Construction method and scheduling method of yard intelligent container receiving gravity model and storage medium

Abstract

The application provides a construction method of a yard intelligent container receiving attraction model, comprising: establishing a UCT-MCTS framework; performing symbolic modeling based on the UCT-MCTS framework; performing target driving design based on the UCT-MCTS framework, and constructing a single-outlet container positioning total target function related to maximum container attraction, minimum repulsion and maximum yard tool attraction; performing hypothesis verification based on the UCT-MCTS framework, and determining the applicable conditions of the corresponding model; and performing constraint design based on the UCT-MCTS framework. The application also provides a scheduling method, which optimizes the positioning of outlet container storage of a container terminal through a yard intelligent container receiving attraction model, wherein the yard intelligent container receiving attraction model is a UCT-MCTS framework constructed by the construction method described above. The application also provides a computer readable storage medium for executing the above method. Through the construction of a multi-dimensional attraction model and an optimization algorithm, the application realizes accurate and efficient container positioning decision-making, and improves the operation efficiency of a container port.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of port intelligent scheduling and control, in particular to a construction method and scheduling method of a yard intelligent container receiving attraction model and a storage medium, which are suitable for export container stacking position optimization of a container terminal. BACKGROUND

[0002] With the rapid development of global container transportation industry, as a key hub of the transportation chain, the intelligent transformation of the port has become an inevitable trend to improve the core competitiveness. In order to realize the construction requirement of "smart port", the port enterprise needs to comprehensively improve the operation efficiency and management level. Among the four core businesses of the container terminal (loading, unloading, port gathering and container picking), the container port business of the export container is faced with significant scheduling challenges due to the characteristics of "scattered port gathering and concentrated loading". Before the ship arrives, the cargo information provided by the ship owner is often not accurate and comprehensive, and the loading operation is strictly time and space constrained, which makes it difficult to develop the export container plan.

[0003] In the prior art, the traditional position selection operation and decision support system cannot meet the development needs of the current shipping industry. Specifically, the wide application of super large container ships makes the number of port gathering containers increase exponentially, and the traditional method has defects such as high container turnover rate, low yard crane operation efficiency and insufficient yard utilization rate when dealing with the position selection problem in a high-density yard environment. For example, when the yard density exceeds 60%, the traditional method cannot improve the utilization rate while considering the loading efficiency, which leads to the extension of the ship waiting time and the increase of the port operation cost.

[0004] In addition, the existing technology lacks a systematic optimization model when dealing with multi-dimensional constraints (such as weight matching, size consistency, voyage purity, etc.), and cannot achieve multi-objective balance (such as maximizing attraction, minimizing repulsion and improving site tool utilization rate). Therefore, an intelligent position selection scheme that can adapt to complex constraint conditions and consider multi-objective optimization is needed to realize the efficient operation of the container terminal. SUMMARY

[0005] In view of the low efficiency, high container turnover rate and insufficient multi-objective optimization of the existing technology in the export container position selection, the present application proposes a construction method and scheduling method of a yard intelligent container receiving attraction model and a storage medium, which realizes accurate and efficient container position selection decision through the construction of a multi-dimensional attraction model and an optimization algorithm, thereby improving the operation efficiency of the container port.

[0006] To achieve the above purpose, the present application adopts the following technical solutions:

[0007] The present application proposes a construction method of a yard intelligent container receiving attraction model, which comprises:

[0008] Establish a UCT-MCTS framework for export container unloading decision-making;

[0009] Symbolic modeling based on the UCT-MCTS framework, defining a set of model dimensions, a set of model parameters, and decision variables x ijk ; The model dimensions include the to-be-landed export container, the in-landed container, the attraction mode, and the berth-related parameters related to symbolic modeling; The model parameters include the weight, size, and matching matrix of the container related to symbolic modeling; The decision variable x ijk represents that the export container i is attracted to the in-landed container j in the attraction mode k.

[0010] Target-driven design based on the UCT-MCTS framework, constructing a single-export container location selection total objective function related to maximizing container attraction, minimizing repulsion, and maximizing yard tool attraction;

[0011] Hypothesis verification based on the UCT-MCTS framework, specifying the applicable conditions of the corresponding model;

[0012] Constraint design based on the UCT-MCTS framework, corresponding constraint conditions including weight constraints, height constraints, and size constraints related to export containers.

[0013] Further, the model parameters include the weight of the export container, the size of the export container, the size matching matrix of the in-landed container, the weight matching matrix , the ship voyage matching matrix , the container height matching matrix , the unloading port matching matrix , the in-landed container height , the maximum height of the yard , the attraction mode preference factor , the repulsion factor , the tonnage difference limit , the container area operation status matrix , the gantry availability matrix , the berth distance , the number of open berths , and the number of open berths ;

[0014] The weight of the export container is used for weight matching constraint and objective function calculation;

[0015] The size of the export container is used for size consistency constraint;

[0016] Weight matching matrix As a quantification of the degree of weight matching between the to-be-entered container and the in-container, when the weight matching matrix The higher the value, the higher the matching degree;

[0017] Ship voyage matching matrix Characterize the consistency of the ship voyage between the to-be-entered container and the in-container, and use it to perform voyage purity constraint;

[0018] Container height matching matrix Used to determine whether the height of the to-be-entered container and the in-container is consistent, thereby avoiding mixed stacking;

[0019] Discharging port matching matrix Used to identify whether the discharging port of the to-be-entered container and the in-container is the same, thereby facilitating grouping and stacking management;

[0020] In-container layer height As the number of stacked layers of the current in-container j;

[0021] Maximum yard layer height As the maximum number of stacked layers allowed by the yard, used for layer height constraint;

[0022] Attractive mode preference factor Used to quantify the priority of the kth attractive mode;

[0023] Repulsion factor Used to quantify the repulsion strength of the kth attractive mode;

[0024] Ton difference limit Used to quantify the maximum weight difference threshold allowed by the kth attractive mode;

[0025] Container area operation state matrix Used to reflect the operation state of the container area where the specified berth is located;

[0026] Yard crane availability matrix Used to reflect the idle state of the yard crane where the specified berth is located;

[0027] Berth distance As the physical distance between the berths i and j, used for inter-berth attraction weakening calculation;

[0028] Number of berths to be opened Used to calculate the theoretical number of berths to be opened according to the quantity of entered containers;

[0029] Number of berths opened As the number of currently opened berths.

[0030] Further, the single-outlet container positioning total target function is:

[0031] (1)

[0032] In formula (1), For the weight difference target, For the repulsive force target, To attract targets to the venue and its tools, , , These are the first weight, the second weight, and the third weight, respectively.

[0033] Furthermore, the first weight Second weight Third weight The AHP-entropy weight combination is determined by the AHP-entropy weight method, and the formula for the AHP-entropy weight combination is: (2)

[0034] In formula (2), =0.4, final weight = [0.56, 0.31, 0.13].

[0035] Furthermore, the exploration coefficient c of the UCT-MCTS framework adopts a dynamic adjustment strategy, and the formula for the exploration coefficient c is:

[0036] (3)

[0037] In formula (3), The initial exploration coefficients serve as the default constants of the UCB1 algorithm. As a decay factor used to control the rate of decrease of the exploration coefficient with stockpile density; The current yard density, which is the ratio of the number of occupied bays to the total number of bays, is used to reflect the degree of yard congestion. The maximum preset density is the maximum density, meaning that when the stockpile density is close to this value, the exploration coefficient approaches 0.

[0038] This invention further proposes a scheduling method that optimizes the location of export container stacking at a container terminal using a yard intelligent container receiving gravity model. This yard intelligent container receiving gravity model is a UCT-MCTS framework constructed using any of the methods described above. The scheduling method includes:

[0039] The selection process involves comparing attractiveness based on the α-RAVE algorithm and selecting the attraction method and location with the highest attractiveness.

[0040] Execute the expansion process: generate new selected child nodes;

[0041] Performing simulation process: simulate the location selection by default strategy until termination;

[0042] Performing back propagation process: updating node reward and access times.

[0043] Further, the node value calculation formula of the α-RAVE algorithm is:

[0044] (4)

[0045] In formula (4), is the current action value, is the historical fast evaluation value; is the fusion coefficient.

[0046] Further, the fusion coefficient of the α-RAVE algorithm Performing dynamic adjustment, the fusion coefficient formula is:

[0047] (5)

[0048] In formula (5), , and when < 10 , when > 50 .

[0049] Further, in the performing simulation process, a constraint priority hierarchical strategy is adopted, first filtering the positions that do not meet the weight, size, and layer height constraints, and then performing roulette selection according to the attraction degree distribution probability, the formula of which is:

[0050] (6)

[0051] In formula (6), the location selection priority calculated based on the total objective function is is the attraction degree score of action ; is a temperature parameter for controlling the randomness of selection; in formula (6), when is higher, the probability of selecting low attraction degree actions is higher, and when is lower, the selection is more biased towards high attraction degree actions; is the compliance candidate set, which is used as the available position set after filtering the weight, size, and layer height constraints.

[0052] Further, in the performing back propagation process, a time discount factor is introduced, and the action value updating formula is

[0053] (7)

[0054] In formula (7), is a learning rate, used to control the value update step size, to avoid parameter oscillation; is a node depth, used to calculate the degree of return decay when the current action is in the level of the search tree; is the return when the simulation is terminated.

[0055] Further, the scheduling method further comprises:

[0056] An exception handling step is performed: when there is no compliant position, unlocking 5% to 10% of the standby berth of the total berth number of the yard or relaxing the constraint by classification; when the device fails, removing the abnormal berth, adjusting the site tool attraction target calculation logic, and increasing the iteration number from 120 to 150.

[0057] The application further provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is characterized by the following: when the computer program is executed by a processor, the steps of the construction method according to any one of the above are realized.

[0058] The application further provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is characterized by the following: when the computer program is executed by a processor, the steps of the scheduling method according to any one of the above are realized.

[0059] The application further provides a computer readable storage medium, the computer readable storage medium stores a computer program system, and the computer program system comprises:

[0060] A model construction module is used to realize the functions of UCT-MCTS framework initialization, symbolic modeling, AHP-entropy weight method weight calculation and constraint verification; in the UCT-MCTS framework initialization process, dynamic exploration coefficient calculation is also performed;

[0061] A scheduling execution module is used to realize the functions of α-RAVE selection, hierarchical simulation, discount back propagation and convergence judgment;

[0062] An exception handling module is used to realize the functions of standby berth calling, constraint relaxation, fault berth screening and iteration number adjustment;

[0063] A visualization module is used to output the results of the site selection scheme, the yard state diagram and the convergence curve;

[0064] When the computer program system is called by a processor, the container site selection decision is made according to the scheduling method according to any one of the above.

[0065] The application has the following beneficial effects:

[0066] The application improves work efficiency: through the multi-objective optimization model and the UCT-MCTS algorithm, the container turnover rate is significantly reduced (expected to be reduced by 15%-25%), the distance of the yard crane is shortened, and the loading speed is accelerated; the dynamic alpha adjustment and the hierarchical simulation further improve the positioning accuracy, and the convergence verification ensures the stability of the decision;

[0067] The application reduces operating costs: optimizing the yard utilization rate (increasing by 10%-15%), reducing the waiting time of the ship (reducing by 20%-30%); the abnormal processing mechanism avoids operation interruption, indirectly reduces emergency costs, thereby reducing the energy consumption and labor cost of the terminal;

[0068] The application enhances adaptability: it can flexibly handle complex constraints (such as size, weight, and voyage purity) in a high-density yard environment, dynamically adapt the exploration coefficient to different yard densities, and adapt the equipment failure strategy to improve system robustness;

[0069] The application provides intelligent decision-making: through the AHP-entropy weight method to objectively determine the weight and data-driven alpha adjustment, the automation and intelligence of the positioning decision are realized, human intervention is reduced, and the consistency of the decision is improved. BRIEF DESCRIPTION OF DRAWINGS

[0070] The accompanying drawings are included to provide a further understanding of the application, and are incorporated into and constitute a part of this specification. The drawings illustrate exemplary embodiments of the application, and together with the specification serve to explain the principles of the application. These drawings are merely illustrative and are not limiting on the scope of the application.

[0071] Figure 1 is a yard container grouping topology diagram in Example 5, showing grouping strategies based on size, voyage, and unloading port dimensions;

[0072] Figure 2 is a schematic diagram of the priority order of container positioning, reflecting the priority logic of different attraction modes;

[0073] Figure 3 is a diagram of the weight difference of the yard container, illustrating the impact of weight matching on the turnover rate;

[0074] Figure 4 is a diagram of the stacking of yard crane operation, showing the relationship between yard crane operation efficiency and position selection;

[0075] Figure 5 is a diagram of the stacking of yard container weight restrictions, illustrating the weight difference constraints of different attraction modes;

[0076] Figure 6 is a diagram of the stacking of yard container attraction modes, visualizing the range of different attraction modes;

[0077] Figure 7 is a diagram of the stacking of yard container size constraints, emphasizing the requirement for size consistency in the same bay;

[0078] Fig. 8 is a schematic diagram of a bay attraction force stack, showing the inter-bay positioning attraction force calculation logic;

[0079] Fig. 9 is a schematic diagram of a stacking scheme comparison stack, comparing the efficiency of the yard crane movement of different positioning schemes;

[0080] Fig. 10 is a schematic diagram of a container size stacking diagram, showing a compliant and non-compliant stacking scenario;

[0081] Fig. 11 is a UCT-MCTS algorithm convergence curve, with the horizontal axis representing the number of iterations, and the vertical axis representing the optimal positioning scheme consistency rate (%) and the comprehensive attraction degree F fluctuation amplitude, reflecting the stability of the algorithm convergence;

[0082] Fig. 12 is a dynamic alpha value change curve of the alpha-RAVE algorithm, showing the adjustment trend of alpha under different historical evaluation times, reflecting the data-driven parameter optimization logic;

[0083] Fig. 13 is a comparison diagram of the constraint filtering effect of the hierarchical simulation strategy, comparing the compliance rate of pure random simulation and constraint priority simulation, verifying the effectiveness of the hierarchical strategy;

[0084] Fig. 14 is a diagram of the availability of the bay when the equipment fails, marking the available / unavailable bay and the coverage of the yard crane, illustrating the bay selection logic for fault adaptation;

[0085] Fig. 15 is a workflow diagram of a method for constructing a container yard intelligent container receiving attraction model according to the present application. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application will be further clearly and completely described below in combination with the embodiments of the present application. It should be noted that the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0087] The terms "first", "second", "third", "fourth" and the like are used only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the definition of "first", "second", "third", "fourth" features can explicitly or implicitly include one or more features.

[0088] The following is a detailed description of embodiments of the application depicted in the accompanying drawings. The embodiments are detailed in order to convey the scope of the application. The number of details provided is not intended to limit the intended variations of the embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the application as defined by the appended claims.

[0089] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the application. It will be apparent, however, to one skilled in the art that the embodiments of the application can be practiced without some or all of these specific details.

[0090] If the specification states a component, or feature "may", "could", "can", "might", or "would" include, have, or have a feature, then it is not required for the specific component or feature to include, have, or have the feature.

[0091] As used in the description of the application and the following claims, the meanings of "a", "an", and "the" include plural references unless the context clearly dictates otherwise. Additionally, as used in the description of the application, the meaning of "in" includes "in" and "on" unless the context clearly dictates otherwise.

[0092] Embodiment 1

[0093] As Figure 15 shown:

[0094] The embodiment proposes a construction method of a yard intelligent container collection gravity model, comprising:

[0095] Establish a UCT-MCTS framework for export container collection site selection decision;

[0096] Symbolic modeling is performed based on the UCT-MCTS framework, defining a set of model dimensions, a set of model parameters, and decision variables x ijk ; the model dimensions include the to-be-landed export container, the in-landed container, the attraction mode, and the berth-related parameters related to symbolic modeling; the model parameters include the weight, size, and matching matrix of the container related to symbolic modeling; the decision variable x ijk represents that the export container i is attracted to the in-landed container j in the attraction mode k; specifically, the decision variable x ijk = 1 is selected, and 0 is not selected;

[0097] Target-driven design is performed based on the UCT-MCTS framework, and a single-export-container site selection total objective function related to maximizing container attraction, minimizing repulsion, and maximizing yard tool attraction is constructed;

[0098] Based on the UCT-MCTS framework, hypothesis verification is performed to determine the applicable conditions of the corresponding model;

[0099] Based on the UCT-MCTS framework, constraint design is performed, and the corresponding constraint conditions include weight constraints, height constraints, and size constraints related to export containers.

[0100] Optimally, the model parameters include the weight of the export container , the size of the export container , the size matching matrix of the on-site container (such as, iff (representing the size of the on-site container j) is 0), the weight matching matrix , the ship voyage matching matrix , the container height matching matrix , the unloading port matching matrix , the on-site container height , the maximum yard layer height , the attraction mode preference factor , the repulsion factor , the tonnage difference limit , the container area operation status matrix , the yard crane availability matrix , the berth distance , the number of berths to be opened , and the number of berths already opened ;

[0101] The weight of the export container (in tons) is used for weight matching constraints and objective function calculations;

[0102] The size of the export container (such as 20 feet, 40 feet) is used to perform size consistency constraints;

[0103] The weight matching matrix quantifies the degree of weight matching between the container to be entered and the on-site container, and the higher the value of the weight matching matrix , the higher the matching degree;

[0104] The ship voyage matching matrix represents the consistency of the ship voyage between the container to be entered and the on-site container, and is used to perform voyage purity constraints;

[0105] The container height matching matrix is used to determine whether the height of the container to be entered and the on-site container is consistent, thereby avoiding mixed stacking;

[0106] The unloading port matching matrix Whether the inbound container is the same as the discharge port of the container in the yard, thereby facilitating the grouping and stacking management;

[0107] Container layer height The number of stacking layers of the current container j in the yard;

[0108] Maximum yard layer height The maximum number of stacking layers allowed by the yard, used for layer height constraints (such as );

[0109] Attractive mode preference factor Used to quantify the priority of the kth attractive mode;

[0110] Repulsion factor Used to quantify the repulsion strength of the kth attractive mode;

[0111] Tonnage difference limit The maximum weight difference threshold allowed for the attractive mode k (i.e. );

[0112] Container area operation status matrix Used to reflect the operation status of the container area where the specified bay is located (for example, indicates that the container area where bay b is located is being operated);

[0113] Yard crane availability matrix Used to reflect the idle state of the yard crane at the specified bay (for example, indicates that bay b has an idle yard crane);

[0114] Bay distance As the physical distance between bay i and j, used for inter-bay attraction weakening calculation;

[0115] Number of bays to be opened Used to calculate the theoretical number of bays to be opened according to the quantity of inbound containers;

[0116] Number of opened bays The number of currently opened bays.

[0117] Optimally, the single-outlet container location selection total objective function is:

[0118] (1)

[0119] In formula (1), is the weight difference target, is the repulsion force target, is the yard tool attraction target, , , are the first weight, second weight, and third weight, respectively.

[0120] Specifically, defined as To minimize the weight difference between the boxes to be delivered and the boxes already on site, thereby reducing the risk of overturning the boxes;

[0121] When performing grading calculations, repulsive forces can arise when dimensions such as weight, container height, and port of discharge are inconsistent. ,in It is an exclusion factor;

[0122] The combination is determined by the AHP-entropy weight method, and the combination formula is as follows: ,in Final weight = [0.56, 0.31, 0.13]; For AHP subjective weights, This refers to the objective weights in the entropy weight method.

[0123] Optimally, the first weight Second weight Third weight The AHP-entropy weight combination is determined by the AHP-entropy weight method, and the formula for the AHP-entropy weight combination is:

[0124] (2)

[0125] In formula (2), =0.4, final weight = [0.56, 0.31, 0.13].

[0126] Ideally, the exploration coefficient c of the UCT-MCTS framework adopts a dynamic adjustment strategy, and the formula for the exploration coefficient c is:

[0127] (3)

[0128] In formula (3), The initial exploration coefficients serve as the default constants of the UCB1 algorithm. As a decay factor used to control the rate of decrease of the exploration coefficient with stockpile density; The current yard density, which is the ratio of the number of occupied bays to the total number of bays, is used to reflect the degree of yard congestion. The maximum preset density is defined as the exploratory coefficient approaching zero when the stockpile density is close to this value; specifically, =1.414, =0.8, =0.9, This is used to make the UCB1 algorithm more biased towards utilizing existing experience.

[0129] Example 2

[0130] Example 2 is a further optimized design of Example 1;

[0131] In this example, the specific process of building the UCT-MCTS framework using this construction method is as follows:

[0132] The site selection decision process is represented by a tree structure, where each node represents a storage yard state and edges represent site selection actions; the UCB1 formula ( ) Assess the value of nodes, among which The average revenue of the child nodes. This represents the number of visits to child nodes. This represents the total number of visits to the parent node. For dynamic exploration coefficients (according to) calculate, =1.414, =0.8, =0.9);

[0133] The formula for optimizing node selection using the UCT strategy is as follows: ,in For the value of the action, For the number of times the action is accessed, This represents the number of times the status has been accessed.

[0134] In this example, the specific process of symbolic modeling based on the UCT-MCTS framework in this construction method is as follows:

[0135] The model dimensions are as follows:

[0136] : Collection of export containers awaiting arrival. ;

[0137] : Containers are collected at the storage yard. ;

[0138] A collection of attraction methods, including precise vertical attraction and mixed horizontal attraction. one way;

[0139] : Collection of bay space in the container area ;

[0140] The model parameters are as follows:

[0141] Export box Weight, in tons;

[0142] : size of the export container (20 feet / 40 feet, etc.) ;

[0143] : size matching matrix, iff ;

[0144] : presence of the container in the current layer height, is the maximum allowed layer height of the yard;

[0145] : attraction mode of the preference factor, repulsion factor;

[0146] : distance between the bay and for inter-bay attraction weakening calculation.

[0147] In this example, the specific process of the construction method based on the UCT-MCTS framework for target-driven design is as follows: wherein the maximum container attraction is as follows:

[0148] Weight difference minimization target: ;

[0149] Wherein the minimum repulsion is as follows:

[0150] Hierarchical repulsion force calculation:

[0151] 1) When the weight, container height, unloading port, and voyage are the same: ;

[0152] 2) Only the weight is different: ;

[0153] 3) When the weight and container height are different: ;

[0154] 4) Comprehensive repulsion force: ;

[0155] Wherein the maximum yard tool attraction is as follows:

[0156] , wherein indicates that the bay has work in the container area; indicates that there is an idle yard bridge;

[0157] Wherein the total objective function is as follows:

[0158] After normalization: wherein the weight is determined by AHP-entropy weight method combination:

[0159] Specifically, AHP subjective weight: construct judgment matrix consistency check =0.052<0.1, get = [0.54, 0.32, 0.14];

[0160] Specifically, entropy weight method objective weight: based on 1000 groups of historical data to calculate information entropy, get = [0.58, 0.29, 0.13];

[0161] Specifically, combination fusion: the final ;

[0162] wherein the constraint conditions are as follows:

[0163] 1) weight constraint: wherein is the maximum ton difference allowed by the attraction mode k;

[0164] 2) layer height constraint: ;

[0165] 3) size constraint: .

[0166] Example 3

[0167] Example 3 is a further optimization of Example 1 or Example 2;

[0168] When the berth selection cannot meet the constraints, the inter-behrt selection model is started; the inter-behrt selection model needs to meet three conditions: performing a maximum inter-behrt attraction, performing a total inter-behrt objective function, and performing a constraint condition combination.

[0169] wherein performing a maximum inter-behrt attraction is specifically as follows:

[0170] voyage and unloading port purity calculation: wherein is the set of present boxes of the box area where the berth b is located;

[0171] distance weakening function: wherein: : calculated according to the layout of the terminal berth, ( is the distance between adjacent berth centers, and the example terminal =15m); : calibrated through field bridge operation experiment, = 0.08 (the attraction force weakens about 30% with every 5m increase in distance);

[0172] Inter-be attraction force: .

[0173] The inter-be total objective function is executed as follows: wherein , verified by AHP-entropy weight method.

[0174] The execution of the combination of constraint conditions includes:

[0175] Be site quantity constraint: the number of opened be sites wherein is the number of future 2-hour inbound containers predicted by the ARIMA model (error < 10%), , is the maximum number of containers in a single be site (example = 8);

[0176] Be site availability constraint: and ( indicates that the be site is empty, indicates that it is not locked).

[0177] Embodiment 4

[0178] As shown in Figures 2-10 and Figures 12-14 :

[0179] The embodiment proposes a scheduling method for optimizing the export container stacking of a container terminal through a stack intelligent container attraction model. The stack intelligent container attraction model is a UCT-MCTS framework constructed by the construction method described in any one of the technical solutions in Embodiment 1 or Embodiment 2 or Embodiment 3. The scheduling method includes:

[0180] Execution of the selection process: based on the α-RAVE algorithm, compare the attraction degrees, and select the attraction mode and position with the largest attraction degree;

[0181] Execution of the expansion process: generate a new site selection child node;

[0182] Execution of the simulation process: simulate the site selection by default strategy until termination;

[0183] Execution of the backpropagation process: update the node revenue and access times.

[0184] Optimally, the node value calculation formula of the α-RAVE algorithm is:

[0185] (4)

[0186] In formula (4), is the current action value, is the historical fast evaluation value; is the fusion coefficient.

[0187] Specifically, the fusion coefficient is dynamically adjusted according to the formula: , , and when when . is the number of historical fast evaluation times, is the current action access times.

[0188] Optimally, the fusion coefficient of the α-RAVE algorithm is dynamically adjusted, and the formula of the fusion coefficient is:

[0189] (5)

[0190] In formula (5), , and when < 10 , when > 50 .

[0191] Optimally, a constraint-priority hierarchical strategy is adopted, that is, first filter the positions that do not meet the weight, size, and layer height constraints, and then perform roulette selection according to the probability of attraction degree distribution. The formula of the probability is:

[0192] (6)

[0193] In formula (6), the site selection priority is calculated based on the total objective function, is the attraction degree score of the action ; is a temperature parameter for controlling the randomness of selection. In formula (6), when is higher, the probability of selecting a low-attraction-degree action is higher, and when is lower, the selection is more biased towards high-attraction-degree actions. is a compliance candidate set, which is used as the available position set after filtering the weight, size, and layer height constraints.

[0194] Preferably, in formula (6), the temperature parameter is initially 1.2 and later is 0.5.

[0195] ​​Optimally, a time discount factor is introduced in the execution of the back propagation process , and the action value update formula is

[0196] (7)

[0197] In formula (7), is the learning rate, used to control the value update step size and avoid parameter oscillation; is the node depth, used to calculate the degree of return decay when the current action is in the search tree; is the return when the simulation is terminated.

[0198] Preferably, in formula (7), = 0.95, = 0.1; Specifically, it is converted according to the loading efficiency x 0.6 + the optimization of the box turnover rate x 0.4, and finally the R value ranges from 0 to 1;

[0199] Optimally, the scheduling method further comprises:

[0200] An exception handling step: when there is no compliance position, unlock 5% to 10% of the standby berth of the total berth number or relax the constraint by classification; when the device fails, remove the abnormal berth, adjust the site tool attraction target calculation logic, and increase the iteration number from 120 to 150.

[0201] Embodiment 5

[0202] Embodiment 5 is an optimized design of any of the technical solutions in Embodiment 4;

[0203] As shown in Figures 1-14 :

[0204] In Figure 14 , the fault-adaptive berth screening logic: mainly based on the coverage range of the yard crane, the priority of the berth (high-priority berth priority logic), the state of the device (preferentially selecting the yard with normal equipment), and further in the order of: within the coverage range of the yard crane > the operation state of the berth (idle priority) > the cost of box turning (low box turnover rate priority) berth.

[0205] In this example, the specific process of the execution selection of the scheduling method is as follows:

[0206] S1, first calculate the comprehensive value of each node based on the a-RAVE algorithm: , wherein is dynamically adjusted ;

[0207] When , , the current action value is prioritized;

[0208] When , , the historical experience reference is enhanced.

[0209] S2, compare the attractiveness of each candidate position ; S2 includes S21-S23:

[0210] S21, if the maximum and the second largest difference < 0.05, start secondary screening:

[0211] S22, calculate the time of the bridge moving ( =30m / min), select the smallest position;

[0212] S23, if consistent, retrieve the box flipping record of the past 1 month, and select the position with the lowest flipping rate.

[0213] In this example, the specific process of the execution expansion of the scheduling method is as follows:

[0214] For nodes that are not fully expanded (number of child nodes < number of selectable attractive ways), generate new child nodes (corresponding to new positioning actions), initialize the access times , cumulative action value , historical rapid evaluation value of the new nodes.

[0215] In this example, the execution simulation process of the scheduling method is specifically a hierarchical simulation process with constraint priority:

[0216] The first layer is: constraint filtering, eliminating positions that do not meet the following conditions to form a compliant candidate set : weight constraint: ; size constraint: ; layer height constraint: ; if is empty, temporarily increase by 10% (maximum not more than 2 tons) and re-screen;

[0217] The second layer is: probability selection, assigning probabilities to positions according to attractiveness: , where is a temperature parameter (before 50 times of simulation =1.2, after 50 times =0.5); select the roulette wheel according to P(a) and generate a simulated path until the yard status is stable (no new containers arrive for three consecutive steps).

[0218] In this example, the specific process of backpropagation in the execution of this scheduling method is as follows:

[0219] Let the profit R at the end of the simulation be (converted according to loading efficiency × 0.6 + container turning rate optimization × 0.4, ranging from [0,1]), and update the following for all nodes (s,a) on the path: Number of visits: Cumulative action value:

[0220] ( , (Node depth); historical fast evaluation values: .

[0221] In this example, the scheduling method also includes algorithm convergence verification, specifically including:

[0222] Experimental conditions were determined as follows: 500 export containers were simulated entering the site, with a stockpile density of 0.6, 10 bays, and 3 attraction methods. The optimal positioning scheme was recorded every 10 iterations. value;

[0223] Determine the convergence criterion: the consistency rate of the scheme is ≥90% for 20 consecutive iterations and Fluctuation range < 0.03;

[0224] Results: After 120 iterations, the convergence criterion was met, and the consistency rate of the proposed solutions was 92%. The fluctuation range is 0.028, proving the algorithm is stable (see...). Figure 11 ).

[0225] In this example, the scheduling method also includes an execution exception handling mechanism after the algorithm convergence verification, including:

[0226] The first step is the downgrade strategy when there is no compliant position; the first step includes sub-steps S10-S20:

[0227] S10, call the spare bay number library (5% of the preset total number of bays in the yard), check the power supply and yard bridge accessibility of the spare bays, and add them if available. Recalculate ;

[0228] S20, if the spare bit is unavailable, relax the constraints according to priority; where, first priority: reserved. ,Will The first priority is to increase 20% and the second priority is to allow 20 feet to be mixed with 40 feet (the same layer needs to be consistent), and the UCT-MCTS positioning is re-executed;

[0229] The second step is an adaptive strategy when the equipment fails. The second step includes the sub-steps S100-S300:

[0230] S100, when the bridge fails ( =0) or the bay sensor is abnormal, the corresponding bay is marked as "temporarily unavailable" and is removed from ;

[0231] S200, adjust the site tool attraction target: ( is the set of available bays);

[0232] S300, increase the number of iterations from 120 to 150 to ensure that the optimal solution is found in the reduced candidate set.

[0233] Embodiment 6

[0234] The embodiment provides a computer readable storage medium storing a computer program, characterized in that: the computer program is executed by a processor to realize the steps of the construction method according to any one of the technical solutions in the embodiments 1 or 2 or 3.

[0235] The embodiment further provides another computer readable storage medium storing a computer program, characterized in that: the computer program is executed by a processor to realize the steps of the scheduling method according to any one of the technical solutions in the embodiments 4 or 5.

[0236] The embodiment further provides a computer readable storage medium storing a computer program system, which comprises:

[0237] A model construction module is configured to realize UCT-MCTS framework initialization, symbolic modeling, AHP-entropy weight calculation and constraint verification functions. In the UCT-MCTS framework initialization process, dynamic exploration coefficient calculation is also performed.

[0238] A scheduling execution module is configured to realize α-RAVE selection, hierarchical simulation, discount back propagation and convergence judgment functions.

[0239] An exception handling module is configured to realize backup bay calling, constraint relaxation, fault bay screening and iteration number adjustment functions.

[0240] A visualization module is configured to output a positioning scheme and a yard state diagram (such as Figures 1-14) and convergence curve, etc.

[0241] The computer program system is called by the processor, and the container positioning decision is executed according to the scheduling method in any one of the technical solutions in the embodiment 4 or the embodiment 5, so that the response time is less than 2 seconds (single container positioning).

[0242] Further optimization, the model construction module executes the steps of the construction method in any one of the technical solutions in the embodiment 1 or the embodiment 2 or the embodiment 3.

[0243] Embodiments of the application include various steps, which will be described below. These steps can be executed by hardware components, or can be included in machine-executable instructions, which can be used to program general-purpose or special-purpose processors using instructions to perform these steps. Alternatively, steps can be performed by a combination of hardware, software, and firmware, and / or human operators.

[0244] The various methods described herein can be practiced by combining one or more machine-readable storage media containing code in accordance with the present application with appropriate standard computer hardware to execute the code contained therein. An apparatus for practicing various embodiments of the present application can include one or more computers (or one or more processors within a single computer) and a network-accessible storage system containing or having access to computer programs coded in accordance with the various methods described herein, and the method steps of the present application can be accomplished through modules, routines, subroutines, or subparts of the computer program product.

[0245] Embodiments of the present application can be provided as a computer program product, which can include a machine-readable storage medium having instructions tangibly embodied on it which can be used to program a computer (or other electronic device) to perform a process. The terms "machine-readable storage medium" or "computer-readable storage medium" includes, but is not limited to, fixed (hard- drive) drives, magnetic tape, floppy diskettes, optical diskettes, compact disk read-only memories (CD-ROMs) and magneto-optical disk drives, semiconductor memories, such as ROMs, PROMs, random access memories (RAMs), programmable read-only memories (PROMs), erasable PROMs (EPROMs), eMMC, electrically erasable PROMs (EEPROMs), SSDs, SD, flash memory, magnetic or optical cards, or other types of media / machine-readable medium suitable for storing electronic instructions (for example, computer programming code, such as software or firmware). The machine-readable medium can include a non-transitory medium where data can be stored and not a carrier wave or other transitory electronic signal propagating wirelessly or over wired connections. Examples of a non-transitory medium can include, but are not limited to, magnetic disks or magnetic tapes, optical storage media such as compact disks (CDs) or digital versatile disks (DVDs), flash memory, memories, or memory devices. The computer program product can include code and / or machine-executable instructions that can represent a procedure, a function, a subprogram, a program, a routine, a subroutine, a module, a software package, a class, or any combination of instructions, data structures, or program statements. A code segment can be coupled to another code segment or a hardware circuit by passing and / or receiving information, data, variables, parameters, or memory contents. Information, variables, parameters, data, etc. can be passed, forwarded, or transmitted by any suitable means including memory sharing, message passing, token passing, network transmission, etc.

[0246] The systems depicted in some of the figures can be provided in various configurations. In some embodiments, the systems can be configured as distributed systems in which one or more components of the system are distributed across one or more networks in a cloud computing system.

[0247] The above-described embodiments are merely illustrative of several embodiments of the present application, which describe in more detail and specifically, but can not be interpreted as limiting the scope of the patent of the present application. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the scope of protection of the patent of the present application should be subject to the appended claims.

Claims

1. A method for constructing a yard intelligent container collecting gravitational model, characterized in that, The application relates to a container yard intelligent container positioning method and system. The UCT-MCTS framework is established for export container positioning decision-making. based on the UCT-MCTS framework for symbolic modeling, defining a set of model dimensions, a set of model parameters and decision variables x ijk ; the model dimensions include the to-be-entered export container, the in-place container, the attraction mode and the berth-related parameters related to symbolic modeling; the model parameters include the weight, size and matching matrix of the container related to symbolic modeling; the decision variables x ijk represent that the export container i is attracted to the in-place container j in the attraction mode k; A target-driven design is performed based on the UCT-MCTS framework to build a single-export-container-positioning total target function related to maximizing container attraction, minimizing repulsion and maximizing yard tool attraction. Hypothesis verification is performed based on the UCT-MCTS framework to determine the applicable conditions of the corresponding model. Constraint design is performed based on the UCT-MCTS framework, and the corresponding constraint conditions include weight constraints, height constraints and size constraints related to export containers. wherein the tree structure is used to represent the yard space selection decision process, each node represents a yard state, and the edge represents the yard space selection action; the node value is evaluated by the formula wherein is the average income of the child node, is the access frequency of the child node, is the total access frequency of the parent node, is the dynamic exploration coefficient, calculated according to ; wherein the UCT policy optimizes the node selection formula as follows: wherein is the action value, is the action visit count, is the state visit count; as an initial exploration coefficient of the default constant of the algorithm, as an attenuation factor for controlling the rate of decline of the exploration coefficient with the yard density; as the current yard density, i.e. the ratio of the number of occupied berths to the total number of berths, for reflecting the degree of yard congestion; is the maximum preset density; The single-export-container-positioning total target function is as follows: （1） In formula (1), is a weight difference target, is a repulsion force target, is a field tool attraction target, , , are a first weight, a second weight, and a third weight, respectively.

2. The method for constructing a yard intelligent container collecting attraction model according to claim 1, characterized in that, said model parameters include weight of export container , size of export container , size matching matrix with on-hand containers , weight matching matrix , ship voyage matching matrix , container height matching matrix , discharge port matching matrix , on-hand container layer height , maximum yard layer height , attraction mode preference factor , repulsion factor , tonnage difference limit , container area operation status matrix , reach stacker availability matrix , berth distance , number of berths to be opened , and number of berths already opened ; Weight of the export container For weight matching constraints and objective function calculation; Size of export container for performing size consistency constraints; Weight matching matrix As a quantification of the degree of weight matching of the incoming box to the boxes already present, the weight matching matrix The higher the value of the weight matching matrix the higher the degree of matching. vessel voyage matching matrix characterizing the vessel voyage consistency of the to-be-entered box with the in-box, for performing voyage purity constraints; Case height matching matrix For judging whether the height of the to-be-entered case is consistent with the height of the in-case, so as to avoid mixed stacking; Discharge port matching matrix For identifying whether the discharge port of the to-be-entered box is same as that of the in-box, thereby facilitating grouping stacking management; Floor height of the presence box is the current number of stacked floors of the presence box j; Maximum yard layer height Maximum number of stacks allowed at the yard as a layer height constraint; Attractiveness mode preference factor for quantifying the priority of the kth attractiveness mode; repulsion factor for quantifying the repulsion strength of the kth attractive mode; Tonnage difference limit Maximum weight difference threshold allowed for the attraction mode k; Box zone operation status matrix for reflecting the operation status of the box zone where the designated berth is located; Field bridge availability matrix for reflecting the idle state of the field bridge where the designated berth is located; Shell position distance As the physical distance between shell positions i and j, used for inter-shell attraction weakening calculation; Number of open bins Theoretical number of open bins for calculation of approach box volume number of open bins as the number of currently open bins.

3. The method for constructing a yard intelligent container collecting attraction model according to claim 1, characterized in that, The first weight The second weight The third weight Determined by AHP-entropy weight method combination, and the formula of AHP-entropy weight method combination is: （2） In equation (2), = 0.4, the minimum weight = [0.56, 0.31, 0.13].

4. A scheduling method characterized by, The export container stacking of a container terminal is positioned and optimized through a container yard intelligent container positioning model, which is constructed by the construction method in any one of claims 1-3. An expansion process is performed to generate new positioning child nodes. A simulation process is performed to simulate positioning until termination through a default strategy. A back propagation process is performed to update node income and visit times. The node value calculation formula of the alpha-RAVE algorithm is as follows:

5. The scheduling method of claim 4, wherein, In the simulation process, a constraint-priority hierarchical strategy is adopted to first filter positions that do not meet weight, size and height constraints, and then perform roulette selection according to attraction degree distribution probability, and the probability formula is as follows: （4） In equation (4), is the current action value, is the historical fast evaluation value; is the fusion coefficient.

6. The scheduling method of claim 5, wherein, The fusion coefficient of the α-RAVE algorithm The dynamic adjustment is performed, and the fusion coefficient formula is: （5） In Equation (5), , and when < 10 , when > 50 .

7. The scheduling method of claim 4, wherein, The scheduling method further includes: （6） In Equation (6), the placement priority is calculated based on the total objective function, The attractiveness score of the action . The temperature parameter is used to control the selection randomness. In Equation (6), when is higher, the probability of selecting a low attractiveness action is higher. When is lower, the selection is more biased towards high attractiveness actions. The compliance candidate set is used as the available location set after filtering the weight, size, and height constraints.

8. The scheduling method of claim 4, wherein, In the execution of the back propagation process, a time discount factor is introduced The action value update formula is （7） In formula (7), is the learning rate, used to control the value update step size, to avoid parameter oscillation; is the node depth, used to calculate the degree of return decay when the current action is in the search tree; is the return when the simulation is terminated.

9. The scheduling method of claim 4, wherein, An exception handling step is performed to unlock 5% to 10% of the total berth number of the yard as backup berths or to relax the constraints in stages when there is no compliant position, and to exclude abnormal berths, adjust the yard tool attraction target calculation logic and increase the iteration number from 120 to 150 when equipment fails. The computer program is executed by a processor to realize the steps of the construction method in any one of claims 1-3.

10. A computer-readable storage medium, the computer-readable storage medium storing a computer program, characterized in that: The computer program is executed by a processor to realize the steps of the scheduling method in any one of claims 4-9. 11.A computer readable storage medium, storing a computer program, and the computer program comprises the following steps: The computer program system includes:

12. A computer readable storage medium having stored a computer program system, characterized in that A model construction module is used to realize UCT-MCTS framework initialization, symbolic modeling, AHP-entropy weight calculation and constraint verification functions, and a dynamic exploration coefficient is calculated during UCT-MCTS framework initialization. A scheduling execution module is used to realize alpha-RAVE selection, hierarchical simulation, discounted back propagation and convergence judgment functions. An exception handling module is used to realize backup berth calling, constraint relaxation, fault berth screening and iteration number adjustment functions. A visualization module is used to output positioning schemes, yard state diagrams and convergence curve results. The computer program system is called by a processor to perform container positioning decision-making according to the scheduling method in any one of claims 4-9. ​

Images