A ship cargo matching system and method based on intelligent optimization algorithm

Abstract

The application discloses a ship cargo matching system based on an intelligent optimization algorithm, which comprises an order management module, a ship management module, a port management module, a route management module, a freight rate management module and a ship draft management module; the input of the order management module comprises basic information of three types of orders, i.e., a marine oil cargo plate, a large customer cargo plate and sporadic cargo sources; the input of the ship management module comprises static information and dynamic information of a ship; the input of the port management module comprises three-level wharf information; the input of the route management module comprises a sailing distance matrix and a sailing time matrix; the input of the freight rate management module comprises freight rate data between different loading and unloading ports; and the input of the ship draft management module comprises a maximum cargo carrying capacity data table of different ship types under different draft depths. The application has strong dynamic adaptability.

Description

Technical Field

[0001] This invention relates to an improvement in ship-cargo matching technology, belonging to the field of shipping logistics, and particularly to a ship-cargo matching system and method based on intelligent optimization algorithms. Background Technology

[0002] Ship scheduling and cargo matching optimization are core research directions in shipping logistics. Existing technologies include heuristic algorithms, ALNS algorithms, and rolling time-domain optimization, but all have certain limitations. Real-world shipping scenarios are highly dynamic and uncertain: at the order level, new orders are constantly being generated, and existing orders may be canceled or modified; at the ship level, location, estimated arrival time, and loading status change in real time; and at the external environment level, weather, port congestion, and waterway restrictions are frequent sources of interference.

[0003] However, existing systems generally rely on static planning approaches: most matching systems lack real-time feedback mechanisms and cannot dynamically adjust the scheduling plan based on actual execution deviations after it is generated; although rolling time-domain optimization is used for dynamic problems, it mostly adopts a fixed-cycle periodic driving mode, which is difficult to respond quickly to sudden events and may lead to the failure of scheduling for unplanned orders or emergency situations; moreover, the lack of a locking mechanism in rolling optimization makes it easy for confirmed ship-cargo matching relationships to be repeatedly adjusted, resulting in unstable scheduling plans, which not only increases resource waste but also increases the actual execution difficulty, seriously restricting its adaptability in real shipping operations.

[0004] Chinese patent application CN202411407880.4, filed on October 10, 2024, discloses a method, apparatus, equipment, and medium for matching bulk cargo to shipping vessels based on a heuristic algorithm. The method includes: acquiring current bulk carrier route loading data; identifying the quantity of return bulk cargo based on the current bulk carrier route loading data; generating a market index based on the quantity of return bulk cargo; analyzing the current market index to obtain cargo solicitation demand data; and upon receiving a cargo solicitation completion message triggered by the cargo solicitation demand data, obtaining a preset... The process involves identifying ship constraints, obtaining feasible ship information based on these constraints, calculating an initial ship scheduling scheme based on this information, and then performing calculations on the initial scheme to obtain a ship scheduling scheme to be matched. Each of the ship scheduling schemes to be matched is simulated, and a ship scheduling scheme is generated based on the simulation results. While this approach addresses the issues of low efficiency and slow decision-making in matching bulk carriers with cargo, it only addresses schemes with no initial dynamic changes and still suffers from poor adaptability under dynamic factors.

[0005] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this patent application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to overcome the problem of insufficient dynamic adaptability in the prior art and to provide a ship-cargo matching system and method based on intelligent optimization algorithm with strong dynamic adaptability.

[0007] To achieve the above objectives, the technical solution of the present invention is: a ship-cargo matching system based on intelligent optimization algorithms, wherein the ship-cargo matching system based on intelligent optimization algorithms includes an order management module, a ship management module, a port management module, a route management module, a freight rate management module, and a ship draft management module;

[0008] The input to the order management module includes basic information on three types of orders: marine oil cargo, large customer cargo, and small-volume cargo.

[0009] The inputs to the ship management module include both static and dynamic information about the ship.

[0010] The input to the port management module includes information on three levels of terminals;

[0011] The inputs to the route management module include the distance matrix and the flight time matrix;

[0012] The input to the freight rate management module includes freight rate data between different loading and unloading ports;

[0013] The input to the ship draft management module includes a table of maximum cargo volume data for different ship types at different draft depths.

[0014] The order management module is used to maintain basic information for three types of orders: marine oil cargo, large customer cargo, and miscellaneous cargo. The basic information includes cargo volume, loading and unloading ports, loading period, specified vessel type, specified vessel, whether heating is required, and whether the vessel is detachable.

[0015] The ship management module is used to maintain the ship's static and dynamic information;

[0016] The port management module is used to manage information about the three-tier terminal, including geographic coordinates, available vessel types, water depth limits, loading and unloading efficiency, average waiting time, and port charges.

[0017] The route management module is used to maintain the flight distance matrix D(n+m)×(n+m) and the flight time matrix T(n+m)×(n+m)×k;

[0018] Where n is the number of port areas, m is the current number of ships, k is the total number of ships in the fleet, and matrix element D[i][j] represents the standard route distance from position i to position j;

[0019] The freight rate management module is used to maintain freight rate data between different loading and unloading ports. It supports manual editing and maintenance, and provides a basis for cost calculation and TCE assessment.

[0020] The vessel draft management module is used to maintain a data table of maximum cargo volume for different vessel types at different draft depths, and is used for constraint verification of order-vessel matching.

[0021] The route management module supports setting vessel priorities. Vessels with higher priorities are assigned tasks more frequently during matching to meet special business needs.

[0022] The tiered division of the three-tier terminal adopts a business-driven static configuration mode, and is divided and maintained according to actual operation and management needs.

[0023] A matching method for a ship-cargo matching system based on an intelligent optimization algorithm, the method comprising the following steps:

[0024] The first step is to load the preprocessed raw input data, then build a fuel consumption-speed function model based on the raw input data, sort the tasks according to the weight scores, and then use a three-layer greedy allocation strategy to generate a greedy initial matching scheme, while generating an assigned task list and an unassigned task list.

[0025] The second step is to initialize the relevant parameters based on the greedy initial matching scheme and the original input data. Then, in each iteration, the assigned tasks are removed according to the preset ratio and added to the list of unassigned tasks. The unassigned tasks are then reinserted into the matching scheme according to the specified strategy. Through multiple rounds of iteration and optimization, the optimized matching scheme is finally obtained.

[0026] The third step is to optimize the speed of the empty flight segment of the assigned task without changing the task allocation relationship and time node, based on the fuel consumption rate formula, construct a multi-objective function to generate multiple types of execution schemes and obtain the corresponding preset evaluation index values.

[0027] The fourth step involves selecting one type of execution plan as the execution benchmark based on multiple execution plans and relevant evaluation indicators, combined with matching requirements. Then, the dynamic data of the execution plans is synchronized through a dual mechanism, and the execution benchmark is updated based on the dynamic data. Matching is then performed based on the updated execution benchmark to obtain the actual execution result. The actual execution result is compared with the preset evaluation indicator value to obtain the deviation value. If the deviation value exceeds the threshold, the matching plan is adjusted based on the dynamic data, and the comparison is performed again until the deviation value meets the threshold requirement. At the same time, all tasks are completed, and the matching method ends.

[0028] In the first step, the raw input data includes ship information, task list, port information, route distance and other constraint data;

[0029] Ship information includes static data and dynamic data;

[0030] Static data includes ship name, ship type, deadweight tonnage, speed, and fuel consumption parameters;

[0031] Dynamic data includes real-time location, estimated arrival time, current loading status, and matching priority;

[0032] The task list, or order data, includes three types of orders: marine oil cargo, major customer cargo, and miscellaneous cargo, along with their cargo volume, loading and unloading ports, loading periods, and specified vessel types.

[0033] Port information includes the geographical coordinates of the third-level terminal, water depth limits, loading and unloading efficiency, and port charges;

[0034] The route distance includes the standard route distance between different locations;

[0035] Other constraint data includes vessel draft restrictions, freight rate data, and order-vessel matching constraints.

[0036] In the first step, the weights include: order type, ship scarcity, and time urgency;

[0037] All tasks are prioritized based on a multi-dimensional scoring mechanism, with scoring factors including:

[0038] Task type weighting: Marine oil tasks > Major client tasks > Small orders;

[0039] Time urgency: Tasks with shorter time windows and earlier start times have higher priority;

[0040] Ship scarcity: The fewer ships available to perform the task, the higher the priority.

[0041] In the second step, relevant parameters are first initialized based on the greedy initial matching scheme and the original input data. Then, in each iteration, assigned tasks are removed according to a preset ratio and added to the unassigned task list. Finally, the unassigned tasks are reinserted into the matching scheme according to a specified strategy. Through multiple rounds of iterative optimization, the optimized matching scheme is finally obtained, as follows:

[0042] The system receives the initial matching scheme generated by the greedy algorithm, converts it into the ALNS internal representation format, including lists of assigned and unassigned tasks, initializes the objective function value, sets the simulated annealing parameters, and sets the initial temperature.

[0043] T0=1000, cooling rate α=0.99, minimum temperature T min =0.1, initialize all destroy and repair operator weights to 1.0, record the current solution and the historical best solution, then re-insert the unassigned tasks into the matching scheme according to the ship priority and global optimization strategy, and then perform deep optimization through the ALNS algorithm to obtain the optimized matching scheme;

[0044] The ALNS algorithm undergoes deep optimization in the following ways:

[0045] A. Operator selection: A roulette wheel selection mechanism is adopted to dynamically select a destroy operator and a repair operator based on the historical performance of the operators;

[0046] B. Destruction Phase: The selected destroy operator removes 20%-40% of the assigned tasks from the current scheme and adds them to the unassigned list, except for rare tasks with ≤3 ships.

[0047] A mission protection mechanism is introduced to skip destruction for scarce missions with ≤3 executable ships;

[0048] C. Repair phase: The selected repair operator will re-insert the unassigned tasks into the matching plan;

[0049] D. Feasibility verification: After each repair, the system performs triple verification: time continuity verification, location continuity verification, and dynamic recalculation;

[0050] E. Scheme Evaluation and Acceptance: Calculate the objective function value of the new scheme. The objective function is designed as follows:

[0051] ;

[0052] in, For total cost, For the set of unassigned tasks, For the task Unallocated penalties;

[0053] The acceptance criterion employs simulated annealing: if the new solution is superior to the historical best solution, it is unconditionally accepted and the score of the best solution is updated. ;

[0054] If the new solution is better than the current solution but not optimal, accept and update the current solution. ;

[0055] If the new solution is worse than the current solution, by probability Accept score No score accepted ;

[0056] F. Weight Update: Every 50 iterations, the system updates the weights of all operators.

[0057] ;

[0058] in, This is the average score of the operator.

[0059] In the third step, the multi-objective function refers to the function that minimizes cost, minimizes the number of ships, and maximizes utilization. The specific objective function automatically generated by the multi-objective scheme is as follows:

[0060] Minimum cost: ;

[0061] Minimum number of ships: ;

[0062] Highest utilization rate: ;

[0063] in, The variable cost of the order includes fuel costs and port charges; For fixed costs of ships, The penalty cost for orders not being scheduled The number of ships used.

[0064] In the fourth step, the dynamic data of the plan to be executed is synchronized through a dual mechanism: by driving real-time triggering and daily timed polling through order change / ship location update events, dynamic data of orders, ships, and ports are synchronized.

[0065] In the fourth step, when adjusting the matching scheme, the lock relationship between the ships and orders specified by the business personnel is retained, and a new round of optimization is triggered only for the unlocked parts.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] 1. In the ship-cargo matching system and method based on intelligent optimization algorithm of the present invention, the system first loads the preprocessed original input data, then constructs a fuel consumption-speed function model based on the original input data, then sorts the tasks according to the weight scores, and then adopts a three-layer greedy allocation strategy to generate a greedy initial matching scheme, while generating an assigned task list and an unassigned task list; first, based on the greedy initial matching scheme and the original input data, the relevant parameters are initialized, then the assigned tasks are removed according to a preset ratio and added to the unassigned task list, and then the unassigned tasks are re-inserted into the matching scheme according to a specified strategy to obtain an updated matching scheme, and then the updated matching scheme is deeply optimized using the ALNS algorithm to obtain an optimized matching scheme; based on the optimized matching scheme... The proposed solution, without altering task allocation relationships or timelines, optimizes flight speed based on the fuel consumption rate formula for empty flight segments with assigned tasks. This constructs a multi-objective function to generate multiple execution plans and corresponding preset evaluation index values. First, based on the multiple execution plans and relevant evaluation indexes, one execution plan is selected as the execution benchmark, combined with matching requirements. Then, dynamic data of the execution plans is synchronized through a dual mechanism, and the execution benchmark is updated based on the dynamic data. Matching is then performed based on the updated execution benchmark to obtain the actual execution result. The actual execution result is compared with the preset evaluation index value to obtain the deviation value. If the deviation value exceeds a threshold, the matching plan is adjusted based on dynamic data, and the comparison is repeated until the deviation value meets the threshold requirement. Simultaneously, all tasks are completed, at which point the matching method ends. The advantages of this design are as follows:

[0068] Firstly, the system receives the latest order requirements, vessel dynamic positions, and port operation status information in real time. Based on the feedback results, it automatically adjusts data parameters and puts the optimized plan into actual execution. It continuously monitors the execution of the plan, identifies deviations in actual execution, and receives the latest data in real time based on the execution feedback and deviation identification results. The rolling optimization closed-loop control mechanism can quickly respond to changes in the external environment and adjust the matching plan in real time. At the same time, the locking mechanism marks the confirmed ship-cargo relationship and automatically skips these relationships during optimization, only recalculating the unlocked part. This ensures the executability of the plan and maintains the ability to adapt to new information.

[0069] Secondly, by prioritizing the allocation of high-weight orders, selecting the nearest vessels, and considering cargo load factors and turnover efficiency, high-quality solutions can be quickly generated, providing a good starting point for subsequent optimization and thus improving efficiency.

[0070] Thirdly, the optimization scheme is refined, including speed optimization for empty flight segments and multi-objective scheme generation, to ensure that the final scheme meets both the optimization objectives and the actual operational requirements.

[0071] Fourthly, the system automatically outputs three distinctive scheduling schemes: minimum cost, fewest vessels, and highest utilization rate. Each scheme is equipped with comprehensive evaluation indicators. The first scheme prioritizes minimizing total operating costs and is suitable for business scenarios that focus on cost control. The second scheme aims to minimize the number of vessels used and improve the load capacity utilization of individual vessels, suitable for scenarios with limited fleet resources or those that wish to optimize fleet size. The third scheme aims to maximize the overall utilization rate of vessels, using as many vessels as possible to improve the overall turnover efficiency of the fleet, suitable for scenarios with abundant fleet resources and those that wish to improve asset utilization. Each scheme is equipped with comprehensive evaluation indicators, providing multi-dimensional selection space and quantitative decision-making basis for decision-making, and achieving intelligent multi-objective optimization.

[0072] Therefore, the present invention has strong dynamic adaptability.

[0073] 2. In this invention, a ship-cargo matching system and method based on intelligent optimization algorithms, the layered and modular system architecture facilitates system expansion and maintenance. The data layer, algorithm optimization layer, and application layer interact through standardized interfaces, allowing each layer to be upgraded independently without affecting other layers. The system can easily support expansion needs such as more ship types, more complex business scenarios, and international route scheduling. The modular design encapsulates different functions in independent modules, reducing coupling between modules. The data layer provides a unified data service interface, and the application layer is responsible for user interaction. Each layer has clear responsibilities, and adding new functions only requires adding or modifying modules in the corresponding layer without affecting the overall architecture. Therefore, this invention's modular processing improves efficiency.

[0074] 3. This invention, a ship-cargo matching system and method based on intelligent optimization algorithms, is refined to the third-level terminal level. It can precisely control the arrival, loading / unloading, and departure times of ships at each port, achieving more refined scheduling management and resource optimization. Different matching strategies and optimization weights are adopted for three different types of cargo: marine oil cargo, large customer cargo, and temporary cargo. Marine oil cargo and large customer orders have higher priority and are allocated first during the optimization process, meeting the special needs of different cargo types. The system supports setting ship priorities; higher-priority ships are assigned tasks first during matching to meet special business needs. The system automatically obtains actual sailing distances and freight rate information between different ports, supporting manual maintenance and adjustment by users, providing accurate basis for cost calculation and scheme evaluation. A Gantt chart visually displays the timeline of ship and cargo arrangements, supporting switching between ship and cargo dimensions for quick understanding and decision-making by dispatchers. The Gantt chart can simultaneously display locked actual execution plans and generated pending ship plans, providing a complete panoramic view of scheduling. Therefore, this invention enables more refined scheduling management and resource optimization.

[0075] 4. In this invention, a ship-cargo matching system and method based on intelligent optimization algorithms, intelligent optimization effectively reduces empty sailing distance and port waiting time, improving ship load factor and turnaround efficiency. The speed optimization function optimizes empty sailing segments based on the cubic relationship between speed and fuel consumption, reducing fuel consumption without affecting timeliness. Initially, it considers ship position and route distance, selecting the closest possible ship. ALNS further reduces the total sailing distance through multiple iterative optimizations. A segmented scoring mechanism for load utilization encourages efficient use of ship capacity. The speed optimization module reduces fuel consumption by lowering speed during empty sailing segments, achieving an optimal balance between fuel cost and time cost. Therefore, this invention improves ship load factor and turnaround efficiency. Attached Figure Description

[0076] Figure 1 This is a system block diagram of the present invention.

[0077] Figure 2 This is a flowchart of the present invention.

[0078] Figure 3 This is a schematic diagram of the greedy strategy in this invention.

[0079] Figure 4 This is a flowchart of the ALNS algorithm in this invention.

[0080] Figure 5 This is a schematic diagram of the random removal operator in this invention.

[0081] Figure 6 This is a schematic diagram of the worst-case cost removal operator in this invention.

[0082] Figure 7 This is a schematic diagram of the related route removal operator in this invention.

[0083] Figure 8 This is a schematic diagram of the time window removal operator in this invention.

[0084] Figure 9 This is a schematic diagram of the ship removal operator in this invention.

[0085] Figure 10 This is a schematic diagram of the optimal position repair operator in this invention.

[0086] Figure 11 This is a schematic diagram of the regret repair operator in this invention.

[0087] Figure 12 This is a flowchart of the three-stage hybrid optimization strategy of the present invention.

[0088] In the diagram: Order Management Module 1, Ship Management Module 2, Port Management Module 3, Route Management Module 4, Freight Rate Management Module 5, Ship Draft Management Module 6. Detailed Implementation

[0089] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0090] See Figures 1 to 12 A ship-cargo matching system based on intelligent optimization algorithm, the ship-cargo matching system based on intelligent optimization algorithm includes order management module 1, ship management module 2, port management module 3, route management module 4, freight rate management module 5 and ship draft management module 6;

[0091] The input to the order management module 1 includes basic information on three types of orders: marine oil cargo, large customer cargo, and miscellaneous cargo.

[0092] The inputs for Ship Management Module 2 include the ship's static and dynamic information;

[0093] The input for Port Management Module 3 includes information on three levels of terminals;

[0094] The inputs to the route management module 4 include the distance matrix and the flight time matrix;

[0095] The input to the freight rate management module 5 includes freight rate data between different loading and unloading ports;

[0096] The input to the ship draft management module 6 includes a table of maximum cargo volume data for different ship types at different draft depths.

[0097] The order management module 1 is used to maintain the basic information of three types of orders: marine oil cargo, large customer cargo, and miscellaneous cargo. The basic information includes cargo volume, loading and unloading ports, loading period, specified ship type, specified vessel, whether heating is required, and whether it is detachable.

[0098] Ship Management Module 2 is used to maintain the static and dynamic information of ships;

[0099] Port Management Module 3 is used to manage information on three-tier terminals, including geographic coordinates, available vessel types, water depth limits, loading and unloading efficiency, average waiting time, and port charges.

[0100] Route management module 4 is used to maintain the flight distance matrix D(n+m)×(n+m) and the flight time matrix T(n+m)×(n+m)×k;

[0101] Where n is the number of port areas, m is the current number of ships, k is the total number of ships in the fleet, and matrix element D[i][j] represents the standard route distance from position i to position j;

[0102] The freight rate management module 5 is used to maintain freight rate data between different loading and unloading ports. It supports manual editing and maintenance, and provides a basis for cost calculation and TCE assessment.

[0103] The Ship Draft Management Module 6 is used to maintain a data table of maximum cargo volume for different ship types at different draft depths, and is used for constraint verification of order-ship matching.

[0104] The route management module 4 supports setting vessel priorities. Vessels with higher priorities are assigned tasks first during matching to meet special business needs.

[0105] The tiered division of the three-tier terminal adopts a business-driven static configuration mode, and is divided and maintained according to actual operation and management needs.

[0106] A matching method for a ship-cargo matching system based on an intelligent optimization algorithm, the method comprising the following steps:

[0107] The first step is to load the preprocessed raw input data, then build a fuel consumption-speed function model based on the raw input data, sort the tasks according to the weight scores, and then use a three-layer greedy allocation strategy to generate a greedy initial matching scheme, while generating an assigned task list and an unassigned task list.

[0108] The second step is to initialize the relevant parameters based on the greedy initial matching scheme and the original input data. Then, in each iteration, the assigned tasks are removed according to the preset ratio and added to the list of unassigned tasks. The unassigned tasks are then reinserted into the matching scheme according to the specified strategy. Through multiple rounds of iteration and optimization, the optimized matching scheme is finally obtained.

[0109] The third step is to optimize the speed of the empty flight segment of the assigned task without changing the task allocation relationship and time node, based on the fuel consumption rate formula, construct a multi-objective function to generate multiple types of execution schemes and obtain the corresponding preset evaluation index values.

[0110] The fourth step involves selecting one type of execution plan as the execution benchmark based on multiple execution plans and relevant evaluation indicators, combined with matching requirements. Then, the dynamic data of the execution plans is synchronized through a dual mechanism, and the execution benchmark is updated based on the dynamic data. Matching is then performed based on the updated execution benchmark to obtain the actual execution result. The actual execution result is compared with the preset evaluation indicator value to obtain the deviation value. If the deviation value exceeds the threshold, the matching plan is adjusted based on the dynamic data, and the comparison is performed again until the deviation value meets the threshold requirement. At the same time, all tasks are completed, and the matching method ends.

[0111] In the first step, the raw input data includes ship information, task list, port information, route distance and other constraint data;

[0112] Ship information includes static data and dynamic data;

[0113] Static data includes ship name, ship type, deadweight tonnage, speed, and fuel consumption parameters;

[0114] Dynamic data includes real-time location, estimated arrival time, current loading status, and matching priority;

[0115] The task list, or order data, includes three types of orders: marine oil cargo, major customer cargo, and miscellaneous cargo, along with their cargo volume, loading and unloading ports, loading periods, and specified vessel types.

[0116] Port information includes the geographical coordinates of the third-level terminal, water depth limits, loading and unloading efficiency, and port charges;

[0117] The route distance includes the standard route distance between different locations;

[0118] Other constraint data includes vessel draft restrictions, freight rate data, and order-vessel matching constraints.

[0119] In the first step, the weights include: order type, ship scarcity, and time urgency;

[0120] All tasks are prioritized based on a multi-dimensional scoring mechanism, with scoring factors including:

[0121] Task type weighting: Marine oil tasks > Major client tasks > Small orders;

[0122] Time urgency: Tasks with shorter time windows and earlier start times have higher priority;

[0123] Ship scarcity: The fewer ships available to perform the task, the higher the priority.

[0124] In the second step, relevant parameters are first initialized based on the greedy initial matching scheme and the original input data. Then, in each iteration, assigned tasks are removed according to a preset ratio and added to the unassigned task list. Finally, the unassigned tasks are reinserted into the matching scheme according to a specified strategy. Through multiple rounds of iterative optimization, the optimized matching scheme is finally obtained, as follows:

[0125] The system receives the initial matching scheme generated by the greedy algorithm, converts it into the ALNS internal representation format, including lists of assigned and unassigned tasks, initializes the objective function value, and sets the simulated annealing parameters: initial temperature T0 = 1000, cooling rate α = 0.99, and minimum temperature T... min=0.1, initialize all destroy and repair operator weights to 1.0, record the current solution and the historical best solution, then re-insert the unassigned tasks into the matching scheme according to the ship priority and global optimization strategy, and then perform deep optimization through the ALNS algorithm to obtain the optimized matching scheme;

[0126] The ALNS algorithm undergoes deep optimization in the following ways:

[0127] A. Operator selection: A roulette wheel selection mechanism is adopted to dynamically select a destroy operator and a repair operator based on the historical performance of the operators;

[0128] B. Destruction Phase: The selected destroy operator removes 20%-40% of the assigned tasks from the current scheme and adds them to the unassigned list, except for rare tasks with ≤3 ships.

[0129] A mission protection mechanism is introduced to skip destruction for scarce missions with ≤3 executable ships;

[0130] C. Repair phase: The selected repair operator will re-insert the unassigned tasks into the matching plan;

[0131] D. Feasibility verification: After each repair, the system performs triple verification: time continuity verification, location continuity verification, and dynamic recalculation;

[0132] E. Scheme Evaluation and Acceptance: Calculate the objective function value of the new scheme. The objective function is designed as follows:

[0133] ;

[0134] in, For total cost, For the set of unassigned tasks, For the task Unallocated penalties;

[0135] The acceptance criterion employs simulated annealing: if the new solution is superior to the historical best solution, it is unconditionally accepted and the score of the best solution is updated. ;

[0136] If the new solution is better than the current solution but not optimal, accept and update the current solution. ;

[0137] If the new solution is worse than the current solution, by probability Accept score No score accepted ;

[0138] F. Weight Update: Every 50 iterations, the system updates the weights of all operators.

[0139] ;

[0140] in, This is the average score of the operator.

[0141] In the third step, the multi-objective function refers to the function that minimizes cost, minimizes the number of ships, and maximizes utilization. The specific objective function automatically generated by the multi-objective scheme is as follows:

[0142] Minimum cost: ;

[0143] Minimum number of ships: ;

[0144] Highest utilization rate: ;

[0145] in, The variable cost of the order includes fuel costs and port charges; For fixed costs of ships, The penalty cost for orders not being scheduled The number of ships used.

[0146] In the fourth step, the dynamic data of the plan to be executed is synchronized through a dual mechanism: by driving real-time triggering and daily timed polling through order change / ship location update events, dynamic data of orders, ships, and ports are synchronized.

[0147] In the fourth step, when adjusting the matching scheme, the lock relationship between the ships and orders specified by the business personnel is retained, and a new round of optimization is triggered only for the unlocked parts.

[0148] The supplementary technical features of this design are as follows:

[0149] This proposal innovatively puts forward a three-stage incremental optimization strategy: greedy heuristic algorithm, ALNS metaheuristic algorithm, and data post-processing.

[0150] The first stage (fast solution): adopts a multi-layered greedy rule heuristic algorithm. Through a three-layer nested greedy strategy of "task priority sorting - ship adaptability assessment - comprehensive solution scoring", a high-quality initial feasible solution is generated quickly within 1-2 minutes, avoiding the inefficient search problem caused by traditional algorithms starting from random solutions.

[0151] The second stage (deep optimization): Based on the initial solution, the ALNS algorithm is used to perform deep optimization through a "break-repair" iterative mechanism, which can effectively escape local optima and significantly reduce the total cost or improve resource utilization while ensuring the quality of the solution.

[0152] The third stage (refined processing) involves post-processing the optimization results, including optimizing the speed of the empty flight segment (minimizing fuel costs based on the cubic relationship between speed and fuel consumption) and generating multi-objective solutions to ensure that the final solution meets both the optimization objectives and the actual operational requirements.

[0153] Greedy algorithms provide a high-quality starting point, ALNS achieves global optimization, and post-processing ensures practicality. The collaboration of these three elements enables the system to achieve the best balance between solution speed and solution quality.

[0154] Example 1:

[0155] A ship-cargo matching system based on intelligent optimization algorithms, the ship-cargo matching system based on intelligent optimization algorithms includes an order management module 1, a ship management module 2, a port management module 3, a route management module 4, a freight rate management module 5, and a ship draft management module 6;

[0156] The input to the order management module 1 includes basic information on three types of orders: marine oil cargo, large customer cargo, and miscellaneous cargo.

[0157] The inputs for Ship Management Module 2 include the ship's static and dynamic information;

[0158] The input for Port Management Module 3 includes information on three levels of terminals;

[0159] The inputs to the route management module 4 include the distance matrix and the flight time matrix;

[0160] The input to the freight rate management module 5 includes freight rate data between different loading and unloading ports;

[0161] The input to the ship draft management module 6 includes a table of maximum cargo volume data for different ship types at different draft depths.

[0162] A matching method for a ship-cargo matching system based on an intelligent optimization algorithm, the method comprising the following steps:

[0163] The first step is to load the preprocessed raw input data, then build a fuel consumption-speed function model based on the raw input data, sort the tasks according to the weight scores, and then use a three-layer greedy allocation strategy to generate a greedy initial matching scheme, while generating an assigned task list and an unassigned task list.

[0164] The second step is to initialize the relevant parameters based on the greedy initial matching scheme and the original input data. Then, in each iteration, the assigned tasks are removed according to the preset ratio and added to the list of unassigned tasks. The unassigned tasks are then reinserted into the matching scheme according to the specified strategy. Through multiple rounds of iteration and optimization, the optimized matching scheme is finally obtained.

[0165] The third step is to optimize the speed of the empty flight segment of the assigned task without changing the task allocation relationship and time node, based on the fuel consumption rate formula, construct a multi-objective function to generate multiple types of execution schemes and obtain the corresponding preset evaluation index values.

[0166] The fourth step involves selecting one type of execution plan as the execution benchmark based on multiple execution plans and relevant evaluation indicators, combined with matching requirements. Then, the dynamic data of the execution plans is synchronized through a dual mechanism, and the execution benchmark is updated based on the dynamic data. Matching is then performed based on the updated execution benchmark to obtain the actual execution result. The actual execution result is compared with the preset evaluation indicator value to obtain the deviation value. If the deviation value exceeds the threshold, the matching plan is adjusted based on the dynamic data, and the comparison is performed again until the deviation value meets the threshold requirement. At the same time, all tasks are completed, and the matching method ends.

[0167] Example 2:

[0168] Example 2 is basically the same as Example 1, except that:

[0169] The data layer is the foundational support of the system, responsible for the unified management and standardized processing of multi-source heterogeneous data, and includes six major management modules:

[0170] The order management module 1 is used to maintain the basic information of three types of orders: marine oil cargo, large customer cargo, and miscellaneous cargo, including cargo volume, loading and unloading ports, loading period, specified ship type, specified vessel, whether heating is required, and whether it is detachable. It realizes the order filtering function and pre-filters the set of executable vessels for each order based on conditions such as cargo volume constraints, ship type constraints, and draft depth constraints.

[0171] Ship Management Module 2 is used to maintain the static and dynamic information of ships;

[0172] Port Management Module 3 is used to manage information on three-tier terminals, including geographic coordinates, available vessel types, water depth limits, loading and unloading efficiency, average waiting time, and port charges.

[0173] Route management module 4 is used to maintain the flight distance matrix D(n+m)×(n+m) and the flight time matrix T(n+m)×(n+m)×k;

[0174] Where n is the number of port areas, m is the current number of ships, k is the total number of ships in the fleet, and matrix element D[i][j] represents the standard route distance from position i to position j;

[0175] The freight rate management module 5 is used to maintain freight rate data between different loading and unloading ports. It supports manual editing and maintenance, and provides a basis for cost calculation and TCE assessment.

[0176] The Ship Draft Management Module 6 is used to maintain a data table of maximum cargo volume for different ship types at different draft depths, and is used for constraint verification of order-ship matching.

[0177] Vessel Draft Management Module: Maintains a data table of maximum cargo volume for different vessel types at different draft depths, used for constraint verification of order-vessel matching.

[0178] The data layer provides data services to the algorithm optimization layer through standardized data interfaces and performs integrity checks (such as checking the integrity of port spacing data and freight rate data) when data is input.

[0179] Overall data flow and inter-layer connection process

[0180] Data layer (including navigation mark system and back-end preprocessing): Reads and obtains raw data (such as ship orders, port information, route plans, etc.) from the business operation system database;

[0181] Specifically, this includes: data acquisition (API retrieval, database synchronization, message queue subscription, etc.); preprocessing such as data cleaning, format standardization, field mapping, and handling of missing values; and building structured intermediate data (such as standardized ship-port-timetable).

[0182] Output method: The preprocessed data is pushed / sent to the algorithm optimization layer via a standardized interface (HTTP protocol);

[0183] Algorithm optimization layer: Receives preprocessed data from the backend (data layer).

[0184] Specifically, this includes: executing core algorithms (such as route optimization, ship scheduling, resource allocation, etc.); potentially calling internal models, rule engines, or optimization solvers; and generating structured computational results (such as optimal route lists, multiple option selections, etc.).

[0185] Output method: The results are returned to the backend via a standard interface (HTTP protocol);

[0186] Application layer (coordinated by the backend and displayed by the frontend): receives the results returned by the algorithm layer;

[0187] Specifically, this includes: performing business-level data matching and fusion (e.g., associating the port ID output by the algorithm with the actual port name); and building a front-end display data model. Front-end roles include: calling the final display interface provided by the back-end; implementing multi-dimensional visualization (such as scheduling matching, Gantt charts, etc.); and supporting interaction (such as re-triggering algorithm calculations after parameter adjustments).

[0188] This proposal puts forward a complete closed-loop control mechanism for planning, execution, feedback, and adjustment, which optimizes scheduling on an hourly basis. It fully utilizes known data such as rigid cargo arrangements, major customer priorities, established vessel scheduling plans, and port operation time windows. It supports a rolling update mechanism, dynamically adjusting subsequent schedules as time progresses and new information is acquired. This allows for the development of optimal plans under deterministic conditions and intelligent adjustments in uncertain environments.

[0189] Example 3:

[0190] Example 3 is basically the same as Example 1, except that:

[0191] The system receives the initial matching scheme generated by the greedy algorithm, converts it into the ALNS internal representation format, including lists of assigned and unassigned tasks, initializes the objective function value, and sets the simulated annealing parameters: initial temperature T0 = 1000, cooling rate α = 0.99, and minimum temperature T... min =0.1, initialize all destroy and repair operator weights to 1.0, record the current solution and the historical best solution, then re-insert the unassigned tasks into the matching scheme according to the ship priority and global optimization strategy, and then perform deep optimization through the ALNS algorithm to obtain the optimized matching scheme;

[0192] The ALNS algorithm undergoes deep optimization in the following ways:

[0193] A. Operator selection: A roulette wheel selection mechanism is adopted to dynamically select a destroy operator and a repair operator based on the historical performance of the operators;

[0194] B. Destruction Phase: The selected destroy operator removes 20%-40% of the assigned tasks from the current scheme and adds them to the unassigned list, except for rare tasks with ≤3 ships.

[0195] <20% is too low: the neighborhood space is too small, making it difficult to escape the local optimum, and the search effect is close to local search; >40% is too high: the destruction is excessive, close to complete reconstruction, losing the high-quality structural information of the existing solution, and the repair difficulty is greatly increased; 20%-40% range: while preserving the main structure of the solution, it provides sufficient space for reorganization.

[0196] A mission protection mechanism is introduced to skip destruction for scarce missions with a limited number of executable ships.

[0197] C. Repair phase: The selected repair operator will re-insert the unassigned tasks into the matching plan;

[0198] D. Feasibility verification: After each repair, the system performs triple verification: time continuity verification, location continuity verification, and dynamic recalculation;

[0199] E. Scheme Evaluation and Acceptance: Calculate the objective function value of the new scheme. The objective function is designed as follows:

[0200] ;

[0201] in, For total cost, For the set of unassigned tasks, For the task Unallocated penalties.

[0202] The acceptance criterion employs simulated annealing: if the new solution is superior to the historical best solution, it is unconditionally accepted and the score of the best solution is updated. ;

[0203] If the new solution is better than the current solution but not optimal, accept and update the current solution. ;

[0204] If the new solution is worse than the current solution, by probability Accept score No score accepted ;

[0205] F. Weight Update: Every 50 iterations, the system updates the weights of all operators.

[0206] ;

[0207] in, This is the average score of the operator.

[0208] A multi-objective function refers to a function that minimizes construction costs, minimizes the number of ships, and maximizes utilization. The specific objective functions automatically generated by the multi-objective scheme are as follows:

[0209] Minimum cost: ;

[0210] Minimum number of ships: ;

[0211] Highest utilization rate: ;

[0212] in, The variable cost of the order includes fuel costs and port charges; For fixed costs of ships, The penalty cost for orders not being scheduled The number of ships used;

[0213] The dynamic data of the pending plan is synchronized through a dual mechanism: real-time triggering and daily scheduled polling are driven by order change / ship location update events to synchronize dynamic data of orders, ships, and ports.

[0214] When adjusting the matching scheme, the lock relationship between the vessels and orders specified by the business personnel is retained, and a new round of optimization is triggered only for the unlocked part.

[0215] When applied: Adaptive weighting mechanism: The system dynamically adjusts the probability of each damage / repair operator being selected based on its historical performance (high or low score). It adopts the decay update formula w_new = 0.8 × w_old + 0.2 × score_avg, which enables the algorithm to automatically learn effective search strategies in different scenarios. Operators with good performance receive higher weights, thereby improving search efficiency.

[0216] Diverse operator combinations: Six destruction operators (random removal, worst-cost removal, related route removal, time window removal, removal by ship, and handling only unassigned tasks) and three repair operators (greedy insertion, best position repair, and regret repair) were designed to achieve large-scale neighborhood search through diverse combinations of operators.

[0217] Task protection mechanism: Automatically skip destructive operations for scarce tasks with ≤3 executable ships to avoid getting stuck in an unsolvable dilemma and improve the robustness of the algorithm;

[0218] Phased Repair Strategy: The repair operator adopts a two-phase strategy of "prioritizing designated vessels and optimizing globally". In the first phase, all orders for designated vessels are forcibly processed to ensure that customer requirements are met. In the second phase, tasks for non-designated vessels are optimized globally to balance costs and resource utilization.

[0219] Example 4:

[0220] Example 4 is basically the same as Example 1, except that:

[0221] The weighting factors include: order type, vessel scarcity, and time urgency;

[0222] All tasks are prioritized based on a multi-dimensional scoring mechanism, with scoring factors including:

[0223] Task type weighting: Marine oil tasks > Major client tasks > Small orders;

[0224] Time urgency: Tasks with shorter time windows and earlier start times have higher priority;

[0225] Ship scarcity: The fewer ships available to perform the task, the higher the priority.

[0226] Designated vessel / mission type: approximately 40% of total weight; Vessel scarcity: approximately 20%; Time urgency: approximately 20%; Scheduling mode adjustment: approximately 20%;

[0227] Example 5:

[0228] Example 5 is basically the same as Example 1, except that:

[0229] A three-stage incremental optimization strategy involving greedy heuristics, ALNS metaheuristics, and data post-processing is employed.

[0230] The first stage (fast solution): adopts a multi-layered greedy rule heuristic algorithm. Through a three-layer nested greedy strategy of "task priority ranking - ship adaptability assessment - comprehensive solution scoring", a high-quality initial feasible solution is generated quickly within 1-2 minutes, avoiding the inefficient search problem caused by traditional algorithms starting from random solutions.

[0231] The second stage (deep optimization): Based on the initial solution, the ALNS algorithm is used to perform deep optimization through a "break-repair" iterative mechanism, which can effectively escape local optima and significantly reduce the total cost or improve resource utilization while ensuring the quality of the solution.

[0232] The third stage (refinement processing): post-processing of the optimization results, including speed optimization of the empty flight segment (minimizing fuel cost based on the cubic relationship between speed and fuel consumption) and generation of multi-objective solutions, to ensure that the final solution meets both the optimization objectives and the actual operational requirements.

[0233] The key innovation of this strategy lies in the organic combination of each stage: the greedy algorithm provides a high-quality starting point, ALNS achieves global optimization, and post-processing ensures practicality. The collaboration of the three enables the system to achieve the best balance between solution speed and solution quality.

[0234] The data transferred from the first phase to the second phase includes:

[0235] First is the scheduling scheme dictionary, which records the task sequence assigned to each vessel. Each task plan includes complete time node information (departure time, arrival time at loading port, loading start time, loading end time, departure time from loading port, arrival time at unloading port, unloading start time, unloading end time, task completion time), location information (starting position, loading port, unloading port, ending position), cost information (fuel cost at each stage, port charges, total cost), as well as cargo volume and task attribute information.

[0236] Secondly, there is the ship status information. Each ship object maintains its current task schedule, which records the ship's usage status (whether it has been dispatched), the current task sequence, and the detailed execution plan for each task.

[0237] The third is the unassigned task list, which contains tasks that could not be successfully assigned during the greedy phase. These tasks will be reassigned during the ALNS phase.

[0238] The data transferred from the second phase to the third phase includes:

[0239] The optimized scheduling scheme dictionary has the same structure as the output of the first stage, but the task allocation relationships and timing have been optimized and adjusted. It also transmits updated ship status information and performance statistics from the optimization process (such as iteration count, improvement magnitude, operator performance, etc.).

[0240] Example 6:

[0241] Example 6 is basically the same as Example 1, except that:

[0242] The system loads static information (load, speed, fuel consumption parameters) and dynamic information (estimated arrival time, current position) of the ship. At the same time, it constructs a fuel consumption-speed function model. For each order, it pre-selects a set of ships that can execute the order based on cargo volume constraints, ship type constraints, and draft constraints, and constructs an order-ship mapping relationship.

[0243] Prioritizing all tasks based on a multi-dimensional scoring mechanism is the core of the first-layer greedy strategy. The scoring factors include: (1) Task type weight: marine oil tasks > major customer tasks > sporadic orders; (2) Time urgency: the shorter the time window and the earlier the start time, the higher the priority of the task; (3) Ship scarcity: the fewer ships that can perform the task, the higher the priority; (4) Scheduling mode adjustment: the weights are dynamically adjusted according to different optimization objectives such as MIN_COST, MIN_VESSELS, and MAX_UTILIZATION.

[0244] For each task, the system first identifies all suitable vessels, then generates multiple feasible execution plans for each vessel (by trying different loading times), calculates the cost and scores each plan, and finally sorts them by score from high to low and attempts to insert them.

[0245] Output optimized solutions and generate detailed scheduling plans, including the task sequence, time nodes, and cost details for each ship. Three types of specialized solutions are also provided for decision-makers to choose from.

[0246] The ALNS (Adaptive Large Neighborhood Search) algorithm is the deep optimization engine of this system. Based on the initial feasible solution generated by the greedy heuristic algorithm, it continuously optimizes the scheduling scheme through an iterative "destruction-repair" mechanism. This algorithm is designed with diverse neighborhood operation operators and adaptive weight adjustment mechanism to address the high-dimensional and strongly constrained characteristics of the ship scheduling problem. It can escape local optima and find a better global solution while ensuring the efficiency of the solution.

[0247] (1) Deeply optimize the initial scheduling scheme to significantly reduce the total cost or improve resource utilization; (2) Dynamically adjust the operator weights through an adaptive mechanism to improve search efficiency; (3) Process unassigned orders and try to re-add tasks that the initial algorithm could not assign to the scheduling plan during the optimization process; (4) Support multi-objective optimization mode and generate differentiated schemes according to different business scenarios (minimum cost, fewest ships, highest utilization rate);

[0248] The system receives the initial matching scheme generated by the greedy algorithm, converts it into the ALNS internal representation format, including lists of assigned and unassigned tasks, initializes the objective function value, and sets the simulated annealing parameters: initial temperature T0 = 1000, cooling rate α = 0.99, and minimum temperature T... min =0.1, initialize all destroy and repair operator weights to 1.0, record the current solution and the historical best solution, then re-insert the unassigned tasks into the matching scheme according to the ship priority and global optimization strategy, and then perform deep optimization through the ALNS algorithm to obtain the optimized matching scheme;

[0249] The ALNS algorithm undergoes deep optimization in the following ways:

[0250] A. Operator selection: A roulette wheel selection mechanism is adopted to dynamically select a destroy operator and a repair operator based on the historical performance of the operators;

[0251] B. Destruction Phase: The selected destroy operator removes 20%-40% of the assigned tasks from the current scheme and adds them to the unassigned list, except for rare tasks with ≤3 ships.

[0252] A mission protection mechanism is introduced to skip destruction for scarce missions with a limited number of executable ships.

[0253] C. Repair phase: The selected repair operator will re-insert the unassigned tasks into the matching plan;

[0254] D. Feasibility Verification: After each repair, the system performs triple verification: time continuity verification, location continuity verification, and dynamic recalculation.

[0255] E. Scheme Evaluation and Acceptance: Calculate the objective function value of the new scheme. The objective function is designed as follows:

[0256] ;

[0257] in, For total cost, For the set of unassigned tasks, For the task Unallocated penalties.

[0258] The acceptance criterion employs simulated annealing: if the new solution is superior to the historical best solution, it is unconditionally accepted and the score of the best solution is updated. ;

[0259] If the new solution is better than the current solution but not optimal, accept and update the current solution. ;

[0260] If the new solution is worse than the current solution, by probability Accept score No score accepted ;

[0261] F. Weight Update: Every 50 iterations, the system updates the weights of all operators.

[0262] ;

[0263] in, This is the average score of the operator.

[0264] In the third step, the multi-objective function refers to the function that minimizes cost, minimizes the number of ships, and maximizes utilization. The specific objective function automatically generated by the multi-objective scheme is as follows:

[0265] Minimum cost: ;

[0266] Minimum number of ships: ;

[0267] Highest utilization rate: ;

[0268] in, The variable cost of the order includes fuel costs and port charges; For fixed costs of ships, The penalty cost for orders not being scheduled The number of ships used.

[0269] The algorithm terminates when any of the following conditions are met: reaching the maximum number of iterations (default 1000), exceeding the time limit (default 300 seconds), or failing to improve after 100 consecutive iterations. It outputs information such as the historical best scheduling scheme, optimized historical trajectories, and operator performance statistics.

[0270] The evaluation indicators for different target schemes are compared in the table below:

[0271] .

[0272] The above description is only a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. A ship-cargo matching system based on intelligent optimization algorithms, characterized in that: The ship-cargo matching system based on intelligent optimization algorithm includes an order management module (1), a ship management module (2), a port management module (3), a route management module (4), a freight rate management module (5), and a ship draft management module (6). The input to the order management module (1) includes basic information on three types of orders: marine oil cargo, large customer cargo, and small-volume cargo. The inputs to the ship management module (2) include the ship's static and dynamic information; The input to the port management module (3) includes information on the three-tier terminal. The inputs to the route management module (4) include the distance matrix and the flight time matrix; The input to the freight rate management module (5) includes freight rate data between different loading and unloading ports; The input to the ship draft management module (6) includes a table of maximum cargo volume data for different ship types at different draft depths.

2. The ship-cargo matching system based on intelligent optimization algorithm according to claim 1, characterized in that: The order management module (1) is used to maintain the basic information of three types of orders: marine oil cargo, large customer cargo and sporadic cargo, including cargo volume, loading and unloading port, loading period, specified ship type, specified vessel, whether heating is required, and whether it is detachable. It realizes the order filtering function and pre-filters the set of executable vessels for each order based on conditions such as cargo volume constraints, ship type constraints, and draft depth constraints. The ship management module (2) is used to maintain the static and dynamic information of the ship; The port management module (3) is used to manage information on the three-level terminals, including geographic coordinates, available ship types, water depth limits, loading and unloading efficiency, average waiting time, and port charges. The route management module (4) is used to maintain the flight distance matrix D(n+m)×(n+m) and the flight time matrix T(n+m)×(n+m)×k; Where n is the number of port areas, m is the current number of ships, k is the total number of ships in the fleet, and matrix element D[i][j] represents the standard route distance from position i to position j; The freight rate management module (5) is used to maintain freight rate data between different loading and unloading ports, supports manual editing and maintenance, and provides a basis for cost calculation and TCE assessment; The ship draft management module (6) is used to maintain a data table of the maximum cargo volume of different ship types at different draft depths, and is used for constraint verification of order-ship matching.

3. The ship-cargo matching system based on intelligent optimization algorithm according to claim 1, characterized in that: The route management module (4) supports setting ship priorities. Ships with higher priorities are assigned tasks first during matching to meet special business needs. The tiered division of the three-tier terminal adopts a business-driven static configuration mode, and is divided and maintained according to actual operation and management needs.

4. A matching method for a ship-cargo matching system based on an intelligent optimization algorithm as described in claim 1, characterized in that: The ship-cargo matching method based on intelligent optimization algorithms includes the following steps: The first step is to load the preprocessed raw input data, then build a fuel consumption-speed function model based on the raw input data, sort the tasks according to the weight scores, and then use a three-layer greedy allocation strategy to generate a greedy initial matching scheme, while generating an assigned task list and an unassigned task list. The second step is to initialize the relevant parameters based on the greedy initial matching scheme and the original input data. Then, in each iteration, the assigned tasks are removed according to the preset ratio and added to the list of unassigned tasks. The unassigned tasks are then reinserted into the matching scheme according to the specified strategy. Through multiple rounds of iteration and optimization, the optimized matching scheme is finally obtained. The third step is to optimize the speed of the empty flight segment of the assigned task without changing the task allocation relationship and time node, based on the fuel consumption rate formula, construct a multi-objective function to generate multiple types of execution schemes and obtain the corresponding preset evaluation index values. The fourth step involves selecting one type of execution plan as the execution benchmark based on multiple execution plans and relevant evaluation indicators, combined with matching requirements. Then, the dynamic data of the execution plans is synchronized through a dual mechanism, and the execution benchmark is updated based on the dynamic data. Matching is then performed based on the updated execution benchmark to obtain the actual execution result. The actual execution result is compared with the preset evaluation indicator value to obtain the deviation value. If the deviation value exceeds the threshold, the matching plan is adjusted based on the dynamic data, and the comparison is performed again until the deviation value meets the threshold requirement. When all tasks are completed, the matching method ends.

5. The matching method of a ship-cargo matching system based on intelligent optimization algorithm according to claim 4, characterized in that: In the first step, the raw input data includes ship information, task list, port information, route distance and other constraint data; Ship information includes static data and dynamic data; Static data includes ship name, ship type, deadweight tonnage, speed, and fuel consumption parameters; Dynamic data includes real-time location, estimated arrival time, current loading status, and matching priority; The task list, or order data, includes three types of orders: marine oil cargo, major customer cargo, and miscellaneous cargo, along with their cargo volume, loading and unloading ports, loading periods, and specified vessel types. Port information includes the geographical coordinates of the third-level terminal, water depth limits, loading and unloading efficiency, and port charges; The route distance includes the standard route distance between different locations; Other constraint data includes vessel draft restrictions, freight rate data, and order-vessel matching constraints.

6. The matching method of a ship-cargo matching system based on intelligent optimization algorithm according to claim 5, characterized in that: In the first step, the weights include: order type, ship scarcity, and time urgency; All tasks are prioritized based on a multi-dimensional scoring mechanism, with scoring factors including: Task type weighting: Marine oil tasks > Major client tasks > Small orders; Time urgency: Tasks with shorter time windows and earlier start times have higher priority; Ship scarcity: The fewer ships available to perform the task, the higher the priority.

7. The matching method of a ship-cargo matching system based on intelligent optimization algorithm according to claim 4, characterized in that: In the second step, relevant parameters are first initialized based on the greedy initial matching scheme and the original input data. Then, in each iteration, assigned tasks are removed according to a preset ratio and added to the unassigned task list. Finally, the unassigned tasks are reinserted into the matching scheme according to a specified strategy. Through multiple rounds of iterative optimization, the optimized matching scheme is finally obtained, as follows: The system receives the initial matching scheme generated by the greedy algorithm, converts it into the ALNS internal representation format, including lists of assigned and unassigned tasks, initializes the objective function value, and sets the simulated annealing parameters: initial temperature T0 = 1000, cooling rate α = 0.99, and minimum temperature T... m ᵢ n =0.1, initialize all destroy and repair operator weights to 1.0, record the current solution and the historical best solution, then re-insert the unassigned tasks into the matching scheme according to the ship priority and global optimization strategy, and then perform deep optimization through the ALNS algorithm to obtain the optimized matching scheme; The ALNS algorithm undergoes deep optimization in the following ways: A. Operator selection: A roulette wheel selection mechanism is adopted to dynamically select a destroy operator and a repair operator based on the historical performance of the operators; B. Destruction Phase: The selected destroy operator removes 20%-40% of the assigned tasks from the current scheme and adds them to the unassigned list, except for rare tasks with ≤3 ships. A mission protection mechanism is introduced to skip destruction for scarce missions with ≤3 executable ships; C. Repair phase: The selected repair operator will re-insert the unassigned tasks into the matching plan; D. Feasibility verification: After each repair, the system performs triple verification: time continuity verification, location continuity verification, and dynamic recalculation; E. Scheme Evaluation and Acceptance: Calculate the objective function value of the new scheme. The objective function is designed as follows: ； in, For total cost, For the set of unassigned tasks, For the task Unallocated penalties; The acceptance criterion employs simulated annealing: if the new solution is superior to the historical best solution, it is unconditionally accepted and the score of the best solution is updated. ; If the new solution is better than the current solution but not optimal, accept and update the current solution. ; If the new solution is worse than the current solution, by probability Accept score No score accepted ; F. Weight Update: Every 50 iterations, the system updates the weights of all operators. ； in, This is the average score of the operator.

8. The matching method of a ship-cargo matching system based on intelligent optimization algorithm according to claim 4, characterized in that: In the third step, the multi-objective function refers to the function that minimizes cost, minimizes the number of ships, and maximizes utilization. The specific objective function automatically generated by the multi-objective scheme is as follows: Minimum cost: ; Minimum number of ships: ; Highest utilization rate: ; in, The variable cost of the order includes fuel costs and port charges; For fixed costs of ships, The penalty cost for orders not being scheduled The number of ships used.

9. The matching method of a ship-cargo matching system based on intelligent optimization algorithm according to claim 4, characterized in that: In the fourth step, the dynamic data of the plan to be executed is synchronized through a dual mechanism: by driving real-time triggering and daily timed polling through order change / ship location update events, dynamic data of orders, ships, and ports are synchronized.

10. The matching method of a ship-cargo matching system based on an intelligent optimization algorithm according to claim 4, characterized in that: In the fourth step, when adjusting the matching scheme, the lock relationship between the ships and orders specified by the business personnel is retained, and a new round of optimization is triggered for the unlocked parts.

Images