An improved ant colony algorithm-based method and system for scheduling marine oil spill emergency resources

Abstract

The application provides an improved marine oil spill emergency resource scheduling method and system based on an ant colony algorithm, and relates to the technical field of data processing.The method comprises the following steps: based on a modified path transition probability function, performing multi-objective ant colony collaborative optimization on a dynamic resource scheduling architecture, initializing a population to construct a scheduling path solution, calculating a time consumption target and an ecological loss target, iteratively updating pheromone distribution until convergence, and generating a multi-agent non-dominated scheduling strategy; according to the multi-agent non-dominated scheduling strategy, calculating a resource scheduling failure condition probability and an ecological sensitive area damage expectation value; if the scheduling failure condition probability exceeds a risk tolerance threshold, taking the ecological damage expectation value as a nonlinear penalty weight, triggering an anti-disturbance rescheduling mechanism to generate a resilience scheduling strategy. The application improves the timeliness and accuracy of emergency resource transportation, and realizes the collaborative optimization of "emergency efficiency-ecological protection".

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a marine oil spill emergency resource scheduling method and system improved by an ant colony algorithm. BACKGROUND

[0002] As a typical environmental disaster, marine oil spill accidents have the characteristics of fast diffusion speed, wide influence range and great ecological damage. The timeliness and accuracy of emergency resource scheduling directly determine the control effect of oil spill pollution and the degree of ecological loss.

[0003] Marine oil spill emergency scheduling needs to balance the "time efficiency" (rapid control of oil spill diffusion) and "ecological protection" (priority coverage of sensitive areas such as mangroves and fishing grounds), and there is a natural conflict between the targets (for example, rapid response may lead to excessive concentration of resources in the near shore, ignoring the sensitive areas in the open sea). Traditional single-target optimization algorithms often focus on a single target (such as the shortest transportation time), resulting in insufficient protection of ecologically sensitive areas and ultimately expanding regional ecological loss. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a marine oil spill emergency resource scheduling method and system improved by an ant colony algorithm, which improves the timeliness and accuracy of emergency resource transportation and realizes the collaborative optimization of "emergency efficiency-ecological protection".

[0005] To solve the above technical problems, the technical scheme of the present application is as follows:

[0006] In a first aspect, a marine oil spill emergency resource scheduling method improved by an ant colony algorithm, the method comprising:

[0007] Step 1, collect the environmental dynamics parameters of the oil spill accident, including the coordinates of the leakage source, the oil pollution diffusion rate, the wind field vector and the ocean current field data;

[0008] Step 2, based on the environmental dynamics parameters, construct a dynamic resource scheduling architecture with emergency resource hubs as spatial nodes and resource transportation channels as directional connection relationships, define resource capacity constraints for node attributes, and encode real-time navigation resistance coefficients for connection relationship weights;

[0009] Step 3, input the navigation resistance coefficient into the ant colony algorithm framework, embed the fluid dynamic resistance correction term in the path transfer probability function, and generate a corrected path transfer probability function;

[0010] Step 4, based on the corrected path transfer probability function, perform multi-objective ant colony collaborative optimization on the dynamic resource scheduling architecture, initialize the population to construct scheduling path solutions, calculate the time consumption target and ecological loss target, iteratively update the pheromone distribution until convergence, and generate multi-agent non-dominated scheduling strategies;

[0011] Step 5: Calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy.

[0012] Step 6: If the probability of scheduling failure exceeds the risk tolerance threshold, the expected value of ecological damage is used as a non-linear penalty weight to trigger the anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy.

[0013] Furthermore, a dynamic resource scheduling architecture is constructed based on environmental dynamic parameters, with emergency resource hubs as spatial nodes and resource transport channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients, including:

[0014] Step 21: Analyze the geodetic coordinates of the leakage source, locate the latitude and longitude coordinates of the emergency resource hub in the electronic nautical chart coordinate system, generate a set of spatial node coordinates, calculate the Euclidean distance between adjacent nodes, and generate a set of basic navigation corridors by combining the ocean current velocity potential field.

[0015] Step 22: Input the basic navigation corridor set into the wind field feasibility filter, remove the corridors with headwind exceeding the threshold according to the wind field gradient vector, and obtain the feasible directional connection relationship set; based on the feasible directional connection relationship set, call the resource management database to match the ship and material inventory of each node to obtain the node resource capacity constraint configuration; calculate the navigation time cost of each connection according to the environmental parameters and generate the navigation impedance coefficient matrix.

[0016] Step 23: Integrate the spatial node coordinate set, feasible directional connection relationship set, node resource capacity constraint configuration and navigation impedance coefficient matrix to construct a dynamic resource scheduling architecture.

[0017] Furthermore, the navigation impedance coefficient is input into the ant colony algorithm framework, and a hydrodynamic drag correction term is embedded in the path transition probability function to generate a corrected path transition probability function, including:

[0018] Step 31: Extract the navigation impedance coefficient matrix and map it to the ant colony algorithm path cost base value set; based on the path cost base value set and combined with wind field vector data, calculate the relative drag coefficient of each segment; input the relative drag coefficient into the ship motion characteristics knowledge base and match the corresponding power loss compensation parameters.

[0019] Step 32: Input the path cost base value set and power loss compensation parameters into the dynamic fusion unit, and perform piecewise linear interpolation to generate a set of dynamic fluid resistance correction terms for each flight segment;

[0020] Step 33: Inject the set of dynamic fluid resistance correction terms into the cost calculation kernel of the path transfer probability function, and reconstruct the probability calculation logic by replacing the standard distance cost variable with the corrected comprehensive cost variable.

[0021] Step 34: Compile and generate an executable corrected path transition probability function entity based on the reconstructed probability calculation logic.

[0022] Furthermore, step 4 includes:

[0023] Step 41: Based on the configuration of spatial node topology and resource capacity constraints of the dynamic resource scheduling architecture, the initial scheduling path solution set is generated by assigning a starting node to each ant and calling the modified path transition probability function to select feasible nodes.

[0024] Step 42: Take the initial scheduling path solution set as input, process the navigation impedance coefficient matrix and power loss compensation parameters of each path in the solution set to accumulate and calculate the total path time consumption target, and process the oil pollution diffusion rate to calculate the increase in the pollution area of ​​the ecologically sensitive area caused by the path time difference as the ecological loss target.

[0025] Step 43: The calculated bi-objective values ​​are used to identify the Pareto front solution set through non-dominated sorting. Based on the front solution set, path pheromones are released and pheromone evaporation operations are performed on the remaining solutions to generate a new generation of scheduling path solution set.

[0026] Step 44: Take the new generation of scheduling path solution set as input. If the Pareto front solution sets of three consecutive generations coincide, it is determined that the process has converged. Decode the final non-dominated solution set to generate a multi-agent scheduling strategy.

[0027] Furthermore, based on the multi-agent non-dominated scheduling strategy, the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas are calculated, including:

[0028] Step 51: Take the time-series path set in the multi-agent scheduling strategy as input, combine wind field vector and ocean current field data for Monte Carlo simulation, and obtain the probability distribution of the sailing time deviation of each path; based on the probability distribution of sailing time deviation, calculate the path failure condition probability through the preset resource scheduling time constraint threshold, and calculate the resource supply interruption probability by integrating the node resource capacity constraint configuration.

[0029] Step 52: Input the conditional probability of resource scheduling failure into the Bayesian risk decision engine, and combine it with the ecological loss target quantification method to obtain the set of failure scenarios for oil pollution diffusion paths; based on the set of failure scenarios, calculate the increase in the pollution area of ​​sensitive areas under each scenario by using the oil pollution diffusion rate and the distribution data of ecologically sensitive areas on the electronic nautical chart.

[0030] Step 53: Using the increase in the polluted area as input, perform a probability-weighted integral based on the failure condition probability to obtain the expected value of damage to the ecologically sensitive area.

[0031] Furthermore, if the probability of scheduling failure exceeds the risk tolerance threshold, an anti-disturbance rescheduling mechanism is triggered, using the expected value of ecological damage as a non-linear penalty weight, to generate a resilient scheduling strategy, including:

[0032] Step 61: Perform a dynamic risk situation comparison operation between the resource scheduling failure condition probability and the preset risk tolerance threshold. If the failure condition probability exceeds the threshold, generate a rescheduling activation instruction. Based on the rescheduling activation instruction, input the expected value of ecological damage into the nonlinear penalty modeling unit to construct a penalty weight function.

[0033] Step 62: Identify the set of flight segments whose failure condition probability exceeds the risk threshold, mark them as high-risk failure path sets, apply the penalty weighting function to the high-risk failure path set, and perform penalty amplification calculation on the flight impedance coefficient of each failure path.

[0034] Step 63: Update the navigation impedance coefficient after penalty amplification to the connection relationship weight matrix of the dynamic resource scheduling architecture to generate the disturbance resistance enhancement architecture; based on the disturbance resistance enhancement architecture, call the modified path transition probability function entity to re-execute multi-objective ant colony cooperative optimization to generate an environmental disturbance resilience scheduling strategy.

[0035] Further, in step 32, the path cost base value set and power loss compensation parameters are input into the dynamic fusion processor, and piecewise linear interpolation is performed to generate a set of dynamic fluid resistance correction terms for each flight segment, including:

[0036] Step 321: Map the ocean current impedance values ​​in the path cost base value set to the preset ocean current velocity classification range, and map the power loss compensation parameters to the preset wind resistance compensation classification level; generate a set of discrete positioning points for each segment in the ocean current-wind resistance two-dimensional parameter space based on the mapping results, and match interpolation calculation units according to the spatial distribution of positioning points; extract the vertex reference values ​​of the matching interpolation calculation units, and calculate the normalized position coefficient of the segment in the parameter space;

[0037] Step 322: First, interpolate the normalized position coefficients along the ocean current dimension to generate transition boundary values, and then interpolate along the wind resistance dimension to generate the comprehensive hydrodynamic correction value for the segment; aggregate the comprehensive hydrodynamic correction values ​​of all segments to generate a set of dynamic fluid resistance correction terms.

[0038] Secondly, an improved ant colony algorithm-based marine oil spill emergency resource scheduling system includes:

[0039] The data acquisition module is used to collect environmental dynamic parameters of oil spill accidents, including the coordinates of the leak source, the oil spill diffusion rate, wind field vectors, and ocean current field data.

[0040] The module is used to construct a dynamic resource scheduling architecture based on environmental dynamic parameters, with emergency resource hubs as spatial nodes and resource transportation channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients.

[0041] The correction module is used to input the navigation impedance coefficient into the ant colony algorithm framework, embed a hydrodynamic drag correction term into the path transition probability function, and generate a corrected path transition probability function.

[0042] The optimization module is used to perform multi-objective ant colony collaborative optimization on a dynamic resource scheduling architecture based on the modified path transition probability function, initialize the population to construct scheduling path solutions, calculate the time consumption objective and the ecological loss objective, iteratively update the pheromone distribution until convergence, and generate a multi-agent non-dominated scheduling strategy.

[0043] The evaluation module is used to calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy.

[0044] The scheduling module is used to trigger an anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy when the probability of scheduling failure exceeds the risk tolerance threshold, with the expected value of ecological damage as a non-linear penalty weight.

[0045] Thirdly, a computing device includes:

[0046] One or more processors;

[0047] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0048] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0049] The above-described solution of the present invention has at least the following beneficial effects:

[0050] By embedding a hydrodynamic drag correction term into the path planning, real-time environmental dynamic parameters such as wind field and ocean current are directly incorporated into the path transition probability calculation of the ant colony algorithm, enabling the navigation impedance coefficient to dynamically respond to changes in environmental drag. Compared to some traditional static weighting algorithms, this effectively avoids path cost assessment deviations caused by environmental dynamism, ensuring a high degree of matching between the planned path and actual sea conditions, and improving the timeliness and accuracy of emergency resource transportation.

[0051] By employing a multi-objective ant colony collaborative optimization mechanism, the dual objectives of time consumption and ecological loss are simultaneously considered, generating a non-dominated scheduling strategy. This solves the imbalance between efficiency and ecological protection in traditional single-objective algorithms. Furthermore, it can dynamically adjust the priority of objectives based on the oil spill spread and the distribution of ecologically sensitive areas. While rapidly controlling the spread of oil spills, it prioritizes the supply of protective resources for key ecological areas, achieving synergistic optimization of "emergency efficiency - ecological protection". Attached Figure Description

[0052] Figure 1 This is a schematic diagram of a marine oil spill emergency resource scheduling method based on an improved ant colony algorithm, provided by an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of an improved ant colony algorithm-based marine oil spill emergency resource scheduling system provided by an embodiment of the present invention. Detailed Implementation

[0054] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0055] like Figure 1 As shown, an embodiment of the present invention proposes an improved ant colony algorithm for emergency resource scheduling in marine oil spills, the method comprising the following steps:

[0056] Step 1: Collect environmental dynamic parameters of the oil spill accident, including the coordinates of the leak source, the oil spill diffusion rate, wind field vectors, and ocean current field data;

[0057] Step 2: Based on environmental dynamic parameters, construct a dynamic resource scheduling architecture with emergency resource hubs as spatial nodes and resource transport channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients.

[0058] Step 3: Input the navigation impedance coefficient into the ant colony algorithm framework, embed the hydrodynamic drag correction term into the path transition probability function, and generate the corrected path transition probability function.

[0059] Step 4: Based on the modified path transition probability function, perform multi-objective ant colony cooperative optimization on the dynamic resource scheduling architecture, initialize the population to construct scheduling path solutions, calculate the time consumption objective and the ecological loss objective, iteratively update the pheromone distribution until convergence, and generate a multi-agent non-dominated scheduling strategy.

[0060] Step 5: Calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy.

[0061] Step 6: If the probability of scheduling failure exceeds the risk tolerance threshold, the expected value of ecological damage is used as a non-linear penalty weight to trigger the anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy.

[0062] In this embodiment of the invention, by embedding a hydrodynamic resistance correction term in the path planning, real-time environmental dynamic parameters such as wind field and ocean current are directly incorporated into the path transition probability calculation of the ant colony algorithm, enabling the navigation impedance coefficient to dynamically respond to changes in environmental resistance. Compared to some traditional static weighting algorithms, this effectively avoids path cost assessment deviations caused by environmental dynamism, ensures a high degree of matching between the planned path and actual sea conditions, and improves the timeliness and accuracy of emergency resource transportation.

[0063] By employing a multi-objective ant colony collaborative optimization mechanism, the dual objectives of time consumption and ecological loss are simultaneously considered, generating a non-dominated scheduling strategy. This solves the imbalance between efficiency and ecological protection in traditional single-objective algorithms. Furthermore, it can dynamically adjust the priority of objectives based on the oil spill spread and the distribution of ecologically sensitive areas. While rapidly controlling the spread of oil spills, it prioritizes the supply of protective resources for key ecological areas, achieving synergistic optimization of "emergency efficiency - ecological protection".

[0064] In a preferred embodiment of the present invention, step 2, constructing a dynamic resource scheduling architecture based on environmental dynamic parameters, with emergency resource hubs as spatial nodes and resource transport channels as directional connections, where node attributes define resource capacity constraints and connection weights encode real-time navigation impedance coefficients, includes:

[0065] Step 21: Analyze the geodetic coordinates of the leakage source, locate the latitude and longitude coordinates of the emergency resource hub in the electronic nautical chart coordinate system, generate a set of spatial node coordinates, calculate the Euclidean distance between adjacent nodes, and generate a set of basic navigation corridors by combining the ocean current velocity potential field.

[0066] Step 22: Input the basic navigation corridor set into the wind field feasibility filter, remove the corridors with headwind exceeding the threshold according to the wind field gradient vector, and obtain the feasible directional connection relationship set; based on the feasible directional connection relationship set, call the resource management database to match the ship and material inventory of each node to obtain the node resource capacity constraint configuration; calculate the navigation time cost of each connection according to the environmental parameters and generate the navigation impedance coefficient matrix.

[0067] Step 23: Integrate the spatial node coordinate set, feasible directional connection relationship set, node resource capacity constraint configuration and navigation impedance coefficient matrix to construct a dynamic resource scheduling architecture.

[0068] In step 21 above, the geodetic coordinates of the leak source are collected in real time using a GNSS positioning device (such as a Beidou or GPS receiver with a positioning accuracy ≤ 0.5 meters), including latitude (range: -90°~90°), longitude (range: -180°~180°), and altitude (0 meters for nearshore waters). Pre-set emergency resource hub registration data is retrieved from the emergency resource management system, including the unique identifier of each hub (e.g., "Hub A", "Hub B") and initial latitude and longitude. The coordinate transformation interface of the electronic nautical chart system (using the WGS-84 coordinate system) is called to map the geodetic coordinates of the leak source and resource hubs to the electronic nautical chart plane coordinate system, generating a spatial node coordinate set containing "hub identifier - latitude - longitude," where the latitude and longitude data are retained to 4 decimal places (accuracy approximately 11 meters), ensuring a spatial positioning error ≤ 20 meters.

[0069] Based on latitude and longitude data from the spatial node coordinate set, the spherical distance formula (referencing the Earth's average radius of 6371 km) is used to calculate the straight-line distance between any two adjacent resource hubs. Specifically, for node i (latitude φ1, longitude λ1) and node j (latitude φ2, longitude λ2), the latitude and longitude are first converted to radians, and then the surface arc distance between the two points is calculated using the cosine theorem. The result is rounded to the nearest integer in meters and used as a reference for the basic navigation path length. Ocean current field data for the accident area is obtained from the marine environmental monitoring database, including the main current direction (range: 0°~360°). The designation is based on the following parameters: 0° north (clockwise increments); average current velocity (range: 0.1–3 m / s); and influence range (width ≤ 5 km). A basic navigation corridor is created by extending along the main ocean current direction, with the straight-line distance between adjacent nodes as the central axis. The corridor width is dynamically set according to the ocean current velocity: 500–1000 m when the velocity is ≤ 1 m / s; and 1000–2000 m when the velocity is > 1 m / s. This ensures the corridor covers the navigable area for ships propelled by the ocean current, forming a basic navigation corridor set encompassing the "start point – end point – corridor range".

[0070] Step 22 above constructs a wind field feasibility filter, the core of which is a rule base model. The inputs are a set of basic navigation corridors, real-time wind field data (wind direction: 0°~360°, wind speed: 0~20 m / s), and a preset headwind threshold. The rule base is defined as follows: when the angle between the corridor's navigation direction and the wind direction is ≥120° (determined as headwind) and the wind speed is ≥10 m / s, it is marked as a "headwind exceeding the threshold" corridor. Wind field data is obtained every 10 minutes through a real-time interface. The direction of the central axis of the basic navigation corridor is compared with the wind direction. Each corridor is checked to see if it meets the headwind exceeding the threshold condition. Corridors exceeding the threshold are removed. The remaining corridors are used as a set of feasible directional connections, containing "start-end-corridor status (feasible)" information. A preset resource management database is used, which includes a node basic table (storing hub identifier and location), a ship table (storing ship ID, type, maximum speed, and load limit), and a material table (storing material type, inventory, and unit weight). The database adopts a MySQL architecture, supports real-time read and write, and the data update delay is ≤5 minutes.

[0071] Based on the node identifiers in the feasible directional connection set, the corresponding ship and material tables in the association database are used to extract data and generate capacity constraints: ship constraints are "number of schedulable ships ≤ total number of registered ships" and "single ship load capacity ≤ load capacity limit"; material constraints are "schedulable material quantity ≤ current inventory quantity", forming a node resource capacity constraint configuration table to clarify the resource scheduling upper limit for each node; navigation impedance coefficient matrix generation: for each connection in the feasible directional connection set, real-time parameters are collected: corridor length (meters), current wind speed (meters / second), ocean current speed (meters / second), and average ship speed (meters / second, preset according to ship type: 10-15 knots for cleaning vessels, 15-20 knots for transport ships); impedance is calculated with "voyage time + comprehensive cost" as the core. Coefficients: Sailing time = Corridor length ÷ (Ship speed + Forward assist - Reverse resistance), where forward assist is the speed bonus when ocean current / wind speed is in the same direction (≤2 m / s), and reverse resistance is the speed loss when the speed is in the opposite direction (≤3 m / s); Comprehensive cost = Sailing time × Fuel consumption cost per unit time + Material transportation loss cost (set according to distance: loss rate ≤1% when ≤50 km, loss rate ≤3% when >50 km); Impedance coefficient = Sailing time × 0.6 + Comprehensive cost × 0.4 (weight dynamically adjusted according to emergency priority); Matrix construction: According to the connection relationship of "node i-node j", the calculated impedance coefficient is filled into the corresponding position of the two-dimensional matrix to generate the navigation impedance coefficient matrix. The matrix value is updated every 30 minutes according to environmental parameters.

[0072] Step 23 above integrates various components using a graph data structure: The set of spatial node coordinates is used as vertices, with vertex attributes bound to node resource capacity constraints; the set of feasible directional connections is used as directed edges, with edge attributes bound to corresponding coefficients in the navigation impedance coefficient matrix; a topology graph is constructed using a Python graph theory library (such as NetworkX), where vertex IDs correspond one-to-one with hub identifiers, the start-end point of edges matches the start-end point of feasible connections, and attribute data is embedded in key-value pairs to ensure that vertex and edge attributes can be accessed in real time; dynamic update trigger conditions are set for the architecture: wind field / ocean current data updates (every 10-30 minutes); changes in node resource inventory (difference ≥ 5%); corridor status changes from feasible to infeasible (when headwind exceeds the threshold); real-time communication is established between the environmental monitoring system, resource management system, and architecture using a message queue (such as RabbitMQ). When the trigger conditions are met, the update interface is automatically called: invalid edges are deleted, new feasible edges are added, and vertex attributes and edge weights are updated, ensuring that the architecture completes data synchronization within 1 minute to reflect the latest status.

[0073] In this embodiment of the invention, high-precision coordinate transformation and dynamic corridor delineation ensure that the spatial matching error between resource hubs and navigation paths is ≤20 meters. Wind field feasibility filters and dynamic impedance coefficient calculations enable navigation paths to respond in real time to changes in wind fields and ocean currents, avoiding inefficient paths against wind and ocean currents, thus improving the timeliness of resource transportation. Node resource capacity constraints and real-time database synchronization ensure that scheduling strategies match actual resource inventory, avoiding execution deviations such as "no resources available for scheduling." The real-time update mechanism enables the scheduling architecture to quickly adapt to environmental and resource changes, providing a real-time and accurate topology foundation for subsequent ant colony optimization and improving emergency scheduling.

[0074] In a preferred embodiment of the present invention, step 3, inputting the navigation impedance coefficient into the ant colony algorithm framework, embedding a hydrodynamic drag correction term into the path transition probability function, and generating a corrected path transition probability function, includes:

[0075] Step 31: Extract the navigation impedance coefficient matrix and map it to the ant colony algorithm path cost base value set; based on the path cost base value set and combined with wind field vector data, calculate the relative drag coefficient of each segment; input the relative drag coefficient into the ship motion characteristics knowledge base and match the corresponding power loss compensation parameters.

[0076] Step 32: Input the path cost base value set and power loss compensation parameters into the dynamic fusion unit, and perform piecewise linear interpolation to generate a set of dynamic fluid resistance correction terms for each flight segment;

[0077] Step 33: Inject the set of dynamic fluid resistance correction terms into the cost calculation kernel of the path transfer probability function, and reconstruct the probability calculation logic by replacing the standard distance cost variable with the corrected comprehensive cost variable.

[0078] Step 34: Compile and generate an executable corrected path transition probability function entity based on the reconstructed probability calculation logic.

[0079] Step 31 above, matrix extraction and mapping: From the navigation impedance coefficient matrix generated in step 22, the impedance coefficient values ​​are extracted one by one according to the directed connection relationship of "node i-node j", invalid connections (impedance coefficient is 0 or null value) are eliminated, and the coefficient values ​​of all feasible connections are retained as the path cost base values ​​to form a path cost base value set; the base value range is 1 to 100 (the larger the value, the higher the cost), where the base value for short-distance near-shore routes is ≤30, and the base value for long-distance offshore routes is ≥50; from the wind field vector data collected in step S1, the wind direction (0° to 360°, due north is 0°) and the wind direction for each route are extracted. Wind speed (0-20 m / s), where the route direction is the axis direction of the "start-end" in the feasible directional connection relationship set (consistent with the electronic chart coordinate system); the wind resistance type is determined by the wind direction angle: when the angle between the route direction and the wind direction is ≤60°, it is a tailwind, and the relative wind resistance coefficient is 0.5-0.8; when the angle is 60°-120°, it is a crosswind, and the value is 0.8-1.2; when the angle is ≥120°, it is a headwind, and the value is 1.2-2.0; for every 2 m / s increase in wind speed, the coefficient increases by 0.1 in the headwind state (maximum not exceeding 2.0), and decreases by 0.1 in the tailwind state (minimum not less than 0.5).

[0080] Ship motion characteristics knowledge base construction and parameter matching:

[0081] Knowledge Base Construction: The ship motion characteristics knowledge base adopts a relational database architecture (such as PostgreSQL), including a ship type table (pollution control vessel, transport vessel, etc.), a speed table (actual speed under different wind speeds), and a power loss table (correspondence between wind resistance and fuel consumption). The data comes from actual ship tests: three typical ships (200-ton pollution control vessel, 500-ton transport vessel, and 1000-ton general-purpose vessel) were selected, and their navigation status under wind speeds of 0-20 m / s was simulated in wind tunnels and water tanks. Power loss data (percentage increase in fuel consumption) was recorded to form the basic dataset.

[0082] Knowledge base training and implementation: The basic dataset is fitted using linear regression to establish a mapping rule of "ship type - relative drag coefficient - power loss compensation parameter" (e.g., when the headwind coefficient of a 200-ton cleaning vessel is 1.5, the power loss compensation parameter is 1.3, which means the cost increases by 30%). The knowledge base supports real-time querying. After inputting the ship type and relative drag coefficient, the corresponding power loss compensation parameter (within the range of 1.0 to 2.5) is output. Based on the ship type involved in the current scheduling (extracted from the resource capacity constraint configuration in step 22) and the calculated relative drag coefficient, the knowledge base query interface is called to match and obtain the power loss compensation parameter for each segment.

[0083] Step 32 above, dynamic fusion construction: The dynamic fusion adopts a modular design, including an input interface (receiving the path cost base value set and power loss compensation parameters), an interpolation processing module, and an output interface (generating a set of correction terms); its core logic is the weighted fusion of "base value × compensation parameter", and the weights are dynamically adjusted according to the flight segment length: when the flight segment length is ≤50 km, the base value weight is 0.6 and the compensation parameter weight is 0.4; when the length is >50 km, the base value weight is 0.5 and the compensation parameter weight is 0.5; piecewise linear interpolation is implemented: the flight segment is divided into 3 segments according to length: short-distance segment (≤30 km), medium-distance segment (30~100 km), and long-distance segment (>100 km), and each segment is set with independent interpolation nodes (5 nodes for short-distance segment, 10 nodes for medium-distance segment, and 15 nodes for long-distance segment).

[0084] Interpolation calculation: For each flight segment, starting from the path cost base value and ending at "base value × compensation parameter", intermediate values ​​are calculated linearly at the interpolation nodes to eliminate abrupt differences between the base value and the compensation parameter. For example, for a short-range flight segment with a base value of 20, a compensation parameter of 1.2, and an endpoint value of 24, the interpolation results for the five nodes are 20, 21, 22, 23, and 24, forming a smooth transition dynamic fluid resistance correction term. The correction term ranges from 1.0 to 2.5 times the path cost base value. The interpolation results for all flight segments are summarized to form a set of dynamic fluid resistance correction terms, with each correction term corresponding to a unique flight segment identifier.

[0085] In step 33 above, the standard function is analyzed as follows: The cost calculation kernel of the original path transition probability function of the ant colony algorithm uses "standard distance cost" as the core variable (i.e., the basic cost corresponding to the actual length of the path) and does not consider the influence of environmental resistance. The correction value in the dynamic fluid resistance correction term set replaces the standard distance cost variable as a new cost input. For example, if the cost of a certain segment in the original function is a distance cost of 30, it is updated to a comprehensive cost (30 × dynamic correction value) after the correction term is injected, so that the cost calculation kernel can reflect the actual power loss caused by wind resistance and ocean currents. The reconstructed probability calculation logic retains the core rule of the ant colony algorithm that "the higher the pheromone concentration and the lower the cost, the greater the probability of path selection", but updates the cost parameter to the corrected comprehensive cost. Specifically, when the comprehensive cost of a certain segment increases, its weight in path selection decreases; when the comprehensive cost decreases, its weight increases, ensuring that the algorithm prioritizes low-resistance and low-cost segments. The probability calculation boundary is set: the minimum probability of a single path selection is ≥0.01 (to avoid completely ignoring potential paths), and the maximum probability is ≤0.8 (to avoid over-concentration on a certain path), ensuring the diversity of the algorithm search.

[0086] Step 34 above converts the reconstructed probability calculation logic (including correction term invocation, comprehensive cost calculation, and probability weight allocation rules) into executable code (such as C++). The code reserves a real-time calling interface for dynamic fluid resistance correction terms, supporting the retrieval of the latest correction term set from step 32 every 30 minutes. The correctness of the function is verified through simulation testing: inputting pheromone concentration and comprehensive cost data for typical flight segments, running the function and outputting the path selection probability, the error must be ≤5% compared with the theoretical expectation; testing the probability reduction effect on headwind flight segments, ensuring that the headwind path selection probability after correction is ≥20% lower than before correction; entity generation: encapsulating the compiled code into an independent function entity, including input parameters (pheromone matrix, dynamic fluid resistance correction term set) and output parameters (path transition probability distribution), with a function response time ≤100 milliseconds to meet real-time scheduling requirements.

[0087] In this embodiment of the invention, the hydrodynamic drag correction term dynamically reflects the effects of wind resistance and ocean currents, making the path transfer probability more closely match actual navigation costs, improving the rationality of headwind segment selection, and avoiding invalid path searches. The ship motion characteristic knowledge base is built based on real ship data, reducing the matching error of power loss compensation parameters, ensuring accurate calculation of the correction term, and improving the practical feasibility of path planning. The reconstructed probability calculation logic retains a reasonable path selection probability boundary, avoiding the algorithm from getting trapped in local optima, and improving the efficiency of global optimal path search. The function entity responds quickly and supports dynamic correction term updates, adapting to real-time changes in the marine environment.

[0088] In a preferred embodiment of the present invention, step 4 involves performing multi-objective ant colony cooperative optimization on a dynamic resource scheduling architecture based on a modified path transition probability function, initializing the population to construct scheduling path solutions, calculating time consumption and ecological loss objectives, iteratively updating pheromone distribution until convergence, and generating a multi-agent non-dominated scheduling strategy, including:

[0089] Step 41: Based on the configuration of spatial node topology and resource capacity constraints of the dynamic resource scheduling architecture, the initial scheduling path solution set is generated by assigning a starting node to each ant and calling the modified path transition probability function to select feasible nodes.

[0090] Step 42: Take the initial scheduling path solution set as input, process the navigation impedance coefficient matrix and power loss compensation parameters of each path in the solution set to accumulate and calculate the total path time consumption target, and process the oil pollution diffusion rate to calculate the increase in the pollution area of ​​the ecologically sensitive area caused by the path time difference as the ecological loss target.

[0091] Step 43: The calculated bi-objective values ​​are used to identify the Pareto front solution set through non-dominated sorting. Based on the front solution set, path pheromones are released and pheromone evaporation operations are performed on the remaining solutions to generate a new generation of scheduling path solution set.

[0092] Step 44: Take the new generation of scheduling path solution set as input. If the Pareto front solution sets of three consecutive generations coincide, it is determined that the process has converged. Decode the final non-dominated solution set to generate a multi-agent scheduling strategy.

[0093] In step 41 above, the number of ants is set according to the node scale of the dynamic resource scheduling architecture: when the number of nodes is ≤10, the number of ants is 20-30; when the number of nodes is >10, the number of ants is 30-50, ensuring that the population covers a sufficient path search space. The starting node is allocated according to the resource inventory of the emergency resource hub: hubs with an inventory ratio of ≥30% are high-priority starting nodes and are allocated 40% of the number of ants; hubs with an inventory ratio of 10%-30% are medium-priority and are allocated 30%; hubs with an inventory ratio of <10% are low-priority and are allocated 30%, to avoid over-searching of idle nodes. During the path generation process, the node resource capacity constraint configuration in step 22 is called in real time to ensure that when each ant selects the next hop node, the number of ships / materials that can be scheduled by the node is ≥ the resource requirements of the current path (e.g., if a path requires the use of 2 cleaning vessels, then the ship inventory of the next hop node must be ≥2). If the condition is not met, the node is skipped.

[0094] Corrected path transition probability function call: Each ant starts from the starting node and calculates the path selection probability (probability value range 0~1) from the current node to other feasible nodes based on the corrected path transition probability function generated in step 34. The next hop node is selected in descending order of probability (nodes with a probability ≥0.1 are preferred to avoid low-probability invalid jumps); Path termination condition: Path generation terminates when the path covers all oil spill areas that need rescue (based on the coordinates of the leak source and the spread range in step 1) or when the resource scheduling amount reaches the node capacity limit; Initial solution set construction: After each ant generates a complete path, it records the "node sequence - resource scheduling amount - flight segment identifier" information contained in the path to form a single scheduling path solution; All path solutions generated by all ants are summarized, and duplicate solutions (solutions with completely identical node sequences) are removed to finally form an initial scheduling path solution set. The size of the solution set is consistent with the number of ants (≥80% of the initial solutions are retained after removing duplicate solutions).

[0095] In step 42 above, for each path in the initial scheduling path solution set, extract all the segments it contains, call the navigation impedance coefficient matrix from step 22 and the power loss compensation parameter from step 31, and accumulate the time consumption in the order of segments: single segment time = segment impedance coefficient × power loss compensation parameter (unit: minutes), where the impedance coefficient already includes the basic navigation time, and the compensation parameter reflects the time increment caused by environmental resistance; total path time consumption = sum of the time of each segment + node resource loading time (fixed value: 5-10 minutes / ship loading, 10-20 minutes / ton of materials loading), and the result is retained to the nearest integer minute.

[0096] Oil spill diffusion model construction and implementation:

[0097] Model construction: A diffusion model is constructed based on the oil spill diffusion rate (unit: square kilometers / hour) in step S1. The model input is time (hours) and the output is the pollution area (square kilometers). The diffusion law is set as follows: linear diffusion within 0 to 6 hours (area = diffusion rate × time), and exponential slowdown after 6 hours (area = diffusion rate × 6 + diffusion rate × 0.5 × (time - 6)), which is adapted to the actual physical characteristics of oil spill diffusion.

[0098] Model Validation: Historical oil spill data (such as a nearshore oil spill in 2019) are used to validate the model's accuracy. After inputting the diffusion rate, the error between the predicted area and the actual monitored area must be ≤10%. Otherwise, the index mitigation coefficient (0.4-0.6) is adjusted until the error meets the standard. Ecologically sensitive area data (such as mangroves and fishing grounds) of the oil spill area are retrieved from the marine ecological database. The boundary coordinates and area of ​​the sensitive areas (range 1-50 square kilometers) are identified and marked as areas requiring priority protection. For each path, the time difference between the path completion time (total path time converted to hours) and the "ideal response time" (preset to 2 hours after the leak) is calculated. If the time difference is ≤0 (the path is completed ahead of schedule or on time), the ecological loss target = 0. If the time difference is >0, the increase in the polluted area within the time difference is calculated according to the diffusion model. The area increase of the overlapping part of the sensitive area is multiplied by a weighting coefficient of 1.5 (the weight of non-sensitive areas is 1.0) as the ecological loss target value (unit: square kilometers) for that path.

[0099] Step 43 above, non-dominated sorting model construction: The non-dominated sorting model is used to identify Pareto front solution sets. The core logic is "hierarchical partitioning under dual-objective optimization": If the time consumption of solution A is less than or equal to that of solution B and the ecological loss is less than or equal to that of solution B, and at least one objective is better, then A dominates B; otherwise, B dominates A; solutions without mutual dominance constitute the same level; for all solutions in the initial scheduling path solution set, compare the time consumption and ecological loss objectives pairwise and mark the dominance relationship; extract solutions without dominance as the first level (Pareto front solution set); after removing the first level, repeat the comparison in the remaining solutions to extract the second level, and so on, until all solutions have completed hierarchical partitioning; for the first level Pareto front solution set, release pheromones according to the rule of "the better the objective, the more pheromones are released": the pheromone release amount of a single solution = the basic release amount (fixed at 10 units) × (1 - normalized value of time consumption). )×(1-normalized value of ecological loss), where normalized value = (target value - minimum value) / (maximum value - minimum value), with a value range of 0 to 1; for the solution set of the second layer and below, perform pheromone volatilization operation, with the volatilization rate set to 10% to 20% (the higher the layer, the higher the volatilization rate: 10% for the second layer, 15% for the third layer, and 20% for the fourth layer and above), that is, the pheromone concentration after volatilization = original concentration × (1-volatilization rate); set the lower limit of pheromone concentration to 1 unit (to avoid pheromone depletion) and the upper limit to 100 units (to avoid path locking due to excessive concentration), and automatically truncate to the boundary value when the concentration exceeds the range; based on the updated pheromone concentration, repeat the path generation process of step 41 (the number of ants remains unchanged, and the allocation of the starting node is dynamically adjusted: nodes with a pheromone concentration ≥ 50 units are given high priority), generate a new generation of scheduling path solution set, and realize population iteration.

[0100] Step 44 above, convergence criterion: The "Pareto front overlap" is used as the convergence index. The similarity of the solution sets of three consecutive Pareto fronts is calculated. If more than 90% of the paths in the solution set are completely identical in node sequence, time consumption, and ecological loss objective (error ≤ 5%), the algorithm is considered converged. If the convergence condition is not met, the upper limit of the number of iterations is set to 50-100 (the higher the number of nodes, the higher the upper limit) to avoid the algorithm getting stuck in infinite iteration. From the final Pareto front solution set, 3-5 representative solutions are selected according to the "time-ecological balance" principle: including... The search identifies three key objectives: minimizing time consumption, minimizing ecological loss, and achieving a bi-objective equilibrium (minimizing the sum of normalized values ​​for time and ecological loss). Each selected solution is decoded to generate a specific scheduling strategy: defining the ship's route sequence from the starting point to the target point, departure time, and estimated arrival time; drone patrol routes: planning drone patrol nodes and time intervals for the oil spill area (once every 30 minutes in sensitive areas and once every hour in non-sensitive areas); and emergency supplies delivery sequence: determining the type, quantity, and order of delivery nodes to ensure priority delivery to sensitive areas.

[0101] In this embodiment of the invention, by weighted allocation of starting nodes and verification of resource constraints, the initial path solution set covers more potential optimization directions, avoiding early local search traps in the algorithm and increasing the probability of finding the global optimal solution. Time consumption is dynamically calculated in conjunction with environmental resistance, and ecological loss is correlated with the oil pollution diffusion model and the characteristics of sensitive areas, improving the matching degree between the target value and actual emergency needs and providing a reliable basis for multi-objective optimization. Non-dominated sorting accurately identifies high-quality solutions, pheromone update rules strengthen the guidance of high-quality paths, and the evaporation mechanism preserves search diversity, improving the algorithm's convergence speed. Finally, strategy decoding focuses on "time-ecological balance," outputting specific executable routes, paths, and delivery sequences, directly supporting emergency command decisions and reducing the difficulty of implementation.

[0102] In a preferred embodiment of the present invention, step 5, calculating the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy, includes:

[0103] Step 51: Take the time-series path set in the multi-agent scheduling strategy as input, combine wind field vector and ocean current field data for Monte Carlo simulation, and obtain the probability distribution of the sailing time deviation of each path; based on the probability distribution of sailing time deviation, calculate the path failure condition probability through the preset resource scheduling time constraint threshold, and calculate the resource supply interruption probability by integrating the node resource capacity constraint configuration.

[0104] Step 52: Input the conditional probability of resource scheduling failure into the Bayesian risk decision engine, and combine it with the ecological loss target quantification method to obtain the set of failure scenarios for oil pollution diffusion paths; based on the set of failure scenarios, calculate the increase in the pollution area of ​​sensitive areas under each scenario by using the oil pollution diffusion rate and the distribution data of ecologically sensitive areas on the electronic nautical chart.

[0105] Step 53: Using the increase in the polluted area as input, perform a probability-weighted integral based on the failure condition probability to obtain the expected value of damage to the ecologically sensitive area.

[0106] In step 51 above, the planned travel time (accurate to the minute), number of segments (3-10 segments, depending on distance), and segment priority (level 1-5, with level 1 being the highest) for each path are extracted from the multi-agent scheduling strategy. Simultaneously, the wind field vectors (wind direction quantized according to 8 azimuth angles: North 360°, Northeast 45°, etc., wind speed divided into 5 levels: 0-4 m / s for light breeze, 5-9 m / s for gentle breeze, 10-14 m / s for moderate breeze, 15-19 m / s for strong wind, and ≥20 m / s for gale) and ocean current field data (current direction according to the same wind vectorization standard, current speed divided into 3 levels: 0.1-1 m / s) are acquired from step S1. (S represents slow current, 1.1-2 m / s represents moderate current, and ≥2.1 m / s represents jet stream). The number of simulations is determined by the path complexity: 1000 simulations for ≤5 segments on a single path, and 2000 simulations for >5 segments. In each simulation, the wind field disturbance is set with deviations according to wind speed levels: light wind ±1 m / s, gentle wind ±2 m / s, strong wind ±3 m / s; the wind direction deviation is uniformly ±30°; ocean current disturbances: slow current ±0.2 m / s, moderate current ±0.3 m / s, jet stream ±0.5 m / s; the flow direction deviation is ±15°; all disturbance values ​​are generated through uniform random sampling to ensure coverage of the possible fluctuation range of environmental parameters.

[0107] The process of generating the probability distribution of flight time deviation:

[0108] Single-segment time recalculation: For each segment of each path, based on the simulated disturbance parameters, the power loss compensation parameter calculation logic from step 31 (tailwind / headwind / crosswind determination and coefficient value) is called to recalculate the actual flight time: If the simulated wind speed is light wind (5~9m / s) and it is determined to be headwind, the power loss compensation parameter is taken as 1.2~1.5 (the higher the wind speed, the larger the value), then the actual segment time = planned segment time × compensation parameter; the total time deviation for each path = Σ (actual time of each segment - planned time); summarize the total time deviation values ​​of all simulations, divide them into intervals according to "-60 minutes to 0 minutes (early)", "0 to 60 minutes (slight delay)", "61 to 120 minutes (moderate delay)" and ">120 minutes (severe delay)", and count the proportion of the occurrence of each interval to the total number of simulations to form a flight time deviation probability distribution table (e.g., the severe delay interval accounts for 15%, that is, the probability is 0.15).

[0109] Quantitative calculation of the conditional probability of path failure:

[0110] Based on the oil spill rate classification: Level 1 accident (leakage rate > 15m³ / h) 3 / h): Resources must arrive within 4 hours; timeliness threshold = 4 × 60 = 240 minutes; Level 2 accident (8-15m) 3 / h): Time threshold = 6 × 60 = 360 minutes; Level 3 accident (<8m) 3 / h): Timeliness threshold = 12 × 60 = 720 minutes; From the flight time deviation probability distribution table, extract all intervals where "total planned time + time deviation > timeliness threshold" (e.g., if the planned time is 300 minutes and the timeliness threshold is 360 minutes, then intervals with deviation > 60 minutes are all failure intervals), sum the probability values ​​of these intervals to obtain the failure condition probability of a single path (value range 0 to 0.8, exceeding 0.8 is counted as 0.8 to avoid the influence of extreme values). Call the node resource capacity constraint configuration in step 22 to clarify the "safety stock threshold" of each node (10% of the stock quantity, such as node A has 10 ships in stock and a safety threshold of 1 ship. When the resource outflow of a node in the scheduling strategy is greater than or equal to (stock - safety threshold), it is marked as a "high-risk node"; when the outflow is greater than or equal to the stock, it is marked as an "extremely high-risk node". The weighted fusion of interruption probabilities is as follows: the base probability of supply interruption for high-risk nodes is 0.2 (historical data statistics), and for extremely high-risk nodes it is 0.5. The probability of resource supply interruption is calculated as follows: (path failure condition probability × 0.4) + (node ​​interruption base probability × 0.6). The weight allocation is based on the following: the impact of path delay on supply accounts for 40%, and the impact of insufficient node stock accounts for 60%.

[0111] Step 52 above: Construction and training of the Bayesian risk decision engine:

[0112] Data Layer: Stores structured historical data, including accident characteristic tables (leakage rate, sea area type, etc.), failure factor tables (wind field intensity, resource delay time, etc.), and consequence tables (pollution area, degree of damage to sensitive areas, etc.). The data comes from multiple oil spill accident reports (part of which is used for training, and the other part is used for validation). Model Layer: Contains a prior probability module and a likelihood function module. The prior probability module obtains the basic probability distribution by statistically analyzing the frequency of occurrence of "path failure → pollution expansion" in historical data (e.g., failure probability of 0.3 within 1 hour delay, 0.5 within 1-2 hours). The likelihood function module establishes the mapping relationship of "failure probability - environmental parameters - consequences" and uses the Logistic regression method for fitting (input is failure probability and wind field level, output is the conditional probability of pollution expansion).

[0113] The specific process of engine training:

[0114] Historical data is normalized, and continuous variables such as leakage rate and delay time are discretized (e.g., delay time is divided into 3 levels: <1 hour, 1-3 hours, >3 hours), and the distance to sensitive areas is divided into three levels: near (<10km), medium (10-20km), and far (>20km). The prior probability mean (e.g., 0.3) and likelihood function coefficient (e.g., wind speed weight 0.2) are initialized and iteratively optimized using the expectation-maximization algorithm (EM algorithm): each iteration uses training data to calculate the prediction error. If the error is >15%, the likelihood function coefficient is adjusted (e.g., the weight of wind speed above medium to high is increased to 0.3) until the prediction accuracy of the case is ≥85%. The trained model is encapsulated into a callable module, with the input being the path failure conditional probability and environmental parameters, and the output being the probability distribution of the failure scenario.

[0115] Based on engine output, failure scenarios are divided into three categories: minor failure: failure probability 0.1–0.3, duration 0.5–1 hour, diffusion acceleration coefficient 1.2–1.5 (i.e., diffusion rate is 1.2–1.5 times the original rate); moderate failure: failure probability 0.3–0.6, duration 1–3 hours, acceleration coefficient 1.5–2.0; severe failure: failure probability > 0.6, duration 3–6 hours, acceleration coefficient 2.0–3.0 (maximum not exceeding 3.0). A unique identifier (e.g., “minor-01”) is assigned to each scenario, and its duration, acceleration coefficient, and corresponding probability (sum of probabilities = 1) are recorded to form a set of failure scenarios for oil spill diffusion paths.

[0116] Spatial analysis of sensitive area data: Extracting the vector boundaries of ecologically sensitive areas (such as the coordinates of mangrove reserve boundaries) from electronic nautical charts, and calculating the area of ​​sensitive areas (1-50 km²) using GIS tools. 2 The pollution risk sub-zones were determined based on the shortest distance (5–50 km) from the leak source and the location of the leak. The sub-zones were divided into a core zone (30% of the area closest to the leak source), a buffer zone (40% in the middle), and a peripheral zone (30% furthest from the leak source). The core zone had a weight of 2.0, the buffer zone 1.5, and the peripheral zone 1.0. For each failure scenario, the total diffusion area was calculated as: total diffusion rate × acceleration factor × duration. The overlapping area between the total diffusion area and the sensitive area was determined using GIS spatial overlay analysis, and the weighted overlapping area was calculated according to the sub-zone weights. The increase in pollution area was calculated as: weighted overlapping area. The result was rounded to one decimal place (km). 2 ).

[0117] Step 53 above, based on the path failure conditional probability calculated in step 51, adjusts the weights according to the severity of the failure scenario: minor failure weight = failure probability × 0.3, moderate × 0.5, severe × 0.2 (sum of weights = 1), ensuring that the impact of high-risk scenarios is reasonably amplified; the possible range of increased pollution area (0-50km) is also considered. 2 ) by 1km 2The interval is divided into 50 intervals, and the midpoint value of each interval is used as the representative value of that interval (e.g., 1-2 km). 2 The interval is 1.5km. 2 For each failure scenario, the incremental value of its polluted area is multiplied by the corresponding weight to obtain the damage contribution value for that scenario; the contribution values ​​of all scenarios are summed to obtain the expected damage value of the ecologically sensitive area (range 0–30 km). 2 More than 30km 2 Based on 30km 2 (Calculation); Select a portion of historical cases similar to the current accident (same sea area, same sensitive area type), extract their actual polluted area and compare it with the expected damage value calculated by the model; if the average error is >10%, adjust the weights according to the error direction (e.g., if the actual pollution is greater than the prediction, increase the weight of the severe failure scenario by 0.1), and recalculate until the error is ≤10% to ensure the reliability of the expected value.

[0118] In this embodiment of the invention, the Bayesian engine trained based on historical data enhances the matching degree of failure scenarios, accurately mapping pollution scenarios corresponding to different risk levels and providing a scientific basis for risk prediction. The probability-weighted integral takes into account the comprehensive impact of light, moderate, and severe failure scenarios, and combined with the weights of sensitive sub-regions, reduces the deviation between the expected damage and the actual pollution. All parameters (such as the number of simulations and weight allocation) have clearly defined value ranges and adjustment rules, and the model training process can be repeatedly verified, facilitating integration into emergency command systems and directly supporting risk decision-making.

[0119] In a preferred embodiment of the present invention, step 6, if the probability of scheduling failure exceeds the risk tolerance threshold, triggers an anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy using the expected value of ecological damage as a non-linear penalty weight, including:

[0120] Step 61: Perform a dynamic risk situation comparison operation between the resource scheduling failure condition probability and the preset risk tolerance threshold. If the failure condition probability exceeds the threshold, generate a rescheduling activation instruction. Based on the rescheduling activation instruction, input the expected value of ecological damage into the nonlinear penalty modeling unit to construct a penalty weight function.

[0121] Step 62: Identify the set of flight segments whose failure condition probability exceeds the risk threshold, mark them as high-risk failure path sets, apply the penalty weighting function to the high-risk failure path set, and perform penalty amplification calculation on the flight impedance coefficient of each failure path.

[0122] Step 63: Update the navigation impedance coefficient after penalty amplification to the connection relationship weight matrix of the dynamic resource scheduling architecture to generate the disturbance resistance enhancement architecture; based on the disturbance resistance enhancement architecture, call the modified path transition probability function entity to re-execute multi-objective ant colony cooperative optimization to generate an environmental disturbance resilience scheduling strategy.

[0123] In step 61 above, combining the protection level of ecologically sensitive areas (national, provincial, and municipal) and the scale of oil spills (small: <50 tons, medium: 50-500 tons, large: >500 tons), a 3×3 risk tolerance threshold matrix is ​​constructed: National sensitive areas: large spill threshold = 0.2, medium = 0.25, small = 0.3; Provincial sensitive areas: large = 0.25, medium = 0.3, small = 0.35; Municipal sensitive areas: large = 0.3, medium = 0.35, small = 0.4; The thresholds are stored in an encrypted risk database, supporting emergency command personnel to dynamically adjust them through authorized authentication (single adjustment range ≤ ±0.05, and an audit log must be generated after adjustment, recording the adjuster, time, and reason).

[0124] Extract the failure condition probability (accurate to 3 decimal places) of each path from the calculation results of step 51, and sort them by path priority (level 1 is the highest); simultaneously retrieve the sensitivity level and leakage scale of the corresponding accident, and match the corresponding threshold in the threshold matrix; perform a "failure probability - threshold" value comparison for each path: if the failure probability of a single path is greater than the threshold, and the path is a level 1 or 2 priority (core path): immediately mark it as a "trigger path"; if the failure probability of level 3 or lower paths is greater than the threshold, but the cumulative proportion of paths exceeding the standard is greater than 30%: also mark it as a "trigger path"; when there is at least one "trigger path", generate a rescheduling activation instruction, including: a unique instruction ID (such as "RS-YYYYMMDD-HHMMSS"), a list of trigger paths (ID and exceeding the standard value), and the current risk level (high / medium / low); otherwise, generate The maintenance instruction includes "Current risk is under control, no adjustment is needed" and details the failure probability of each path; Hardware architecture: Industrial-grade servers (CPU ≥ 8 cores, memory ≥ 16GB), equipped with redundant power supplies to ensure continuous operation, and communicate in real time with the risk decision database and dynamic resource scheduling architecture via Ethernet interface (communication latency ≤ 100ms); Software architecture: Includes a data receiving module (parses the expected value of ecological damage), a rule engine module (executes nonlinear mapping), and a result output module (generates penalty weights), developed in C++ language, supporting ≥ 100 computation requests per second; Training dataset construction: Collected handling records of multiple marine oil spill accidents, extracted data pairs of "expected value of ecological damage (measured value) - expert suggested penalty coefficient", of which 80% is used as the training set and 20% as the validation set; Expected values ​​range from 0 to 5km. 2 5-10km 2 , ..., 25~30km 2 Divide into 6 intervals, each containing at least 15 data points.

[0125] Low expected value range (0-10km) 2The least squares method was used to fit the linear relationship. The initial slope was set to 0.05. Through iterative optimization using the training set, the average error between the predicted weights and the expert-suggested values ​​was ≤3%, and the final slope was determined to be 0.048 (adjustment range ±0.005); (10~20km) 2 A quadratic function was used for fitting, with the initial coefficients set to 0.001. Gradient descent was used for optimization to ensure the error was ≤4%, and the final coefficient was determined to be 0.0012. (High expectation value range (20-30km)) 2 ): A saturation function was used for fitting, with an upper limit set to 3.0. The inflection point (25km) was calibrated using the validation set. 2 (Weight = 2.9), ensuring a distance exceeding 25km. 2 Growth slowed down afterward; Validation criteria: In 120 data points, the deviation of more than 95% of the predicted weights from the expert's suggested value is ≤5%; otherwise, increase the amount of training data (supplement with similar accident cases) and retrain.

[0126] The specific calculation process of "fitting a linear relationship using the least squares method" is as follows: From multiple historical accident data, extract 80 data pairs (as the training set) of "expected ecological damage value - expert-suggested penalty weight". The selection criteria are: expected ecological damage value ∈ [0, 10km]. 2 (Low expected value range); penalty weight ∈ [1.5, 2.0] (a reasonable penalty range corresponding to low loss scenarios); when the weight value of a data pair deviates from the average value of the expected value range by more than 20%, it is marked as an outlier and removed (finally 75 valid data points are retained); 0–10km 2 Expected value based on 1km 2 The interval is divided into 10 sections (0-1km) 2 1-2km 2 ..., 9-10km 2 Each interval contains 7-8 data points; calculate the midpoint of the expected value for each interval (e.g., 0-1km). 2 The midpoint is 0.5km. 2 The average weights recommended by experts (e.g., the average weight for this interval is 1.55) are used to form 10 simplified data pairs of "interval midpoint - average weight" to reduce computational complexity. The linear relationship model for the low expected value segment is set as follows: penalty weight = basic weight + slope × expected value of ecological damage. The basic weight is the initial weight when the expected value = 0, and is set to 1.5 based on historical data (a fixed value that does not participate in optimization). The slope is the parameter to be optimized (representing the rate of increase of the weight with the expected value), and the initial value range is set to 0.04 to 0.06 (referring to a reasonable range based on expert experience). The initial value is tentatively set to 0.05.

[0127] The deviation between the model's predicted value and the expert's suggested value is quantified using the "mean absolute error" method. The calculation steps are as follows:

[0128] For the midpoint of each interval in the 10 simplified data pairs (e.g., 0.5km) 2 Substitute the values ​​into the model to calculate the predicted weights: Predicted value = 1.5 + slope × 0.5; Calculate the absolute error of a single data set = |predicted value - expert average weight|; Calculate the average absolute error of all 10 data sets = (Σ single-set absolute error) / 10, with the error unit being the weight value (range 0~0.1); If the average absolute error > 5% (i.e. > 0.05 × average weight value), then adjust the slope: When the overall predicted value is lower than the expert suggested value (error is positive): the slope increases by 0.002; when the overall predicted value is higher than the expert suggested value (error is negative): The slope is reduced by 0.002; after each adjustment, the predicted values ​​and mean absolute error of 10 sets of data are recalculated, and the iteration is repeated; the mean absolute error is ≤5%, or the number of iterations reaches 50 (to avoid infinite iteration); 20 validation data points (without overlap with the training set data) in the low expected value range are extracted from historical data and substituted into the optimized model to calculate the prediction weights. If the mean absolute error of the validation data is ≤5%, the current slope is determined as the final parameter; if the validation error is >5%, 5 similar data points (e.g., expected value 5-6km) are added from the training set. 2 (A typical case), repeat step 3 iterative optimization until the verification error meets the standard; after iterative optimization, the linear model parameters for the low expected value segment are: basic weight = 1.5 (fixed); slope = 0.048 (the value after iterative adjustment, within the initial range of 0.04 to 0.06); model expression: penalty weight = 1.5 + 0.048 × expected value of ecological damage (applicable to 0 to 10 km) 2 Input the expected value of ecological damage calculated in step 53 (accurate to 0.1 km). 2 The weight values ​​are calculated using mapping rules (rounded to two decimal places), for example: expected value 3.2km. 2 →Lower section calculation: 1.5 + 0.048 × 3.2 ≈ 1.65; Expected value: 14.7km 2 →Middle section calculation: 2.0 + 0.0012 × (14.7 - 10) 2 ≈2.27; Expected value 27.5km 2 →High-level calculation: 2.9 + 0.01 × (27.5 - 25) ≈ 2.93; The calculation logic is encapsulated into a callable function, with the expected value as the input and the penalty weight as the output, and the response time is ≤ 50ms.

[0129] In step 62 above, all "activated" flight segments (excluding "backup" paths) in the dynamic resource scheduling architecture are traversed, and their failure condition probabilities are compared with the corresponding thresholds one by one: if the path failure probability is greater than the threshold, and the resource share of the path in the scheduling strategy is greater than or equal to 15% (resource share = path scheduling resource amount / total resource amount): it is marked as a "Level 1 High-Risk Path"; if the failure probability is greater than the threshold, but the resource share is less than 15%, it is marked as a "Level 2 High-Risk Path"; all Level 1 and Level 2 high-risk paths are summarized to generate a high-risk failure path set, which includes the following for each path: unique ID, origin-end node, original impedance coefficient, failure probability, and risk level (Level 1 / Level 2); in the electronic chart visualization interface, Level 1 paths are marked with a bold red dashed line, Level 2 paths are marked with an orange dashed line, and normal paths are marked with a blue solid line, making it easy for command personnel to identify them intuitively; from the navigation impedance coefficient matrix generated in step 22, the original impedance coefficient of each path in the high-risk failure path set is extracted (the value range is 10 to 10). 0), Classified by risk level: Level 1 paths are grouped separately, and Level 2 paths are grouped together; Level 1 path: Penalized resistance coefficient = original coefficient × penalty weight × 1.2 (an additional 20% penalty); Level 2 path: Penalized resistance coefficient = original coefficient × penalty weight; Constraints: Regardless of the calculation result, the penalized coefficient shall not exceed 3 times the original coefficient (upper limit = original coefficient × 3), if it exceeds, the upper limit shall be used; at the same time, it shall not be lower than 1.5 times the original coefficient (lower limit = original coefficient × 1.5). Ensure the penalty is effective; Calculation examples: Primary path original coefficient = 50, penalty weight = 2.0 → calculated value = 50 × 2.0 × 1.2 = 120, original coefficient multiplied by 3 = 150 → final value is 120; Secondary path original coefficient = 30, penalty weight = 1.6 → calculated value = 30 × 1.6 = 48, original coefficient multiplied by 1.5 = 45 → final value is 48; Secondary path original coefficient = 20, penalty weight = 3.2 → calculated value = 20 × 3.2 = 64, original coefficient multiplied by 3 = 60 → final value is 60.

[0130] Step 63 above, weight matrix update process: Open the connection relationship weight matrix of the dynamic resource scheduling architecture (two-dimensional table structure, rows / columns are node IDs, cells are impedance coefficients); for each path (node ​​i → node j) in the high-risk failure path set, write the penalized impedance coefficient calculated in step 62 into the matrix at position (i, j), overwriting the original coefficient; the coefficients of unmarked paths remain unchanged, generating a new weight matrix; architecture connectivity verification: use the depth-first search algorithm (DFS) to traverse the node connection relationship of the new architecture to check whether there is an "isolated node" (no feasible path connection with other nodes); if there is an isolated node, retrieve the original impedance coefficients of the node and the three nearest nodes from the original weight matrix, restore the connection by 1.2 times the original coefficient (slight penalty), mark it as an "emergency backup path", and ensure that all nodes in the architecture are in a connected state.

[0131] Re-execution of multi-objective ant colony collaborative optimization: Increase the number of ants by 50% (e.g., from 20 to 30, from 40 to 60), but the upper limit is 100 (to avoid excessive computation); the evaporation rate for first-level risk scenarios (with first-level high-risk paths) is 5%, and for second-level risk scenarios (only second-level high-risk paths) it is 10% (the original evaporation rate was 10% to 20%); multiply the original upper limit by 1.5 (e.g., from 60 times to 90 times), but the maximum is no more than 120 times. At the same time, set a convergence judgment condition: terminate early when the overlap of the Pareto front solution set for 5 consecutive generations is ≥90%; call the modified path transition probability function entity in step 34, re-execute the path search based on the anti-disturbance enhancement architecture, and generate a new generation of scheduling path solution set; select solutions from the solution set that meet the following conditions: the proportion of first-level high-risk paths is 0, the proportion of second-level high-risk paths is ≤5%, and both the time consumption and ecological loss objectives are better than 110% of the original strategy (i.e., not significantly worse than the original strategy), and finally retain 3 to 5 optimal solutions.

[0132] The selected optimal solution is decoded to generate a resilient scheduling policy that includes the following elements:

[0133] Main and Backup Path List: Each main path is matched with two backup paths (arranged in ascending order of impedance coefficient after penalty), and the activation conditions for backup paths are clearly defined (such as a main path delay exceeding 20 minutes, a sudden increase in wind speed ≥5m / s, etc.); Resource Redundancy Configuration Scheme: At core nodes (such as the nodes closest to sensitive areas), an additional 15%–20% of emergency supplies (absorbent materials, oil booms, etc.) and 10%–15% of vessel capacity are reserved, and the activation permissions of redundant resources are marked (requiring confirmation from on-site command personnel); Dynamic Monitoring Scheme: For wind field and ocean current parameters along high-risk paths, the monitoring frequency is increased from 30 minutes / time to 10 minutes / time, and the data is transmitted to the optimization module in real time. The path risk is reassessed every 30 minutes; A strategy document in PDF format (including path map, resource list, and monitoring point distribution) and structured data (XML format) that can be imported into the emergency command system are generated to ensure that command personnel can directly call and execute it.

[0134] In this embodiment of the invention, a refined threshold matrix and hierarchical triggering mechanism avoid missed or false alarms; reduced instruction generation and transmission latency buys valuable time for emergency response. Connectivity verification and emergency backup path settings ensure the architecture remains connected even after high-risk paths are penalized; reduced weight matrix update response time meets real-time scheduling requirements. Ant colony optimization with parameter adjustments reduces the probability of new strategy failure, and redundant resource configuration and dynamic monitoring further enhance the strategy's adaptability to environmental changes, resulting in improved emergency resource availability in actual execution.

[0135] like Figure 2As shown, embodiments of the present invention also provide an improved ant colony algorithm-based marine oil spill emergency resource scheduling system, comprising:

[0136] The data acquisition module is used to collect environmental dynamic parameters of oil spill accidents, including the coordinates of the leak source, the oil spill diffusion rate, wind field vectors, and ocean current field data.

[0137] The module is used to construct a dynamic resource scheduling architecture based on environmental dynamic parameters, with emergency resource hubs as spatial nodes and resource transportation channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients.

[0138] The correction module is used to input the navigation impedance coefficient into the ant colony algorithm framework, embed a hydrodynamic drag correction term into the path transition probability function, and generate a corrected path transition probability function.

[0139] The optimization module is used to perform multi-objective ant colony collaborative optimization on a dynamic resource scheduling architecture based on the modified path transition probability function, initialize the population to construct scheduling path solutions, calculate the time consumption objective and the ecological loss objective, iteratively update the pheromone distribution until convergence, and generate a multi-agent non-dominated scheduling strategy.

[0140] The evaluation module is used to calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy.

[0141] The scheduling module is used to trigger an anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy when the probability of scheduling failure exceeds the risk tolerance threshold, with the expected value of ecological damage as a non-linear penalty weight.

[0142] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0143] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0144] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0145] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An improved ant colony algorithm for emergency resource scheduling in marine oil spills, characterized in that, The method includes: Step 1: Collect environmental dynamic parameters of the oil spill accident, including the coordinates of the leak source, the oil spill diffusion rate, wind field vectors, and ocean current field data; Step 2: Based on environmental dynamic parameters, construct a dynamic resource scheduling architecture with emergency resource hubs as spatial nodes and resource transport channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients. Step 3 involves inputting the navigation impedance coefficient into the ant colony algorithm framework and embedding a hydrodynamic resistance correction term into the path transition probability function to generate a modified path transition probability function. This includes: extracting the navigation impedance coefficient matrix and mapping it to the ant colony algorithm path cost base value set; calculating the relative drag coefficient for each segment based on the path cost base value set and wind field vector data; inputting the relative drag coefficient into the ship motion characteristic knowledge base and matching the corresponding power loss compensation parameters; inputting the path cost base value set and power loss compensation parameters into a dynamic fusion processor and performing piecewise linear interpolation to generate a dynamic hydrodynamic resistance correction term set for each segment; injecting the dynamic hydrodynamic resistance correction term set into the cost calculation kernel of the path transition probability function, reconstructing the probability calculation logic by replacing the standard distance cost variable with the corrected comprehensive cost variable; and compiling and generating an executable modified path transition probability function entity based on the reconstructed probability calculation logic. Step 4: Based on the modified path transition probability function, perform multi-objective ant colony cooperative optimization on the dynamic resource scheduling architecture, initialize the population to construct scheduling path solutions, calculate the time consumption objective and the ecological loss objective, iteratively update the pheromone distribution until convergence, and generate a multi-agent non-dominated scheduling strategy. Step 5: Based on the multi-agent non-dominated scheduling strategy, calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas. This includes: taking the time-series path set in the multi-agent scheduling strategy as input, combining wind field vectors and ocean current field data for Monte Carlo simulation to obtain the probability distribution of navigation time deviation for each path; based on the navigation time deviation probability distribution, calculating the conditional probability of path failure through a preset resource scheduling timeliness constraint threshold, and integrating node resource capacity constraint configuration to calculate the probability of resource supply interruption; inputting the conditional probability of resource scheduling failure into a Bayesian risk decision engine, combining it with an ecological loss target quantification method to obtain a set of oil spill diffusion path failure scenarios; based on the failure scenario set, calculating the increase in the polluted area of ​​sensitive areas under each scenario using oil spill diffusion rate and electronic nautical chart data on the distribution of ecologically sensitive areas; taking the increase in the polluted area as input, performing a probability-weighted integral based on the failure conditional probability to obtain the expected value of damage to ecologically sensitive areas. Step 6: If the probability of scheduling failure exceeds the risk tolerance threshold, the expected value of ecological damage is used as a non-linear penalty weight to trigger the anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy.

2. The method for emergency resource scheduling in marine oil spill response based on an improved ant colony algorithm according to claim 1, characterized in that, Step 2 includes: Step 21: Analyze the geodetic coordinates of the leakage source, locate the latitude and longitude coordinates of the emergency resource hub in the electronic nautical chart coordinate system, generate a set of spatial node coordinates, calculate the Euclidean distance between adjacent nodes, and generate a set of basic navigation corridors by combining the ocean current velocity potential field. Step 22: Input the basic navigation corridor set into the wind field feasibility filter, remove the corridors with headwind exceeding the threshold according to the wind field gradient vector, and obtain the feasible directional connection relationship set; based on the feasible directional connection relationship set, call the resource management database to match the ship and material inventory of each node to obtain the node resource capacity constraint configuration; calculate the navigation time cost of each connection according to the environmental parameters and generate the navigation impedance coefficient matrix. Step 23: Integrate the spatial node coordinate set, feasible directional connection relationship set, node resource capacity constraint configuration and navigation impedance coefficient matrix to construct a dynamic resource scheduling architecture.

3. The method for emergency resource scheduling in marine oil spill response based on an improved ant colony algorithm according to claim 2, characterized in that, Step 4 includes: Step 41: Based on the configuration of spatial node topology and resource capacity constraints of the dynamic resource scheduling architecture, the initial scheduling path solution set is generated by assigning a starting node to each ant and calling the modified path transition probability function to select feasible nodes. Step 42: Take the initial scheduling path solution set as input, process the navigation impedance coefficient matrix and power loss compensation parameters of each path in the solution set to accumulate and calculate the total path time consumption target, and process the oil pollution diffusion rate to calculate the increase in the pollution area of ​​the ecologically sensitive area caused by the path time difference as the ecological loss target. Step 43: The calculated bi-objective values ​​are used to identify the Pareto front solution set through non-dominated sorting. Based on the front solution set, path pheromones are released and pheromone evaporation operations are performed on the remaining solutions to generate a new generation of scheduling path solution set. Step 44: Take the new generation of scheduling path solution set as input. If the Pareto front solution sets of three consecutive generations coincide, it is determined that the process has converged. Decode the final non-dominated solution set to generate a multi-agent scheduling strategy.

4. The marine oil spill emergency resource scheduling method based on an improved ant colony algorithm according to claim 3, characterized in that, If the probability of scheduling failure exceeds the risk tolerance threshold, an anti-disturbance rescheduling mechanism is triggered, using the expected value of ecological damage as a non-linear penalty weight, to generate a resilient scheduling strategy, including: Step 61: Perform a dynamic risk situation comparison operation between the resource scheduling failure condition probability and the preset risk tolerance threshold. If the failure condition probability exceeds the threshold, generate a rescheduling activation instruction. Based on the rescheduling activation instruction, input the expected value of ecological damage into the nonlinear penalty modeling unit to construct a penalty weight function. Step 62: Identify the set of flight segments whose failure condition probability exceeds the risk threshold, mark them as high-risk failure path sets, apply the penalty weighting function to the high-risk failure path set, and perform penalty amplification calculation on the flight impedance coefficient of each failure path. Step 63: Update the navigation impedance coefficient after penalty amplification to the connection relationship weight matrix of the dynamic resource scheduling architecture to generate the disturbance resistance enhancement architecture; based on the disturbance resistance enhancement architecture, call the modified path transition probability function entity to re-execute multi-objective ant colony cooperative optimization to generate an environmental disturbance resilience scheduling strategy.

5. A method for emergency resource scheduling in marine oil spill response based on an improved ant colony algorithm according to claim 4, characterized in that, Step 32: Input the path cost base value set and power loss compensation parameters into the dynamic fusion processor, and perform piecewise linear interpolation to generate a set of dynamic fluid resistance correction terms for each flight segment, including: Step 321: Map the ocean current impedance values ​​in the path cost base value set to the preset ocean current velocity classification range, and map the power loss compensation parameters to the preset wind resistance compensation classification level; generate a set of discrete positioning points for each segment in the ocean current-wind resistance two-dimensional parameter space based on the mapping results, and match interpolation calculation units according to the spatial distribution of positioning points; extract the vertex reference values ​​of the matching interpolation calculation units, and calculate the normalized position coefficient of the segment in the parameter space; Step 322: First, interpolate the normalized position coefficients along the ocean current dimension to generate transition boundary values, and then interpolate along the wind resistance dimension to generate the comprehensive hydrodynamic correction value for the segment; aggregate the comprehensive hydrodynamic correction values ​​of all segments to generate a set of dynamic fluid resistance correction terms.

6. An improved ant colony algorithm-based marine oil spill emergency resource scheduling system, wherein the system implements the method as described in any one of claims 1 to 5, characterized in that, include: The data acquisition module is used to collect environmental dynamic parameters of oil spill accidents, including the coordinates of the leak source, the oil spill diffusion rate, wind field vectors, and ocean current field data. The module is used to construct a dynamic resource scheduling architecture based on environmental dynamic parameters, with emergency resource hubs as spatial nodes and resource transportation channels as directional connections. Node attributes define resource capacity constraints, and connection weights encode real-time navigation impedance coefficients. The correction module is used to input the navigation impedance coefficient into the ant colony algorithm framework, embed a hydrodynamic drag correction term into the path transition probability function, and generate a corrected path transition probability function. The optimization module is used to perform multi-objective ant colony collaborative optimization on a dynamic resource scheduling architecture based on the modified path transition probability function, initialize the population to construct scheduling path solutions, calculate the time consumption objective and the ecological loss objective, iteratively update the pheromone distribution until convergence, and generate a multi-agent non-dominated scheduling strategy. The evaluation module is used to calculate the conditional probability of resource scheduling failure and the expected value of damage to ecologically sensitive areas based on the multi-agent non-dominated scheduling strategy. The scheduling module is used to trigger an anti-disturbance rescheduling mechanism to generate a resilient scheduling strategy when the probability of scheduling failure exceeds the risk tolerance threshold, with the expected value of ecological damage as a non-linear penalty weight.

7. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images