A marine emergency decision-making method and system based on dynamic phase perception

Abstract

This invention discloses a maritime emergency decision-making method and system based on dynamic stage perception, comprising the following steps: constructing a first-stage objective function for accident rescue with the goal of minimizing the total emergency response time; solving the first-stage objective function to obtain a first-stage maritime accident resource allocation scheme; judging the rescue time of the accident and the overall completion status of the rescue mission; when entering the second stage, constructing a second-stage overall objective function with the goals of minimizing the total emergency response time, maximizing mission completion efficiency, and minimizing the switching cost of resource allocation schemes; and solving the second-stage adjusted maritime accident resource allocation scheme based on an improved non-dominated sorting genetic algorithm. This invention, through accident state recognition and transformation and rolling time-domain optimization mechanisms, enables the emergency decision-making scheme to adaptively adjust with the accident process, improving robustness in dealing with maritime uncertainties.

Description

Technical Field

[0001] This invention belongs to the field of marine emergency technology and relates to a marine emergency decision-making method and system based on dynamic phase perception. Background Technology

[0002] Maritime emergency decision-making is dynamic, uncertain, and complex with multiple objectives, presenting certain limitations in current research and practice. First, existing models often focus on static, single events. For example, in maritime search and rescue accidents, decision-making models may only focus on improving the success rate of the search and rescue operation. Even as the rescue mission nears its end, improving the success rate remains the sole optimization objective, neglecting economic costs and potentially leading to significant resource waste. In oil spill emergencies, even if the model considers the oil slick weathering process, the structure of existing models remains fixed. Therefore, existing models fail to abstract the common phased patterns of maritime emergency accidents, resulting in insufficient model universality. Second, the development of existing decision support systems has transitioned from manual to intelligent decision-making. While no longer relying entirely on the decision-maker's subjective experience, excessive reliance on algorithms to automatically generate emergency plans, excluding the decision-maker's opinion from the decision-making process, often leads to emergency plans that do not meet real-world needs or are overly idealistic. Therefore, existing decision support systems fail to fully leverage the advantages of combining the objective computational capabilities of algorithms with the subjective experience of decision-makers through human-machine collaboration. Finally, while existing metaheuristic algorithms have been widely applied in various fields of maritime emergency response, integrating elements such as rolling time-domain optimization and phase transitions into a universal emergency decision-making framework remains an underexplored area. Therefore, this invention aims to construct a systematic decision-making framework to address the challenges posed by existing technologies in terms of lack of dynamic adaptability, human-machine collaboration, and universality.

[0003] Existing maritime emergency decision-making methods are mostly static and single-stage, which makes it difficult for the decision-making process to adapt to the dynamic evolution of the accident and lacks a phased approach. At the same time, existing decision-making methods often cannot adjust the decision plan according to the rolling updates of real-time on-site information. The decision-making methods are mostly set for specific scenarios, and the plan has poor universality. In addition, plans that are entirely generated automatically by algorithms or judged by expert experience will result in low decision-making efficiency and credibility. Summary of the Invention

[0004] To address the aforementioned problems, the technical solution adopted by this invention is: a maritime emergency decision-making method based on dynamic phase perception, comprising the following steps:

[0005] To acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; Based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident, a first-stage objective function for accident rescue is constructed with the goal of minimizing the total emergency response time. Based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved to obtain the allocation scheme of dispatchable rescue resources for the first stage of maritime accidents. The rescue time and overall completion status of the rescue mission for the accident are compared with the threshold for the rescue time and the overall completion status of the rescue mission for the accident to determine whether to enter the second stage of accident rescue and to adjust the allocation plan of the available rescue resources for maritime accidents. When entering the second phase, construct a second-phase overall objective function that includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. When not entering the second phase, continue to execute the rescue resource allocation scheme of the first phase of accident rescue. Based on the improved non-dominated sorting genetic algorithm, the objective function of the second stage is solved to obtain the second stage of the rescue resource allocation scheme after the maritime accident adjustment.

[0006] Furthermore: The expression for minimizing the total emergency response time of the objective function in the first stage is as follows:

[0007] Where M represents the total number of available rescue units. Spatial coordinates of the base location where the rescue unit or emergency resource is stationed. Target: The spatial coordinates of the location where the target event requiring rescue occurred. The travel time for rescue unit J from its base to the target point. : 0-1 decision variable, indicating whether to dispatch unit j.

[0008] Furthermore, the expression for maximizing task completion efficiency is as follows:

[0009] Where M represents the total number of available emergency resource units; N: The total number of incidents or tasks that need to be addressed (which may include multiple concurrent incidents of different types).

[0010] L: Types and number of emergency resource units; in:

[0011] : Indicates the expected effectiveness that resource unit j can generate in handling accident point i; : Match score; : The maximum recoverable loss or highest completion rate of task point i; : Indicates the type of resource unit j; : Indicates the main accident type that needs to be addressed at accident point i; Furthermore: The expression for minimizing the total emergency response time in the second-stage objective function is as follows:

[0012] : Represents a function that calculates the estimated travel or journey time required to travel from position A to position B; : The current position of resource unit j before executing the plan for this stage; The geographical location of the accident point or mission point i after drifting and spreading over time; : is a 0-1 decision variable. =1 indicates that resource unit j is scheduled and assigned to task point i for processing. =0 indicates that this allocation will not be performed.

[0013] Furthermore, the expression for minimizing the switching cost of the resource allocation scheme is as follows:

[0014] : Represents a function that calculates the total cost of moving resources from location A to location B; : The current actual location of resource unit j.

[0015] : The location of accident point i.

[0016] Furthermore, the process of comparing the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively to determine whether to enter the second stage of accident rescue is as follows: Set the rescue time threshold for the accident to X hours and the overall completion status of the rescue mission to Y%. When an accident has occurred and there are people who need to be rescued, and the rescue time is ≥X hours, then the second stage of accident rescue begins. If an accident has occurred and there are materials and resources that need to be rescued, and the overall completion rate of the rescue mission is ≥Y%, then the second stage of accident rescue will begin. Otherwise, the accident rescue will continue with the first phase of the rescue plan.

[0017] Furthermore: Based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved, and the process of obtaining the allocation scheme of dispatchable rescue resources for the first stage of maritime accidents is as follows: S31: Initialize the population: Randomly initialize the population P and set relevant parameters, including population size, number of iterations and mutation operator. S32: Calculate fitness value: Based on the objective function of the first stage, calculate the fitness response time of each individual in the population, which serves as a standard for evaluating the merits of the first stage's maritime accident relocation plan. S33: Mutation and Crossover Operations: The mutation and crossover operators of the differential evolution algorithm are used to operate on the individuals in the population to generate the experimental population U; S34: Greedy selection update: Compare the fitness values ​​of individuals in the original population P with those in the experimental population U, and execute a greedy selection strategy to update population P; S35: Read the current best individual: Identify and extract the best individual X_best in the current generation from the updated population P; S36: Simulated Annealing Local Search: Based on the current best individual, a simulated annealing local search mechanism is introduced to generate a new neighborhood solution through small step-size flip perturbation; S37: Neighborhood Solution Judgment and Replacement: Determine whether the function value of the new solution after perturbation is less than the function value of the original solution. If it is less, the new solution is better, and the current best solution in the population is replaced; otherwise, the original solution remains unchanged. S38: Termination Condition Determination: Check if the preset termination condition is met. If not, return to S33 to perform mutation and crossover operations. If met, stop the iteration and output the globally optimal rescue resource allocation plan.

[0018] Furthermore, the process of solving the objective function of the second stage based on the improved non-dominated sorting genetic algorithm to obtain the second-stage adjusted rescue resource allocation scheme for maritime accidents is as follows: S61 Initialization and Scheme Inheritance: Initialize the parameters of the improved non-dominated sorting genetic algorithm and inherit the optimal scheme output in the first stage as the basis for solving the second stage; S62 Multi-Objective Evaluation: For the objective function of the second stage, the function values ​​of three objectives—response time, task efficiency, and execution stability—are evaluated simultaneously. S63 Non-dominated ranking and crowding calculation: Perform fast non-dominated ranking on individuals in the population and calculate the crowding of individuals in the target space to establish dominance relationships between individuals; S64 evolutionary operator operations: generate offspring population Q through selection, crossover, and mutation operations; S65 Population Merging and Selection: The parent population P and the offspring population Q are merged into a combined population R. An elite retention strategy is used to select high-performing individuals from R to form a new generation population P. S66 extracts non-dominant individuals of Rank 1 from the current new population P, forming the current Pareto front. S67 Front Extension Search: Introduces a front extension strategy based on simulated annealing to perform perturbation extension search on individuals on the front. S68 Update Non-Dominated Solution Set: Update the non-dominated solution set based on the extended search results to ensure the quality of Pareto optimal solutions in the non-dominated solution set; S69 Iteration and Output: Determine if the set maximum number of iterations has been reached. If not, return to the non-dominated sorting step to continue the loop. If it has been reached, output the final Pareto optimal solution set as a reference scheme for the adjusted allocation of rescue resources.

[0019] A maritime emergency decision-making system based on dynamic phase perception includes: Information perception module: used to acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; The first-stage construction module is used to construct the first-stage objective function for accident rescue, which aims to minimize the total emergency response time, based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident. The first-stage solution module is used to solve the objective function of the first stage of accident rescue based on the improved differential evolution algorithm, so as to obtain the first-stage allocation plan of the available rescue resources for maritime accidents. The transition phase judgment module is used to compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively, to determine whether to enter the second phase of accident rescue and to adjust the allocation plan of available rescue resources for maritime accidents. The second-stage judgment module is used to construct a second-stage overall objective function when entering the second stage, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. If the second phase is not initiated, the rescue resource allocation plan for the first phase of the accident rescue will continue to be implemented. The second-stage solution module is used to solve the objective function of the second stage based on the improved non-dominated sorting genetic algorithm, so as to obtain the second-stage adjusted rescue resource allocation plan for maritime accidents.

[0020] The present invention provides a maritime emergency decision-making method and system based on dynamic phase perception, which has the following advantages and benefits: Enhancing dynamic adaptability: Through accident state identification and transformation and rolling time-domain optimization mechanism, emergency decision-making schemes are adaptively adjusted as the accident progresses, effectively improving the system's robustness in dealing with maritime uncertainties.

[0021] A two-stage optimization structure is adopted: the response and optimization are separated into two stages. The first stage aims to minimize the emergency response time, seize the critical window of the incident, and control the escalation of the situation. The second stage coordinates resource scheduling under multiple constraints to maximize the effectiveness of the response. This design conforms to the inherent timing requirements of maritime emergency operations.

[0022] Constructing a human-machine collaborative decision-making mechanism: This mechanism involves building an intelligent optimization algorithm to generate decision-making schemes, which are then ultimately approved by the decision-maker. While leveraging the efficiency of the algorithm in solving multiple constraints and evaluating schemes, this mechanism retains the decision-maker's decision-making power based on experience and the overall situation. It integrates the objectivity of the intelligent optimization algorithm with the decision-maker's experience, ensuring the scientific nature of the decision-making process while maintaining the integrity of the decision-making command system and improving the feasibility of the schemes.

[0023] Modular and scalable: It adopts a modular design, and its phase division, objective function setting, optimization algorithm selection and other parts can be flexibly configured and expanded according to specific emergency scenarios (such as search and rescue, oil spill, fire, etc.), which makes it applicable to diverse maritime emergency missions. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of the method described in this application; Figure 2 This is the solution process for the first phase of the maritime accident rescue resource allocation plan; Figure 3 This is the solution process for the second phase of the maritime accident relocation plan; Figure 4It is a comprehensive diagram of the system's overall architecture and human-machine collaborative decision-making; Figure 5 This is a stage transition judgment logic diagram; Figure 6 This is a schematic diagram of the Pareto front for the second-stage multi-objective optimization; where (a) is a diagram of response time versus efficiency, and (b) is a diagram of efficiency versus stability. Figure 7 This is a schematic diagram of the three-dimensional Pareto front for the second-stage multi-objective optimization. Figure 8 This is a schematic diagram of the human-machine collaborative decision-making interaction interface, where (a) is the initial interface and (b) is the weight adjustment interface. Detailed Implementation

[0026] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Figure 1 This is a flowchart of the method described in this application; A maritime emergency decision-making method based on dynamic phase perception includes the following steps: S1: Obtain marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; S2: Based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident, construct the first-stage objective function for accident rescue with the goal of minimizing the total emergency response time; S3: A rolling time-domain optimization strategy with a first-step length (improved differential evolution algorithm) is adopted to solve the objective function of the first stage of accident rescue. Based on the dynamic evolution of the environment and mission status, the allocation scheme of the available rescue resources for the first stage of maritime accidents is periodically updated; the first-step length can be 0.5h. S4: Compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively to determine whether to enter the second stage of accident rescue and adjust the allocation plan of available rescue resources for maritime accidents. S5: When entering the second phase, construct the overall objective function for the second phase, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. When not entering the second phase, continue to execute the rescue resource allocation scheme of the first phase of accident rescue. S6: A rolling time-domain optimization strategy with a second step length (based on an improved non-dominated sorting genetic algorithm) is adopted to solve the objective function of the second stage. The rescue plan is dynamically corrected and smoothly switched through a continuously sliding optimization window, resulting in the adjusted rescue resource allocation plan for the second stage of the maritime accident. The second step length is 2 hours. Steps S1 / S2 / S3 / S4 / S5 / S6 are executed sequentially; Marine environmental data, maritime accident data, and available rescue resources in the event of a maritime accident: The marine environmental data includes: Real-time meteorological and hydrological data and forecasts: including wind field, current field, wave data, and visibility data for the accident area. Geospatial data: including shoreline data, channel distribution, water depth data, and distribution of navigational obstructions near the accident site; Environmental impact factors: Seawater temperature and dynamic change parameters of the search area used to calculate the survival probability of people who fall into the water.

[0029] The maritime accident data: Basic characteristic information: including the exact latitude and longitude coordinates of the accident, the initial time of the accident, and the type of accident; Attributes of distressed targets: including the number of distressed targets, the current estimated survival probability of distressed targets, and the drift and diffusion trend of targets; Key threshold parameters include the response time threshold for transitioning from the first stage to the second stage and preset indicator thresholds for the completion of the rescue mission.

[0030] The data on the available rescue resources includes: Dynamic location and status information: including the current latitude and longitude of the rescue unit and its current mission status; Mobility and load parameters: including the cruising speed of each rescue unit, maximum search and rescue speed, and preparation time from receiving instructions to deployment; Rescue effectiveness parameters include the effective search width, maximum passenger capacity, search and rescue equipment configuration, and continuous operation capability of each unit.

[0031] Furthermore, in steps S3 and S6, the solution process adopts a rolling time-domain optimization mechanism to update the scheme within each stage: in the first stage, the decision step size is set to 0.5h, and high-frequency rolling is used to cope with the uncertainty in the early stage of the accident; in the second stage, the decision step size is set to 2h, and medium-frequency rolling is used to reduce the resource consumption caused by scheme switching while ensuring rescue efficiency.

[0032] Furthermore, the expression for minimizing the total emergency response time of the objective function in the first stage is as follows:

[0033] Where M represents the total number of available rescue units. Spatial coordinates of the base location where rescue unit j is stationed. Target: The spatial coordinates of the location where the target event requiring rescue occurred. The travel time for rescue unit J from its base to the target point. : 0-1 decision variable, indicating whether to dispatch unit j.

[0034] For any unit j, if it is dispatched ,So: Its sailing time will be included in the total time. If it is not dispatched ( ),So It does not contribute to the total time. Therefore, the summation of this objective function ultimately calculates the sum of the flight times of all selected units. The goal of the model is to select a set of units that minimizes their total arrival time.

[0035] Furthermore, the expression for maximizing task completion efficiency is as follows:

[0036] Where M represents the total number of available emergency resource units (e.g., rescue boats, fireboats, debris removal vessels, helicopters). N: The total number of incidents that need to be handled / the total number of task points (which may include multiple concurrent incidents of different types).

[0037] L: Types and quantities of emergency resource units; (e.g., 1 = professional search and rescue vessel, 2 = fireboat, 3 = oil spill recovery vessel); in:

[0038] : represents the expected efficiency that resource unit j can generate in handling accident point i; it is an interval [0, 1] or a positive real number.

[0039] : Match score; when the resource type and task type match perfectly (e.g., fireboat to fire), the score is less than 1 when they do not match perfectly (e.g., ordinary boat to oil spill), and the score is 0 when they do not match at all. : The maximum recoverable loss or highest completion rate at task point i; this is a normalized value used to unify the effectiveness of different types of tasks to a comparable scale (e.g., a search and rescue task might be "maximum number of people that can be saved", while an oil spill task might be "maximum amount of oil that can be recovered").

[0040] : Indicates the type of resource unit j; : Indicates the main accident type that needs to be addressed at accident point i; for example: 1 = search and rescue, 2 = fire, 3 = oil spill). The objective function E aims to maximize the expected total effectiveness generated by all allocated resources. For any combination of resource unit j and incident point i, if they are matched and allocated (x{ij}=1), then the expected effectiveness (Effectiveness_{ij}) of that unit in handling this incident is fully factored into the total effectiveness E. If they are not allocated (x{ij}=0), then the combination contributes nothing to the total effectiveness. This function guides the model to precisely allocate the most suitable resources to the tasks that best utilize their value, thereby optimizing the handling effect at the global level.

[0041] Furthermore: The expression for minimizing the total emergency response time in the second-stage objective function is as follows:

[0042] : Represents a function that calculates the estimated travel or journey time required to travel from position A to position B; : The current position of resource unit j before executing the current phase plan; (usually the position after the first phase ends, or the base).

[0043] The geographical location of the accident site or mission point i; : is a 0-1 decision variable. =1 indicates that resource unit j is scheduled and assigned to task point i for processing. =0 indicates that this allocation will not be performed.

[0044] The objective function T aims to minimize the total travel time of all scheduled resources to their new task points. For any unit j, if it is reassigned to a new task point i ( =1), then it starts from its current position. Go to sailing time This will be fully included in the total response time T. If the unit is not reallocated (i.e., for all i, ( If the value is 0 or remains in its original position, it does not generate new travel time. This function prompts the model to prioritize scheduling idle or mobile resources closer to the incident point to reduce overall response latency.

[0045] Furthermore, the expression for minimizing the switching cost of the resource allocation scheme or maximizing the stability of scheme execution is as follows:

[0046] : Represents a function that calculates the total cost of moving resources from location A to location B; its core is usually distance cost, such as Distance(A,B), but it can also be extended to include other costs, such as the risk cost of crossing complex sea areas, fuel consumption cost, etc.

[0047] : The current actual location of resource unit j.

[0048] : The location of accident point i.

[0049] The objective function S aims to minimize the total resource relocation cost incurred by implementing the new scheduling scheme. For any unit j, if it is required to leave its current position in this round of the scheme... Head to the new mission point i ( =1), then the cost of this transfer This will be included in the total variable cost S. If the unit is not redistributed, its mobilization cost is 0. This function encourages the model to prioritize the proximity or standby of resources while meeting efficiency and response time requirements, reducing large-scale, long-distance redeployment of supplies, thereby reducing the complexity of emergency decision-making and command, and improving the feasibility of the plan and the continuity of the emergency decision-making process.

[0050] Furthermore, the process of comparing the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively to determine whether to enter the second stage of accident rescue is as follows: Set the rescue time threshold for the accident to X hours and the overall completion rate of the rescue mission to X%; When an accident has occurred and there are personnel requiring rescue, and the rescue time is ≥X hours, then the second stage of accident rescue begins; the optimal value for X is 4 hours, which is based on the consensus on the "golden rescue time" in international maritime search and rescue practice.

[0051] When an accident has occurred and there are materials and resources that need to be rescued, and the overall completion rate of the rescue mission is ≥Y%, then the second stage of accident rescue is entered; Y is 95, which means that when the overall mission completion rate predicted by the system reaches 95%, the rapid response goal is considered to have been basically achieved. Otherwise, the accident rescue will continue with the first phase of the rescue plan.

[0052] The decision-making framework of this invention can configure different algorithms at the optimization algorithm level according to different scenario requirements. Each optimization module does not depend on a specific algorithm, but rather configures a corresponding metaheuristic optimization algorithm according to the scenario requirements. The following uses the maritime search and rescue scenario as an example to give an algorithm configuration scheme.

[0053] ① Algorithm type configuration The first stage of optimization process: The core requirement of this stage is to generate an executable solution in a short time. Therefore, a single-objective meta-heuristic algorithm with fast convergence speed and simple operation can be configured. In this example, the differential evolution algorithm is configured as the intelligent optimization algorithm for this stage.

[0054] The second stage of optimization involves dealing with multiple conflicting optimization objectives. Therefore, a multi-objective meta-heuristic algorithm that can effectively explore the Pareto frontier can be configured. In this example, a non-dominated sorting genetic algorithm is configured as the intelligent optimization algorithm for this stage.

[0055] Table 1. Optimization Algorithm Parameter Configuration Table (Taking Maritime Search and Rescue Scenario as an Example)

[0056] Figure 2 This is the solution process for the first phase of the maritime accident rescue resource allocation plan; Furthermore, based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved, and the process of obtaining the first stage of the allocation scheme for dispatchable rescue resources for maritime accidents is as follows: S31: Initialize the population: Randomly initialize the population P and set the relevant parameters of the algorithm, including the population size, number of iterations and mutation operator. S32: Calculate fitness value: Based on the objective function of the first stage, calculate the fitness response time of each individual in the population, which serves as the standard for evaluating the merits of the scheme; S33: Mutation and Crossover Operations: The mutation and crossover operators of the differential evolution (DE) algorithm are used to operate on the individuals in the population to generate the experimental population U; S34: Greedy selection update: Compare the fitness values ​​of individuals in the original population P with those in the experimental population U, and execute a greedy selection strategy to update population P; S35: Read the current best individual: Identify and extract the best individual X_best in the current generation from the updated population P; S36: Simulated Annealing (SA) Local Search: Based on the current best individual, a simulated annealing local search mechanism is introduced to generate new solutions in the neighborhood through small step-size flip perturbation, so as to enhance the local optimization ability of the algorithm. S37: Neighborhood Solution Judgment and Replacement: Determine whether the function value of the new solution after perturbation is less than the function value of the original solution. If it is less, the new solution is better, and the current best solution in the population is replaced; otherwise, the original solution remains unchanged. S38: Termination Condition Determination: Check if the preset termination condition is met (such as reaching the maximum number of iterations). If not met, return to perform mutation and crossover operations; if met, stop iteration and output the globally optimal rescue resource allocation plan.

[0057] The sole criterion for determining whether a new solution is superior to the original solution is the magnitude of the objective function value in the first stage (i.e., the total emergency response time). Since the objective of the first stage is to minimize the total emergency response time, the judgment logic is: if the total response time calculated by the new solution is less than the total response time of the original solution, then the new solution is considered "superior," as detailed below: First, calculate the total response time T corresponding to the 'new solution (neighborhood solution)' generated after the perturbation. new ; Then, it is compared with the total response time T corresponding to the 'original solution (current best individual)'. old Compare them.

[0058] If T new <T old If the new solution is determined to be better, the replacement operation is performed.

[0059] If T new ≥Told: Determines that the new solution is not better than the original solution, and keeps the original solution unchanged.

[0060] Figure 3 This is the solution process for the second phase of the maritime accident relocation plan; Furthermore, based on the improved non-dominated sorting genetic algorithm, the objective function for the second stage is solved, and the process of obtaining the second-stage adjusted rescue resource allocation scheme is as follows: S61 Initialization and Scheme Inheritance: Initialize the parameters of the non-dominated sorting genetic algorithm and inherit the optimal scheme output in the first stage as the basis for solving the second stage; S62 Multi-Objective Evaluation: For the objective function of the second stage, the function values ​​of three objectives—response time, task efficiency, and execution stability—are evaluated simultaneously; thus establishing the basis for comparison (three objectives). S63 Non-dominated sorting and crowding calculation: Performs fast non-dominated sorting on individuals in the population and calculates the crowding of individuals in the target space to establish dominance relationships between individuals; the mention of "establishing dominance relationships between individuals" is the core step of Rank1 determination; S64 evolutionary operator operations: generate offspring population Q through selection, crossover, and mutation operations; S65 Population Merging and Selection: The parent population P and the offspring population Q are merged into a combined population R. An elite retention strategy is used to select high-performing individuals from R to form a new generation population P. S66 extracts non-dominated individuals of Rank 1 from the current new population to form the current Pareto front; the determination of Rank 1 is based on the Pareto Dominance rule in the Non-Dominated Sorting Genetic Algorithm (NSGA-II).

[0061] S67 Front Extension Search: Introduces a front extension strategy based on simulated annealing (SA) to perform perturbation extension search on individuals on the front, aiming to broaden the distribution range of the solution set and improve the diversity of solutions; S68 Update Non-Dominated Solution Set: Update the non-dominated solution set based on the extended search results to ensure the quality of Pareto optimal solutions; S69 Iteration and Output: Determine if the set maximum number of iterations has been reached. If not, return to the non-dominated sorting step and continue the loop; if so, output the final Pareto optimal solution set as a reference scheme for the adjusted allocation of rescue resources.

[0062] In steps S62 and S63, all individuals in the population are compared pairwise for three objective functions (response time, task efficiency, and execution stability).

[0063] A maritime emergency decision-making system based on dynamic phase perception includes: Information perception module: used to acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; The first-stage construction module is used to construct the first-stage objective function for accident rescue, which aims to minimize the total emergency response time, based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident. The first-stage solution module is used to solve the objective function of the first stage of accident rescue based on the improved differential evolution algorithm, so as to obtain the first-stage allocation plan of the available rescue resources for maritime accidents. The transition phase judgment module is used to compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively, to determine whether to enter the second phase of accident rescue and to adjust the allocation plan of available rescue resources for maritime accidents. The second-stage judgment module is used to construct a second-stage overall objective function when entering the second stage, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. If the second phase is not initiated, the rescue resource allocation plan for the first phase of the accident rescue will continue to be implemented. The second-stage solution module is used to solve the objective function of the second stage based on the improved non-dominated sorting genetic algorithm, so as to obtain the second-stage adjusted rescue resource allocation plan for maritime accidents.

[0064] Example 1 A maritime emergency decision-making method based on dynamic phase perception includes the following steps: S1: Obtain marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; S2: Based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident, construct the first-stage objective function for accident rescue with the goal of minimizing the total emergency response time; S3: Based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved to obtain the allocation scheme of the available rescue resources for the first stage of maritime accidents. S4: Compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively to determine whether to enter the second stage of accident rescue and adjust the allocation plan of the available rescue resources for maritime accidents; determine whether to trigger the stage switch within the rolling cycle; assuming that 2.5 hours after the accident, the system starts a new round of rolling time domain optimization.

[0065] S5: When entering the second phase, construct the overall objective function for the second phase, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. When not entering the second phase, continue to execute the rescue resource allocation scheme of the first phase of accident rescue. S6: Based on the improved non-dominated sorting genetic algorithm, the objective function of the second stage is solved to obtain the second stage of the rescue resource allocation scheme after the maritime accident adjustment; it is used to perform multi-objective optimization during the fine optimization period and output the Pareto optimal solution set.

[0066] This method also includes integrating optimization results with expert judgment, and the decision-making process is as follows: Figure 4 As shown S7: First, the entropy weight TOPSIS method is used to recommend a solution from the Pareto solution set. If the decision-maker deems it feasible, it is adopted directly. If the decision-maker deems it infeasible, the decision-maker inputs the weights of each objective, and the system recalculates based on the new weights, and finally outputs the selected execution solution.

[0067] A maritime emergency decision-making system based on dynamic phase perception includes: Information perception module: used to acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; The first-stage construction module is used to construct the first-stage objective function for accident rescue, which aims to minimize the total emergency response time, based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident. The first-stage solution module is used to solve the objective function of the first stage of accident rescue based on the improved differential evolution algorithm, so as to obtain the first-stage allocation plan of the available rescue resources for maritime accidents. The transition phase judgment module is used to compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively, to determine whether to enter the second phase of accident rescue and to adjust the allocation plan of available rescue resources for maritime accidents. The second-stage judgment module is used to construct a second-stage overall objective function when entering the second stage, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. If the second phase is not initiated, the rescue resource allocation plan for the first phase of the accident rescue will continue to be implemented. The second-stage solution module is used to solve the objective function of the second stage based on the improved non-dominated sorting genetic algorithm, so as to obtain the second-stage adjusted rescue resource allocation plan for maritime accidents.

[0068] Human-machine collaboration module: It first uses the entropy weight TOPSIS method to recommend a solution from the Pareto solution set. If the decision-maker thinks it is feasible, it will be adopted directly. If the decision-maker thinks it is not feasible, the decision-maker will input the weights of each objective, and the system will recalculate based on the new weights and finally output the selected execution solution. Rolling Time Domain Controller: Its output is connected to the above modules and is used to periodically trigger the operation of the above system according to a preset time interval.

[0069] Figure 1 System overall architecture and human-machine collaborative decision-making integrated diagram; This diagram illustrates the overall system architecture and human-machine collaborative decision-making process of this invention: ① Horizontal data flow: Information starts from the perception module, enters the corresponding optimization module after stage judgment, and outputs the final solution through the human-machine collaborative decision-making module. During this process, the rolling time domain controller periodically triggers this process. ② Vertical control flow: The rolling time domain controller controls the operating rhythm of each module, so that the process of accident evolution and decision-making are synchronized. ③ Human-machine collaborative process: The stage output solution is judged by the human-machine collaborative decision-making module, which incorporates the decision-maker's experience judgment while ensuring the computational efficiency of the algorithm.

[0070] Figure 5 This is a stage transition judgment logic diagram; Following the accident, the first phase of accident rescue, namely the rapid response phase, begins: Core objective: Minimize the total response time of the emergency response system. This objective function aims to mobilize initial resources to the scene as quickly as possible to control the situation.

[0071] Optimization method: Efficient metaheuristic optimization algorithms (such as differential evolution algorithm DE) can be used to solve the problem, so as to ensure that a feasible solution can be obtained quickly in a short time.

[0072] Automatic transition conditions between the first phase of accident rescue and the first phase of accident rescue: During each rolling time-domain cycle, the following conditions are checked. To balance response speed and optimize accuracy, the system switches from the first stage of accident rescue to the second stage when any of the following conditions are met, ensuring that the system can quickly adapt to the dynamic environment: Condition A (Schedule Condition): Estimated task completion rate (t) ≥ Threshold1 The estimated task completion rate (t) is a proxy indicator calculated using real-time data, and Th1 is a preset progress threshold, set based on the experience-based completion standards for this type of emergency task. For example: Regarding search and rescue: The survival probability is a function that decays over time.

[0073] Regarding the oil spill: Th1 is a preset threshold, such as 95%.

[0074] For various emergency scenarios, the estimated task completion rate is calculated based on a specific scenario assessment model. Input real-time data such as environmental conditions and resource status, and output a normalized completion rate index and threshold. and Set according to the stage transition threshold configuration table (Table 1).

[0075] Condition B (Time Condition): . The pre-defined critical timeframes for this type of accident are based on the industry consensus regarding the "golden rescue / response window" for such accidents. For example: Regarding personnel search and rescue: It can be set to 4 hours (based on the "golden 4 hours" principle of maritime search and rescue, emphasizing early and rapid response).

[0076] Regarding the handling of the oil spill: It can be set to 24 hours (based on the "golden 24-hour" window period, after which oil film weathering intensifies and recovery efficiency decreases).

[0077] Table 2 Stage Conversion Threshold Configuration Table

[0078] Phase Two of Accident Rescue (Refinement and Optimization Period): Core objectives: Simultaneously optimize multiple objectives, including: a) minimizing total response time; b) maximizing mission completion efficiency (e.g., maximizing the total survival probability of search and rescue operations or the total amount of oil spill recovery); c) minimizing the switching costs of resource allocation schemes (ensuring the smooth execution of the schemes).

[0079] Optimization method: Use a multi-objective metaheuristic algorithm (such as non-dominated sorting genetic algorithm) to generate a set of Pareto optimal solutions.

[0080] Figure 6 This is a schematic diagram of the Pareto front for the second-stage multi-objective optimization; (a) In the figure, the horizontal axis represents response time, the vertical axis represents task completion efficiency, and the color represents the stability of the scheme execution. The lower the corresponding value, the smaller the scheme fluctuation. (b) The figure shows the trade-off relationship between the three objectives from another perspective. The horizontal axis represents task completion efficiency, the vertical axis represents the stability of the scheme execution, and the color represents response time. The lower the corresponding value, the shorter the response time.

[0081] Figure 7 This is a schematic diagram of the three-dimensional Pareto front for the second-stage multi-objective optimization. The diagram shows the overall distribution of the Pareto optimal solution set in three dimensions, and comprehensively demonstrates the trade-offs between the three objectives of response time, task completion efficiency, and scheme execution stability.

[0082] Note: This figure, generated based on simulated data, illustrates the distribution of the Pareto optimal solution set output by the second-stage optimization module. The assumptions underlying the data in the figure are: response time is negatively correlated with task completion efficiency; the stability of the solution execution is a quadratic function of response time, with optimal stability at a moderate response time; the simulated response time range is 5-30 minutes, the task completion efficiency range is 60%-95%, and the solution execution stability range is 65%-92%. This diagram simulates the distribution of 80 solutions. In actual maritime accident applications, the system calculates the objective function value based on real-time environmental and resource data to generate the corresponding Pareto optimal solution set.

[0083] Furthermore, in the human-machine collaboration phase, there are two ways to determine the solution: Automatic Recommendation: An objective weighting method is used to calculate the weights of each objective, and based on these weights, the TOPSIS method is used to automatically select a recommended solution from the Pareto solution set. Simultaneously, the objective weights of each objective are obtained. .

[0084] in: The percentage of response time. The percentage of task completion efficiency. The percentage representing the stability of the plan's implementation; Expert decision-making: The recommended solutions and their objective weights are presented to the decision-maker. If the decision-maker approves, the solution is adopted directly; if not, the decision-maker inputs custom subjective weights, and the system re-runs the standard TOPSIS based on the Pareto solution set used by the entropy-weighted TOPSIS to select the final solution.

[0085] Figure 8 This is a schematic diagram of the human-machine collaborative decision-making interaction interface; where (a) is the initial interface and (b) is the weight adjustment interface. This diagram illustrates the user interface design of the human-machine collaborative decision-making module, which includes two sub-diagrams: (a) Initial interface: The system recommends a solution. The left side is a visualization of the Pareto solution set, the right side is the detailed objective function value and objective weight of the selected solution, and the bottom provides decision-makers with two options: "Adopt the recommended solution" and "Adjust the weights and recalculate".

[0086] (b) Weight Adjustment Interface: Decision-makers can input subjective weights here based on the situation on-site, and adjust the weights of the three objectives—efficiency, time, and stability—using sliders. This diagram is a schematic of the interface design, mainly demonstrating the interaction between decision-makers and the system.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A maritime emergency decision-making method based on dynamic phase perception, characterized in that: Includes the following steps: To acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; Based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident, a first-stage objective function for accident rescue is constructed with the goal of minimizing the total emergency response time. Based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved to obtain the allocation scheme of dispatchable rescue resources for the first stage of maritime accidents. The rescue time and overall completion status of the rescue mission for the accident are compared with the threshold for the rescue time and the overall completion status of the rescue mission for the accident to determine whether to enter the second stage of accident rescue and to adjust the allocation plan of the available rescue resources for maritime accidents. When entering the second phase, construct a second-phase overall objective function that includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. When not entering the second phase, continue to execute the rescue resource allocation scheme of the first phase of accident rescue. Based on the improved non-dominated sorting genetic algorithm, the objective function of the second stage is solved to obtain the second stage of the rescue resource allocation scheme after the maritime accident adjustment.

2. The maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The objective function of the first stage aims to minimize the total emergency response time, and the expression for minimizing the total emergency response time is as follows: Where M represents the total number of available rescue units. Spatial coordinates of the base location where the rescue unit or emergency resource is stationed. Target: The spatial coordinates of the location where the target event requiring rescue occurred. The travel time for rescue unit J from its base to the target point. : 0-1 decision variable, indicating whether to dispatch unit j.

3. The maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The expression for maximizing task completion efficiency is as follows: Where M represents the total number of available emergency resource units; N: The total number of incidents or tasks that need to be addressed. L: Types and number of emergency resource units; in: : Indicates the expected effectiveness that resource unit j can generate in handling accident point i; : Match score; : The maximum recoverable loss or highest completion rate of task point i; : Indicates the type of resource unit j; : Indicates the main accident type that needs to be addressed at accident point i.

4. The maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The expression for minimizing the total emergency response time in the second-stage objective function is as follows: : Represents a function that calculates the estimated travel or journey time required to travel from position A to position B; : The current position of resource unit j before executing the plan for this stage; The geographical location of the accident point or mission point i after drifting and spreading over time; : is a 0-1 decision variable. =1 indicates that resource unit j is scheduled and assigned to task point i for processing. =0 indicates that this allocation will not be performed.

5. A maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The expression for minimizing the switching cost of the resource allocation scheme is as follows: : Represents a function that calculates the total cost of moving resources from location A to location B; : The current actual location of resource unit j. : The location of accident point i.

6. The maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The process of comparing the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively to determine whether to enter the second stage of accident rescue is as follows: Set the rescue time threshold for the accident to X hours and the overall completion status of the rescue mission to Y%. When an accident has occurred and there are people who need to be rescued, and the rescue time is ≥X hours, then the second stage of accident rescue begins. If an accident has occurred and there are materials and resources that need to be rescued, and the overall completion rate of the rescue mission is ≥Y%, then the second stage of accident rescue will begin. Otherwise, the accident rescue will continue with the first phase of the rescue plan.

7. A maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: Based on the improved differential evolution algorithm, the objective function of the first stage of accident rescue is solved, and the process of obtaining the allocation scheme of dispatchable rescue resources for the first stage of maritime accidents is as follows: S31: Initialize the population: Randomly initialize the population P and set relevant parameters, including population size, number of iterations and mutation operator. S32: Calculate fitness value: Based on the objective function of the first stage, calculate the fitness response time of each individual in the population, which serves as a standard for evaluating the merits of the first stage's maritime accident relocation plan. S33: Mutation and Crossover Operations: The mutation and crossover operators of the differential evolution algorithm are used to operate on the individuals in the population to generate the experimental population U; S34: Greedy selection update: Compare the fitness values ​​of individuals in the original population P with those in the experimental population U, and execute a greedy selection strategy to update population P; S35: Read the current best individual: Identify and extract the best individual X_best in the current generation from the updated population P; S36: Simulated Annealing Local Search: Based on the current best individual, a simulated annealing local search mechanism is introduced to generate a new neighborhood solution through small step-size flip perturbation; S37: Neighborhood Solution Judgment and Replacement: Determine whether the function value of the new solution after perturbation is less than the function value of the original solution. If it is less, the new solution is better, and the current best solution in the population is replaced; otherwise, the original solution remains unchanged. S38: Termination Condition Determination: Check if the preset termination condition is met. If not, return to S33 to perform mutation and crossover operations. If met, stop the iteration and output the globally optimal rescue resource allocation plan.

8. A maritime emergency decision-making method based on dynamic phase perception according to claim 1, characterized in that: The process of solving the objective function of the second stage based on the improved non-dominated sorting genetic algorithm to obtain the second-stage adjusted rescue resource allocation scheme for maritime accidents is as follows: S61 Initialization and Scheme Inheritance: Initialize the parameters of the improved non-dominated sorting genetic algorithm and inherit the optimal scheme output in the first stage as the basis for solving the second stage; S62 Multi-Objective Evaluation: For the objective function of the second stage, the function values ​​of three objectives—response time, task efficiency, and execution stability—are evaluated simultaneously. S63 Non-dominated ranking and crowding calculation: Perform fast non-dominated ranking on individuals in the population and calculate the crowding of individuals in the target space to establish dominance relationships between individuals; S64 evolutionary operator operations: generate offspring population Q through selection, crossover, and mutation operations; S65 Population Merging and Selection: The parent population P and the offspring population Q are merged into a combined population R. An elite retention strategy is used to select high-performing individuals from R to form a new generation population P. S66 extracts non-dominant individuals of Rank 1 from the current new population P, forming the current Pareto front. S67 Front Extension Search: Introduces a front extension strategy based on simulated annealing to perform perturbation extension search on individuals on the front. S68 Update Non-Dominated Solution Set: Update the non-dominated solution set based on the extended search results to ensure the quality of Pareto optimal solutions in the non-dominated solution set; S69 Iteration and Output: Determine if the set maximum number of iterations has been reached. If not, return to the non-dominated sorting step to continue the loop. If it has been reached, output the final Pareto optimal solution set as a reference scheme for the adjusted allocation of rescue resources.

9. A maritime emergency decision-making system based on dynamic phase perception, characterized in that: include: Information perception module: used to acquire marine environmental data, marine accident data, and available rescue resources in the event of a marine accident; The first-stage construction module is used to construct the first-stage objective function for accident rescue, which aims to minimize the total emergency response time, based on marine environmental data, marine accident data, and available rescue resources in the event of a marine accident. The first-stage solution module is used to solve the objective function of the first stage of accident rescue based on the improved differential evolution algorithm, so as to obtain the first-stage allocation plan of the available rescue resources for maritime accidents. The transition phase judgment module is used to compare the rescue time and overall completion status of the rescue mission with the rescue time threshold and overall completion status threshold of the rescue mission respectively, to determine whether to enter the second phase of accident rescue and to adjust the allocation plan of available rescue resources for maritime accidents. The second-stage judgment module is used to construct a second-stage overall objective function when entering the second stage, which includes minimizing the total emergency response time, maximizing task completion efficiency, and minimizing the cost of switching resource allocation schemes. If the second phase is not initiated, the rescue resource allocation plan for the first phase of the accident rescue will continue to be implemented. The second-stage solution module is used to solve the objective function of the second stage based on the improved non-dominated sorting genetic algorithm, so as to obtain the second-stage adjusted rescue resource allocation plan for maritime accidents.

Images