Container terminal intelligent management method and system based on space-time coordinate mapping

Abstract

The invention relates to the technical field of container terminal intelligent scheduling, in particular to a container terminal intelligent management method and system based on space-time coordinate mapping, and the method comprises the steps: obtaining ship operation associated multi-source data, and carrying out the data fusion; establishing a control management coordinate system; generating a ship distribution thermodynamic diagram, and mapping each ship into a rectangular color block; loading container operation feature data, embedding the container operation feature data into color blocks, and performing partition identification; analyzing container space distribution characteristics from the partition identifiers; generating a crane deployment scheme and issuing the crane deployment scheme to the wharf operation terminal; determining whether to trigger an early warning mechanism; if not, the current crane deployment scheme is maintained; if yes, an early warning mechanism is triggered, and the current crane deployment scheme is dynamically optimized; and pushing the optimized crane deployment scheme to an operation terminal in real time. The method provided by the invention has the effect of improving the problems of poor dynamic environment adaptability and low multi-target cooperation efficiency of the wharf scheduling method.

Description

Technical Field

[0001] This application relates to the technical field of intelligent scheduling of container terminals, and in particular to an intelligent management method and system for container terminals based on spatiotemporal coordinate mapping. Background Technology

[0002] Container terminal scheduling is the central nervous system of a port logistics system. Through scientific planning and real-time coordination, it enables the dynamic matching of resources such as ship loading and unloading, container storage, horizontal transport by trucks, and quay cranes and yard cranes to minimize ship time in port, reduce operating costs, and improve overall throughput efficiency. Specifically, scheduling needs to address the order of ship arrivals and berth allocation, determining the optimal berthing location based on ship size, cargo type, and arrival time; optimizing loading and unloading processes, coordinating the operating routes of quay cranes, yard cranes, and trucks to reduce equipment idle time and waiting time; and dynamically managing yard space, allocating storage areas based on container import / export status, container type, destination port, and other attributes to avoid excessive container turnover.

[0003] In related technologies, dock scheduling is mainly implemented through mathematical modeling and agent systems. Mathematical modeling methods are based on linear programming or integer programming to establish objective functions such as minimizing total operation time, and then combined with solvers such as CPLEX to generate static scheduling schemes. Agent systems mainly use metaheuristic methods such as genetic algorithms and ant colony algorithms to generate approximate optimal solutions.

[0004] Regarding the aforementioned technologies, the terminal scheduling method based on mathematical modeling suffers from high model complexity, making it difficult to handle dynamic emergencies such as ship delays and equipment failures, and also exhibits poor real-time performance. The terminal scheduling method based on agent systems relies on a fixed rule base, making it unsuitable for complex scenarios involving multiple objective trade-offs, and also suffers from low efficiency in cross-device collaboration. Summary of the Invention

[0005] To address the issues of poor adaptability to dynamic environments and low efficiency in multi-objective coordination in terminal scheduling methods, this application provides an intelligent management method and system for container terminals based on spatiotemporal coordinate mapping.

[0006] Firstly, this application provides a method for intelligent management of container terminals based on spatiotemporal coordinate mapping, employing the following technical solution: A smart management method for container terminals based on spatiotemporal coordinate mapping includes: Acquire multi-source data related to ship operations, and fuse ship arrival time, occupancy length, and container distribution data through a preset spatiotemporal fusion algorithm; A control and management coordinate system is established based on the fused and associated multi-source data, with the physical location of the wharf shoreline as the X-axis, which extends from the 0-meter reference point to the end of the wharf, and the planned arrival time of the ships as the Y-axis. In the control and management coordinate system, a ship distribution heat map is dynamically generated, and each ship is mapped as a rectangular color block. The length of the color block along the X-axis represents the ship's occupancy length, and the width of the color block along the Y-axis represents the planned berthing duration. Container operation characteristic data is loaded in real time and embedded inside the color blocks. The container operation characteristic data includes the current total loading volume, the ratio of large to small standard containers, and the partition identification of imported and exported containers. The spatial distribution characteristics of containers are analyzed from the zoning identifiers. The spatial distribution characteristics of containers include the concentrated coordinate set of imported containers in ship holds and the distribution density matrix of exported containers in terminal yards. Based on preset crane efficiency parameters and container spatial distribution characteristics, a crane deployment plan is generated and then distributed to the terminal operation terminal.

[0007] By adopting the above technical solutions, spatiotemporal coordinate mapping and heat map visualization significantly improve the dynamic adaptability and multi-objective collaborative efficiency of container terminal scheduling. The spatiotemporal fusion algorithm integrates multi-source data to establish an intuitive control and management coordinate system, dynamically mapping vessel positions to arrival times, facilitating real-time monitoring and rapid response to emergencies such as vessel delays. Embedded container operation feature analysis analyzes spatial distribution patterns, optimizing crane paths and yard allocation, reducing empty runs and container handling operations, thereby lowering operating costs. Simultaneously, this method enhances cross-equipment collaboration capabilities, using visualized heat maps to assist decision-making, achieving efficient resource matching, shortening vessel port time, and comprehensively improving terminal throughput efficiency and overall operational flexibility.

[0008] Optionally, methods for generating crane deployment plans include: Determine the optimization objectives for crane deployment, including minimizing crane travel distance, balancing the workload of each crane, and reducing container loading and unloading delays; A weighted distance calculation model is constructed, with the current position of the crane as the origin, to calculate the weighted sum of the absolute value of the X-axis coordinate difference and the Y-axis time difference of the berth where the target container is located. The weight parameters include the crane operation efficiency coefficient and the container urgency coefficient. Based on the spatial distribution characteristics of containers, the coordinate set of import container hold locations and the density matrix of export container yards are extracted to establish a container priority queue, in which import containers have higher priority than export containers, and the priority of containers in high-density yard areas is increased by a preset ratio. The weighted distance model is iteratively optimized by improving the genetic algorithm, setting a preset range for crossover probability and mutation probability, and retaining a preset proportion of candidate schemes before the fitness value. The candidate solutions were verified by generating a preset number of random scenarios through Monte Carlo simulation to evaluate the job delay fluctuations within the confidence interval. When the fitness value fluctuates less than a preset threshold after a preset number of consecutive iterations, the optimization is terminated and the optimal crane deployment path is output.

[0009] By adopting the above technical solutions, a weighted distance model integrating spatiotemporal factors and operational priorities is constructed to improve the accuracy and practicality of the scheduling scheme. By integrating container spatial distribution characteristics with multi-objective optimization, the system minimizes crane travel distance, balances operational load, and reduces loading and unloading delays. An improved combination of genetic algorithms and Monte Carlo simulations ensures the robustness and stability of the scheme in dynamic environments, effectively addressing unforeseen circumstances such as equipment failures and ship delays. The final output scheme significantly reduces crane idle time and waiting time, improves loading and unloading efficiency, and enhances the overall coordination and reliability of terminal operations.

[0010] Optionally, methods for optimizing the crane deployment scheme include: Using the position of the ship's color block in the heat map coordinate system as a spatial constraint, the real-time throughput rate deviation of each crane is extracted based on the operation monitoring chain. The association weight of the ship color blocks is calculated by combining the X-axis position coordinates of the ship color blocks in the heat map. The association weight of the color blocks is calculated by the center point coordinate of the ship color block on the X-axis. The ships corresponding to the first N color blocks in descending order of association weight are selected as the initial solution set for optimization. A multi-objective optimization function is constructed, which includes a travel distance reduction term, a load balancing term, and a rate deviation compensation term. The travel distance reduction term is determined by calculating the total travel distance of the cranes, the load balancing term is determined by calculating the total difference between the load of each crane and the maximum load, and the rate deviation compensation term is determined by calculating the sum of squares of the real-time throughput rate deviation. Each sub-item is configured with a preset weight coefficient. A hybrid optimization process combining tabu search and simulated annealing is performed. In the tabu search stage, a candidate solution set is generated in the neighborhood of the current solution. Based on the tabu list, recently occurring solutions are excluded and the objective function is selected to improve the optimal non-tabu solution. In the simulated annealing stage, some degraded solutions are allowed according to the probability acceptance criterion based on the annealing temperature decay strategy. The tabu search and simulated annealing phases are executed alternately. After each tabu search generates a candidate solution, the number of consecutive unimproved solutions is checked. If the preset number is reached, the simulated annealing phase is triggered to update and optimize the state parameters, including the current optimal solution, the tabu list, and the annealing temperature. The improvement rate of the objective function is calculated. The improvement rate is determined by dividing the absolute value of the difference between the current iteration objective value and the previous iteration objective value by the previous objective value. The iteration is terminated when the improvement rate is lower than the preset convergence threshold. The final solution set is output from the Pareto optimal solution set maintained by the hybrid optimization process. Retrieve cases from the historical operation database whose environmental parameter similarity is not lower than a preset threshold, and use the final solution set as the verification object to verify the feasibility of the final solution set in terms of throughput achievement rate and equipment utilization fluctuation index. The verified optimal solution is mapped to the heatmap coordinate system, and crane coordinate update instructions are generated based on the crane position and container distribution characteristics in the solution set.

[0011] By adopting the above technical solutions and combining real-time monitoring data with heatmap spatial constraints, a highly efficient multi-objective hybrid optimization framework is constructed. Utilizing the collaborative mechanism of tabu search and simulated annealing, global exploration and local search capabilities are balanced, premature convergence is avoided, and the optimization quality and solution efficiency are improved. Historical case verification ensures the high feasibility and robustness of the output solution set in actual operating environments. The final dynamic scheduling commands can accurately match the real-time operating conditions of the terminal, reduce crane travel distance, balance equipment load, and stabilize operating rates, thereby comprehensively improving loading and unloading efficiency, resource utilization, and the system's adaptability to unexpected situations.

[0012] Optionally, after verifying the feasibility of the final solution set in terms of throughput achievement rate and equipment utilization fluctuation metrics, the following steps are also included: Extract the top K extreme scenario cases with the highest similarity from the historical operation database. Extreme scenarios include at least one of typhoon weather, crane failure, and ship congestion at port. Construct a digital twin emergency simulation model and inject the current heat map coordinate system parameters, real-time crane status data, and extreme scenario characteristic parameters into the emergency simulation model; In the emergency simulation model, the scheduling scheme corresponding to the final solution set is executed, and key performance indicators are extracted. Key performance indicators include maximum operation delay time, number of ships congested at port, and equipment idle rate. When the maximum job delay exceeds the preset tolerance threshold, the strategy fusion and optimization module is activated to extract a set of successful coping strategies from extreme scenario cases. Based on the tabu search algorithm, the coping strategies are combined and optimized with the current solution set to generate an anti-interference scheduling scheme. Map the anti-interference scheduling scheme to the heat map coordinate system and update the crane coordinate update command.

[0013] By adopting the above technical solutions, digital twin emergency simulations and strategy fusion optimization under extreme scenarios are introduced to enhance the robustness and anti-interference capabilities of the scheduling system. A high-fidelity simulation environment is constructed using historical extreme cases to accurately predict the impact of emergencies such as typhoons, malfunctions, and port congestion on the scheduling plan, and to quantitatively evaluate key performance indicators. When operational delays are predicted to exceed limits, the strategy fusion module is triggered to generate anti-interference solutions incorporating historical successful experiences, ensuring that the system can maintain efficient and reliable operation under extreme pressure. This reduces the negative impact of emergencies, avoids operational interruptions and sudden drops in efficiency, and improves the continuity and stability of terminal operations through forward-looking dynamic command updates, achieving a leap from passive response to proactive defense.

[0014] Optionally, methods for acquiring multi-source data related to ship operations and fusing ship arrival time, berthing length, and container distribution data using a preset spatiotemporal fusion algorithm include: Outlier removal and missing value filling are performed on the acquired multi-source data related to ship operations; The planned, estimated, and actual ship arrival times are aligned chronologically, and the time confidence level is calculated by weighted average, with the weight increasing exponentially as the data update time approaches. The design and measured values ​​of the ship's occupancy length, as well as the container distribution coordinates of the hold and yard, are transformed to a unified spatial reference. This unified spatial reference is made consistent with the X-axis reference of the control and management coordinate system, and the systematic errors of different measurement systems are eliminated through the coordinate mapping matrix. A spatiotemporal correlation matrix is ​​constructed based on the confidence level in the time dimension and the spatial coordinate deviation. The covariance between the confidence level in the time dimension and the spatial coordinate deviation is calculated. Data with a correlation level higher than a preset threshold are subjected to weighted fusion, and the weights are taken as the eigenvalues ​​of the inverse matrix of the covariance. Output the fused standardized spatiotemporal data frame, which includes the ship's unique identifier, the fused arrival timestamp, the ratio of the normalized occupancy length to the total length of the wharf shoreline, and the spatiotemporal index code of the container distribution. The spatiotemporal index code is generated by combining the X-axis physical segment number and the Y-axis planned time window number. The X-axis division rule is based on the preset physical segment length, and the Y-axis division rule is based on the preset time window length. The normalized space length is used as the calculation basis for the X-axis extension of the heatmap color block; The merged arrival timestamp is used as the basis for calculating the Y-axis extension of the heatmap color block. When dynamic optimization is triggered, the fusion weights of the spatiotemporal correlation matrix are updated based on real-time monitoring data.

[0015] By adopting the above technical solutions, systematic data cleaning, spatiotemporal alignment, and dynamic weighted fusion improve the accuracy and reliability of ship operation data. It eliminates the temporal and spatial biases of multi-source heterogeneous data, constructing high-confidence standardized data frames to provide a unified and accurate data foundation for subsequent scheduling decisions. Dynamically adjusting the fusion weights through a spatiotemporal correlation matrix enables the system to adaptively respond to real-time changes, enhancing the adaptability and robustness of data processing. The resulting spatiotemporal index encoding and normalized parameters provide an efficient and consistent computational benchmark for heatmap visualization and dynamic optimization, improving the entire terminal management system's perception and decision-making capabilities in complex dynamic environments.

[0016] Secondly, this application provides a container terminal intelligent management system based on spatiotemporal coordinate mapping, which adopts the following technical solution: A container terminal intelligent management system based on spatiotemporal coordinate mapping includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, it implements the container terminal intelligent management method based on spatiotemporal coordinate mapping as described in the first aspect.

[0017] By adopting the above technical solutions, the system can stably operate the intelligent management method for container terminals based on spatiotemporal coordinate mapping, implement intelligent management technology, ensure efficient collaboration in terminal operation data processing, crane scheduling, emergency response and other aspects, and effectively improve the intelligent management level of container terminals. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an intelligent management method for container terminals based on spatiotemporal coordinate mapping, according to an embodiment of this application.

[0019] Figure 2 This is a flowchart illustrating the method for generating a crane deployment scheme in an embodiment of this application.

[0020] Figure 3 This is a flowchart illustrating the method for optimizing the crane deployment scheme in this embodiment of the application.

[0021] Figure 4 This is a flowchart illustrating the process after verifying the feasibility of the final solution set in terms of throughput achievement rate and equipment utilization fluctuation indicators in this application embodiment.

[0022] Figure 5 This is a flowchart illustrating a method for obtaining multi-source data related to ship operations and fusing ship arrival time, occupancy length, and container distribution data using a preset spatiotemporal fusion algorithm, as described in this application embodiment. Detailed Implementation

[0023] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-5 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0024] This application discloses an intelligent management method for container terminals based on spatiotemporal coordinate mapping. (Refer to...) Figure 1 Intelligent management methods for container terminals based on spatiotemporal coordinate mapping include: Step S100: Obtain multi-source data related to ship operations, and fuse ship arrival time, occupancy length, and container distribution data using a preset spatiotemporal fusion algorithm.

[0025] Multi-source data related to vessel operations: A dynamic data set covering the entire process of container terminal operations, including but not limited to vessel dynamic data, container distribution data, and environmental auxiliary data. Vessel dynamic data includes planned arrival time, actual arrival timestamp, designed berthing length, and measured berthing length; container distribution data includes import / export container identifiers, large / small TEU quantity ratio, vessel hold container coordinate set, and yard container density matrix; environmental auxiliary data includes the influence coefficient of tides on berthing duration and real-time occupancy status of quay cranes.

[0026] Spatiotemporal fusion algorithm: A multi-source data integration method based on spatiotemporal correlation. It eliminates system-level errors and generates standardized spatiotemporal data frames through coordinate system normalization, confidence weighting and covariance analysis.

[0027] The data fusion process includes data acquisition and preprocessing, spatiotemporal benchmark alignment, spatiotemporal correlation modeling, standardized data output, and dynamic optimization and linkage.

[0028] Step S200: Establish a control and management coordinate system based on the fused associated multi-source data, with the physical location of the wharf shoreline as the X-axis, extending from the 0-meter reference point to the end of the wharf, and the planned arrival time of the ship as the Y-axis.

[0029] Control and Management Coordinate System: A two-dimensional spatiotemporal coordinate system specifically designed for container terminal operations. Through orthogonal mapping between physical space and time dimensions, it transforms ship operation data into quantifiable and analyzable geometric relationships, including an X-axis constructed in the spatial dimension and a Y-axis constructed in the time dimension.

[0030] The X-axis is based on the actual physical location of the wharf shoreline, with the 0-meter reference point usually set at the beginning of wharf operations (port entrance), extending forward to the end of the wharf (end of loading and unloading area); the Y-axis is based on the planned arrival time of the ship, using a standard timestamp or relative time offset (hours).

[0031] The general process is as follows: 1. Benchmark point calibration: Install physical benchmark piles (RFID tag pile number CT-001) at the dock site as the 0-meter benchmark point of the coordinate system; obtain the terminal coordinates of the dock shoreline by measuring with a total station, calculate the total shoreline length, such as L=2500 meters; convert the normalized occupancy length output in step S100 into the actual physical length, X actual = normalized occupancy length × L.

[0032] 2. Spatiotemporal axis construction: X-axis generation: Using the 0-meter reference point as the origin, establish a positive direction along the extension direction of the wharf shoreline. According to the preset segmentation rules, divide the physical interval into segments of 50 meters each. The segment numbering format is X + serial number, for example, 0-50 meters → X001, 50-100 meters → X002; Y-axis generation: Using 00:00:00 of the day as the time origin (T0), convert the ship's planned arrival timestamp into minute offsets. Y offset = (planned arrival hour × 60) + planned arrival minute. Divide the time period according to the preset time window (1 hour). The time window numbering format is Y + serial number, for example, 08:00-09:00 → Y008.

[0033] 3. Ship operation data mapping: The standardized spatiotemporal data frame is parsed into coordinate parameters to generate a spatial-temporal positioning frame of the ship in the coordinate system. The X-axis range is Xstart, Xstart+Xlength; the Y-axis range is Ystart, Ystart+planned berthing duration.

[0034] 4. Dynamic maintenance of the coordinate system: When a new ship berthing plan is added, the Y-axis time window distribution in the coordinate system is updated in real time. If the wharf shoreline is modified (new berths are added), the X-axis extension parameters are recalibrated through the reference pile position. The coordinate system parameters are stored as a spatiotemporal topology graph structure, which supports fast spatial query (finding all ships in the X-axis 120-180 meter interval).

[0035] Step S300: Dynamically generate a ship distribution heat map in the control and management coordinate system, mapping each ship as a rectangular color block. The length of the color block along the X-axis represents the ship's occupancy length, and the width of the color block along the Y-axis represents the planned berthing duration.

[0036] Ship distribution heatmap: A visualization model of ship operations based on a spatiotemporal coordinate system, intuitively displayed through color temperature gradients and geometric parameters. The color of the blocks indicates the urgency of the ship operations; for example, red = high urgency, blue = routine. The length of the blocks is proportional to the actual length occupied by the ship (X-axis physical scale); the width of the blocks is proportional to the planned berthing duration (Y-axis time scale).

[0037] Step S400: Load container operation feature data in real time and embed it inside the color block. The container operation feature data includes the current total loading volume, the ratio of large to small standard containers, and the partition identification of imported and exported containers.

[0038] Container operation characteristic data: A set of core dynamic parameters for ship loading and unloading operations, including: Current total load: The total number of standard containers currently loaded on the ship, with real-time updates on loading and unloading progress; Large / small standard container ratio: The ratio of 40-foot containers to 20-foot containers. Import / export container zone identification: Import containers are identified using IM + destination area code, such as IM-NA indicating North America; export containers are identified using EX + origin area code, such as EX-AS indicating Asia.

[0039] Embedded within color blocks: Feature data is converted into visual elements and integrated into heatmap color blocks to achieve multi-dimensional data presentation in a single image.

[0040] Step S500: Analyze the spatial distribution characteristics of containers from the zoning identifiers. The spatial distribution characteristics of containers include the concentrated coordinate set of imported containers in the ship's hold and the distribution density matrix of exported containers in the terminal yard.

[0041] Container spatial distribution characteristics: A quantitative model of container physical distribution extracted based on zone identifiers (IM / EX) includes a concentrated coordinate set of import containers and a distribution density matrix of export containers. The concentrated coordinate set of import containers is a set of three-dimensional coordinates of all import containers (IM identifiers) within the ship's hold; the coordinate format is (cargo hold bay / row / layer); the concentration threshold is that a zone with ≥5 containers per square meter is marked as a hot zone.

[0042] Export container distribution density matrix: a planar distribution density grid of export containers (EX markings) within the terminal yard. Matrix dimension: divided by yard blocks (e.g., 50m×50m grid). Density value: number of containers per unit area (TEU / m²).

[0043] Step S600: Generate a crane deployment plan based on preset crane efficiency parameters and container spatial distribution characteristics, and send the crane deployment plan to the terminal operation terminal.

[0044] Crane efficiency parameters include the single-machine operating efficiency benchmark, efficiency decay coefficient, movement conversion time cost, and minimum safe distance constraint for multi-machine collaborative operation. These parameters together constitute the basic input of the resource optimization model.

[0045] Crane deployment scheme: Cranes include yard cranes and quay cranes. Based on a multi-objective optimization resource allocation strategy, it includes a quay crane allocation table (crane number → ship hot zone coordinates), a yard crane scheduling matrix (yard block → number of yard cranes), movement path planning (shortest crane transfer path sequence), and time window constraints (operation start / end timestamp).

[0046] The general process is as follows: First, based on the concentrated coordinate set of imported containers in the ship's hold, quay cranes are preferentially allocated to container-dense, high-density container bays, forming a quay crane allocation table. Simultaneously, combined with the distribution density matrix of export containers in the yard, yard cranes are dynamically scheduled to yard blocks with high export container density, generating a yard crane scheduling matrix to minimize empty travel distances and balance equipment load. The system plans specific movement path sequences and operation time windows for each crane, forming a structured, time-sequential crane deployment plan, including quay crane and yard crane allocation information, path planning, and a global operation timeline. Finally, the plan is transmitted in real-time via the terminal's IoT communication network in a standardized data format to the onboard control terminals of quay cranes and yard cranes, the central dispatch system, and the mobile terminals of management personnel to guide actual operations.

[0047] Reference Figure 2 The methods for generating crane deployment plans include: Step S601: Determine the optimization objectives for crane deployment, including minimizing crane travel distance, balancing the workload of each crane, and reducing container loading and unloading delays.

[0048] Minimize crane travel distance: The minimum total path length for crane movement across berths, used to reduce energy consumption and time loss during idle runs. Historical average travel distance is used as a baseline; a distance penalty mechanism is triggered when a single displacement exceeds 100 meters. Balance crane workload: Strictly control the difference in standard container workload within 15%, quantified and evaluated using a balance formula (target value ≥ 0.85), avoiding localized overload or resource idleness. Reduce container loading and unloading delay rate: The delay probability in normal scenarios should be less than 5%, relaxed to 12% in extreme weather such as typhoons, with emergency containers having double the delay weight in the calculation.

[0049] An adaptive weight allocation mechanism is established, with balanced weights for the three objectives (0.4:0.3:0.3) during routine operations; for ship congestion scenarios, delay control is prioritized (weight increased to 0.5); and during equipment failure recovery periods, mobility efficiency is prioritized (weight 0.5). The three-dimensional objective system incorporates physical displacement, equipment load, and time constraints into a unified optimization framework, providing multi-dimensional decision guidance for the weighted distance calculation model and addressing the systemic imbalances caused by traditional single-objective optimization.

[0050] Step S602: Construct a weighted distance calculation model. With the current position of the crane as the origin, calculate the weighted sum of the absolute value of the X-axis coordinate difference and the Y-axis time difference of the berth where the target container is located. The weight parameters include the crane operation efficiency coefficient and the container urgency coefficient.

[0051] Weighted distance calculation model: A decision-making tool that integrates spatial displacement and time cost to quantify the comprehensive cost of dispatching a crane to a target container. Using the real-time position of the crane as the coordinate origin, the X-axis physical distance deviation (m) and Y-axis planned time window deviation (min) of the target container's berth in the control and management coordinate system are calculated respectively. Weighting parameters are then used to normalize these two types of heterogeneous data into comparable distance evaluation values.

[0052] Crane operation efficiency coefficient: The crane operation efficiency coefficient is the weight of the X-axis distance deviation, representing the inherent performance and real-time status of the equipment. It is jointly determined by the rated loading and unloading efficiency (e.g., 30 boxes / hour) and the actual operational stability. High-performance equipment is given priority for long-distance tasks. When the equipment is well maintained and the historical failure rate is ≤2%, a baseline value of 0.7 is taken; if efficiency degradation or repair warnings occur recently, the coefficient is reduced to 0.4 to lower the scheduling priority.

[0053] The container urgency coefficient is derived from the entropy value calculation result ξ, reflecting the urgency of the operation. For regular containers, ξ∈[0.3,0.7], with a weight of 1.0; for high-urgency containers (ξ greater than 0.8, such as cold chain medicine containers), the weight is increased to 1.2, significantly amplifying the time deviation cost. This coefficient serves as the weight for the Y-axis time window deviation, ensuring that tasks with high timeliness requirements receive a rapid response.

[0054] The construction method of the weighted distance calculation model is as follows: First, obtain the current position coordinates (x0, y0) of the crane and the target berth coordinates (x1, y1) from the terminal digital twin system. Calculate the absolute distance difference on the X-axis |Δx|=|x1-x0|, and calculate the planned time difference on the Y-axis |Δt|=|planned operation start time-current system time|. The weighted distance Dw is equal to |Δx|×crane efficiency coefficient+|Δt|×container urgency coefficient.

[0055] For example, when the crane efficiency coefficient is 0.7 and the urgency coefficient is 1.2, the weighted distance Dw for handling a task 150 meters away that starts in 20 minutes is 150 × 0.7 + 20 × 1.2 = 126 distance units. This weighted distance calculation model breaks through the limitations of traditional geometric distances, incorporating equipment performance losses and time opportunity costs into scheduling decisions, thus providing a fitness evaluation benchmark for genetic algorithms.

[0056] Step S603: Based on the spatial distribution characteristics of containers, extract the coordinate set of import container bays and the density matrix of export container yards, and establish a container priority queue, in which import containers have higher priority than export containers, and the priority of containers in high-density yard areas is increased by a preset ratio.

[0057] Container Spatial Distribution Characteristics: This section outlines the physical location patterns of containers, including two key data points: hotspots for imported containers on ships and the density distribution of exported containers in the yard. Imported Container Hold Coordinate Set: This is the three-dimensional clustering result of imported containers on the ship's deck (e.g., a hotspot centered on Bay22-Row12-Tier1 with a radius of 5 meters). The coordinates are quantified using the ship's local coordinate system (origin at the port bow). Exported Container Yard Density Matrix: This is the container density value (range 0~1.0) per 50m×50m grid unit within the terminal yard. High-density areas (≥0.8) indicate areas prone to container congestion. Container Priority: This is a scheduling weight value generated based on operational rules and spatial characteristics. The baseline rule assigns an inherent priority to imported containers (base value 1.2) that is higher than that to exported containers (base value 1.0). Furthermore, containers in high-density yard areas receive a proportional increase in priority based on their density (each 0.1 density increase raises the priority by 0.15).

[0058] The general process is as follows: First, the coordinate set of the hot zone for imported containers and the density matrix of exported containers are loaded. All imported container coordinate clusters are placed at the top and arranged in ascending order of hot zone radius (smaller hot zones are given priority). Export containers are inserted into the queue in descending order of density matrix values. For yard grid cells with a density ≥ 0.8, the priority of containers within them is increased. For example, the priority of an exported container in a grid with a density of 0.92 is 1.0 + (0.92 - 0.8) / 0.1 × 0.15 = 1.18. The grid coordinates are then associated with specific container operation instructions through a spatial mapping engine. The queue continuously receives real-time updates of the urgency coefficient ξ. When the ξ value of a container in a hot zone suddenly increases by more than 0.3, the priority of containers in that cluster is collectively increased by 30% and moved forward three queue ranks. The resulting dynamic priority queue, such as ["Bay15 hot zone (priority 1.8)", "A-02 high-density container (priority 1.18)", "regular export container (priority 1.0)"], is directly fed into the genetic algorithm to drive crane resources to prioritize covering high-value units. For example, 80% of the quay cranes are scheduled to the high-priority import container hot zone (Bay30), and only 20% of the yard cranes handle the high-priority export container dense zone.

[0059] Step S604: Iteratively optimize the weighted distance model, set the preset range of crossover probability and mutation probability, and retain the candidate schemes with a preset proportion before the fitness value.

[0060] The weighted distance model is iteratively optimized using an improved genetic algorithm: the initial solution set output by the weighted distance model is used as the parent population, and each solution represents a complete crane deployment scheme, including quay crane location, yard crane allocation, and path sequence. Crossover probability: dynamically maintained between 0.65 and 0.8; if fitness does not improve for three consecutive generations, the probability increases to 0.8 to enhance exploration. Mutation probability: set between 0.05 and 0.15; the probability doubles when encountering a local optimum to break through stagnation. Fitness evaluation directly uses the weighted distance value; the smaller the value, the lower the overall cost of the scheme. The dynamic adjustment formula for crossover probability is: Pc = 0.65 + 0.15 × (1 − e) −0.3t ), where t is the number of consecutive unimproved algebras, takes effect when t≥3, and the mutation probability doubles when encountering a local optimum.

[0061] Optimization Process: An initial population of 50 schemes is randomly generated, and 30 high-quality parent schemes are selected. A sequential crossover operation is performed, randomly selecting the path node sequences of two parent schemes (the movement path of the QC12 quay crane), swapping their X-axis coordinate segments (100-200 meter range), and repairing conflict nodes. During the mutation phase, key parameters of the schemes are fine-tuned with preset probabilities, such as randomly switching the target grid of the quay crane (Block B-03 to B-07) or adjusting the operation time window by ±10%. In each iteration, the top 15% of the elite schemes in terms of fitness are directly promoted to the next generation, while the remaining schemes are refreshed through crossover and mutation. Optimization terminates after 100 iterations or when the fitness fluctuation rate is below 0.5%, outputting the three candidate schemes with the smallest weighted distance on the Pareto front, providing input for Monte Carlo validation.

[0062] Step S605: Generate a preset number of random scenarios to verify the job delay fluctuation of the candidate solution within the confidence interval.

[0063] Monte Carlo simulations are used to generate a preset number of random scenarios: the robustness of the evaluation scheme is assessed by injecting random scenarios, using the three candidate crane deployment schemes output from the previous steps as the verification objects, and 1000 random scenario simulations are preset (each simulation represents a possible operating environment state). Random scenarios are constructed based on the terminal's historical operation database, covering sudden equipment failures (quay crane failure rate set at the historical average of 5%), weather deterioration (10% probability of wind speed > level 14), delayed ship arrivals (15% probability of berth occupancy exceeding time limit), and emergency container replenishment (8% probability of a sudden 50% increase in the emergency coefficient ξ). Each type of disturbance factor generates random intensity values ​​according to a normal distribution (e.g., wind speed average of 25 m / s, standard deviation of 4).

[0064] The simulation process quantifies operational delay fluctuations: Candidate solutions are executed in each random scenario, and the deviation (in minutes) between the actual operation time and the planned value for key nodes is recorded. Three types of delay indicators are monitored: path delay for the crane to move to the target hot zone (e.g., an additional 10 minutes due to truck congestion), average loading and unloading time per container (e.g., efficiency reduction of 30% due to malfunction), and overall vessel departure delay (e.g., delays exceeding 3 hours during typhoons). After 1000 iterations, the delay fluctuation range of each solution within a 90% confidence interval is statistically analyzed. For example, solution A controls vessel departure delays within [-18 minutes, +42 minutes], and the probability of exceeding the delay limit is calculated (e.g., solution B has a delay >60 minutes in 7.3% of scenarios). Finally, the solution with the smallest confidence interval width and a probability of exceeding the limit <5% is selected as the optimized output. If all solutions exceed the limit, a tabu search with weight 3 is triggered for re-optimization. This verification process ensures that the scheduling solution remains stable in the real operational environment, reducing the expected delay rate during typhoons from 22% to 11%.

[0065] Step S606: When the fitness value fluctuation of a preset number of consecutive iterations is less than a preset threshold, the optimization is terminated and the optimal crane deployment path is output.

[0066] During the iterative improvement of the genetic algorithm, the system continuously monitors the fluctuations in the population fitness value. When the standard deviation of the optimal fitness value is less than a preset threshold of 0.5% within 10 consecutive generations (a preset number of times), i.e., the fluctuation range is ≤ ±0.63 distance units, the algorithm is determined to have converged to a stable solution domain. At this point, the optimization process is immediately terminated to avoid unnecessary computational waste. The threshold is derived from statistical analysis of historical optimization cases from 10 large wharves, effectively distinguishing between algorithm stagnation and true convergence.

[0067] From the final generation population, the three schemes with the smallest weighted distance on the Pareto front are extracted, and a second screening is performed based on the Monte Carlo verification results from step S605. The scheme that simultaneously meets the following conditions is selected as the optimal deployment path: operation delay fluctuation ≤ ±35 minutes in the 90% confidence interval; movement distance reduced by ≥15% compared to the initial scheme; load balancing improvement rate >10%.

[0068] The optimal solution is compiled into an executable instruction set, and the core path parameters include: the target coordinates of the quay crane (e.g., QC08 is located to X=175.3), the density response coefficient of the yard crane (e.g., BlockA-02 coefficient 0.9), the sequence of critical path nodes (e.g., "POS-A1→POS-B3→HS-229"), and the time window tolerance parameter (loading and unloading time flexibility ±8%).

[0069] Reference Figure 3 Methods for dynamically optimizing the current crane deployment plan include: Step S301: Using the position of the ship's color block in the heat map coordinate system as a spatial constraint, extract the real-time throughput rate deviation value of each crane based on the operation monitoring chain.

[0070] Based on the ship distribution heatmap obtained in step S300, the geometric attributes of the ship color blocks are transformed into spatial constraints. The coordinates of the lower left corner (Xmin, Ymin) and the upper right corner (Xmax, Ymax) of each color block constitute a rectangular optimization boundary. The length of the color block (X-axis) maps to the physical occupancy range of the ship, and the width of the color block (Y-axis) is associated with the planned berthing time window. Within this spatial constraint, the system extracts the real-time throughput rate deviation value ΔV of each crane from the operation monitoring chain in step S700.

[0071] The deviation extraction process incorporates a dynamic weighting mechanism. For cranes at different locations within the heatmap, deviation correction is applied based on the urgency of their associated color blocks. When the urgency ξ of a color block is greater than 0.8 (red warning zone), the ΔV value is amplified by 1.3 times to strengthen the issue weight. The ΔV of cranes in overlapping color block areas (multiple ships moored together) is calculated based on the overlap ratio. For cranes at the boundary (<10 meters from the edge of the color block), a 20% safety tolerance is added to the ΔV. The final output is a deviation dataset with spatial labels.

[0072] Step S302: Calculate the association weight of the color blocks by combining the X-axis position coordinates of the ship color blocks in the heat map. The association weight of the color blocks is calculated by the center point coordinate of the ship color blocks on the X-axis. Select the ships corresponding to the first N color blocks in descending order of association weight as the initial solution set for optimization.

[0073] The color block association weights are dynamically calculated based on spatial location and operational urgency, using the X-axis center point coordinates of the ship color block in the heat map coordinate system as a reference, and assigning differentiated weights in combination with dock operation rules.

[0074] The specific calculation rules are as follows: Location weight: The closer the color block is to the dock entrance (X=0), the higher the weight (base weight = 1 + 0.003 × |Xc-0|); Urgency correction: The urgency coefficient ξ (red ξ > 0.8) of the color mapping of the color block is proportionally amplified (final weight = location weight × ξ); Size compensation: Large ships with a color block length > 200 meters are given an additional 20% weight premium.

[0075] Initial solution set generation process: First, all ship color blocks within the heatmap are traversed, and the association weight value of each color block is calculated. For example, the weight of a large red color block near the edge is (1 + 0.003 × 50) × 0.9 × 1.2 ≈ 1.48. After sorting by weight in descending order, the system dynamically determines the number of blocks to be extracted, N. In normal scenarios, the top 5 are selected (N=5), and when the total number of color blocks exceeds 20, it is expanded to the top 8 (N=8). The real-time deployment schemes (including crane coordinates and path sequences) of the ships corresponding to the selected high-weight color blocks are directly extracted as the optimized initial solution to ensure that resources are preferentially tilted towards high-urgency large ships in the dock entrance area. For example, the set of current crane configuration schemes for the top 5 ships by weight.

[0076] Step S303: Construct a multi-objective optimization function, which includes a travel distance reduction term, a load balancing term, and a rate deviation compensation term. The travel distance reduction term is determined by calculating the total travel distance of the cranes, the load balancing term is determined by calculating the sum of the differences between the load of each crane and the maximum load, and the rate deviation compensation term is determined by calculating the sum of the squares of the real-time throughput rate deviations. Each sub-item is configured with a preset weight coefficient.

[0077] **Travel Distance Reduction Item:** Calculates the total displacement path of all cranes (in meters), with a focus on penalizing long-distance travel exceeding 100 meters. For example, when the scheme includes QC08 moving 150 meters and QC12 moving 80 meters, this item has a value of 230. **Load Balancing Item:** Based on the sum of the absolute differences between the allocated container capacity (standard container) and the maximum load for each crane. For example, when three cranes have a load of [120, 90, 110] containers, and the maximum load is 120 containers, the sum of the differences is |120-120|+|90-120|+|110-120|=40. **Deviation Compensation Item:** Takes the sum of the squares of the real-time throughput rate deviation value ΔV (enhancing the impact of high deviations). For example, if the ΔV values ​​for two cranes are -18% and -5% respectively, the item value is (-18)²+(-5)²=349.

[0078] Weighting coefficients: For standard scenarios, the weights are: movement distance (0.5), load balancing (0.3), and rate deviation (0.2). For ship congestion scenarios, these are adjusted to (0.3, 0.2, 0.5). For equipment failure scenarios, they are switched to (0.6, 0.3, 0.1). The final function expression is: Total cost = (0.5 × total movement distance) + (0.3 × total load difference) + (0.2 × Σ(ΔV²)). This breaks through the limitations of a single objective and achieves three-dimensional objective balance in forbidden annealing hybrid optimization. For example, in a certain scheme with a total movement distance of 500 meters, a load difference of 300 containers, and a ΔV² sum of squares of 400, the cost = 0.5 × 500 + 0.3 × 300 + 0.2 × 400 = 250 + 90 + 80 = 420 cost units.

[0079] Step S304: Perform a hybrid optimization process that combines tabu search and simulated annealing. In the tabu search stage, a candidate solution set is generated in the neighborhood of the current solution. Based on the tabu list, recently occurring solutions are excluded and the optimal non-tabuation solution is selected to improve the objective function. In the simulated annealing stage, some degraded solutions are allowed according to the probability acceptance criterion based on the annealing temperature decay strategy.

[0080] During the tabu search phase, a neighborhood candidate solution set is generated centered on the current crane deployment scheme. This is achieved by fine-tuning the crane coordinates (within ±50 meters), exchanging the order of path nodes, or resetting the field bridge allocation density. Each generated candidate solution needs to be filtered through a tabu list, which records the features of solutions adopted in the last 10 iterations, such as the movement pattern of QC08 coordinates [120,480]→[125,475]. If the new solution features match any record in the list with a degree greater than 85%, it is directly discarded. From the remaining non-taboo solutions, the scheme with the greatest improvement in the objective function (e.g., reducing the total cost from 420 to 398) is selected. If there is no improvement for 3 consecutive rounds, the simulated annealing phase is forcibly activated.

[0081] A probabilistic degraded solution acceptance mechanism is introduced during the simulated annealing stage. The solution acceptance probability is calculated based on the current temperature Tk, with an initial T0=1000, and decays exponentially according to Tk=T0×0.95^k. When the cost of a new solution increases by ΔE, the degraded solution is accepted with probability P=e^(-ΔE / Tk). During the high-temperature period (Tk>600), there is a 97% probability of escaping the local optimum (e.g., allowing QC12 to move a distance of 15% to test new hot zones). During the low-temperature period (Tk<400), only minor adjustments with a cost increase of <2% are accepted. The two stages are executed alternately until the convergence condition is met.

[0082] Step S305: Alternately execute the tabu search and simulated annealing stages. After each tabu search generates a candidate solution, check the number of consecutive unimproved solutions. If the preset number is reached, trigger the simulated annealing stage and update the optimization state parameters, including the current optimal solution, the tabu list, and the annealing temperature.

[0083] After generating a candidate solution set during the tabu search phase, the system immediately checks the number of consecutive iterations in which the objective function has not been improved. When this number accumulates to a preset threshold of 3 times, for example, if the optimal cost value fluctuates by less than 0.5% for 3 consecutive rounds, the simulated annealing phase is automatically triggered.

[0084] The phase switch involves three key state updates: First, the solution with the lowest cost in the current population is marked as the temporary optimal solution; second, the features of historical solutions stored in the taboo list for more than 5 iterations are cleared, and only the 2 most recent valid solutions are retained; finally, the temperature parameters are updated according to the annealing temperature decay formula.

[0085] Simulated annealing runs for a fixed two iterations at the current temperature Tk, during which a suboptimal solution is allowed to be accepted with probability P = e^(-ΔE / Tk). At the end of annealing, the process switches back to tabu search, injecting new feasible solutions discovered during the annealing phase. This process is repeated until a global termination condition is met (e.g., a total of 50 iterations), ultimately outputting a hybrid optimized crane deployment scheme.

[0086] Step S306: Calculate the improvement rate of the objective function. The improvement rate is determined by dividing the absolute value of the difference between the current iteration objective value and the previous iteration objective value by the previous objective value. When the improvement rate is lower than the preset convergence threshold, the iteration is terminated, and the final solution set is output from the Pareto optimal solution set maintained by the hybrid optimization process.

[0087] The objective function improvement rate serves as a convergence criterion for real-time monitoring of the optimization process. After each iteration, the absolute difference between the current objective function value (total cost) and the previous iteration value is immediately calculated and divided by the historical value to obtain the percentage improvement rate. For example, when the cost decreases from 398 to 397.5, the improvement rate is |397.5-398| / 398≈0.125%. The system presets a convergence threshold of 0.5%. If the improvement rate is below this threshold for three consecutive iterations (e.g., 0.12%, 0.08%, 0.15%), the algorithm is considered to have fully converged and the optimization is terminated.

[0088] The output strategy is based on a fine-tuning of the Pareto optimal solution set: from the Pareto front solution set maintained by the hybrid optimization process, the solution that simultaneously satisfies the following three conditions is selected as the final solution: moving distance ≤ 105% of the historical best value, load balancing degree ≥ 0.82, and Monte Carlo verification delay rate < 8%.

[0089] For example, three schemes with a cost value in the range of 395-402 were selected. After digital twin simulation, the scheme with a total cost of 398.7 was determined to be the optimal solution (22% reduction in travel distance, 0.85 load balancing, and 5.7% delay rate).

[0090] Step S307: Retrieve cases from the historical operation database whose environmental parameter similarity is not lower than a preset threshold, and use the final solution set as the verification object to verify the feasibility of the final solution set in terms of throughput achievement rate and equipment utilization fluctuation index.

[0091] The solution set parameters were injected into the operational scenarios of historical cases. After simulation, the throughput achievement rate was statistically analyzed. It was required that the achievement rate of 90% of cases be >92%, and the lower limit of the typhoon scenario was relaxed to 87%. The equipment utilization fluctuation rate was analyzed. The fluctuation rate of the normal scenario should be <8%, and the tolerance of the high load scenario (utilization rate >90%) was increased to 12%. Spearman correlation analysis was used to confirm the statistical consistency between the solution set performance and the historical best solution (R² > 0.75).

[0092] If two of the three schemes pass the verification (e.g., Scheme B achieves 94% success rate / volatility 7.2%, and Scheme C achieves 91% success rate / volatility 8.5%), then the scheme with the lowest overall cost (Scheme B) is selected as the output; if none of them meet the criteria, then the extreme scenario simulation module is activated to reconstruct the solution set.

[0093] Step S308: Map the verified optimal solution to the heat map coordinate system, and generate a crane coordinate update command based on the crane position and container distribution characteristics in the solution set.

[0094] Crane coordinate update command: An executable command that converts spatial parameters to a heatmap coordinate system based on the final solution set output by the hybrid optimization algorithm, driving the dynamic relocation of crane resources.

[0095] The main components include: a location mapping engine: aligning the crane target coordinates (x0, y0) in the solution set with the XY axis of the heat map, mapping the x0 value to the physical location on the quay (e.g., X=185.3 meters), and converting the y0 value to an absolute timestamp (e.g., Y=08:45); path sequence compilation: generating a discrete coordinate sequence of the crane's movement trajectory based on the movement path nodes in the solution set (e.g., POS-A1→HS-229), with the distance between adjacent nodes ≤10 meters; and operation boundary redrawing: dynamically adjusting the crane's operation range based on the container distribution density matrix, where the new boundary = original boundary - k × (current delay time / remaining docking time), and k is the crane's maneuverability correction factor (k=0.7 for quay cranes and k=0.9 for yard cranes).

[0096] Reference Figure 4 After verifying the feasibility of the final solution set in terms of throughput achievement rate and equipment utilization fluctuation indicators, the following steps are also included: Step S3071: Extract the top K extreme scenario cases with the highest similarity from the historical operation database. Extreme scenarios include at least one of typhoon weather, crane failure, and ship congestion at port.

[0097] Extreme Scenario Case Library: An emergency response knowledge base built on a historical operational database, storing typical abnormal operating conditions that have occurred within the past five years. Each case contains a multi-dimensional feature vector: Ci=[EventType,WindLevel,CraneFaultRate,ShipDelayRatio,ThroughputDropRate], where EventType is the scenario classification code (1=typhoon, 2=crane failure, 3=ship congestion); WindLevel is the typhoon level (based on Beaufort wind force level, ≥12 is an extreme typhoon); CraneFaultRate is the crane failure rate (number of failed cranes / total number of cranes); ShipDelayRatio is the ship delay ratio (number of delayed ships / number of ships in port); and ThroughputDropRate is the throughput decay rate (peak efficiency / current efficiency - 1).

[0098] Step S3072: Construct a digital twin emergency simulation model by injecting the current heat map coordinate system parameters, real-time crane status data, and extreme scenario characteristic parameters into the emergency simulation model.

[0099] The core of the digital twin emergency simulation model lies in establishing a network of temporal dependencies that maps virtual and real operations. The model uses ship loading and unloading events as the basic unit to construct a temporal dependency graph. Each event node contains both physical attributes (such as the type of loading and unloading equipment and the coordinates of the work area) and temporal attributes (planned start time and latest completion time). Spatiotemporal constraints between events are established through directed edges. The weight of the edges is obtained from regression analysis of historical operation data. For example, the minimum interval between the hatch cover opening and the gantry crane positioning event is usually 8-12 minutes. This value is obtained by analyzing the operation data of 8,000 ships in the past three years.

[0100] In the parameter injection phase, the system first extracts spatiotemporal constraint parameters from the heatmap coordinate system. The Y-axis position of the ship's color block corresponds to the time window constraint of the event. For example, if a ship's color block covers the Y-axis time window Y080-Y095 (corresponding to 08:00-09:15), the latest completion time of the hatch cover opening event is set to the end of the Y095 time window. The physical position on the X-axis is mapped to spatial constraints. If the color block covers the X-axis area X150-X200 (corresponding to 150-200 meters of shoreline at the wharf), the relevant loading and unloading events are limited to execution within this physical area.

[0101] Real-time data on the crane's status is accessed to the model via an IoT interface. For example, parameters such as the current coordinates (X=180, Y=220), remaining working time of 3.5 hours, and equipment fault status of the quay crane QC12 are converted into hard constraints on the equipment's available timeline.

[0102] Extreme scenario feature parameters are extracted from historical case databases. For example, the wind speed variation parameters during typhoon weather follow a Weibull distribution (shape parameter 2.1, scale parameter 15), and the crane failure rate follows a Poisson process (mean failure interval 48 hours). These parameters are used to generate random disturbances through the Monte Carlo method.

[0103] The model employs a layered architecture. The bottom layer is the spatiotemporal reference layer, establishing a mapping relationship between the physical space of the wharf and time coordinates. The middle layer is the event network layer, constructing a temporal dependency graph containing 23 types of loading and unloading events. The top layer is the simulation and deduction layer, integrating equipment kinematics models and operational efficiency models. Taking a typical unloading operation as an example, the system first extracts the ship's color block parameters (X-axis X150-X200, Y-axis Y080-Y095) from the heatmap coordinate system. Combined with the real-time position of crane QC12 (X=180, Y=220) and the remaining operation time of 3.5 hours, it automatically derives the gantry crane's positioning path and operation time window. When typhoon scenario parameters (wind speed 28m / s, failure rate increased to 0.03 times / hour) are injected, the model automatically adjusts the event time window constraints; for example, increasing the buffer time for hatch cover opening events from 12 minutes to 25 minutes.

[0104] During parameter injection, the system employs a dynamic verification mechanism to ensure data consistency. For example, when a conflict occurs between the Y-axis time window (Y080-Y095) of the heatmap color block and the available timeline (Y075-Y090) of crane QC12, the system automatically triggers an early warning and generates an alternative solution: either adjust the operation time window to Y090-Y105, or schedule the standby crane QC15 to take over part of the operation tasks, and compare the spatiotemporal constraint parameters with the equipment status parameters in real time to ensure the feasibility of the simulation results.

[0105] The parameter injection for extreme scenarios employs a hierarchical processing strategy. For common abnormal scenarios (such as single crane failure), the system directly calls the preset disturbance parameter template; for complex extreme scenarios (such as typhoon + multiple ships congesting at port), the system generates combined disturbance parameters using the Monte Carlo method. For example, when simulating the "Fireworks" typhoon (historical peak wind speed 42m / s) scenario, the system simultaneously injects wind speed disturbance parameters (normal distribution, μ=35m / s, σ=5m / s), crane failure rate parameters (Poisson distribution, λ=0.05 times / hour), and ship arrival delay parameters (uniform distribution, delay duration 5-15 minutes), forming a multi-dimensional abnormal input.

[0106] Model validation employs a combination of historical playback and stress testing. In a typical typhoon scenario, the system incorporates complete meteorological data from Typhoon Muifa in 2022 (24 hours, maximum wind speed 42 m / s), combined with real-time port operation data (12 ships in port, 32 cranes operating). Through 2000 Monte Carlo iterations, a work chain containing 327 delay events is generated. By comparing the scheduling schemes before and after optimization, it is found that the scheme using a hybrid algorithm of tabu search and simulated annealing can reduce the maximum operation delay time from 98 minutes to 42 minutes, and reduce the number of ships congested at port by 37%. Validated scheduling schemes will be marked as reliable contingency plans and stored in a digital twin knowledge base to provide rapid response strategies for similar scenarios in the future.

[0107] Step S3073: Execute the scheduling scheme corresponding to the final solution set in the emergency simulation model and extract key performance indicators, including maximum operation delay time, number of ships stuck in port, and equipment idle rate.

[0108] The system first maps the crane deployment scheme output by the hybrid optimization algorithm to an operation time-series dependency graph. It then converts the target coordinates (X=185.3, Y=280) of the quay crane QC12 into time window constraints for event nodes, requiring the crane to complete the loading and unloading tasks at the designated hatches within the Y-axis time window Y085-Y095 (corresponding to 08:30-09:35). Simultaneously, it maps the density response coefficient of the yard crane (e.g., Block A-02 coefficient 0.9) to the operation priority weight of the yard area, assigning higher weight areas (e.g., blocks with density ≥ 0.8) higher path planning priority.

[0109] The system collects crane operating status parameters through a real-time data interface, including GPS positioning coordinates, lifting mechanism pressure sensor data, and inverter frequency, and performs dynamic verification by combining these with spatiotemporal constraints in the heatmap coordinate system. For example, when the actual movement trajectory of the quay crane QC08 deviates from the planned path by more than 50 meters, the system automatically triggers a path replanning mechanism, calling a tabu search algorithm to generate a sequence of alternative paths in the adjacent berth area. During this process, crane status data is updated to the model at a frequency of 10Hz, forming a dynamic operation monitoring chain.

[0110] The extraction of key performance indicators is based on the state transition data of event nodes. The maximum operation delay time is calculated by comparing the theoretical latest completion time of the event with the actual start time. For example, the theoretical latest completion time of a hatch cover opening event is the end of the Y090 time window (09:00), while the actual start time is Y092 (09:12). Then the delay time of the event is 12 minutes. The system will continuously track the delay value of all event nodes and retain the maximum value.

[0111] The calculation of the number of ships congested at the port is based on the dwell time of the ship color blocks in the heat map. When the actual dwell time of a ship color block in the Y-axis time window exceeds 150% of the planned berthing time, it is determined to be a congested ship. The system calculates the real-time congestion scale by counting the number of ships covered by the color block.

[0112] The calculation of equipment vacancy rate integrates equipment status monitoring data and spatiotemporal occupancy analysis. Taking yard cranes as an example, the system reads the current location of the equipment through RFID tags and, combined with the thermal distribution of the yard density matrix, calculates the matching degree between the actual operating area and the theoretical operating area of ​​the equipment. When the actual operating area (Block B-05) of yard crane FT03 within the Y-axis time window Y080-Y090 is inconsistent with the theoretically allocated area (Block A-02), the system dynamically adjusts the vacancy rate calculation weight according to the container density ratio of the two areas (Block B-05 density 0.6 / Block A-02 density 0.9), avoiding misjudgments caused by simply calculating based on equipment idle time. The final equipment vacancy rate is calculated using the formula: (Theoretical maximum operating time - Actual effective operating time) / Theoretical maximum operating time × Area density correction coefficient.

[0113] The extraction process of performance indicators is deeply integrated with the dynamic characteristics of the digital twin model. For example, in typhoon scenario simulation, the system adjusts the crane's moving speed parameters in real time by injecting wind speed variation parameters (normal distribution, μ=35m / s, σ=5m / s), so that the calculated maximum delay time can truly reflect the reduction in operational capacity under severe weather conditions. Simultaneously, the statistics of the number of ships congested at port trigger a linked response of the early warning mechanism. When the number of ships congested exceeds a preset threshold, the system automatically calls the strategy fusion and optimization module in step S3074 to generate an anti-interference scheduling scheme.

[0114] Step S3074: When the maximum job delay exceeds the preset tolerance threshold, activate the strategy fusion and optimization module, extract the successful response strategy set from extreme scenario cases, and combine and optimize the response strategy with the current solution set based on the tabu search algorithm to generate an anti-interference scheduling scheme.

[0115] When the digital twin model detects that the maximum operation delay exceeds a preset tolerance threshold (e.g., 60 minutes in a typhoon scenario), the system immediately activates the strategy fusion and optimization module. The module first extracts a set of successful response strategies that highly match the characteristics of the current scenario from the extreme scenario case library. Case matching employs a multi-dimensional similarity calculation model, comprehensively considering parameters such as event type, spatiotemporal constraints, and optimization objectives, and uses a cosine similarity algorithm to select the top 5 similar cases.

[0116] In the strategy extraction phase, the system deconstructs the handling solutions of historical cases, separating them into three core strategies: spatiotemporal resource reallocation strategies (such as prioritizing operations in high-density areas), equipment scheduling compensation strategies (such as quay crane cross-berth support mechanisms), and operational process optimization strategies (such as dynamically adjusting container loading and unloading sequences). For example, in a historical typhoon case, strategy ID S-023 was extracted: wind speed > 30 m / s, port congestion vessels > 5, quay crane cross-berth support; strategy ID S-041: crane failure rate > 0.05, yard density reallocation. By increasing the yard crane configuration coefficient from 0.9 to 1.2 in areas with a yard density > 0.8, the delay time was reduced by 25%, and this type of strategy was marked as a high-priority strategy fragment.

[0117] The initial solution set of the tabu search algorithm is generated by combining the current crane deployment scheme with historical strategy fragments. The algorithm uses a tabu list to remember invalid strategy combinations from the last five iterations, avoiding redundant exploration. For example, if the strategy combination of "quay crane cross-berth support + yard density compensation" has caused load imbalance for three consecutive iterations in the past, the system automatically adds this combination to the tabu list. During the neighborhood search, the algorithm prioritizes trying strategy combinations that are not prohibited, and evaluates the effectiveness of the strategy by calculating the objective function improvement rate (Δ = total delay reduction / computation time increment).

[0118] The strategy integration employs a layered optimization mechanism: the bottom layer performs rapid strategy matching, the middle layer verifies strategy combinations, and the top layer optimizes strategy parameters. Taking a real-world typhoon scenario as an example, the system completed a Monte Carlo simulation of 127 strategy combinations within 2 minutes, selecting the combination with the highest overall score. This temporarily increased the movement speed of the QC12 quay crane by 15%, raised the weighting coefficient of the outlet container zone identifier to 1.5, and shortened the equipment failure response time to 2 minutes. This solution, verified by a digital twin model, reduced the maximum operational delay from 72 minutes to 45 minutes while maintaining 92% equipment utilization, meeting the tolerance threshold requirements.

[0119] The final anti-interference scheduling scheme is designed to ensure feasibility through a triple verification mechanism. First, it verifies compliance with spatiotemporal constraints (e.g., the crane's movement path does not conflict with the ship's color block boundaries). Second, it conducts equipment load stress tests (the continuous operation time of a single piece of equipment does not exceed 80% of its rated load). Finally, it performs an economic evaluation (the increase in implementation cost is ≤15%). The verified scheme is encapsulated into a standardized instruction set and distributed to the terminal operation terminal through a protocol conversion gateway. Simultaneously, it updates the crane status parameters and container distribution characteristics in the heat map coordinate system, forming a closed-loop feedback mechanism.

[0120] Step S3075: Map the anti-interference scheduling scheme to the heat map coordinate system and update the crane coordinate update command.

[0121] The process of mapping the anti-interference scheduling scheme to the heatmap coordinate system is to establish a dynamic mapping relationship between the abstract strategy and the physical space. Based on the crane coordinate sequence output by the digital twin model (such as QC12 from X=185.3→X=220.7) and the yard crane density adjustment parameters (BlockA-02 coefficient from 0.9→1.2), the equipment movement trajectory is mapped to the XY spatiotemporal reference of the heatmap through a coordinate transformation algorithm. In this process, the crane movement path is decomposed into a series of discrete coordinate nodes, and the distance between adjacent nodes is controlled within 10 meters to match the accuracy requirements of terminal operations; the yard crane operation boundary is dynamically adjusted through the heatmap color block edge clipping algorithm to ensure spatial coupling with the container distribution hot zone.

[0122] The instruction update adopts a layered transmission mechanism. The bottom layer compiles the coordinate sequence into a binary instruction set through a protocol conversion gateway, which includes 12 parameters such as crane number, target position, and movement speed. The middle layer distributes instructions through a 5G private network. The quay crane instructions are directly connected to the operator's HMI through a high-priority channel, while the yard crane instructions are preloaded into the PLC cache through edge computing nodes. The top layer provides visual guidance through an AR navigation system. The crane operator's HMI interface synchronously displays movement path arrows with timestamps, and the yard crane operation screen automatically highlights container stacking areas with increased priority.

[0123] After the instruction update is completed, the system starts a closed-loop verification mechanism. On the one hand, the position data is fed back in real time through the crane encoder to perform sub-meter level deviation correction with the theoretical path. On the other hand, the equipment status parameters in the heat map coordinate system are updated, the adjusted crane position is mapped to the new color block coordinates, and the urgency coefficient of the associated containers is recalculated to form a dynamically updated visual operation status map.

[0124] Reference Figure 5 Methods for acquiring multi-source data related to ship operations and fusing ship arrival time, berthing length, and container distribution data using a preset spatiotemporal fusion algorithm include: Step S101: Remove outliers and fill in missing values ​​in the acquired multi-source data associated with ship operations.

[0125] A data cleaning mechanism combining physical rule verification and statistical models ensures the input quality of the spatiotemporal fusion algorithm. For missing data, when ship arrival times are missing, a weighted average of historical data from the same ship is used for imputation, with the weight decreasing exponentially with the route similarity index. Missing berth lengths are calculated using the ship's design length multiplied by the current draft correction factor, with a draft depth weight of 0.8. Missing container distribution coordinates are obtained using bay space interpolation, prioritizing the coordinates of adjacent containers with higher cargo value to generate substitute values. Data quality is dynamically monitored using a reliability index Qscore = (effective data volume / total data volume) × (1 - imputation error rate). When Qscore < 90%, a manual review process is activated; when it > 95%, the geofence boundary is automatically widened by 5%. After preprocessing, the spatiotemporal error of the data is controlled within ±2 minutes in time and ±1.5 meters in space, providing standardized input for subsequent fusion.

[0126] Step S102: The planned, expected and actual values ​​of the ship's arrival time are aligned in time sequence, and the time confidence is calculated by weighted average. The weight increases exponentially as the data update time approaches.

[0127] Based on a time reference axis, multi-source vessel time data is dynamically integrated. First, the planned arrival time (from the terminal scheduling system), estimated arrival time (based on AIS trajectory prediction), and actual arrival timestamp (collected by the Automatic Identification System) are uniformly converted to a minute-offset coordinate system with 00:00 of the current day as the origin. A sliding time window mechanism aligns the three time series. When a timestamp deviation exceeds ±15 minutes, a Kalman filter algorithm is automatically triggered to correct clock drift. Time confidence is calculated using an exponential weighted model, assigning the highest weight to the most recent data. Data T hours from the current time has a weight of e^(-0.5T). For example, data from one hour ago has a weight of 0.6, while real-time data has a weight of 1.0. The final output is the fused time value and a confidence index (0-1 range). Time data with a confidence score > 0.85 is directly used for heatmap Y-axis mapping, while low-confidence data triggers AIS trajectory re-verification.

[0128] Step S103: Transform the design value and measured value of the ship's occupancy length, as well as the container distribution's hold coordinates and yard coordinates, to a unified spatial reference, so that the unified spatial reference is consistent with the X-axis reference of the control and management coordinate system, and eliminate the systematic errors of different measurement systems through the coordinate mapping matrix.

[0129] Using the physical reference stake (RFID tag CT-001) buried at the starting point of the wharf operation as the absolute origin, a spatial coordinate system for the wharf shoreline was established through total station measurements. Ship occupancy length data were uniformly converted using affine transformation. The design value was replaced with a rigid body displacement model to eliminate drawing scale errors, while the measured value was fitted with a tidal draft correction coefficient using the least squares method. Container hold coordinates were constructed using a rotation matrix based on the ship's berthing azimuth, converting the ship's local coordinate system (origin at the port bow) into the wharf's global coordinate system. Yard coordinates were registered using grid control points, and a Helmholtz transformation was performed on the GPS / BeiDou differential positioning data to eliminate measurement benchmark deviations in different areas. The conversion process used a weighted coordinate mapping matrix M = α·R + β·T (R is the rotation matrix, T is the translation vector, and α and β are equipment error weights). Residual analysis ensured that the spatial alignment error was ≤ ±0.5 meters.

[0130] Step S104: Construct a spatiotemporal correlation matrix based on the confidence level of the time dimension and the spatial coordinate deviation, calculate the covariance of the confidence level of the time dimension and the spatial coordinate deviation, and perform weighted fusion on data with a correlation level higher than a preset threshold, with the weights being the eigenvalues ​​of the inverse matrix of the covariance.

[0131] A two-dimensional spatiotemporal correlation matrix is ​​constructed to achieve dynamic data fusion. The row vectors of the matrix correspond to the time confidence sequence (from step S102), and the column vectors are the spatial coordinate residual sequence (from step S103). A sliding time window mechanism is used to calculate the covariance matrix Cov(T,S), where the covariance between the time dimension confidence T and the spatial deviation S reflects the spatiotemporal coupling strength. Eigenvalue decomposition is performed on the covariance matrix, and the eigenvector corresponding to the largest eigenvalue λmax is extracted as the main correlation direction. When the correlation degree of matrix elements is >0.8 (preset threshold), weighted fusion is triggered: weight w=1 / (λi+ε) (ε is a constant divided by zero), and data with higher weights participate in fusion first. The fusion process adopts spatiotemporal adaptive weighted averaging. For example, when the confidence degree of a ship's arrival time is 0.92 and the spatial residual is 0.3m, its fusion weight is increased by 40% compared to the benchmark value. Standardized spatiotemporal data frames are generated, carrying the ship's unique identifier, fusion spatiotemporal stamp, and normalized spatial label. The spatiotemporal alignment error is controlled within ±1.2 meters / ±1.8 minutes.

[0132] Step S105: Output the fused standardized spatiotemporal data frame. This data frame includes the ship's unique identifier, the fused arrival timestamp, the ratio of the normalized occupancy length to the total length of the wharf shoreline, and the spatiotemporal index code of the container distribution.

[0133] The unique identifier for a vessel uses a combination of IMO number and berthing sequence number (e.g., IMO9456789-CT05) to ensure global terminal resolution. The merged arrival timestamp format is the ISO8601 extended format, with time accuracy linked to data confidence: High confidence (confidence > 0.9): Timestamp = weighted average ± (1 / 2 × time window width); Medium confidence (0.5 < confidence ≤ 0.9): Timestamp = median ± (1 × time window width); Low confidence (confidence ≤ 0.5): Triggers AIS trajectory reverification. Normalized berth length is calculated as measured berth length / total shoreline length (e.g., 300m / 2500m = 0.12), eliminating the impact of differences in terminal size. The spatiotemporal index coding of container distribution is generated by hashing the physical segment number on the X-axis (50-meter segment, such as X012) and the planned time window number on the Y-axis (1-hour window, such as Y008) to create a unique geographic label (such as X012Y008#).

[0134] Step S106: The spatiotemporal index code is generated by combining the X-axis physical segment number and the Y-axis planned time window number. The X-axis division rule is the preset physical segment length, and the Y-axis division rule is the preset time window length.

[0135] A unique geographic identifier is constructed based on the spatiotemporal gridding rules of the control and management coordinate system. The X-axis discretizes and encodes the wharf shoreline according to preset 50-meter physical segments. The segment number format is "X" + three-digit serial number (e.g., 0-50 meters → X001, 1200-1250 meters → X024). Centimeter-level positioning is achieved by matching the measured coordinates of the total station with the center point of the segment. The Y-axis uses a 1-hour time window unit to encode the planned operation period of the vessel as "Y" + three-digit time offset (e.g., 08:00-09:00 → Y008, 13:30-14:30 → Y013).

[0136] Step S107: The normalized occupancy length is used as the calculation basis for the X-axis extension of the heatmap color block.

[0137] The spatial mapping engine enables the visualization conversion of ship physical dimensions into heat maps. The normalized occupancy length is obtained by dividing the ship's measured length by the total length of the wharf shoreline (e.g., 300m / 2500m=0.12). The X-axis extension of the color block = normalized value × total display width of the heat map X-axis × (1 + 0.005 × draft), where the draft is in meters and is corrected using real-time tide sensor data (high tide compensation +3%, low tide compensation -2%). The dynamic calibration module combines real-time tide data to correct for the influence of ship draft; for example, if the draft increases by 2 meters during high tide, the extension automatically increases by 1%. When a color block spans multiple physical segments (e.g., the extension corresponds to segment X012-X014), the system triggers a multi-segment color block segmentation algorithm to maintain a consistent length ratio for each segment.

[0138] Step S108: Use the merged arrival timestamp as the basis for calculating the Y-axis extension of the heatmap color block.

[0139] A time mapping engine is used to visualize the transformation of ship operation periods into heat maps. Using the merged arrival timestamp output in step S105 as the reference point (ISO8601 format), the Y-axis extension of the color blocks is calculated based on the planned berthing duration of the ship. The Y-axis extension is calculated as: berthing duration (minutes) × time axis scaling factor + tidal compensation value. The scaling factor is 0.8 pixels / minute by default, and the tidal compensation value is dynamically adjusted according to the daily tide curve (high tide compensation +3%, low tide compensation -2%). When a ship spans multiple time windows (e.g., 08:30-10:45), the system automatically segments the color blocks while maintaining temporal continuity, with a gap of ≤0.5 pixels between adjacent color blocks. The dynamic calibration module is linked to AIS real-time positioning data; if the time deviation is >5 minutes, the color block width is recalculated (if the arrival is delayed, the extension increases by 8%).

[0140] Step S109: When dynamic optimization is triggered, update the fusion weights of the spatiotemporal correlation matrix based on real-time monitoring data.

[0141] After the early warning mechanism is activated, the data fusion strategy is dynamically adjusted. When the operation monitoring chain detects that the rate deviation ΔV exceeds the threshold (e.g., quay crane ΔV < -15%), the system collects real-time data on crane position, container displacement, and weather changes. Through the covariance matrix eigenvalue recalculation engine, parameters such as real-time throughput rate deviation and ship drift distance are injected into the spatiotemporal correlation matrix to generate a dynamic weight update factor: Wnew = Worigin × (1 + 0.2|ΔV|) × η, where η is the equipment status coefficient (normal η = 1.0, fault η = 0.6). The weight update frequency increases with the warning level, updating every 5 minutes for yellow warnings and every second for red warnings.

[0142] Based on the same inventive concept, this application provides a container terminal intelligent management system based on spatiotemporal coordinate mapping, including a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, it can realize a container terminal intelligent management method based on spatiotemporal coordinate mapping.

[0143] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for intelligent management of a container terminal based on space-time coordinate mapping, characterized in that, The method comprises the following steps: Obtain ship operation related multi-source data, and perform data fusion on ship arrival time, berth length and container distribution data through a preset space-time fusion algorithm; Establish a control and management coordinate system based on the fused multi-source data, taking the physical position of the wharf as the X-axis and the planned arrival time of the ship as the Y-axis; Generate a ship distribution heat map in the control and management coordinate system, map each ship to a rectangular color block, the length of the color block along the X-axis represents the berth length of the ship, and the width of the color block along the Y-axis represents the planned docking time; Load the container operation feature data in real time and embed it in the color block, the container operation feature data includes the current total loading capacity, the ratio of large / small standard containers, and the partition identification of import and export containers; Analyze the spatial distribution characteristics of the containers from the partition identification, including the concentrated coordinate set of import containers in the ship's cabin and the distribution density matrix of export containers in the wharf yard; Generate a crane deployment scheme based on the preset crane efficiency parameters and the spatial distribution characteristics of the containers, and send the crane deployment scheme to the terminal of the wharf operation.

2. The intelligent management method of the container terminal based on the space-time coordinate mapping according to claim 1, characterized in that, The method for generating the crane deployment scheme comprises the following steps: Determine the optimization target of the crane deployment, including minimizing the crane moving distance, balancing the load of each crane, and reducing the container loading and unloading delay rate; Construct a weighted distance calculation model, taking the current position of the crane as the origin, calculate the weighted sum of the absolute value of the X-axis coordinate difference and the Y-axis time difference of the target container in the berth, and the weight parameters include the crane operation efficiency coefficient and the container emergency degree coefficient; Based on the spatial distribution characteristics of the containers, extract the cabin coordinate set of the import containers and the yard density matrix of the export containers, and establish a container priority queue, wherein the priority of the import containers is higher than that of the export containers, and the priority of the containers in the high-density yard area is improved by a preset proportion; Iteratively optimize the weighted distance model, set the preset range of the crossover probability and the mutation probability, and retain the candidate schemes with a preset proportion of the fitness values; Generate a preset number of random scenarios to verify the operation delay fluctuation of the candidate schemes within the confidence interval; When the fluctuation of the iteration fitness value is less than a preset threshold for a continuous preset number of times, terminate the optimization and output the optimal crane deployment path.

3. The intelligent management method for container wharf based on space-time coordinate mapping according to claim 1, characterized in that, The method for optimizing the crane deployment scheme comprises the following steps: Take the position of the ship color block in the heat map coordinate system as a spatial constraint, and extract the real-time throughput rate deviation value of each crane based on the operation monitoring chain; Calculate the color block correlation weight based on the X-axis position coordinate of the ship color block in the heat map, the color block correlation weight is calculated based on the center point coordinate of the ship color block in the X-axis, and the ships corresponding to the first N color blocks are selected as the initial solution set according to the descending order of the correlation weight; Construct a multi-objective optimization function, including a moving distance reduction term, a load balancing term and a rate deviation compensation term, wherein the moving distance reduction term is determined by calculating the total moving distance of the crane, the load balancing term is determined by calculating the difference between the load of each crane and the maximum load, and the rate deviation compensation term is determined by calculating the square sum of the real-time throughput rate deviation, and each sub-term is configured with a preset weight coefficient. A hybrid optimization process of fusion Tabu search and simulated annealing is performed, the Tabu search stage generates a candidate solution set in the neighborhood of the current solution, and the non-tabu solution with the optimal target function improvement is selected based on the Tabu list to exclude recently appeared solutions, and the simulated annealing stage allows part of the deteriorated solutions to be accepted according to the probability acceptance criterion according to the annealing temperature decay strategy; The Tabu search and simulated annealing stages are alternately performed, and after each Tabu search generates a candidate solution, the number of consecutive non-improvements is checked, and when the preset number of times is reached, the simulated annealing stage is triggered, and the optimization state parameters including the current optimal solution, the Tabu list and the annealing temperature are updated; The target function improvement rate is calculated, the improvement rate is determined by the absolute value of the difference between the current iteration target value and the previous iteration target value divided by the previous iteration target value, and when the improvement rate is lower than the preset convergence threshold, the iteration is terminated, and the final solution set is output from the Pareto optimal solution set maintained by the hybrid optimization process. The final solution set is used as a verification object to verify the feasibility of the final solution set in the throughput achievement rate and the equipment utilization rate fluctuation index. The verified optimal solution is mapped to the heat map coordinate system, and the crane coordinate update instruction is generated according to the crane position and container distribution characteristics in the solution set.

4. The intelligent management method for container wharf based on space-time coordinate mapping according to claim 3, characterized in that, After verifying the feasibility of the final solution set in the throughput achievement rate and the equipment utilization rate fluctuation index, it further includes: Extract the top K extreme scene cases with the highest similarity from the historical operation database, the extreme scene includes at least one of typhoon weather, crane failure and ship port pressure; A digital twin emergency deduction model is constructed, and the current heat map coordinate system parameters, crane state real-time data and extreme scene characteristic parameters are injected into the emergency deduction model; The final solution set is executed in the emergency deduction model to extract the key performance indicators, including the maximum operation delay time, the number of ship port pressure and the equipment vacancy rate. When the maximum operation delay time exceeds the preset tolerance threshold, activate the strategy fusion optimization module, extract the successful coping strategy set from the extreme scene cases, and combine and optimize the coping strategy and the current solution set based on the Tabu search algorithm to generate an anti-interference scheduling scheme. The anti-interference scheduling scheme is mapped to the heat map coordinate system to update the crane coordinate update instruction.

5. The intelligent management method for container wharf based on space-time coordinate mapping according to claim 1, characterized in that, Obtain ship operation associated multi-source data, and perform data fusion on ship arrival time, berth length and container distribution data through a preset spatio-temporal fusion algorithm, including: Remove outliers and fill missing values from the obtained ship operation associated multi-source data; Align the planned value, estimated value and actual value of the ship arrival time in time sequence, calculate the time confidence by weighted average, and the weight increases exponentially as the data update time approaches; Convert the design value, measured value of the ship berth length and the container distribution cabin coordinates and yard coordinates to a unified space reference, so that the unified space reference is consistent with the X-axis reference of the control and management coordinate system, and eliminate the system error of different measurement systems through the coordinate mapping matrix; A time-dimension confidence and spatial coordinate deviation based spatio-temporal correlation degree matrix is constructed, the covariance of the time-dimension confidence and the spatial coordinate deviation is calculated, the data with a correlation degree higher than a preset threshold is subjected to weighted fusion, and the weight is an inverse matrix eigenvalue of the covariance; An output fused normalized spatio-temporal data frame is outputted, which contains a unique ship identification, a fused arrival timestamp, a normalized berth length ratio relative to a total length of a wharf shoreline, and a spatio-temporal index code of container distribution; The spatio-temporal index code is generated by combining an X-axis physical section number and a Y-axis planned time window number, the X-axis division rule is a preset physical section length, and the Y-axis division rule is a preset time window length; The normalized berth length is taken as a calculation reference of an X-axis extension of a heat map color block; The fused arrival timestamp is taken as a calculation reference of a Y-axis extension of the heat map color block; When triggering dynamic optimization, the fusion weight of the spatio-temporal correlation degree matrix is updated based on real-time monitoring data.

6. A container wharf intelligent management system based on space-time coordinate mapping, characterized in that, The program can be loaded and executed by the processor to implement the method of claim 1 to 5.

Images