Ship-yard-vehicle-engine cooperative scheduling optimization method for dry bulk cargo port

Abstract

The invention discloses a ship-yard-vehicle-machine cooperative scheduling optimization method for a dry bulk cargo port, and relates to the technical field of port logistics scheduling or the field of intelligent traffic systems, and the method comprises the following steps: S1, obtaining the shipping date, business order and operation plan data of a to-be-arrived ship, and carrying out the data preprocessing; and S2, on the basis of the preprocessed data, predicting a shipping mode, a shipping period and a stack average shipping period of the pre-arrived goods through a combined prediction model, and generating ship loading and unloading operation data. The core of the method is to construct a full-process optimization system which takes data prediction as a drive and multi-module intelligent collaboration as a venation; comprising but not limited to a prospective dispatching behavior prediction method based on time sequence prediction, a collaborative allocation mechanism of berth and yard resources, a mechanical dispatching and reshipment dispatching strategy based on multi-objective optimization, and an automobile transportation departure dispatching method.

Description

Technical Field

[0001] The present invention relates to the technical field of port logistics scheduling or the field of intelligent transportation systems, and particularly relates to a ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk ports. Background Art

[0002] In the field of dry bulk port operation, the traditional scheduling management mode usually regards links such as ship berthing, yard planning, horizontal transportation, and mechanical operation as independent modules for management. Each module often uses an independent scheduling system or relies on the experience of dispatchers for decision-making. Although the segmented management mode simplifies the operation processes of each link to a certain extent, it causes the collaborative breakage of the overall operation chain. There is no effective information interaction mechanism between each operation unit, forming a data barrier, resulting in the difficulty of achieving an overall breakthrough in port operation efficiency.

[0003] In the actual operation process, the traditional method often focuses on improving the equipment utilization rate of a single link, while ignoring the overall efficiency of multi-equipment collaborative operation. Due to the lack of an effective data sharing and collaboration mechanism between systems, the yard plan often fails to fully consider the operation efficiency of ship unloaders or the transportation capacity of trucks, easily leading to the occurrence of ship congestion. At the same time, there is no linkage between the evacuation plan and the yard inventory location and the loader status, often causing trucks to wait for a long time or travel empty, seriously reducing the resource utilization efficiency. Existing systems mostly formulate plans based on static plans and are difficult to cope with the inherent uncertainties in the port operation environment. The ship arrival time and the quantity of goods often deviate from the expectations, while the yard space allocation plan cannot be dynamically adjusted accordingly, resulting in the infeasibility of the pre-allocation plan or serious waste of yard space resources. This contradiction between local optimization and overall efficiency directly leads to the decline of operation resource utilization rate and the increase of operation costs.

[0004] Although some ports have begun to introduce information management systems in recent years, most systems remain at the level of independent scheduling and management within individual modules, lacking a holistic collaborative scheduling approach. While existing solutions achieve digital monitoring of individual aspects, such as ship berthing management, yard zoning monitoring, vehicle transportation scheduling, or machinery resource allocation, they fail to break down data silos between modules and cannot form a unified, optimized decision-making system. These systems are often limited to statistical analysis of historical operational data, lacking the ability to model the interrelationships and constraints of multiple aspects, and are even less able to cope with the real-time, dynamically changing port operating environment. When faced with complex conditions such as fluctuations in ship arrival times, cargo volume deviations, and sudden yard congestion, they cannot achieve rapid reconfiguration of cross-module resources, nor can they provide globally optimal emergency scheduling solutions, ultimately hindering further improvements in overall port operational efficiency. Therefore, developing an intelligent scheduling method that can achieve deep integration and collaborative optimization of ships, yards, vehicles, and machinery has become an urgent need to promote the intelligent and green transformation and upgrading of dry bulk ports.

[0005] Existing technological solutions: In the field of smart port scheduling, existing research has proposed various technical solutions to improve operational efficiency. For example, the patent "A vehicle-port-ship collaborative scheduling method and system" (CN111785028A) published by Shanghai Maritime University connects five operational links—berths, quay cranes, automated guided vehicles, yard cranes, and storage yards—through a chain structure, and adjusts the scheduling plan based on the predicted arrival times of ships and vehicles. Although this method achieves multi-link linkage to a certain extent, its chain model is relatively fixed, and the links are tightly coupled. When facing uncertainties common in dry bulk port operations (such as ship arrival delays and changes in cargo types), it lacks sufficient dynamic adjustment capabilities and a global optimization perspective.

[0006] Another publicly disclosed project, "A Smart Port Operation Vehicle Dispatch System and Dispatch Robot" (CN120746080A), by Sun Zhijun et al., employs a bipartite graph matching and multi-objective optimization method, aiming to reduce vehicle empty load rates and improve dispatch fairness. This solution focuses on optimizing vehicle resources, but its optimization scope is mainly limited to vehicle dispatching itself. It fails to deeply integrate and collaboratively optimize upstream processes such as berth allocation, yard planning, and machinery deployment, making it difficult to maximize efficiency throughout the entire process from ship arrival to cargo clearance.

[0007] In summary, existing technologies mostly focus on optimizing specific aspects of a port or employ relatively fixed operational models. When dealing with the complex and dynamically changing operating environment of dry bulk ports, they fall short in terms of global resource coordination, dynamic real-time adjustments, and end-to-end cost control. Therefore, there is an urgent need for an optimization method that can achieve integrated collaborative scheduling of "ship-yard-vehicle-machinery" to comprehensively improve the overall operational efficiency of dry bulk ports.

[0008] Disadvantages of the prior art: Most existing scheduling technologies manage ships, yards, machinery, and vehicles as independent modules, lacking effective data interaction and collaboration mechanisms between modules. This isolated decision-making mode leads to the disconnection of the port operation chain and fails to achieve the overall coordination and collaborative optimization of ship, yard, vehicle, and machine resources.

[0009] Although existing patented technologies have achieved automated scheduling in some local links, they lack an integrated optimization model for the entire process. When dealing with the uncertainty of ship arrivals, dynamic yard allocation, and machinery collaborative constraints, they often adopt static or segmented optimization strategies and are difficult to make dynamic adjustments and adaptive responses according to the actual port operation situation.

[0010] Existing optimization methods mostly focus on improving the cost or efficiency of a single link and fail to establish a global optimization model that comprehensively considers multiple objectives such as operation cost, equipment utilization rate, and carbon emissions. This local optimal strategy is likely to lead to low overall system efficiency and cannot achieve the optimization of the overall operation cost of dry bulk ports.

[0011] When existing patented technologies solve large-scale collaborative scheduling problems, they often have to simplify constraint conditions or adopt inefficient solution strategies due to the high complexity of the model. This results in scheduling plans that either take too long to calculate and cannot meet the real-time requirements or have insufficient plan quality and are difficult to be effectively applied in actual complex operation environments. Summary of the Invention

[0012] The purpose of the present invention is to provide a ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk ports to solve the problems raised in the above background technology.

[0013] Overall idea: The core innovation of the present invention is to introduce an "iterative collaborative optimization" mechanism to achieve data sharing, model interaction, and dynamic re-optimization in links such as ship scheduling, yard stacking, transfer dispatching, and trucking for port evacuation.

[0014] Different from the traditional linear scheduling mode, the present invention establishes a closed-loop system of prediction-driven, collaborative solution, real-time feedback, and re-iterative correction, enabling each sub-module to continuously evolve to obtain the global optimal solution under a unified model framework.

[0015] The system, as Figure 1 shown, consists of five core modules: 1. Data prediction module; 2. Collaborative decision-making module (including berth-yard collaboration); 3. Resource scheduling module (including machinery and transfer dispatching); 4. Port evacuation scheduling module; 5. Real-time feedback and re-optimization module.

[0016] The modules mentioned above interact through a unified data bus, and the system has a built-in iterative solution engine, enabling each module to have linkage characteristics in both time and space dimensions.

[0017] (1) Prediction-driven stage

[0018] The system first uses a combined prediction model (GM-BP model) to make multi-level predictions of ship arrival time, cargo flow and yard dispatch cycle.

[0019] The prediction results not only serve as the initial input, but are also dynamically corrected in each subsequent iteration.

[0020] The model recalculates the residuals and automatically adjusts the weights in each feedback loop, enabling data self-learning.

[0021] (2) Collaborative decision-making stage like Figure 2 As shown, the core of the system's collaboration lies in the integrated decision-making between the berth and the yard.

[0022] The two modules run in parallel: The stacking module assesses yard space utilization and future relocation distances to generate a set of feasible stacking locations. The scheduling module generates candidate berthing schemes based on constraints such as berth compatibility, tides, and yard distance.

[0023] The two interact with each other through a collaborative algorithm engine—when berth resources or yard distribution are adjusted, the other party automatically re-solves the problem; This process may trigger multiple iterations within a single computational cycle until the berth-reactor coupling relationship converges.

[0024] (3) Resource scheduling phase Once the collaborative results stabilize, the mechanical resource scheduling module is activated.

[0025] Based on the previous round of yard-berth decisions, this module constructs a multi-objective optimization model with the goal of minimizing overall operating costs and equipment idle rate.

[0026] A hybrid integer programming approach combined with heuristic algorithms is used to solve the problem, generating fixed-machine and mobile-machine scheduling schemes through a combination of local search and global correction.

[0027] Subsequently, the relocation and dispatch module takes over the task allocation.

[0028] If the condition of a vehicle or machine changes (e.g., equipment failure, road congestion), the system will automatically trigger a rescheduling.

[0029] (4) Port opening coordination phase The trucking and port clearance module receives information on yard inventory, vehicle queue length, and loader status. A hybrid solution system combining queuing theory and optimization model is constructed to dynamically generate vehicle release sequences and loader allocation instructions.

[0030] Its core lies in the fact that when the port clearance plan changes, the system immediately feeds back to the yard and transshipment modules, triggering the upstream plan to be recalculated, thereby forming a multi-layered, mutually supportive collaborative closed loop.

[0031] (5) Iterative feedback mechanism The key innovation of the entire system lies in the "global iteration and local correction fusion mechanism".

[0032] like Figure 2 As shown, the system is based on Benders decomposition, separating the main problem from its subproblems: Main problem: dealing with integer decision variables, such as berth allocation, pallet assignment, and machine matching; Sub-problems: Handling continuous variables, such as operation time and transportation routes.

[0033] When the field data is updated or the output of any module deviates, the system will generate new cutting constraints and return to the main problem to solve it again.

[0034] Each iteration causes the upper and lower bounds of the model to gradually converge until the global optimality condition is met.

[0035] The system achieves this through this mechanism: Dynamically respond to emergencies at the port; Automatically adjust calculation weights and constraints; Ensure that the algorithm achieves a balance between real-time performance and accuracy.

[0036] To achieve the above objectives, the present invention adopts the following technical solution: A method for optimizing the coordinated scheduling of ships, yards, vehicles, and machinery at dry bulk ports includes the following steps: S1: Obtain the expected vessel schedule, commercial order, and work plan data, and perform data preprocessing; S2, based on the preprocessed data, uses a combined prediction model to predict the shipping method, shipping cycle and average shipping cycle of the pre-arrived goods, and generates ship loading and unloading operation data. S3, input the predicted data obtained from S2 into the intelligent stacking module and the ship scheduling module in parallel to jointly generate the optimal unloading stacking scheme and berth allocation scheme. S4, based on the stacking and unloading scheme generated in S3, formulate the work ratio and scheduling strategy for fixed machines and flow machines; S5, the transfer dispatching is based on the transfer plan generated by S3 and the mechanical scheduling plan of S4, combined with the real-time vehicle status, to carry out intelligent assignment and dynamic scheduling of transfer machinery. S6 automatically executes vehicle release and loader dispatch instructions according to the port clearance plan, and monitors and outputs key operational indicators in real time.

[0037] In this invention, in the data acquisition and preprocessing step S1, the system connects to the port business system to acquire the raw data of the expected arriving ships, and cleans, normalizes, and processes missing values ​​in the data to provide high-quality data input for the subsequent prediction model.

[0038] In this invention, in the collaborative decision-making step S3 between the storage yard and berth: In the intelligent stacking module, the available space in the yard is first assessed, and then the specific unloading stacking position is recommended for the expected ship through the unloading stacking position recommendation algorithm, and the standard stacking information and transshipment plan are output. In the vessel scheduling module, the berth-yard collaborative scheduling algorithm integrates the stackable berth information output by the intelligent stacking module with the current available berth information to allocate suitable berths for expected arrival vessels, thereby achieving efficient coordination between berth and yard resources.

[0039] In this invention, in the mechanical resource scheduling and planning step S4, with multiple objectives such as operating cost and efficiency as the optimization direction, detailed allocation schemes and scheduling strategies are generated for equipment such as gantry cranes, stacker-reclaimers, and loaders, and fixed machine allocation schemes and mobile machine allocation schemes are output.

[0040] In this invention, in the transshipment dispatch and execution step S5, the system uses an optimization algorithm to achieve one-click intelligent dispatch under various schemes based on the predetermined transshipment plan (including unloading and transshipment) and real-time vehicle information, and can make real-time adjustments according to the dynamic situation on site.

[0041] In this invention, in the port transport monitoring and management step S6, the system automatically controls the release rhythm of port vehicles based on the real-time status of port traffic and operations, and simultaneously issues loader dispatch instructions. At the same time, it continuously collects and analyzes key performance indicators such as operation efficiency, cost, and vehicle dwell time in the port.

[0042] As a preferred embodiment of the present invention, the combined prediction model adopts a two-layer structure: the first layer uses grey prediction, BP neural network and multiple linear regression model for parallel prediction respectively; the second layer calculates dynamic weights based on the prediction error of the first layer and outputs weighted prediction results.

[0043] As a preferred technical solution of the present invention, the berth-yard collaborative scheduling algorithm takes minimizing ship waiting time and maximizing berth utilization as optimization objectives, and considers constraints such as ship size, water depth, cargo type and yard distance to achieve bidirectional matching between berths and stacking positions.

[0044] As a preferred technical solution of the present invention, the stacking location planning algorithm comprehensively considers the available area of ​​the storage yard, stacking height limit, relocation distance, storage time and unloading sequence constraints, and uses simulated annealing or genetic algorithm for solution.

[0045] As a preferred technical solution of the present invention, the mechanical resource scheduling plan adopts a mixed integer programming model, and the constraints include mechanical capacity, working time period, regional accessibility and work allocation. It outputs multiple alternative plans and makes the optimal selection based on cost-efficiency weights.

[0046] As a preferred technical solution of the present invention, the transport dispatch module allocates resources through a two-layer optimization model: the first layer completes the matching and binding of vehicles and tasks; the second layer optimizes the operation route and time window based on the principle of proximity allocation.

[0047] As a preferred technical solution of the present invention, the port transportation scheduling model combines real-time data such as vehicle queue length, road congestion level, and loader working status, and uses dynamic release strategies to adjust the pace of vehicle entry into the port, thereby optimizing the port's traffic efficiency.

[0048] As a preferred technical solution of the present invention, the overall collaborative model is solved by the Benders decomposition algorithm, wherein the main problem is used to handle integer variables such as berth allocation and berth assignment, and the sub-problems are used to solve continuous variables such as work process and resource allocation.

[0049] As a preferred embodiment of the present invention, the system dynamically generates cut constraints based on real-time job data after each Benders iteration, updates the master problem, and achieves real-time convergence and adaptive adjustment of the model.

[0050] As a preferred technical solution of the present invention, the system has a multi-scheme decision-making mechanism, which can output cost-optimal, efficiency-optimal or balanced scheduling schemes according to weight parameters for the scheduler to choose to execute.

[0051] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention constructs a highly integrated data flow closed loop, starting with accurate prediction to drive intelligent decision-making in all downstream stages. From prediction of dispatch behavior to collaborative decision-making, and then to resource scheduling and execution, each stage is closely connected and data interacts in real time, ensuring the scientific nature and consistency of the scheduling scheme. This achieves a fundamental shift from passive response to proactive optimization, significantly improving the accuracy and reliability of the overall plan.

[0052] 2. By defining clearly defined module interfaces and data flows, deep collaboration among the four core components—ships, yards, vehicles, and machinery—was achieved on a unified data platform. In particular, the berth-yard collaborative scheduling algorithm fundamentally solved the efficiency bottleneck caused by the mismatch between berth and yard locations, forming a closed-loop optimization system from prediction to execution.

[0053] 3. Each core module uses multi-constraint, multi-objective intelligent algorithms for decision-making, and realizes one-click assignment and dynamic scheduling, which greatly reduces manual intervention, improves response speed and decision quality, and reduces labor costs.

[0054] 4. Improved efficiency across the entire process: Through collaborative optimization across the entire chain, the port's dwell time and vehicle waiting time have been effectively shortened, and the utilization rate of machinery and the turnover rate of the storage yard have been improved. Ultimately, this has resulted in a leapfrog improvement in the overall operational efficiency of the dry bulk port and a significant reduction in operating costs.

[0055] In summary, this invention provides a collaborative scheduling optimization method for ship-yard-vehicle-machinery systems in dry bulk ports. Its core lies in constructing a full-process optimization system driven by data prediction and based on intelligent collaboration of multiple modules, including but not limited to a forward-looking departure behavior prediction method based on time-series prediction, a collaborative allocation mechanism for berth and yard resources, a machinery scheduling and transshipment dispatch strategy based on multi-objective optimization, and a port clearance scheduling method based on road transport. Attached Figure Description

[0056] Figure 1 The system architecture diagram shows the interaction relationships between the data prediction module, collaborative decision-making module, resource scheduling module, port dispatching module, and feedback re-optimization module.

[0057] Figure 2 The system iterative collaboration flowchart shows the cyclical iterative path of prediction, collaboration, execution, feedback, and re-optimization, as well as the solution logic block diagram of the main problem and sub-problems. Detailed Implementation

[0058] 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. 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.

[0059] Please see Figure 1-2 This invention provides a technical solution for a ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports, comprising the following steps: S1: Obtain the expected vessel schedule, commercial order, and work plan data, and perform data preprocessing; S2, based on the preprocessed data, uses a combined prediction model to predict the shipping method, shipping cycle and average shipping cycle of the pre-arrived goods, and generates ship loading and unloading operation data. S3, input the predicted data obtained from S2 into the intelligent stacking module and the ship scheduling module in parallel to jointly generate the optimal unloading stacking scheme and berth allocation scheme. S4, based on the stacking and unloading scheme generated in S3, formulate the work ratio and scheduling strategy for fixed machines and flow machines; S5, the transfer dispatching is based on the transfer plan generated by S3 and the mechanical scheduling plan of S4, combined with the real-time vehicle status, to carry out intelligent assignment and dynamic scheduling of transfer machinery. S6 automatically executes vehicle release and loader dispatch instructions based on the port clearance plan and port traffic information, and monitors and outputs key operational indicators in real time.

[0060] In this invention, in the data acquisition and preprocessing step S1, the system connects to the port business system to acquire the raw data of the expected arriving ships, and cleans, normalizes, and processes missing values ​​in the data to provide high-quality data input for the subsequent prediction model.

[0061] In this invention, in the collaborative decision-making step S3 between the storage yard and berth: In the intelligent stacking module, the available space in the yard is first assessed. Then, the stacking location, size and stacking situation of the goods are intelligently determined through a multi-constraint, multi-objective stacking planning algorithm. Finally, the unloading stacking location recommendation algorithm recommends specific unloading stacking locations for the expected ships and outputs standard stacking information and transshipment plan. In the vessel scheduling module, the berth-yard collaborative scheduling algorithm integrates the stackable berth information output by the intelligent stacking module with the current available berth information to allocate suitable berths for expected arrival vessels, thereby achieving efficient coordination between berth and yard resources.

[0062] In this invention, in the mechanical resource scheduling and planning step S4, with multiple objectives such as operating cost and efficiency as the optimization direction, detailed allocation schemes and scheduling strategies are generated for equipment such as gantry cranes, stacker-reclaimers, and loaders, and fixed machine allocation schemes and mobile machine allocation schemes are output.

[0063] In this invention, in the transshipment dispatch and execution step S5, the system uses an optimization algorithm to achieve one-click intelligent dispatch under various schemes based on the predetermined transshipment plan (including unloading and transshipment) and real-time vehicle information, and can make real-time adjustments according to the dynamic situation on site.

[0064] In this invention, in the port transport monitoring and management step S6, the system automatically controls the release rhythm of port vehicles based on the real-time status of port traffic and operations, and simultaneously issues loader dispatch instructions. At the same time, it continuously collects and analyzes key performance indicators such as operation efficiency, cost, and vehicle dwell time in the port.

[0065] Specifically: (1) Data acquisition and preprocessing: Connect to the port production management system to obtain data on expected vessel arrival schedules, business orders, work plans, equipment status, yard capacity, and traffic flow; clean, normalize, and fill in missing values ​​for the data. Missing value handling: ① For empty data, if the missing value exceeds 50%, all data in that row will be deleted; ② For cases where only a small portion of the data is missing, the mean of the other data in that category will be used to fill the missing values, in order to ensure the accuracy of the constructed indicator system and the stability of the model training.

[0066] Outlier handling: For outlier repair, this paper also replaces outliers by filling in the mean, in order to reduce noise and inconsistency in the data.

[0067] Zero value handling: For cases where the data table contains a value of 0, since 0 values ​​will not have any impact during data processing and analysis, all data in that row will be deleted.

[0068] The historical data is randomly divided into 70% and 30% using the most common method. The 70% is used as the training set, where the model learns and adjusts its parameters to optimize its predictive ability. The remaining 30% is used as the test set to evaluate the performance of the trained model on unseen data. The test set data must have never been used during training to ensure the fairness and accuracy of the evaluation.

[0069] (2) Prediction of issuance behavior: A combined prediction model is constructed based on the grey prediction model, BP neural network and multiple linear regression model to make multi-level predictions on cargo shipping method, shipping cycle and average shipping cycle of stacking location, and output the prediction results. Step 1: Network initialization: Randomly assign initial weights to neurons in the hidden layer and input layer to determine the initial values.

[0070] Step 2: Iterative Training: To ensure the final result meets the accuracy requirements... Repeat the following calculation until the convergence condition is met:

[0071] Forward process: Calculate each layer , and , ; Reverse process: for each layer calculate ; Adjust the weights: ,in, This is the learning rate.

[0072] Step 3: Output the results.

[0073] (3) Coordinated decision-making between berths and storage yards: Using the prediction results, the intelligent stacking module determines the stacking location and stacking scheme of goods through a multi-constraint, multi-objective stacking location planning algorithm; the ship scheduling module generates the optimal berth allocation scheme and unloading stacking location scheme by integrating the storage information of the storage yard and the berth resource information based on the berth-yard collaborative scheduling algorithm. In the implementation process, the complex scheduling problem influenced by multiple factors is first transformed into a computable optimization model. This model is constructed by introducing decision variables, input parameters, an objective function, and constraints. The ship set is defined as follows: berths are grouped as The core decision variables include the ship. berthing start time And Boolean variables characterizing the ship-berth allocation relationship. When the ship Arranged to berth Its value is 1 when the task is completed, and 0 otherwise.

[0074] During the optimization process, the system needs to weigh multiple operational objectives. A typical objective function aims to minimize the total time all ships spend in port, and its expression is: ,in The estimated arrival time of the vessel. The planned operating hours are determined accordingly. This objective directly contributes to improving the overall operational efficiency of the port.

[0075] To ensure the effectiveness of the model and the practical feasibility of the scheduling scheme, a series of strict constraints must be incorporated. These constraints mainly include: The non-overlapping berth constraint ensures that any berth serves at most one vessel at any given time; Berth compatibility constraints require that the physical parameters of a vessel, such as its length and draft, match the corresponding capabilities of the assigned berth. Tidal time window constraints restrict deep-draft vessels to berthing during tidal periods that are safe for their operation; Operational readiness constraints ensure that vessel scheduling meets pre-requisites such as business, yard, and equipment readiness requirements; Safety interval constraints ensure necessary buffers in time or space between vessels operating continuously; and scheduling priority constraints incorporate specific operational rules such as "first-come, first-served" or priority for key vessels. These constraints are integrated into the model in the form of mathematical inequalities or logical relationships, collectively defining the feasible domain of understanding.

[0076] In summary, the ship scheduling problem addressed by this invention is characterized by its large scale, multiple constraints, and dynamic nature, classifying it as an NP-hard problem. Therefore, its implementation requires combining operations research theory with intelligent algorithms. This involves constructing the aforementioned scalable mathematical model and designing efficient heuristic or metaheuristic solution algorithms to achieve an evolution from experience-based decision-making to intelligent computation in the scheduling process.

[0077] Based on the aforementioned ship scheduling scheme, further collaborative optimization of yard operations is performed. By constructing an ore cargo storage demand allocation model, with the objective of minimizing yard congestion and turnaround time, the objective function can be expressed as:

[0078] in, Indicates time period The upper limit of the storage yard capacity, For goods During the period The planned entry volume, Let 0-1 be the decision variable, representing the goods Is it during the time period? Entry. The system uses a simulated annealing algorithm to solve the problem and dynamically plans the entry sequence of each batch of goods to ensure that the real-time inventory in the yard is always below the set threshold.

[0079] For non-process-oriented pallet allocation, the system acquires real-time information on goods to be stacked and pallet status, and establishes a pallet selection optimization model. Let... Let be the decision variable, representing the goods. Should it be assigned to a stack location? The objective function is a comprehensive indicator that maximizes the utilization rate of the stack space and minimizes the handling distance:

[0080] in, Indicates goods Place in stack Space utilization rate For the corresponding transport distance, and These are the weighting coefficients. The model is solved using a Skyline-based heuristic algorithm to generate stacking location assignment schemes and scheduling instructions.

[0081] For streamlined stacking operations, a linear stacking model is established. This is achieved by introducing auxiliary variables. Indicates stack position Should an arrangement method be used? And the Big M method is used to transform nonlinear constraints into linear forms, for example:

[0082] in, For stack position Arrangement The length occupied below, The length is set to an upper limit, and M is a sufficiently large positive number. The model integrates multiple objectives into a single objective using a weighting method and calls a mathematical programming solver to find the optimal cargo arrangement and stacking location allocation scheme.

[0083] Finally, based on the outputs of the storage demand assessment model and the stacking optimization model, the system constructs a expedited pickup and handling decision model. Let... For goods The emergency storage threshold, if the current storage amount satisfy:

[0084] Then, a reminder task is generated and relocation resources are allocated. Its quantity is determined by the following formula:

[0085] in This is the standard capacity for a single transport operation. Based on this, the system automatically generates work instructions and sends them to relevant departments and customer terminals.

[0086] (4) Mechanical resource scheduling planning: Based on the berth allocation and stacking scheme, a mixed integer programming model is constructed with the goal of minimizing operating costs and maximizing equipment utilization to generate fixed and flow machine ratio schemes for equipment such as gantry cranes, stacker-reclaimers, and loaders.

[0087] The input information integrates operational requirements and system data, mainly including: the workload, location, deadline and priority of each work point; the quantity, efficiency and process accessibility of various types of machinery (gantry cranes, stacker-reclaimers, belt conveyors); current and predicted ore unloading needs; additional revenue or expenses related to dispatch and demurrage; and requirements for machinery collaboration.

[0088] The model uses a dynamic programming algorithm for recursive decision-making. At each decision time... According to the current mechanical state vector and the set of tasks to be assigned Determine an allocation action This action specifies the exact workload of each task assigned to each machine. After assignment, the status is updated. The immediate cost of each decision. This considers both time consumption and travel distance costs. The overall optimization objective is to minimize the total scheduling cost, which can be expressed as a function:

[0089] Simultaneously, the model needs to maximize the overall utilization rate of mechanical resources. To achieve multi-objective trade-offs and multi-solution generation, a weighted sum method is used to integrate multiple objectives, and penalty terms are added to the constraints. Let the original objective cost be... Utilization rate The degree of constraint violation is Then the weighted single-objective function is:

[0090] The different weight combinations are systematically traversed through a grid search. This can generate a series of scheduling schemes, each with its own emphasis on key indicators such as operational efficiency and cost, for decision-makers to choose from.

[0091] The model strictly incorporates the following key constraints: Equipment capacity constraints: The workload of each machine must not exceed its maximum load or speed limit.

[0092] Time window and priority constraints: Tasks must be completed within their allowed time period, and high-priority or urgent tasks are scheduled first.

[0093] Work allocation constraints: The workload among gantry cranes, stacker-reclaimers, and belt conveyors needs to be dynamically balanced to adapt to changes in demand at different stages of unloading and avoid equipment idleness or bottlenecks.

[0094] Environmental constraints: When weather conditions such as wind speed and rainfall affect equipment availability and operational safety, the scheduling plan will be adjusted accordingly.

[0095] (5) Relocation, dispatching, and execution: Based on the aforementioned transport scheme and machinery scheduling scheme, combined with real-time vehicle status, a multi-objective optimization algorithm is used to achieve one-click intelligent assignment of transport machinery and tasks, and to make real-time adjustments according to the dynamic situation on site. The main constraints include: ① Constraints on the allocation of dump trucks and loaders: The planned number of dump trucks and loaders cannot exceed the total number of dump trucks and loaders in the work team; ② Coordination quantity constraint: The minimum number of dump trucks and loaders allocated to each plan must meet a certain quantity relationship, and the number of dump trucks must not be less than the number of loaders; ③ Planned work group constraint: Each plan can be assigned to a maximum of one work group; ④ Restrictions between decision variables; ⑤ The loader meets the safe distance requirements.

[0096] ⑥ Planned work group constraints: Each plan only dispatches vehicles within the work group; ⑦ Dump trucks and loaders can only serve one program at a time.

[0097] (6) Port transport scheduling and monitoring by road: Based on the port clearance plan and real-time traffic information within the port, a scheduling model combining queuing theory and mixed integer programming is constructed to dynamically generate vehicle release sequences and loader scheduling instructions, and output key performance indicators such as operational efficiency, cost, and vehicle dwell time within the port. When the input process is a Poisson distribution, The probability that at least one customer arrives within the interval is

[0098] This probability can be expressed as...

[0099] Therefore, the independent and identical negative exponential distribution of the arrival intervals is equivalent to the input process being a Poisson distribution, and both are labeled as follows: To express.

[0100] Service Hours The distribution of the time taken for each vehicle entering the port also follows a negative exponential distribution in this scenario, with its distribution function and probability density function being as follows:

[0101] ,

[0102] in, The number of vehicles that can be serviced per unit of time is called the average service rate, or in this scenario, the average workload rate. This represents the average service time for a customer.

[0103] In the scenario of automatic vehicle release for port transport vehicles, the queuing model studied is a multi-server negative exponential distribution system with limited system capacity.

[0104] Assume the maximum capacity limit of the system is , The number of operating vehicles in the system The state probabilities and operational metrics of the entire queuing system at this point are as follows:

[0105]

[0106] in .

[0107] In this process, steps (3) to (6) involve real-time data interaction and feedback through standardized data interfaces to form a closed-loop scheduling system for the entire process, thereby achieving collaborative optimization of port vessels, storage yards, vehicles and machinery.

[0108] The implementation process of the overall technical solution of this invention: The implementation of this invention is based on the core principle of data-driven global collaboration. First, a combined prediction model is used to accurately predict the dispatch behavior of arriving cargo, providing a scientific basis for all subsequent scheduling decisions. Based on this, the system performs multi-level collaborative optimization: First, the intelligent stacking and ship scheduling module, based on the prediction results, collaboratively generates the optimal stacking and berth allocation scheme with the goal of minimizing operational conflicts and transportation distances. Next, the machinery resource scheduling module establishes an independent mixed-integer programming model to formulate cost- and efficiency-optimized allocation and scheduling strategies for equipment such as gantry cranes and stacker-reclaimers. Based on the machinery scheduling scheme, the transshipment dispatch module establishes a corresponding optimization model for transshipment tasks within the port, realizing one-click intelligent assignment and dynamic scheduling of transshipment machinery. Subsequently, the trucking port clearance module, based on the port clearance plan and real-time operating conditions, constructs a multi-objective mixed-integer programming model with the core objective of improving the efficiency of external truck operations and shortening port dwell time, automatically generating and executing vehicle release and loader scheduling instructions. Finally, the system forms a closed-loop intelligent scheduling system from prediction and planning to execution and feedback.

[0109] Specifically, when a ship is about to arrive at the port, the system obtains the pre-arrival berth data, uses predictive analysis methods to accurately predict the dispatch behavior of the arriving cargo, and outputs ship loading and unloading operation data.

[0110] Based on ship loading and unloading operation data, the intelligent yard stacking module first assesses the available space in the yard. If there is available space, it plans reasonable stacking locations using a stacking location planning algorithm and outputs the available stacking location information.

[0111] Meanwhile, after receiving the ship loading and unloading operation data, the ship scheduling module, if there are currently available berths, coordinates with the stackable stacking information output by the intelligent stacking module, and outputs available berth information through the berth-yard collaborative scheduling algorithm to allocate berths to arriving ships; if there are currently no available berths, it allocates temporary anchorage for the expected arriving ships, thus realizing the berth allocation for the expected arriving ships.

[0112] Simultaneously, the available berth information data from the ship scheduling module will be transmitted to the intelligent stacking module. Together with the available stacking location information, a stacking location recommendation algorithm will output standard stacking information and quantitative indicators for the optimal stacking process. Based on the information output by this algorithm, the machinery resource scheduling module will allocate machinery resources for the stacking process, determining both fixed machinery allocation and mobile machinery allocation schemes. Machinery will then be allocated accordingly to complete the stacking process.

[0113] Returning to the intelligent stacking module, if no available storage yard is available, the overall stacking layout needs to be optimized to generate a stacking and relocation plan. This involves merging and relocating stacks in the current storage yard to create vacant stacking spaces. During this process, merging and relocation plans will be generated. These plans are then input into the machinery resource scheduling module to generate the required fixed machinery allocation and mobile machinery allocation plans. Working in conjunction with the merging and relocation plans output by the intelligent stacking module and the fixed machinery and mobile machinery allocation plans output by the machinery resource scheduling module, the relocation dispatch module will use an optimization model to assign relocation teams and machinery to create vacant stacking spaces in the storage yard. Finally, the intelligent stacking module will allocate stacking spaces to the arriving goods to complete the stacking process.

[0114] After the machinery is allocated by the machinery resource scheduling module, the machinery information will be input into the port transportation module. At the same time, based on the current port waiting vehicle information and port traffic information, the queuing vehicles will be released to enter the port through the port vehicle scheduling optimization algorithm.

[0115] Meanwhile, when the transshipment dispatching is optimized, the vehicles and machinery dispatched will also affect the dispatching decisions of the road transport port module.

[0116] In the above multi-module collaborative scheduling, the system uses the Benders decomposition algorithm as the iterative solution engine to decompose the complex multi-stage scheduling problem into a main problem and multiple sub-problems. In each iteration, new constraint information (cut constraints) is generated based on real-time feedback data to correct the solution results of the previous round.

[0117] The entire iterative loop consists of the following three levels: Data feedback layer (outer iteration): This layer collects execution data in real time from modules such as berth operations, yard sensing, vehicle scheduling, and machinery status. It is used to detect plan deviations, such as ship arrival delays, yard congestion, and abnormal vehicle queuing, and triggers model input updates.

[0118] Model Reconstruction Layer (Mid-Level Iteration): Based on data deviations, the system automatically adjusts the parameter weights and constraint boundaries of each module. For example, if the yard turnover rate is below a threshold, the system automatically increases the constraint weight of "yard distance"; if ship waiting time is too long, a penalty term is added to the berth scheduling model; if vehicle empty-running rate increases, the objective function of the transshipment dispatch sub-model will increase the path optimization weight. The adjusted constraints are synchronized to all sub-modules via the data bus.

[0119] Optimize the solution layer (inner iteration): After each model reconstruction, the Benders engine resolves the main problem and subproblems: the main problem determines integer decisions, such as berth allocation, racking assignment, and machine grouping; the subproblems solve for continuous variables, such as operation time, resource utilization, and transportation routes. When the subproblem solution violates the constraints of the main problem or has a large objective deviation, the system automatically generates new Benders cuts (including feasibility cuts and optimality cuts) and returns the updated constraint set to the main problem.

[0120] This process continues until the lower bound of the main problem and the upper bound of the subproblems converge to within the allowable error range.

[0121] The feedback mechanism of this invention is not limited to adjustments within a single module, but rather forms cross-module collaboration at the system level: When the berth allocation scheme is updated, the yard stacking module will adjust the transfer path accordingly. Once the relocation work is redistributed, the trucking and port clearance module immediately updates the vehicle release schedule. When the pace of port clearance changes too much, it will in turn affect the rate of cargo outflow from the yard, thereby triggering a new round of berth-yard re-optimization.

[0122] This iterative structure of "horizontal parallelism + vertical nesting" enables multi-module mutual feedback and full-process collaborative optimization, avoiding the delay and inconsistency problems of traditional segmented scheduling.

[0123] The specific operation methods of each module are as follows: Data Acquisition and Shipment Behavior Forecasting Module: This module integrates multi-source data such as expected arrival dates, business orders, and work plans, combined with external market indicators and historical operation records, to construct a complete forecasting data foundation. Employing a combined forecasting method, through multi-model parallel forecasting and error-weighted integration strategies, it achieves accurate predictions of cargo shipment methods and cycles, providing data support for subsequent scheduling decisions.

[0124] Intelligent Palletizing Module: Based on prediction results and real-time yard status data, this module uses intelligent optimization algorithms to assess yard storage needs and generates palletizing solutions with the goals of optimal space utilization, operational efficiency, and cost. According to the characteristics of different yard types, it adaptively selects spatial layout optimization methods to determine the best stacking location and palletizing strategy for goods, ultimately outputting a complete palletizing solution that includes comparison indicators for multiple solutions.

[0125] Vessel Scheduling Module: This module comprehensively considers vessel characteristics, berth conditions, and operational requirements, establishing a multi-factor comprehensive evaluation system for berth adaptability analysis. By integrating berthing and departure resource calculation, operation time estimation, and navigation condition verification, a complete vessel scheduling constraint system is constructed, and optimization algorithms are used to generate the optimal berthing and departure sequences and berth allocation schemes.

[0126] Machinery Resource Scheduling Module: Based on stacking schemes and transportation requirements, this module transforms work tasks into specific equipment scheduling instructions through a multi-objective optimization model. Taking into account equipment status, operational constraints, and business requirements, it uses intelligent optimization methods to generate multiple alternative scheduling schemes and provides quantitative indicators of efficiency and cost for each scheme to aid in decision-making.

[0127] Intelligent Dispatch Module for Transport: This module optimizes transport plan combinations through intelligent binding algorithms, generating task grouping schemes with the core objective of reducing empty runs. It employs a hierarchical optimization strategy, first completing macro-level matching of vehicles and plans, then performing fine-grained scheduling, and finally outputting a complete scheduling scheme including work routes and time arrangements.

[0128] Port Transportation Scheduling Module: This module utilizes queuing theory to construct a vehicle scheduling model. Through system performance analysis, it dynamically adjusts vehicle release strategies. By combining real-time traffic conditions and equipment status, it collaboratively optimizes vehicle release and loader allocation, forming a closed-loop control port transportation scheduling system.

[0129] Global Cooperative Scheduling Framework: Based on the independent optimization of each module, this invention constructs a global cooperative scheduling framework through a systematic integration method, establishes a unified optimization model to coordinate the decisions of each module, and uses a decomposition and coordination algorithm to achieve full-process optimization solution, ensuring efficient coordination and dynamic adjustment of each link of ship-yard-vehicle-machine.

[0130] Dynamic Adjustment and Multi-Option Decision-Making: The overall collaborative model and solution algorithm together constitute the core of the system's intelligent decision-making. The system dynamically responds to changes in the port operating environment by driving the iterative process of the Benders decomposition algorithm through real-time data. Simultaneously, by adjusting model parameters and weights, the algorithm can generate multiple scheduling schemes with different focuses, such as the cost-optimal scheme and the efficiency-optimal scheme, providing managers with diversified decision-making options and significantly enhancing the adaptability of the scheduling plan.

[0131] The technical solution proposed in this invention has the following beneficial effects: 1. This invention constructs a highly integrated data flow closed loop, starting with accurate prediction to drive intelligent decision-making in all downstream stages. From prediction of dispatch behavior to collaborative decision-making, and then to resource scheduling and execution, each stage is closely connected and data interacts in real time, ensuring the scientific nature and consistency of the scheduling scheme. This achieves a fundamental shift from passive response to proactive optimization, significantly improving the accuracy and reliability of the overall plan.

[0132] 2. By defining clearly defined module interfaces and data flows, deep collaboration among the four core components—ships, yards, vehicles, and machinery—was achieved on a unified data platform. In particular, the berth-yard collaborative scheduling algorithm fundamentally solved the efficiency bottleneck caused by the mismatch between berth and yard locations, forming a closed-loop optimization system from prediction to execution.

[0133] 3. Each core module uses multi-constraint, multi-objective intelligent algorithms for decision-making, and realizes one-click assignment and dynamic scheduling, which greatly reduces manual intervention, improves response speed and decision quality, and reduces labor costs.

[0134] 4. Improved efficiency across the entire process: Through collaborative optimization across the entire chain, the port's dwell time and vehicle waiting time have been effectively shortened, and the utilization rate of machinery and the turnover rate of the storage yard have been improved. Ultimately, this has resulted in a leapfrog improvement in the overall operational efficiency of the dry bulk port and a significant reduction in operating costs.

[0135] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0136] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the coordinated scheduling of ships, yards, vehicles, and machinery in dry bulk cargo ports, characterized in that: Includes the following steps: (1) Data acquisition and preprocessing: Connect to the port production management system to obtain data on expected vessel arrival schedules, business orders, work plans, equipment status, yard capacity, and traffic flow; clean, normalize, and fill in missing values ​​for the data. (2) Prediction of issuance behavior: A combined prediction model is constructed based on the grey prediction model, BP neural network and multiple linear regression model to make multi-level predictions on cargo shipping method, shipping cycle and average shipping cycle of stacking location, and output the prediction results. (3) Coordinated decision-making between berths and storage yards: Using the prediction results, the intelligent stacking module determines the stacking location and stacking scheme of goods through a multi-constraint, multi-objective stacking location planning algorithm; the ship scheduling module generates the optimal berth allocation scheme and unloading stacking location scheme by integrating the storage information of the storage yard and the berth resource information based on the berth-yard collaborative scheduling algorithm. (4) Mechanical resource scheduling planning: Based on the berth allocation and stacking scheme, a mixed integer programming model is constructed with the goal of minimizing operating costs and maximizing equipment utilization to generate a fixed-machine and flow-machine ratio scheme for gantry cranes, stacker-reclaimers, and loaders. (5) Relocation, dispatching, and execution: Based on the aforementioned transport scheme and machinery scheduling scheme, combined with real-time vehicle status, a multi-objective optimization algorithm is used to achieve one-click intelligent assignment of transport machinery and tasks, and to make real-time adjustments according to the dynamic situation on site. (6) Port transport scheduling and monitoring by road: Based on the port clearance plan and real-time traffic information within the port, a scheduling model combining queuing theory and mixed integer programming is constructed to dynamically generate vehicle release sequences and loader scheduling instructions, and output performance indicators such as operation efficiency, cost, and vehicle dwell time within the port. In this process, steps (3) to (6) involve real-time data interaction and feedback through standardized data interfaces to form a closed-loop scheduling system for the entire process, thereby achieving collaborative optimization of port vessels, storage yards, vehicles and machinery.

2. The ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports according to claim 1, characterized in that: The combined prediction model adopts a two-layer structure: the first layer uses grey prediction, BP neural network and multiple linear regression model for parallel prediction; the second layer calculates dynamic weights based on the prediction error of the first layer and outputs weighted prediction results.

3. The ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports according to claim 1, characterized in that: The berth-yard collaborative scheduling algorithm aims to minimize ship waiting time and maximize berth utilization. It takes into account constraints such as ship size, water depth, cargo type and yard distance to achieve two-way matching between berths and stacking positions.

4. The ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports according to claim 1, characterized in that: The stacking location planning algorithm comprehensively considers the available area of ​​the storage yard, stacking height limit, relocation distance, storage time and unloading sequence constraints, and uses simulated annealing or genetic algorithm for solution.

5. The ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports according to claim 1, characterized in that: The mechanical resource scheduling plan adopts a mixed integer programming model, with constraints including mechanical capacity, operating time period, regional accessibility and operation ratio. It outputs multiple alternative plans and makes the optimal selection based on cost-efficiency weights.

6. The method for optimizing ship-yard-vehicle-machine collaborative scheduling for dry bulk cargo ports according to claim 1, characterized in that: The dispatching module allocates resources through a two-layer optimization model: the first layer matches and binds vehicles and tasks; the second layer optimizes the work route and time window based on the principle of proximity allocation.

7. The method for optimizing ship-yard-vehicle-machine collaborative scheduling for dry bulk cargo ports according to claim 1, characterized in that: The port transportation scheduling model combines real-time data on vehicle queue length, road congestion level, and loader working status, and uses dynamic release strategies to adjust the pace of vehicle entry into the port, thereby optimizing the efficiency of traffic flow within the port.

8. The method for optimizing ship-yard-vehicle-machine collaborative scheduling for dry bulk cargo ports according to claim 1, characterized in that: The overall collaborative model is solved using the Benders decomposition algorithm, where the main problem is used to handle the integer variables of berth allocation and berth assignment, and the sub-problems are used to solve the continuous variables of work process and resource allocation.

9. The method for optimizing ship-yard-vehicle-machine collaborative scheduling for dry bulk cargo ports according to claim 1, characterized in that: The system dynamically generates cut constraints based on real-time job data after each Benders iteration, updates the main problem, and achieves real-time convergence and adaptive adjustment of the model.

10. The ship-yard-vehicle-machine collaborative scheduling optimization method for dry bulk cargo ports according to claim 1, characterized in that: The system has a multi-scheme decision-making mechanism, which can output cost-optimal, efficiency-optimal, or balanced scheduling schemes based on weight parameters, for schedulers to choose from and execute.

Images