations, the material configuration of rescue ships, and the distribution of available rescue forces; The fourth category is site facility data, which is obtained based on the BIM+GIS 3D model of the wind farm. It includes the location coordinates, physical boundaries, and real-time operation status data of wind turbines, underwater cables, and substations within the site.

[0021] The collected raw data were preprocessed as follows: Kalman filtering was used to eliminate random noise in ship positions and meteorological and hydrological data; linear interpolation was used to complete missing values ​​that occurred during data collection; isolated forest anomaly detection algorithm was used to remove erroneous data that exceeded the reasonable range; after data cleaning, the format of all data was unified, the coordinate system was changed to WGS84 coordinate system, all units of measurement were converted to international standard units, and finally a standardized two-dimensional data matrix was output as the basis for subsequent fusion and planning.

[0022] Step S2: Multi-source data hierarchical fusion. The standardized data matrix is ​​subjected to three-layer fusion processing in sequence, specifically: (1) Data-level fusion: Integrate multi-source data of the same type. For example, wave height data from the same area from buoys and satellites are weighted and summed according to the credibility of the data acquisition equipment. At the same time, the DS evidence theory is used to process conflicting data between different data sources, eliminate data contradictions, and improve the reliability and consistency of single-dimensional data.

[0023] (2) Feature-level fusion: The CNN-LSTM hybrid deep learning model is used for feature extraction. The input of the model is a standardized data matrix. First, the data is input into two CNN convolutional layers. After convolution and pooling operations, spatial features such as the relative position of ships and wind turbine facilities and the spatial distribution of marine environmental parameters are extracted, and the output is a 128-dimensional spatial feature vector. At the same time, the time series data of three consecutive time steps are input into two LSTM layers to extract time features such as meteorological parameters and the changing trend of ship dynamics, and the output is a 128-dimensional time feature vector. The spatial feature vector and the time feature vector are concatenated to obtain a 256-dimensional multi-dimensional feature vector, thus completing the feature-level fusion.

[0024] (3) Decision-level fusion: The judgment matrix is ​​constructed by the Analytic Hierarchy Process (AHP). For this personnel falling into the water accident, the weight of the rescue efficiency dimension is increased. Finally, the weights of the four dimensions of safety, efficiency, resources and facilities are determined to be 0.3, 0.4, 0.1 and 0.2 respectively. Based on the determined weight system, combined with the multi-dimensional feature vector output by feature-level fusion, the comprehensive cost coefficient corresponding to each grid in the target sea area is calculated. The higher the comprehensive cost coefficient, the higher the passage cost of the grid.

[0025] The structure and connection relationships of the CNN-LSTM hybrid deep learning model are as follows: The model consists of three core parts: a CNN spatial feature extraction module, an LSTM temporal feature extraction module, and a feature concatenation module. The CNN spatial feature extraction module consists of two convolutional layers, one max-pooling layer, and one flattening layer connected in series. The first convolutional layer uses 32 3×3 convolutional kernels with a stride of 1, same padding, and ReLU activation. The second convolutional layer uses 64 3×3 convolutional kernels with a stride of 1, same padding, and ReLU activation. The max pooling layer has a 2×2 pooling window with a stride of 2. The flattening layer converts the two-dimensional feature map output from the pooling layer into a one-dimensional vector, outputting a 128-dimensional spatial feature vector. The LSTM time feature extraction module consists of two serially connected LSTM layers and one fully connected layer. Both the first and second LSTM layers have 128 hidden units, a dropout rate of 0.2, an activation function of tanh, and a recurrent activation function of sigmoid. The fully connected layer maps the output of the LSTM layer to a fixed-dimensional vector, outputting a 128-dimensional temporal feature vector; The feature concatenation module uses the concat concatenation method to concatenate the 128-dimensional spatial feature vector output by the CNN module and the 128-dimensional temporal feature vector output by the LSTM module along the feature dimensions, and finally outputs a 256-dimensional multi-dimensional feature vector. The input of the model is the standardized data matrix output in step S1, with the input dimensions being [3, number of grids, 8]. The output of the model is a 256-dimensional multi-dimensional feature vector. The output result is directly input to the decision-level fusion module in step S2 to calculate the comprehensive cost coefficient of each grid in the sea area.

[0026] Furthermore, the training steps of the CNN-LSTM hybrid deep learning model are as follows: (1) Training dataset construction: Collect historical data of the target marine new energy power station and surrounding sea area over the past 3 years, including ship AIS dynamic data, buoy and satellite meteorological and hydrological data, power station facility operation and maintenance data, and historical emergency rescue case data. Perform the same preprocessing as in step S1 on the data, and label the actual passage cost of the sea area grid in each historical case as the training label. A total of 1200 effective samples are constructed and divided into training set, validation set and test set in a ratio of 8:1:1. (2) Model initialization: The weight parameters of each layer of the model are initialized using the Xavier uniform distribution, the bias parameters are initialized to 0, the batch size is set to 32, the optimizer is the Adam optimizer, the initial learning rate is 0.001, the learning rate decay coefficient is 0.9, it decays once every 10 iterations, the maximum number of iterations is 100, and the loss function is the mean squared error (MSE). The loss function expression is: ; in, The number of samples within the batch. For the first The actual passage cost label for each sample. The model outputs the predicted passage cost corresponding to the features; (3) Model training: Input the training set data into the model for forward propagation, calculate the output features and the predicted pass value, and calculate the error between the predicted value and the label through the loss function; execute the back propagation algorithm, and update the weights and bias parameters of each layer of the model according to the error; after each iteration cycle, use the validation set data to verify the model performance. When the validation set loss value no longer decreases for 10 consecutive iteration cycles, trigger the early stopping mechanism, stop training, and save the current optimal model parameters. (4) Model testing: The accuracy of the trained model is tested using test set data. When the determination coefficient R2 of the test set is greater than or equal to 0.92, the model is deemed to meet the requirements of this scenario and can be used for feature extraction through feature-level fusion.

[0027] Furthermore, the CNN-LSTM hybrid deep learning model is integrated with this scenario in the following way: The model's input is a standardized data matrix specific to emergency scenarios at offshore new energy power stations, containing features across four core dimensions required for route planning: vessels, environment, stations, and resources, perfectly matching the emergency needs of this scenario. The spatial features extracted by the model include the relative positions of vessels and station facilities, and the spatial distribution of marine environmental parameters, which can be directly used to quantify facility avoidance costs and spatial environmental risks along the route. The temporal features extracted by the model include meteorological parameters and the changing trends of vessel dynamics, which can be directly used to quantify the temporal environmental risks and traffic efficiency along the route. The multi-dimensional feature vector output by the model is the core input for the decision-level fusion calculation of the comprehensive cost coefficient, which is the improved A... The core parameters of the algorithm's heuristic function ultimately determine the planning result of the emergency path, forming a complete link process of "scenario data input - model feature extraction - cost coefficient calculation - path planning output", which is very suitable for the emergency path planning scenario of offshore new energy power stations.

[0028] Step S3: Construct a scenario-based multi-level rigid constraint model. For this emergency scenario, a four-level rigid constraint system is constructed: Absolutely rigid constraints: The physical boundaries of the underwater cable burial area and the wind turbine tower base within the wind farm are not allowed to be breached under any circumstances; breaching them will result in an invalid path. High rigidity constraints: A standard safety avoidance zone of 50m around each wind turbine, with a preset maximum allowable ratio. That is, the minimum avoidance distance is not less than 30m, and the basic risk coefficient is... ; Medium rigid constraints: Temporary maintenance work areas within wind farms can have their boundaries dynamically adjusted according to the work progress, with a basic risk coefficient. ; Soft constraints: shortest total rescue time and lowest ship fuel consumption.

[0029] The system includes a constraint adjustment mapping rule: High / medium rigidity constraints can only be relaxed when the initial path rescue time gap exceeds 2 minutes. The mapping rule is preset to allow for a 10m increase in the wind turbine avoidance distance for every 1 minute reduction in rescue time. A safety cost compensation coefficient is also calculated for the relaxed area. This forms a dynamically interactive constraint model and reserves a two-way data interface with the path planning module.

[0030] Step S4: Initial emergency path generation through bidirectional constraint-path iteration: (1) Coupled differentiated grid division: Based on the constraint level in step S3, the underwater cable area with absolute rigid constraint adopts a 10m×10m ultra-fine grid, the wind turbine avoidance area with high rigid constraint adopts a 50m×50m fine grid, and the open sea area outside the wind farm adopts a 200m×200m coarse grid.

[0031] (2) First round of baseline path search: Using the default highest rigidity constraint (wind turbine avoidance 50m, all constraints not relaxed), by improving A The algorithm searches for an initial baseline path and calculates that the total rescue time for the initial path is 18 minutes, which exceeds the golden rescue threshold of 15 minutes, resulting in a 3-minute time gap. The time gap and the information on the optimizable path segments between the two wind turbines with the most severe detours in the wind farm are fed back to the constraint model in step S3.

[0032] (3) Dynamic adjustment of constraints and calculation of safety cost coefficient: After receiving feedback, the constraint model in step S3 verifies the trigger threshold of a time gap of 3 minutes > 2 minutes. According to the mapping rule, the avoidance distance between the two wind turbines is relaxed from 50m to 20m, and the relaxation ratio is calculated. After verification, the maximum allowable distance was adjusted to 40%, meaning the avoidance distance was increased to 30 meters. Calculate the safety cost compensation coefficient for the relaxed area. Areas that have not been relaxed Output the adjusted constraint model and the corresponding mesh. value.

[0033] (4) Second round of path search optimization: Based on the adjusted constraint model, an improved A The algorithm re-searches for paths, improving A. The heuristic function of the algorithm is ; in, , , The weight sum is 1; the search yields an optimized path that passes through the 30m gap between the two wind turbines, with a total rescue time of 14 minutes, meeting the 15-minute time threshold. At the same time, it is verified that the path completely avoids the underwater cable area with absolute rigid constraints. The final output is an initial emergency path that includes waypoint coordinates, suggested speed, and constraint adjustment instructions.

[0034] Step S5: Dynamic path optimization and iteration. Real-time collection of new data such as meteorological and ship dynamics in the target sea area. Comparison of the new data with the data used in historical planning. When the wind speed change exceeds 3m / s, the wave height change exceeds 0.5m, or the ship trajectory deviation exceeds 500m, local path replanning is triggered. Only the grid area affected by the data change is updated, while the path in the remaining area remains unchanged.

[0035] Meanwhile, this method reserves a multi-ship collaborative communication protocol interface to support real-time data exchange among rescue vessels, operation vessels, and command vessels. When multiple vessels are conducting emergency rescue collaboratively, the planned paths of each vessel are compared in real time through a path conflict detection mechanism. When there is a risk of collision, the waypoints and speeds of the corresponding vessels are automatically adjusted to avoid collisions.

[0036] The deviation between the data and the planned values ​​during the actual execution of the emergency path is recorded. The gradient descent method is used to optimize the weight parameters in the algorithm, including the dimension weights in the fusion stage and the cost weights of the heuristic function, so as to continuously improve the accuracy of path planning and form a closed-loop iterative mechanism of "data monitoring - local replanning - deviation feedback - parameter optimization".

[0037] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A ship emergency route planning method based on multi-source maritime data fusion, characterized in that, Includes the following steps: Step S1: Perform multi-source data acquisition and preprocessing to obtain a standardized data matrix corresponding to the multi-source core raw data of the emergency scenario of offshore new energy power stations; Step S2: Perform data-level fusion, feature-level fusion, and decision-level fusion on the standardized data matrix in sequence to generate the comprehensive cost coefficients corresponding to each grid in the sea area; Step S3: Construct a multi-level rigid constraint model for emergency scenarios of offshore new energy power stations, classify constraint types of different rigidity levels, set constraint elastic adjustment mapping rules and safety cost compensation coefficient calculation rules, and form a path planning constraint model that can be dynamically interacted. Step S4: Divide the rigidity level based on the constraint model into differentiated meshes, and improve A through constraint-path bidirectional iteration. The algorithm searches for a low-cost path that meets the constraints, starting from the location of the rescue ship and ending at the accident point, and outputs an initial emergency path that includes waypoints and speed prompts. Step S5: Collect new field data in real time and compare it with historical data. When the data change triggers a preset threshold, start local path replanning. At the same time, record the deviation between the actual path execution data and the planned value, optimize the algorithm parameters, and form a closed-loop iteration of the whole process.

2. The ship emergency route planning method based on multi-source maritime data fusion according to claim 1, characterized in that, In step S3, the constraints are divided into four levels: absolute rigid constraints, high rigid constraints, medium rigid constraints, and soft constraints. The absolute rigid constraint is an inviolable safety red line that cannot be adjusted under any circumstances; the high rigid constraint and medium rigid constraint can be relaxed to a limited extent according to the timeliness requirements of emergency scenarios and the mapping rules. At the same time, the corresponding safety cost compensation coefficient is calculated and superimposed for the relaxed area.

3. The ship emergency route planning method based on multi-source maritime data fusion according to claim 2, characterized in that, In the constraint elastic adjustment mapping rule of step S3, the relaxation of high / medium rigid constraints is only allowed when the time gap of path rescue exceeds the preset threshold of emergency scenario. The relaxation range is proportional to the rescue time benefit, and the maximum relaxation ratio boundary of each constraint is preset.

4. The ship emergency route planning method based on multi-source maritime data fusion according to claim 2, characterized in that, In step S4, the division of the differentiated grid is coupled with the constraint rigidity level: the region corresponding to the absolute rigid constraint adopts the ultra-fine grid, the region corresponding to the high rigidity constraint adopts the fine grid, and the open sea area corresponding to the medium rigidity constraint and the soft constraint adopts the coarse grid.

5. The ship emergency route planning method based on multi-source maritime data fusion according to claim 3, characterized in that, The constraint-path bidirectional iteration process in step S4 is as follows: First, the initial baseline path is searched using the default highest rigidity constraint; If the baseline path does not meet the emergency response time requirements, the time gap and information on the optimizable path segments will be fed back to the constraint model. The constraint model dynamically adjusts the constraint boundaries and calculates the safety cost compensation coefficient for the corresponding region. The optimization path is re-searched based on the adjusted constraint model and safety cost compensation coefficient.

6. The ship emergency route planning method based on multi-source maritime data fusion according to claim 5, characterized in that, Improvement A in step S4 The algorithm's heuristic function incorporates a safety cost compensation coefficient for constraint adjustment, and the expression for the heuristic function is: ； in, The heuristic cost of the current node n; The comprehensive cost coefficient of the sea area grid corresponding to node n is derived from the fusion result of step S2; Let n be the Euclidean distance from node n to the target accident node; The compensation coefficient for the safety cost resulting from the relaxation of constraints corresponding to node n; , , These are the comprehensive cost weight, distance weight, and safety cost weight, respectively, and satisfy the following conditions: .

7. The ship emergency route planning method based on multi-source maritime data fusion according to claim 1, characterized in that, The formula for calculating the safety cost compensation coefficient S(n) is as follows: ； in, The basic risk coefficient corresponding to the constraint type, and the high rigidity constraint. Rigid constraints Areas where restrictions have not been relaxed ; To constrain the relaxation of proportions, , The range of values ​​is , This is the preset maximum relaxation ratio.

8. The ship emergency route planning method based on multi-source maritime data fusion according to claim 1, characterized in that, In step S1, the multi-source core raw data includes four categories: ship dynamic data, meteorological environment data, rescue resource data, and station facility data.

9. The ship emergency route planning method based on multi-source maritime data fusion according to claim 1, characterized in that, The data on the facilities at the site is obtained based on the BIM+GIS model of the offshore new energy power station, and includes the location coordinates, physical boundaries and real-time operation status data of the wind turbines, photovoltaic arrays and underwater cables at the site.

10. The ship emergency route planning method based on multi-source maritime data fusion according to claim 1, characterized in that, In the decision-level fusion of step S2, the weight of each dimension of data is calculated using the analytic hierarchy process. The weight is dynamically adjusted according to the type of emergency: for accidents involving people falling into water, the weight of the rescue efficiency dimension is increased; for accidents involving facility collisions, the weight of the facility protection dimension is increased.

Images