A port container dispatching method and system

Abstract

The present invention discloses a method and system for dispatching containers at a port, and belongs to the technical field of dispatching resources at a port. The method comprises: using a quantum particle swarm algorithm to analyze the real-time data in the port after preprocessing, and generating an initial dispatching plan after solving; based on the initial dispatching plan, using a deep deterministic policy gradient algorithm to locally optimize the dispatching strategy in combination with real-time data to generate an optimized dispatching plan; converting the optimized dispatching plan into specific operation instructions and sending them to the equipment control system, instructing each device to perform container loading and unloading tasks as planned; monitoring the real-time status of the port, and using a deep deterministic policy gradient algorithm to re-optimize the dispatching strategy based on abnormal feedback. The system based on the method comprises a data acquisition module, an optimization algorithm module, a task scheduling module, a real-time monitoring module, and a user interface module. The present invention can quickly generate an optimal dispatching plan and make real-time adjustments, effectively improving equipment utilization, shortening operation time, and reducing energy consumption.

Description

Technical Field

[0001] The present invention relates to a port container dispatching method, and in particular to a port container dispatching method and system based on a QPSO and DDPG double-layer optimization structure. Background Art

[0002] With the advancement of smart port construction, container dispatching systems play an important role in improving port operation efficiency, optimizing equipment utilization and reducing operating costs. Traditional container dispatching methods mostly use static optimization algorithms, such as genetic algorithms and simulated annealing algorithms. These methods show obvious limitations when dealing with complex environments that change in real time, and it is difficult to dynamically adapt to emergencies in port operations. In addition, traditional dispatching methods usually face the challenges of multi-objective optimization, such as minimizing operation time, reducing energy consumption and reducing task delays. Existing methods are inefficient in dealing with these multiple objectives, which often leads to low equipment utilization, long operation time and high energy consumption.

[0003] The static optimization algorithms in the existing technologies mainly rely on preset scheduling rules and fixed working modes, lack flexible adjustment mechanisms, and are difficult to cope with the frequently changing task allocation and equipment scheduling needs in the port. At the same time, these methods lack effective global search capabilities and local adaptability when facing dynamically changing port environments (such as equipment failures, changes in task urgency, etc.), resulting in poor overall scheduling efficiency and robustness.

[0004] In order to solve the above problems, there is an urgent need for a new type of container scheduling system and its operation method, which can combine the advantages of global optimization and local optimization, dynamically adapt to the real-time changes of the port, optimize the scheduling strategy of equipment, and improve operational efficiency and scheduling stability. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a port container scheduling method to solve the problems of low overall scheduling efficiency and poor robustness of existing port scheduling algorithms. Another purpose of the present invention is to propose a port container scheduling system to solve the problem of how to improve scheduling efficiency and system stability and realize real-time optimization scheduling of container loading and unloading tasks.

[0006] Technical solution: A port container dispatching method according to the present invention comprises the following steps:

[0007] The quantum particle swarm algorithm is used to analyze the pre-processed real-time data in the port, search for the optimal solution in the multi-dimensional parameter space, and generate the initial scheduling plan;

[0008] Based on the initial scheduling plan, the deep deterministic policy gradient algorithm is used to locally optimize the scheduling strategy in combination with real-time data to generate an optimized scheduling plan;

[0009] Convert the optimized scheduling plan into specific operation instructions and send them to the equipment control system, instructing each device to carry out container loading and unloading tasks as planned;

[0010] Monitor the real-time status of the port and perceive changes in the real-time status of the port. When abnormal changes occur, use the deep deterministic policy gradient algorithm to re-optimize the scheduling strategy based on abnormal feedback, dynamically adjust the task scheduling plan, and update the operation instructions.

[0011] The present invention uses the quantum particle swarm optimization (QPSO) algorithm to globally optimize the initial scheduling plan, and combines the deep deterministic policy gradient (DDPG) algorithm to achieve local optimization and real-time adjustment, effectively improving the efficiency and adaptability of port container scheduling.

[0012] In the present invention, the QPSO algorithm is responsible for generating the initial scheduling plan and providing the direction and basis for global optimization; the DDPG algorithm further optimizes the local scheduling strategy based on real-time data to cope with changes in the dynamic environment of the port. The DDPG algorithm focuses on local optimization and dynamic adjustment. The DDPG algorithm learns the optimal scheduling strategy in the real-time environment of the port by constructing a deep neural network, and dynamically optimizes the loading and unloading sequence, task allocation and equipment utilization. The DDPG algorithm uses real-time acquired data to dynamically adjust the local scheduling strategy according to the real-time changes in the status of port equipment and the progress of tasks to cope with emergencies and uncertainties in port operations. Through continuous learning and adaptive adjustment, the DDPG algorithm realizes dynamic optimization and enhanced adaptability of the strategy in actual scheduling. The feedback and real-time adjustment mechanism can ensure the continuity and efficiency of container loading and unloading tasks. The two algorithms work together to ensure that the system always maintains the optimal scheduling state and equipment performance in a complex and changing environment.

[0013] Preferably, the real-time data in the port includes equipment status, task progress, traffic flow and environmental status in the port, the preprocessing method includes denoising, formatting and standardization processing, and the initial scheduling plan includes the loading and unloading sequence of containers, equipment allocation and task priority, with the goal of minimizing total operation time, minimizing equipment energy consumption and maximizing equipment utilization.

[0014] Preferably, the quantum particle swarm algorithm searches for the optimal solution in the multidimensional parameter space including:

[0015] Set the size N of the particle swarm and the maximum number of iterations As well as the parameters of the quantum behavior model, the initial position of each particle represents a possible global scheduling scheme, which includes the initial equipment assignment for each task;

[0016] The fitness value of each particle is calculated to minimize the global objective function:

[0017]

[0018] in, is the total duration of all loading and unloading tasks, is the total energy consumption of all devices, and are the weight coefficients of time and energy consumption respectively;

[0019] Based on the quantum behavior model, update the position of each particle:

[0020]

[0021] in, represents the current position of the i-th particle at the t-th iteration, is the historical optimal position of the particle, is the global optimal position, and u is a random number used to introduce randomness and diversity in the search;

[0022] The process continues to iterate until the preset maximum number of iterations is reached. Or until the fitness value no longer improves significantly, the final global optimal solution is output as the initial scheduling plan.

[0023] Preferably, the local optimization scheduling strategy using a deep deterministic policy gradient algorithm includes:

[0024] Define the structural parameters of the Actor and Critic networks, including the number of nodes in the input layer, hidden layer, and output layer, and initialize the maximum capacity of the experience pool , learning rate α and discount factor γ, set the initial weights of the Actor network and the Critic network to ensure that the network is gradually optimized during the learning process;

[0025] Actor network based on current state Output Scheduling Strategy , the Critic network estimates the strategy output by the Actor network The value function , and based on the actual reward after execution t To update the strategy; at each time step t, the system samples the current state , the Actor network outputs the corresponding scheduling action , then execute the action and observe the new state and instant rewards t :

[0026]

[0027] in, Cost function representing task delay; empirical samples It is stored in the experience pool D, from which the system randomly extracts batch samples and uses the gradient descent method to update the parameters of the Actor and Critic networks. The Critic network uses the time difference error to update its parameters:

[0028]

[0029] The Actor network updates parameters through the policy gradient method:

[0030]

[0031] The global optimal solution output by the quantum particle swarm algorithm is used as the initial input, and the local optimization process continues until the maximum number of iterations is reached. Or until the reward function converges, the optimized local scheduling strategy is output.

[0032] Preferably, the real-time status of the port includes equipment failure information, weather change information, task progress information and task priority information;

[0033] When an abnormal change occurs, the deep deterministic policy gradient algorithm is used to re-optimize the scheduling strategy based on abnormal feedback, including:

[0034] When equipment failure, task delay or other abnormal conditions are detected, the abnormal information is fed back to the deep deterministic policy gradient algorithm to locally optimize the scheduling strategy, adjust the scheduling strategy in real time, reallocate tasks and optimize the operation path;

[0035] If the deep deterministic policy gradient algorithm detects that the initial solution generated by the quantum particle swarm algorithm deviates significantly from the actual operating environment, the information is passed back to the quantum particle swarm algorithm through the feedback mechanism, triggering the re-search and global optimization process of the quantum particle swarm algorithm.

[0036] The feedback mechanism continuously evaluates and adjusts the effectiveness of the scheduling strategy by comparing the scheduling execution results with the expected goals. When deviations or inefficient scheduling are detected, the system automatically triggers the optimization process and adjusts the scheduling plan in real time to ensure the stability and efficiency of the port scheduling task. Through the collaborative work of QPSO and DDPG, the mechanism realizes the continuous optimization of the scheduling strategy and improves the overall operating efficiency of the port container scheduling system.

[0037] Preferably, the above-mentioned port container scheduling method further comprises the following steps:

[0038] Provide operators with a scheduling information display and control operation interface to show the current task status and equipment conditions; operators can view the scheduling execution results in real time through the interface, and manually intervene in abnormal situations or adjust scheduling parameters; based on the scheduling execution results and operator feedback, further adjust the optimization strategy to provide reference and improvement for subsequent scheduling tasks.

[0039] Another aspect of the present invention discloses a port container dispatching system, comprising a data acquisition module, a task dispatching module, an optimization algorithm module, a real-time monitoring module and a user interface module;

[0040] The data acquisition module is used to collect real-time data from port equipment and environmental sensors;

[0041] The optimization algorithm module includes the QPSO algorithm submodule for global optimization and the DDPG algorithm submodule for local dynamic optimization. Based on the real-time data obtained by the data acquisition module, it adopts a method combining the quantum particle swarm algorithm and the deep deterministic policy gradient algorithm to achieve global optimization and local dynamic adjustment of port container scheduling tasks.

[0042] The task scheduling module is used to execute the scheduling plan output by the optimization algorithm module and perform specific task allocation and equipment scheduling operations;

[0043] The real-time monitoring module is used to monitor the device status and task execution, handle exceptions and dynamically adjust the scheduling strategy;

[0044] The user interface module is used to display scheduling information, monitor status and receive user instructions.

[0045] Furthermore, the data acquisition module is used to obtain real-time data from various equipment and environmental monitoring systems within the port to support the optimization decision of container scheduling. The data acquisition module integrates a variety of sensors and data interfaces, including equipment status sensors, task progress sensors, traffic flow sensors, and environmental status sensors. Equipment status sensors such as load sensors, position sensors, and temperature sensors are used to monitor the workload, location information, and equipment temperature of port equipment such as quay cranes, yard cranes, AGVs, and container trucks. Task progress sensors include RFID readers, GPS modules, and cameras, etc., which are used to track the movement of containers and task completion. Port traffic flow sensors are deployed in key sections within the port, and use laser radars, millimeter-wave radars, and traffic video monitoring equipment to obtain internal traffic conditions and transportation path information. Environmental status sensors such as temperature, humidity, and wind speed sensors are used to collect meteorological data within the port. All collected data are filtered, formatted, and standardized by the processing unit in the data acquisition module, and then transmitted to the optimization algorithm module to provide real-time and reliable data support for the QPSO and DDPG double-layer optimization algorithms, and realize efficient optimization and dynamic adjustment of container scheduling.

[0046] Furthermore, the optimization algorithm module continuously evaluates and adjusts the effectiveness of the scheduling strategy by comparing the scheduling execution results with the expected goals; when deviations or scheduling inefficiencies are detected, the system automatically triggers the optimization process and adjusts the scheduling plan in real time; when the task scheduling module detects task execution deviations or abnormal equipment status, it can reallocate or adjust tasks based on real-time feedback information.

[0047] Based on the real-time data obtained by the data acquisition module, the optimization algorithm module adopts a method combining the quantum particle swarm algorithm (QPSO) and the deep deterministic policy gradient (DDPG) algorithm to achieve global optimization and local dynamic adjustment of the port container scheduling task. The QPSO algorithm is used to optimize the global scheduling strategy. By simulating the superposition and entanglement principles in quantum mechanics, it explores multiple solutions in the multi-dimensional parameter space and generates an initial scheduling plan, including the container loading and unloading sequence and the initial allocation of equipment. The QPSO algorithm combines the real-time status data of the port to dynamically adjust the scheduling strategy to minimize the total operation time and equipment energy consumption. In actual operation, the QPSO algorithm can quickly converge to the global optimal solution to ensure the overall scheduling efficiency of the system.

[0048] Furthermore, the task scheduling module first receives the optimal scheduling plan generated by the QPSO and DDPG algorithms, which contains information such as the specific container loading and unloading sequence, task allocation, and equipment usage plan. The task scheduling module converts these scheduling instructions into executable operation commands and sends them to the relevant equipment control systems, such as quay cranes, yard cranes, and automatic guided vehicles (AGVs). The module communicates data through standardized interfaces (such as Ethernet, RS485, 5G, etc.) to ensure that each device performs container loading and unloading tasks in the specified order and time, and continuously tracks the operation progress of the equipment and records the completion of the task. When the task scheduling module detects a deviation in task execution or an abnormal equipment status, it can reallocate or adjust the task based on real-time feedback information to ensure the continuity and efficiency of container loading and unloading tasks.

[0049] Furthermore, the real-time monitoring module is connected to the data acquisition module to obtain the equipment status, task progress and environmental data in real time, and analyze these data to identify abnormal situations during task execution, such as equipment failure, task delay or traffic jam. The real-time monitoring module cooperates with the task scheduling module to feedback abnormal situations to the optimization algorithm module to trigger corresponding adjustment strategies; the real-time monitoring module can also dynamically adjust the task scheduling plan and update the equipment instructions according to the changes in real-time data to ensure the stability of task execution and the effectiveness of the scheduling strategy; the real-time monitoring module realizes intelligent monitoring and dynamic optimization of the scheduling process through a closed-loop feedback control mechanism to ensure the optimal use of port equipment and resources.

[0050] Furthermore, the user interface module provides operators with an interface for displaying scheduling information and controlling operations, and feedbacks the real-time status of container loading and unloading tasks. The user interface module displays the current status of port equipment, task progress, traffic flow, and real-time loading and unloading task execution through an intuitive graphical interface, ensuring that operators can grasp the execution status of all scheduling operations in real time. The module has a feedback function for task completion. When a task is delayed or abnormal, it will automatically trigger an alarm notification to prompt the operator to intervene or adjust. In addition, the user interface module allows operators to manually adjust scheduling parameters or directly control equipment to deal with emergencies. The module supports the query and analysis of historical data, helping managers optimize scheduling decisions and improve the overall operational efficiency and equipment utilization of the port. By interacting with other modules of the scheduling system for data, the real-time and accuracy of the displayed information is ensured.

[0051] Based on the above method, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program implements the steps of the above method when executed by a processor.

[0052] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0053] (1) The present invention realizes global and local dynamic optimization of port container scheduling by combining the QPSO and DDPG double-layer optimization algorithms. It can quickly generate the optimal scheduling plan and make real-time adjustments, effectively improve equipment utilization, shorten operation time, reduce energy consumption, and enhance the dynamic adaptability and robustness of the scheduling strategy.

[0054] (2) The data acquisition module of the present invention integrates a variety of sensors and data interfaces to ensure the real-time acquisition and processing of various equipment and environmental data in the port, provide accurate input data for the optimization algorithm, and improve the scientificity and reliability of scheduling decisions.

[0055] (3) By optimizing the feedback mechanism of the algorithm module, the present invention can continuously evaluate and optimize the scheduling strategy, dynamically respond to emergencies in port operations, such as equipment failures and task delays, and ensure the stability and continuity of container loading and unloading tasks.

[0056] (4) The real-time monitoring module of the present invention works in coordination with the task scheduling module to realize intelligent monitoring and dynamic optimization of the scheduling process, timely handle abnormal situations in the scheduling process, and ensure efficient execution of the scheduling plan and stable operation of the system.

[0057] (5) The user interface module of the present invention provides real-time information display and control operation functions, so that operators can understand the scheduling execution status in real time and intervene and adjust when necessary, thereby improving the operational flexibility and response speed of the system.

[0058] (6) The present invention realizes efficient data exchange and system integration between modules through standardized interfaces, ensures the stability and accuracy of data transmission, supports rapid deployment and maintenance of the system, is suitable for container scheduling optimization of smart ports, and significantly improves the overall operational efficiency and management level of the port. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is the overall structural diagram of the port container dispatching system;

[0060] Figure 2 This is the structure diagram of the data acquisition module;

[0061] Figure 3 The workflow diagram for the optimization algorithm module;

[0062] Figure 4 It is the execution flow chart of the task scheduling module;

[0063] Figure 5 It is the monitoring and feedback flow chart of the real-time monitoring module;

[0064] Figure 6This is a schematic diagram of the user interface module interface. DETAILED DESCRIPTION

[0065] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0066] The present invention proposes a port container dispatching system and an operation method thereof based on a double-layer optimization structure of QPSO and DDPG, aiming to improve the efficiency and resource utilization of port container loading and unloading. Figure 1 The overall architecture of the system is demonstrated, including data acquisition module, optimization algorithm module, task scheduling module, real-time monitoring module and user interface module.

[0067] The core of the port container dispatching system lies in the collaborative work between various modules. The data acquisition module is located at the front end of the system and is responsible for obtaining real-time data from the equipment and environmental monitoring systems in the port to provide support for dispatching decisions. The collected data is transmitted to the optimization algorithm module through standardized interfaces (such as Ethernet, USB, RS485, etc.). The optimization algorithm module uses QPSO and DDPG algorithms for global and local optimization to generate a dispatching plan. The task scheduling module converts the optimization results into specific operating instructions and transmits them to various equipment control systems, such as quay cranes, yard cranes and AGVs. The real-time monitoring module monitors the task execution process, detects the equipment status and task progress, and feeds back anomalies to the optimization algorithm module to ensure the continuous optimization and execution stability of the dispatching strategy. The user interface module provides operators with dispatching information display, equipment control and abnormal alarm functions, and supports manual adjustment and real-time monitoring of dispatching tasks.

[0068] Data acquisition module such as Figure 2 As shown in the figure, it includes equipment status sensors, task progress sensors, traffic flow sensors and environmental status sensors, etc., integrating multiple sensors and data interfaces. Equipment status sensors (such as load sensors, position sensors and equipment temperature sensors) are used to monitor the workload and status of various types of equipment; task progress sensors (such as RFID readers, GPS modules and yard cameras) are used to track the movement of containers and task completion; traffic flow sensors (such as lidar, millimeter wave radar and traffic video monitoring equipment) are used to obtain internal traffic conditions; environmental status sensors (such as temperature, humidity, wind speed sensors) are used to collect meteorological data of the port. The data acquisition module transmits data to the optimization algorithm module through interfaces (such as RS232, I2C, CAN, Ethernet, etc.) for scheduling optimization.

[0069] The optimization algorithm module is the core of this system. Figure 3This module adopts a two-layer optimization method combining QPSO (Quantum Particle Swarm Optimization) and DDPG (Deep Deterministic Policy Gradient Algorithm) to achieve global and local optimization of port container scheduling.

[0070] First, the QPSO algorithm is used for global optimization. It generates an initial scheduling plan by performing a global search in a multi-dimensional parameter space, including loading and unloading sequence and equipment allocation. The algorithm combines the superposition and entanglement principles in quantum mechanics to quickly find the global optimal solution and ensure that the scheduling plan is optimal overall. The optimization plan takes into account multiple objectives, including minimizing the total operation time, reducing equipment energy consumption, and optimizing equipment utilization.

[0071] On the basis of global optimization, the DDPG algorithm performs local optimization through deep neural network learning, and dynamically adjusts the execution strategy of loading and unloading tasks according to real-time environmental data. The DDPG algorithm is mainly responsible for refining the optimized scheduling plan, such as dynamically adjusting according to data such as equipment status, task progress, and traffic flow in the port, optimizing task allocation and equipment operation, and improving the flexibility and adaptability of scheduling. The algorithm can quickly respond to real-time environmental changes and ensure that the system's scheduling decisions are sufficiently flexible.

[0072] The optimization algorithm module forms a closed-loop feedback system through continuous data input, global optimization, local optimization, self-learning mechanism and feedback adjustment, and continuously adjusts and optimizes the scheduling strategy to cope with the complex and changing port environment and dynamic task requirements. The module design ensures the efficient operation and real-time response capability of the port scheduling system, thereby improving the overall operating efficiency and stability.

[0073] The task scheduling module is the execution core of this system. Figure 4 As shown in the figure. Based on the scheduling scheme generated by the optimization algorithm module, this module converts the results of global and local optimization into specific equipment instructions and performs actual container loading and unloading and transportation operations. The task scheduling module first receives the optimal scheduling scheme generated by the QPSO and DDPG algorithms, which includes information such as container loading and unloading sequence, equipment usage plan, and task allocation.

[0074] The task scheduling module breaks down these scheduling plans into specific operating instructions to ensure accurate execution of each device. Specifically, the module is responsible for breaking down the optimized loading and unloading sequence and equipment usage plan into instructions for various types of port equipment, including quay cranes, yard cranes, and automated guided vehicles (AGVs). Figure 4 As shown in the figure, these instructions are transmitted to the control systems of various equipment through a variety of communication methods: the instructions of the quay crane and yard crane are usually transmitted through RS485 and Ethernet interfaces to ensure stability and real-time performance; the instructions of the automatic guided vehicle (AGV) are transmitted at high speed through the 5G network to cope with dynamically changing task requirements.

[0075] After receiving the instructions, each equipment control system (such as quay cranes, yard cranes and AGVs) executes the container loading and unloading tasks according to the plan. The task scheduling module is not only responsible for issuing initial instructions, but also continuously monitors the equipment's operating progress and task completion, and dynamically adjusts task allocation based on real-time feedback information. Through the feedback mechanism, the task scheduling module can quickly adjust the equipment's scheduling plan when it detects abnormal equipment status, task delays or other emergencies, to ensure the efficient execution and continuity of container loading and unloading tasks.

[0076] In quay crane operations, the task scheduling module can dynamically adjust the order of loading and unloading containers and the equipment priority to avoid equipment idleness and improve overall operational efficiency; for AGVs, the module can update their path planning and speed in real time to ensure the optimal route and time for transportation. In yard crane operations, the task scheduling module dynamically adjusts the task allocation and movement path of the yard crane according to the container location in the yard and the priority of the loading and unloading tasks.

[0077] The intelligent management of the task scheduling module not only ensures the coordination and efficient operation of various devices, but also achieves the optimal execution of scheduling tasks through real-time adjustment and feedback mechanisms, fully utilizing the performance of port equipment and improving the overall efficiency and resource utilization of port container scheduling. Through such an intelligent scheduling mechanism, the system can dynamically adapt to the complex and changing environmental conditions and task requirements within the port, significantly improving scheduling efficiency and system stability.

[0078] The real-time monitoring module is used to ensure the stable execution and dynamic adjustment of tasks, such as Figure 5 As shown in the figure. This module monitors abnormal conditions (such as equipment failure, task delay or traffic jam) during task execution by collecting equipment status, task progress, port traffic flow and environmental data in real time. When an abnormality is detected, the module uses built-in analysis and classification functions to process the abnormality and feeds back the abnormal information to the optimization algorithm module to trigger re-optimization or adjustment of the scheduling strategy. By analyzing the latest data, the real-time monitoring module can dynamically adjust the task scheduling plan to ensure that the task is completed smoothly as planned, thereby improving the overall stability and responsiveness of the scheduling system.

[0079] The real-time monitoring module works closely with the data acquisition module to ensure the timeliness and accuracy of information flow, monitor the operating status and task progress of each device through a closed-loop control mechanism, and update the scheduling plan in real time. The module is designed to quickly identify and respond to abnormal situations, effectively reduce the impact of delays and failures on the overall scheduling tasks, and ensure the efficiency and continuity of port operations.

[0080] The user interface module provides an intuitive and friendly graphical interface for operators, such as Figure 6As shown. This interface is used to monitor the system status and task progress in real time, including displaying key information such as equipment status, task execution, port traffic flow, etc. Operators can view the various parameters of the current dispatch through the interface and manually intervene when abnormal conditions are detected, such as triggering an emergency stop of the equipment or rescheduling. The user interface module also supports real-time adjustment of dispatch parameters and provides historical data analysis functions, enabling operators to optimize dispatch strategies based on data feedback and improve the flexibility and response speed of the system. The module is designed as a multi-functional operating platform with functions such as alarm notification, equipment control, data query and log recording, supporting operators to manage the system in all aspects. Through the user interface module, operators can not only respond and handle emergencies quickly, but also make more scientific and efficient dispatch decisions based on historical data analysis to ensure the stability and efficiency of port container dispatch tasks.

[0081] The present invention models the port container scheduling problem as a multi-objective optimization problem, with the goal of effectively utilizing port resources, shortening container loading and unloading operation time, and reducing overall energy consumption. The optimization objectives of the model include minimizing the total working time and energy consumption of the equipment, and maximizing the utilization rate of the equipment, etc., aiming to ensure the efficiency and economy of port operations through reasonable scheduling strategies.

[0082] During port operations, container loading and unloading tasks have different priorities and time limits, and each task needs to be completed within a specific time window. These tasks are performed by various equipment such as quay cranes (QC), yard cranes (YC), automatic guided vehicles (AGVs) and trucks. These equipment have their own working capacity, load capacity and energy consumption characteristics, so the scheduling system needs to allocate tasks among different equipment, taking into account the use efficiency and energy consumption of various types of equipment to ensure efficient completion of loading and unloading tasks.

[0083] Specifically, the port scheduling problem can be described as: assigning tasks to appropriate equipment within a given time window to ensure that all tasks are completed efficiently according to the priority and performance characteristics of the equipment. Due to the dynamic environment of the port, including equipment status changes, emergencies (such as equipment failures, weather changes), etc., the scheduling system must be flexible and adaptable. To this end, the present invention adopts a two-layer optimization structure, combined with the quantum particle swarm algorithm (QPSO) and the deep deterministic policy gradient algorithm (DDPG) to achieve the goals of global optimization and local optimization.

[0084] The global optimization layer uses the QPSO algorithm to conduct a comprehensive search in the multi-dimensional parameter space to generate an initial scheduling plan. The algorithm uses the superposition and entanglement principles in quantum mechanics to ensure that the global optimal solution is found in a complex multi-objective environment and balance the allocation of tasks on different devices. The local optimization layer uses the DDPG algorithm for real-time adjustment. DDPG uses deep reinforcement learning methods to continuously learn and adapt to environmental changes (such as changes in equipment status and task progress) during port operations to optimally allocate equipment and adjust scheduling strategies to ensure flexibility and efficiency in scheduling.

[0085] The optimization objectives of the model are divided into three points: minimizing the total operation time: improving the port's operating efficiency by reducing the total completion time of the task; minimizing equipment energy consumption: reducing the total energy consumption of various types of equipment by reasonably allocating equipment and optimizing the operation sequence; maximizing equipment utilization: improving the equipment's utilization efficiency by maximizing the ratio of the equipment's actual working time to its total available time.

[0086] In addition, the model needs to meet multiple constraints, including equipment capacity, task time window, priority, internal port traffic flow, etc., to ensure the feasibility and effectiveness of the model in practical applications.

[0087] The establishment of this mathematical model provides a solid theoretical basis for the subsequent algorithm implementation and provides guidance for the efficient operation of the scheduling system in a complex and dynamic port environment. By using a two-layer optimization structure, the present invention can not only find the global optimal solution for the scheduling task, but also make flexible adjustments according to the actual operation situation, thereby improving the adaptability and robustness of the scheduling system in a changing environment. The symbols, formulas and solution methods in the model are described in detail below.

[0088] First, define the following symbols and sets. Let J be the equipment set, including quay cranes (QC), yard cranes (YC), automated guided vehicles (AGVs), and trucks. Let I be the task set, including all loading and unloading tasks that need to be scheduled.

[0089] , where N I represents the number of tasks. Let T be the time window set, , where N T Indicates the total number of time periods.

[0090] At the global optimization level, the objective function mainly includes three aspects: minimizing the total operation time T total , minimize equipment energy consumption C energy and minimize the task delay rate D delay Among them, the total operation time T total is defined as the sum of the completion times of all tasks:

[0091]

[0092] Among them, s i is the start time of task i, e i is the end time of task i. Equipment energy consumption C energy The calculation formula is:

[0093]

[0094] Among them, x ij = 1 means task i is executed by device j, x ij = 0 means task i is not executed by device j, E j is the energy consumption per unit time of device j, is the operation time of device j. Task delay rate D delay Indicates that the task is not completed within the specified time, which is defined as:

[0095]

[0096] in, represents the time window of task i, a i is the earliest start time, b i is the latest end time. The global optimization objective function can be expressed as:

[0097]

[0098] At the local optimization layer, the model further optimizes task delay through the DDPG algorithm, and the local objective function is:

[0099]

[0100] The above optimization objective function needs to be solved under a series of constraints, including equipment capacity constraints, time window constraints, traffic flow constraints, operation safety constraints and operation sequence constraints. The specific constraints are as follows:

[0101] Device capacity constraints ensure that each device's usage does not exceed its maximum capacity:

[0102]

[0103] The time window constraint ensures that each task is completed within its specified time range:

[0104]

[0105] Traffic flow constraints limit the vehicle flow within the port to not exceed the designed flow limit:

[0106]

[0107] Operational safety constraints ensure that the load of each device does not exceed its maximum safe load at any time:

[0108]

[0109] Operation sequence constraints ensure that equipment tasks are performed in order and avoid conflicts:

[0110]

[0111] In order to solve the above mathematical model, the present invention adopts a two-layer optimization strategy. First, in the global optimization layer, the QPSO algorithm is used to perform a global search in the multidimensional parameter space to generate an initial scheduling plan to ensure that the global optimal solution is found. Then, in the local optimization layer, the DDPG algorithm is used to refine and adjust the initial plan based on real-time data. The DDPG algorithm learns the real-time status information of the port through a deep neural network, dynamically adjusts task allocation and equipment operation, and improves the adaptability and flexibility of the scheduling plan. The specific operation steps are as follows:

[0112] Step 1: Initialization phase, first define and set the parameters of QPSO and DDPG algorithms. For QPSO algorithm, you need to set the size of particle swarm N, the maximum number of iterations As well as the parameters of the quantum behavior model (such as search space, acceleration factor, etc.). The initial position of each particle represents a possible global scheduling scheme, which includes the initial device allocation for each task (device type, task order, etc.). At the same time, the structural parameters of the Actor and Critic networks are defined for the DDPG algorithm, including the number of nodes in the input layer, hidden layer, and output layer, and the maximum capacity of the experience pool is initialized. , learning rate α and discount factor γ. Set the initial weights of the Actor network and the Critic network to ensure that the network is gradually optimized during the learning process.

[0113] Step 2: In the global optimization stage, the QPSO algorithm is used to perform a global search in the multidimensional parameter space to generate the initial optimal solution to the port scheduling problem. First, the fitness value of each particle (i.e., scheduling solution) is calculated, with the goal of minimizing the global objective function:

[0114]

[0115] Among them, T total is the total duration of all loading and unloading tasks, C energy is the total energy consumption of all devices, w1 and w2 are the weight coefficients of time and energy consumption respectively. Then, based on the quantum behavior model, the position of each particle is updated:

[0116]

[0117] in, represents the current position of the i-th particle at the t-th iteration, is the historical optimal position of the particle, is the global optimal position, and u is a random number used to introduce randomness and diversity in the search. The process is continuously iterated until the preset maximum number of iterations is reached. Or the fitness value is no longer significantly improved. The final output global optimal solution is used as the initial input of the local optimization stage.

[0118] Step 3: In the local optimization phase, the DDPG algorithm is used to perform further local optimization and real-time adjustment based on the global optimal solution generated by QPSO. The port scheduling task is modeled as a Markov decision process (MDP), where the state space S includes information such as the current equipment load, task status, and real-time traffic conditions, and the action space A includes various scheduling operations (such as task allocation, adjusting task order, etc.).

[0119] The DDPG algorithm uses the Actor-Critic architecture to optimize the scheduling strategy. The Actor network is based on the current state. Output Scheduling Strategy , the Critic network estimates the strategy output by the Actor network The value function , and based on the actual reward after execution t To update the strategy. At each time step t, the system samples the current state S t , the Actor network outputs the corresponding scheduling action , then execute the action and observe the new state S t +1 and instant rewards t :

[0120]

[0121] Among them, D delay Represents the cost function of task delay. Empirical samples It is stored in the experience pool D, from which the system randomly extracts batch samples and uses the gradient descent method to update the parameters of the Actor and Critic networks. The Critic network uses the time difference error (TD error) to update its parameters:

[0122]

[0123] The Actor network updates parameters through the policy gradient method:

[0124]

[0125] The local optimization process continues until the maximum number of iterations is reached. Or until the reward function converges, the optimized local scheduling strategy is output.

[0126] Step 4: Real-time adjustment and feedback mechanism. In actual operation, the DDPG algorithm continuously monitors the real-time status of the port, such as equipment failures, weather changes, task priorities, and other changes. The real-time monitoring module senses these changes, feeds back information to the local optimization stage, and dynamically adjusts the scheduling strategy to ensure that the optimization scheme is robust and adaptable. At the same time, if DDPG detects that the initial solution generated by QPSO deviates significantly from the actual operating environment, the information is passed back to the global optimization stage through the feedback mechanism, triggering the re-search and optimization process of QPSO to ensure the continuous and efficient operation of the scheduling system.

[0127] Through this double-layer optimization structure, the present invention can achieve global and local efficient scheduling optimization, ensure the maximization of port equipment utilization, minimize operation time, and minimize energy consumption, thereby improving port operation efficiency and management level.

[0128] Based on the above system, the present invention further provides a port container scheduling method based on the double-layer optimization structure of QPSO and DDPG, such as Figures 1 to 6 As shown, the specific implementation steps are as follows:

[0129] Step 1: Data collection and preprocessing, such as Figure 2 As shown in the figure, the data acquisition module acquires data such as equipment status, task progress and traffic flow in the port in real time, involving multiple sensors such as load sensors, GPS modules and traffic monitoring equipment. The collected data is transmitted to the optimization algorithm module through standardized interfaces such as RS485, Ethernet and 5G. The data acquisition module performs preliminary processing on the data, including denoising, formatting and standardization, to ensure the accuracy and real-time nature of the data, and provide high-quality input for the subsequent optimization process.

[0130] Step 2: Generate a global scheduling solution, such as Figure 3 As shown in the figure, the optimization algorithm module first uses the quantum particle swarm algorithm (QPSO) for global optimization. In the multidimensional parameter space, QPSO searches for the global objective function of port container scheduling and generates the optimal initial scheduling plan. The plan includes loading and unloading sequence, equipment allocation, and task priority, with the goal of minimizing total operation time and equipment energy consumption. QPSO continuously optimizes the scheduling plan by iteratively updating the particle position until the global optimal solution is found or the termination condition is reached.

[0131] Step 3: Local optimization and dynamic adjustment, based on the initial scheduling solution generated by QPSO, such as Figure 3As shown in the figure, the system further uses the deep deterministic policy gradient algorithm (DDPG) for local optimization. DDPG builds a deep neural network to analyze the current equipment status and task progress of the port in real time, and dynamically adjusts task allocation and loading and unloading order to improve the flexibility and adaptability of the scheduling plan. With the support of real-time data, the DDPG algorithm continuously optimizes the scheduling strategy for emergencies such as equipment failures and traffic jams to ensure the efficient completion of scheduling tasks.

[0132] Step 4: Task execution and monitoring, such as Figure 4 As shown in the figure, the task scheduling module converts the optimized scheduling plan into specific operation instructions and sends them to the equipment control system through standardized interfaces (such as RS485, Ethernet, Wi-Fi and 5G). After receiving the instructions, each device (such as quay crane, yard crane and AGV) performs the container loading and unloading tasks as planned. The real-time monitoring module continuously monitors the task execution and equipment status to ensure that the scheduling plan is implemented as planned and records the task completion status.

[0133] Step 5: Exception handling and feedback optimization, such as Figure 5 As shown in the figure, when the real-time monitoring module detects equipment failure, task delay or other abnormal conditions, the abnormal information is fed back to the optimization algorithm module. Based on the abnormal feedback, the DDPG algorithm adjusts the scheduling strategy in real time, reallocates tasks and optimizes the operation path to ensure the continuity and efficiency of container loading and unloading tasks. The system dynamically adjusts the strategy through the feedback mechanism to adapt to the real-time changing environment of the port.

[0134] Step 6: Result analysis and adjustment, such as Figure 6 As shown in the figure, the user interface module provides operators with a scheduling information display and control operation interface, showing the current task status and equipment status. Operators can view the scheduling execution results in real time through the interface, and manually intervene in abnormal situations or adjust scheduling parameters. The system further adjusts the optimization strategy based on the scheduling execution results and the feedback from operators, providing improvement references for subsequent scheduling tasks. Through continuous monitoring and feedback adjustment, the efficient operation and stability of the port scheduling system are ensured.

[0135] In summary, the present invention combines the QPSO and DDPG double-layer optimization algorithms (such as Figure 3 As shown in Figure 2), the global and local dynamic optimization of port container scheduling is realized, and the optimal scheduling plan is quickly generated and adjusted in real time (as shown in Figure 2). Figure 5 As shown in the figure, it significantly improves equipment utilization, shortens operation time, reduces energy consumption, and enhances the dynamic adaptability and robustness of the scheduling strategy.

[0136] The data acquisition module of the present invention integrates a variety of sensors and data interfaces (such as Figure 2As shown in Figure 1), it can obtain various data of equipment and environment in the port in real time, providing accurate input data for the optimization algorithm. The close integration of the data acquisition module and the optimization algorithm module (as shown in Figure 1) Figure 3 As shown in the figure, it improves the scientificity and reliability of scheduling decisions and ensures the stability of the system in complex environments.

[0137] The optimization algorithm module of the present invention introduces a feedback mechanism (such as Figure 5 As shown in Figure 1), it can continuously evaluate and optimize scheduling strategies, dynamically respond to emergencies in port operations, such as equipment failures and task delays, and ensure the stability and continuity of container loading and unloading tasks. This design improves the intelligence of the system and ensures the efficient execution of scheduling tasks.

[0138] The real-time monitoring module and the task scheduling module work together (such as Figure 4 and Figure 5 As shown in the figure, it realizes intelligent monitoring and dynamic optimization of the dispatching process, can handle abnormal situations in the dispatching process in a timely manner, ensure the efficient execution of the dispatching plan and the stable operation of the system, and effectively improve the overall operational efficiency of the port.

[0139] The user interface module provides real-time information display and control operation functions (such as Figure 6 The module enables operators to understand the scheduling execution status in real time and intervene and adjust when necessary. This module improves the operational flexibility and response speed of the system through a visual interface and user-friendly operation design.

[0140] The present invention uses standardized interfaces to achieve efficient data exchange and system integration between modules (such as Figure 1 As shown in the figure, it ensures the stability and accuracy of data transmission and supports the rapid deployment and maintenance of the system. Its modular design and interface application enable the system to quickly adapt to the needs of different port environments and significantly improve the level of port management and intelligence.

Claims

1. A port container dispatching method, characterized in that: The steps include: The quantum particle swarm algorithm is used to analyze the pre-processed real-time data in the port, search for the optimal solution in the multi-dimensional parameter space, and generate the initial scheduling plan; Based on the initial scheduling plan, the deep deterministic policy gradient algorithm is used to locally optimize the scheduling strategy in combination with real-time data to generate an optimized scheduling plan; Convert the optimized scheduling plan into specific operation instructions and send them to the equipment control system, instructing each device to carry out container loading and unloading tasks as planned; Monitor the real-time status of the port and perceive changes in the real-time status of the port. When abnormal changes occur, use the deep deterministic policy gradient algorithm to re-optimize the scheduling strategy based on abnormal feedback, dynamically adjust the task scheduling plan, and update the operation instructions; The quantum particle swarm algorithm searches for the optimal solution in the multidimensional parameter space including: The size N of the particle swarm, the maximum number of iterations, and the parameters of the quantum behavior model are set. The initial position of each particle represents a possible global scheduling scheme, which includes the initial equipment allocation for each task. The fitness value of each particle is calculated to minimize the global objective function: in, is the total duration of all loading and unloading tasks, is the total energy consumption of all devices, and are the weight coefficients of time and energy consumption respectively; Based on the quantum behavior model, update the position of each particle: in, Indicates The particle in The current position of the iteration, is the historical optimal position of the particle, is the global optimal position, is a random number used to introduce randomness and diversity in the search, β is a constant, which represents the adjustment factor of the step between the global optimal solution and the current position of the particle during the particle update process; The process continues to iterate until the preset maximum number of iterations is reached. Or until the fitness value is no longer significantly improved, the final output global optimal solution is used as the initial scheduling solution; The local optimization scheduling strategy using the deep deterministic policy gradient algorithm includes: Define the structural parameters of the Actor and Critic networks, including the number of nodes in the input layer, hidden layer, and output layer, and initialize the maximum capacity of the experience pool , learning rate and discount factor , set the initial weights of the Actor network and the Critic network to ensure that the network is gradually optimized during the learning process; Actor network based on current state Output Scheduling Strategy , the Critic network estimates the strategy output by the Actor network The value function , and based on the actual rewards after implementation To update the policy; at each time step , the system samples the current state , the Actor network outputs the corresponding scheduling action , then execute the action and observe the new state and instant rewards : in, It is a weighting factor used to weigh different objectives during the optimization process; Cost function representing task delay; empirical samples It is stored in the experience pool D, from which the system randomly extracts batch samples and uses the gradient descent method to update the parameters of the Actor and Critic networks. The Critic network uses the time difference error to update its parameters: in, It is the loss function of the Critic network, which is used to evaluate the gap between the output of the Critic network and the actual reward; It’s an instant reward. is the discount factor, are the parameters of the Critic network, is the target value of the next state, is the current value estimate; The Actor network updates parameters through the policy gradient method: in, is the gradient of the objective function of the Actor network parameter optimization, representing the strategy function About Actor Network Parameters The gradient of are the parameters of the Actor network, Represents the gradient of the Actor network parameters; N is the number of samples, that is, the number of samples in a batch; is the value function of the Critic network Actor Network Parameters The gradient of The global optimal solution output by the quantum particle swarm algorithm is used as the initial input, and the local optimization process continues until the maximum number of iterations is reached. Or until the reward function converges, the optimized local scheduling strategy is output.

2. The port container dispatching method according to claim 1, characterized in that: The real-time data in the port includes the equipment status, task progress, traffic flow and environmental status in the port. The preprocessing method includes denoising, formatting and standardization. The initial scheduling plan includes the loading and unloading sequence of containers, equipment allocation and task priority, with the goal of minimizing total operation time, minimizing equipment energy consumption and maximizing equipment utilization.

3. The port container dispatching method according to claim 1, characterized in that: The real-time status of the port includes equipment failure information, weather change information, task progress information and task priority information; When the change is abnormal, the deep deterministic policy gradient algorithm is used to re-optimize the scheduling strategy based on the abnormal feedback, including: When equipment failure, task delay or other abnormal conditions are detected, the abnormal information is fed back to the deep deterministic policy gradient algorithm to locally optimize the scheduling strategy, adjust the scheduling strategy in real time, reallocate tasks and optimize the operation path; If the deep deterministic policy gradient algorithm detects that the initial solution generated by the quantum particle swarm algorithm deviates significantly from the actual operating environment, the information is passed back to the quantum particle swarm algorithm through the feedback mechanism, triggering the re-search and global optimization process of the quantum particle swarm algorithm.

4. The port container dispatching method according to claim 1, characterized in that: The following steps are also included: Provide operators with a scheduling information display and control operation interface to show the current task status and equipment conditions; operators can view the scheduling execution results in real time through the interface, and manually intervene in abnormal situations or adjust scheduling parameters; based on the scheduling execution results and operator feedback, further adjust the optimization strategy to provide reference and improvement for subsequent scheduling tasks.

5. A port container dispatching system, characterized in that: It includes data acquisition module, task scheduling module, optimization algorithm module, real-time monitoring module and user interface module; The data acquisition module is used to collect real-time data from port equipment and environmental sensors; The optimization algorithm module includes the QPSO algorithm submodule for global optimization and the DDPG algorithm submodule for local dynamic optimization. Based on the real-time data obtained by the data acquisition module, it adopts a method combining the quantum particle swarm algorithm and the deep deterministic policy gradient algorithm to achieve global optimization and local dynamic adjustment of port container scheduling tasks. The task scheduling module is used to execute the scheduling plan output by the optimization algorithm module and perform specific task allocation and equipment scheduling operations; The real-time monitoring module is used to monitor the device status and task execution, handle exceptions and dynamically adjust the scheduling strategy; The user interface module is used to display scheduling information, monitor status and receive user instructions; The QPSO algorithm for global optimization includes: The size N of the particle swarm, the maximum number of iterations, and the parameters of the quantum behavior model are set. The initial position of each particle represents a possible global scheduling scheme, which includes the initial equipment allocation for each task. The fitness value of each particle is calculated to minimize the global objective function: in, is the total duration of all loading and unloading tasks, is the total energy consumption of all devices, and are the weight coefficients of time and energy consumption respectively; Based on the quantum behavior model, update the position of each particle: in, Indicates The particle in The current position of the iteration, is the historical optimal position of the particle, is the global optimal position, is a random number used to introduce randomness and diversity in the search, β is a constant, which represents the adjustment factor of the step between the global optimal solution and the current position of the particle during the particle update process; The process continues to iterate until the preset maximum number of iterations is reached. Or until the fitness value is no longer significantly improved, the final output global optimal solution is used as the initial scheduling solution; The DDPG algorithm for local dynamic optimization includes: Define the structural parameters of the Actor and Critic networks, including the number of nodes in the input layer, hidden layer, and output layer, and initialize the maximum capacity of the experience pool , learning rate and discount factor , set the initial weights of the Actor network and the Critic network to ensure that the network is gradually optimized during the learning process; Actor network based on current state Output Scheduling Strategy , the Critic network estimates the strategy output by the Actor network The value function , and based on the actual rewards after implementation To update the policy; at each time step , the system samples the current state , the Actor network outputs the corresponding scheduling action , then execute the action and observe the new state and instant rewards : in, It is a weighting factor used to weigh different objectives during the optimization process; Cost function representing task delay; empirical samples It is stored in the experience pool D, from which the system randomly extracts batch samples and uses the gradient descent method to update the parameters of the Actor and Critic networks. The Critic network uses the time difference error to update its parameters: in, It is the loss function of the Critic network, which is used to evaluate the gap between the output of the Critic network and the actual reward; It’s an instant reward. is the discount factor, are the parameters of the Critic network, is the target value of the next state, is the current value estimate; The Actor network updates parameters through the policy gradient method: in, is the gradient of the objective function of the Actor network parameter optimization, representing the strategy function About Actor Network Parameters The gradient of are the parameters of the Actor network, Represents the gradient of the Actor network parameters; N is the number of samples, that is, the number of samples in a batch; is the value function of the Critic network Actor Network Parameters The gradient of The global optimal solution output by the quantum particle swarm algorithm is used as the initial input, and the local optimization process continues until the maximum number of iterations is reached. Or until the reward function converges, the optimized local scheduling strategy is output.

6. The port container dispatching system according to claim 5, characterized in that: The optimization algorithm module continuously evaluates and adjusts the effectiveness of the scheduling strategy by comparing the scheduling execution results with the expected goals; when deviations or scheduling inefficiencies are detected, the system automatically triggers the optimization process and adjusts the scheduling plan in real time; When the task execution deviation or device status abnormality is detected, the task scheduling module can reallocate or adjust the task according to the real-time feedback information.

7. The port container dispatching system according to claim 5, characterized in that: The real-time monitoring module is connected to the data acquisition module to obtain the equipment status, task progress and environmental data in real time, analyze these data and identify abnormal situations during task execution; the real-time monitoring module collaborates with the task scheduling module to feed back abnormal situations to the optimization algorithm module to trigger corresponding adjustment strategies; the real-time monitoring module can also dynamically adjust the task scheduling plan and update the equipment instructions according to the changes in real-time data to ensure the stability of task execution and the effectiveness of the scheduling strategy; the real-time monitoring module realizes intelligent monitoring and dynamic optimization of the scheduling process through a closed-loop feedback control mechanism.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images