Automatic wharf storage yard container turnover position selection random optimization method

Abstract

The invention discloses a stochastic optimization method for container turnover and position selection of an automatic wharf container yard, and the method comprises the steps: constructing and training a yard container lifting probability prediction model; and calculating a container lifting probability matrix # imgabs1 # formed by the probability # imgabs0 # of the container i in the storage yard in the time period t on the basis of the probability of all containers and the complete time period of the storage yard, so as to realize random optimization of the container turnover position selection decision of the container storage yard. The container lifting probability prediction model of the storage yard container is based on a multilayer perceptron (MLP) neural network structure. And during training of the yard container lifting probability prediction model, obtaining historical operation records of the wharf container yard, and taking at least container attribute information, shore bridge working road records and ship stowage in the records as input of the model.

Description

Technical Field

[0001] The invention belongs to the technical field of logistics, and in particular relates to a random optimization method for selecting locations for container turnover in an automated terminal yard. Background Art

[0002] Ports are basic and hub facilities and are an important support for economic development. With the development of technologies such as big data and automation, automated container terminals have obvious advantages in improving loading and unloading efficiency, enhancing production safety, and improving operating conditions, and have become the development direction of future ports.

[0003] like Figure 1 The figure shows the layout of an automated container terminal. In the terminal yard, containers are stacked vertically in different container areas. During loading operations, containers are first moved from the container area to the inner container truck by rail cranes, then transported to the quay crane, and finally placed in the designated location on the ship by the quay crane. Containers are different and cannot be replaced. The order of picking up containers is generally determined by the quay crane according to the ship's loading plan. Figure 1 In the container yard, the term "be" refers to the horizontal arrangement of containers in the yard. The term "row" usually refers to the unit in which containers are arranged in a direction perpendicular to the front of the wharf. The numbering of the columns usually starts from the front of the wharf and goes sequentially inside the yard. The term "layer" refers to the number of layers of containers stacked in the vertical direction. In the yard, the term "layer" refers to the height of the stacked containers, usually counting from the ground upwards.

[0004] Due to the dynamics and uncertainties in the terminal operation process, there will inevitably be inconsistencies between the order of container storage in the yard and the order of container extraction. When other containers are stacked on top of the container to be extracted, the obstructing container that is blocking the extraction of the target container needs to be moved to another position first. This process is called container flipping. The determination of the position of the obstructing container is called container flipping selection.

[0005] Therefore, in the container terminal yard, the obstruction box is a container located above the target container, which prevents the target container from being picked up. In order to be able to pick up the target container, these obstruction boxes must first be moved to other locations, a process called turning the box. The turning position selection is the process of determining which obstruction boxes need to be moved and where they should be moved in the yard so that the target container can be picked up smoothly. Since the turning operation greatly reduces the efficiency of the terminal yard's container dispatch, if the turning position selection strategy is inappropriate, it will also lead to repeated turning of boxes, which will further reduce the turning efficiency of the yard. Summary of the invention

[0006] One of the embodiments of the present invention is a random optimization method for selecting a container site for a terminal container yard. By constructing and training a container yard container pick-up probability prediction model, the probability of container i in the yard being picked up in time period t is calculated. The container pick-up probability matrix based on the probability of all containers in the yard and the complete time period Realize random optimization of container yard rotation location selection decision.

[0007] Preferably, the container yard pick-up probability prediction model is based on a multi-layer perceptron MLP neural network structure.

[0008] In the training of the container yard container pick-up probability prediction model, the historical operation records of the terminal container yard are obtained, and at least container attribute information, quay crane working path records, and ship loading in the records are used as inputs of the model.

[0009] The container ship operation time corresponding to the container pickup operation is divided into a plurality of equally spaced time periods, and the time period sequence number including the container pickup time is the category of the container.

[0010] The disclosed embodiments aim at the container turning operation at a container terminal, take into account the dynamic uncertainties existing in the terminal operation process, optimize the selection of container turning locations, determine the placement of obstructing containers, avoid the occurrence of repeated container turning, and improve the terminal yard efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0012] Figure 1 Schematic diagram of the layout of an automated container terminal in the prior art.

[0013] Figure 2 A schematic diagram of training and deployment of a probability prediction model for container pickup sequence in a yard according to one embodiment of the present invention.

[0014] Figure 3 A schematic diagram comparing container turnover rates using the SSF method and the LSGP method according to one embodiment of the present invention.

[0015] Figure 4 Schematic diagram of the probability accuracy sensitivity test results of the LSGP method according to one embodiment of the present invention. 1 - container area, 2 - shell, 3 - row, 4 - layer, 5 - container, 6 - quay crane, 7 - rail crane, 8 - internal container truck, 9 - external container truck. DETAILED DESCRIPTION

[0016] In order to solve the problem of container yard roaming location selection, existing solutions often adopt the shortest stack first (SSF) rule. Obstruction BoxMove to the row with the shortest layer height. This method is simple and fast, but it is easy to cause repeated box turning and cannot effectively improve the efficiency of box delivery in the yard.

[0017] The academic community has proposed mathematical programming models, heuristics, tree search and other methods around the problem of container selection, but these methods are all based on a given complete container pickup sequence. The container pickup sequence here refers to the order in which containers are picked up in the terminal yard. Due to factors such as extreme weather, equipment failures, and ship delays, there are a lot of uncertainties in the actual operation of the terminal, and the complete container pickup sequence cannot be determined in advance.

[0018] Therefore, in order to solve the problem that the sequence of container pickup at the yard cannot be determined due to high random uncertainty in the terminal operation process, the present invention proposes a new concept: The probability matrix of container pickup at the yard , to describe the uncertainty of the order of container pickup at the yard, and to obtain a real solution to the problem through an accurate description of the problem. The present invention collects historical data of terminal operation, extracts the time of container pickup at the yard and its corresponding container attribute information, ship stowage, quay crane working path arrangement and other information, and establishes a machine learning prediction model for the probability matrix of container pickup at the yard.

[0019] Based on the container picking probability matrix, the present disclosure also proposes a rule called LSGP (Local Safety and Global Potential), which comprehensively considers the local safety and global optimization potential of each container picking location to avoid repeated container picking. Finally, the present disclosure verifies the effectiveness and robustness of the proposed method through experiments.

[0020] According to one or more embodiments, a random optimization method for selecting a container turnaround location in an automated terminal yard is provided. Here, random optimization is performed to maximize or minimize the expectation of a target. In the case of probability, the target expectation is maximized or minimized through decision making. The random optimization method for selecting a container turnaround location in an automated terminal yard of this embodiment mainly includes two steps, namely, predicting the probability matrix of the container pickup sequence in the yard and determining the container turnaround location.

[0021] (1) Prediction The probability matrix of the container pickup sequence at the yard.

[0022] To capture Extraction timeThe nonlinear relationship between the factors including the quay crane work route plan, ship loading and container attribute information is estimated by using a multilayer perceptron (MLP). A multilayer perceptron is a feedforward artificial neural network that contains at least three layers of nodes: an input layer, one or more hidden layers, and an output layer. Except for the input node, each node is a neuron with a nonlinear activation function. Each connection of the MLP has a weight, and each neuron has a bias. These parameters are continuously updated during the training process through the back-propagation algorithm. During the training process, the network minimizes the difference between the output and the actual value by adjusting the weights and biases. This process is usually quantified using mean square error or other cost functions. Multilayer perceptrons are used to capture and learn complex mapping relationships between input data. The steps to establish a multi-classification MLP model are as follows: Figure 2 The specific description is as follows:

[0023] Step S101: feature engineering.

[0024] Collect terminal operation history records. Each record contains three aspects: container attribute information, quay crane work route record, and ship loading. Among them, container attribute information includes container ID, weight, type, size, destination, storage location, and extraction time. The quay crane work route record includes the quay crane number assigned to each container and its operation start and end time. Ship loading includes the position assigned to each container on the ship.

[0025] Step S102: Label definition.

[0026] The service time of the ship is divided into 12 equal time intervals, and the time segment number containing the container pickup time is used as the category of the container to obtain the label of the multi-classification model.

[0027] Step S103: model training.

[0028] Use the MLP model. The model has 5 layers of neurons, with 64, 128, 256, 128, and 64 neurons in each layer. The MLP model uses softmax to calculate probabilities, cross entropy to calculate losses, and back propagation to update weights in the fully connected network. The MLP model outputs the probability of each category and considers the category with the highest probability as the prediction result.

[0029] Step S104: Extraction of extraction probability.

[0030] Once the model training is completed, the network structure and weights are maintained. When the model is deployed, other information about the container to be predicted is collected in addition to the extraction time, and the probability of each container category is calculated using the trained network. This probability is the probability of the container pickup sequence at the yard. The probability of the container pickup sequence of all containers constitutes the probability matrix of the container pickup at the yard. This matrix is ​​a two-dimensional matrix Its elements represents the probability that container i is picked up in time period t.

[0031] like Figure 2 The figure below is a diagram of the training and deployment of the probability prediction model for container yard pickup sequence. The model is trained by analyzing historical operation data, and the container pickup operation at the terminal is predicted and optimized. The training steps for the model include:

[0032] S201, first collect historical operation data of the terminal, including quay crane operation route records, ship loading information, container attribute information, etc. as feature inputs of the model, and the serial number of the container extraction time period in the historical operation data of the terminal as the output of the model;

[0033] S202, using the softmax function as a probability calculation function, converting the output of the model into a probability distribution for multi-classification problems, and defining the input layer, hidden layer, output layer and the connection between them, and optimizing and adjusting the model structure and parameter weights to minimize the prediction error;

[0034] S203, at the model output layer, the category with the highest prediction probability is selected using the argmax function.

[0035] The steps for deploying the model include:

[0036] S205, obtaining information of the container to be predicted as a model input, including a quay crane working route plan, ship loading information, and container attribute information;

[0037] S206, the model outputs the container extraction probability, and assigns an extraction probability to each container in each time period.

[0038] Therefore, the container pickup probability matrix is ​​a two-dimensional matrix that represents the probability of a container being picked up in each time period in the terminal yard. This matrix is ​​constructed based on the probability data predicted by the multi-layer perceptron (MLP) model and is used to optimize the container picking location decision at the terminal yard.

[0039] (2) Determine the location for container turnover.

[0040] When it is necessary to determine the turning position for the obstruction box, it can be determined by calculating the LSGP value of other obstruction boxes that can be placed. The LSGP calculation method is as follows:

[0041] LSGP(k)=P(x→*)+α|N(k)| 2

[0042] Wherein, k represents the column id where the target location is located; P(x→*) represents the probability that the blocking box x will become the first box to be extracted after being placed in the k column; N(k) represents the set of all containers in the k column; |N(k)| represents the number of existing containers in the kth column; α is a coefficient, which is 0.05 in this embodiment.

[0043] The calculation method of P(i→*) is as follows:

[0044] S301, firstly, based on the container pick-up probability matrix at the container yard, calculate the priority pick-up probability matrix M between all containers = (m i,j ), where m i,j represents the probability that container i is picked up before container j,

[0045]

[0046] S302, calculating the approximate priority extraction probability of the blocked container x compared to all existing containers in the k column

[0047]

[0048] S303, normalize the probability and calculate the final P(x→*) estimate.

[0049]

[0050] The larger the LSGP value, the lower the possibility of repeated box turning caused by the column being used to place the obstruction box, and more space is left to prepare for the box turning that may be needed in the future. Therefore, when actually selecting, choose the column with the largest LSGP value, and the final box turning position is above the current highest box in the column.

[0051] Therefore, in the disclosed embodiment, the LSGP value is an indicator for evaluating the quality of container tumbling position selection. The larger the LSGP value, the lower the possibility of future tumbling operations caused by placing the obstruction box at this position, and more space is reserved for possible future tumbling operations. When actually selecting a position, the column with the largest LSGP value should be selected as the tumbling position.

[0052] At the same time, since the calculation of LSGP only involves a small amount of direct calculation, the calculation time can be ignored. However, the traditional way of solving this problem often requires a long solution search. When the problem scale becomes larger, the time required for the search increases exponentially.

[0053] In order to verify the effectiveness of the proposed method, 56 examples of different scales were designed for experiments. The experimental results are as follows: Figure 3 shown. Figure 3 The points in the figure represent the container turnover rate of the existing SSF method and the LSGP method proposed in the present disclosure under the same example. The figure shows that the results obtained by the method proposed in the present disclosure are better than those of the existing methods in all examples. The average container turnover rate of LSGP in 56 examples is 47.27%, which is 6.48% lower than that of SSF's 53.75%.

[0054] In view of the possible inaccuracy of the yard extraction probability, the LSGP method was tested for probability accuracy sensitivity under different prediction accuracies. The results are as follows: Figure 4 . Figure 4 The relationship between the accuracy of the probability of picking up a container at the yard and the average probability of turning over a container is shown. The horizontal axis (X-axis) represents the accuracy of the probability of picking up a container at the yard, ranging from 0.2 to 1.0, reflecting the accuracy of predicting the probability of picking up a container. The vertical axis (Y-axis) represents the average probability of turning over a container (%), ranging from 47% to 55%, which is used to measure the effect of the model on problem optimization. The smaller the value, the fewer operations are used to complete the task.

[0055] It can be found that the method proposed in the present disclosure does not require high accuracy in the prediction of the probability of picking up containers. When the prediction accuracy is above 0.3, the LSGP method can achieve better results. This means that even if the prediction of the probability of picking up containers is not very accurate, the LSGP method can still work effectively. This shows that the LSGP method has strong robustness in dealing with uncertainty and inaccuracy. Even in the face of inaccurate predictions of the probability of picking up containers, it can effectively guide the selection of container positions, thereby optimizing the terminal operation process, and has good practicality.

[0056] Therefore, the beneficial effects of the present disclosure include:

[0057] 1. The probability matrix of container pickup at the yard is used to describe the uncertainty of the order of container pickup at the yard. By reducing the expectation of container tumbling, repeated container tumbling operations can be avoided, and the optimization effect is good;

[0058] 2. Calculate the probability matrix of container pickup at the yard by establishing a machine learning model. This method has strong versatility and can be applied to different scenarios by establishing different machine learning models.

[0059] 3. The proposed method for solving the container picking location based on the container yard pick-up probability matrix has high computational efficiency and can be solved quickly in scenarios with strict time requirements;

[0060] 4. The proposed method for solving the container picking location based on the container yard pickup probability matrix does not have high requirements on the accuracy of the container yard pickup probability matrix. Even when the accuracy of the container yard pickup probability matrix is ​​not high due to the inaccuracy of the machine learning model, it can still achieve good optimization results.

[0061] It should be understood that in the embodiments of the present invention, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0062] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0063] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0064] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A random optimization method for selecting container positions for container yard turnover at a terminal, characterized in that: By building and training a container pick-up probability prediction model, the probability of container i in the yard being picked up in time period t is calculated. The probability matrix of container pick-up is formed based on the probability of all containers in the yard and the complete time period Realize random optimization of container yard rotation location selection decision.

2. The method according to claim 1, characterized in that The container yard pick-up probability prediction model is based on a multi-layer perceptron (MLP) neural network structure.

3. The method according to claim 2, characterized in that In the training of the container yard container pick-up probability prediction model, the historical operation records of the terminal container yard are obtained, and at least container attribute information, quay crane working path records, and ship loading in the records are used as inputs of the model.

4. The method according to claim 2, characterized in that: The container ship operation time corresponding to the container pickup operation is divided into a plurality of equally spaced time periods, and the time period sequence number including the container pickup time is the category of the container.

5. The method according to claim 1, characterized in that The LSGP value is used as an indicator to evaluate the quality of container roaming location selection. LSGP(k)=P(x→*)+α|N(k)| 2 Among them, k represents the column id where the target location is located; P(x→*) represents the probability that the blocking box x will become the first box to be extracted after being placed in the k column; N(k) represents the set of all containers in column k; |N(k)| represents the number of existing containers in the kth column; α is the preset coefficient.

6. The method according to claim 5, characterized in that The calculation methods of P(i→*) include: S301, firstly, based on the container pick-up probability matrix at the container yard, calculate the priority pick-up probability matrix M between all containers = (m i,i ), where m i,j represents the probability that container i is picked up before container j; S302, calculate the approximate priority extraction probability of the blocked container x compared to all existing containers in the k column S303, normalize the probability and calculate the final P(x→*), 7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor runs the computer program to implement the method according to any one of claims 1 to 6.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer program product, comprising a computer program, characterized in that The computer program is executed by a processor to implement the method according to any one of claims 1 to 6.

Images