Container ship operation duration prediction method and system

Abstract

The invention relates to the field of deep learning, and provides a container ship loading and unloading operation time prediction method and system for solving the problems that an existing container ship operation time estimation method is rough, and operation data has noise in a real loading and unloading operation scene. On the basis of a deep residual shrinkage network DRSN-CW (Deep Residual Shrinkage Net) model with feature extraction and denoising capabilities under a ResNet-18 architecture, a fixed soft threshold function is changed into a more flexible adaptive threshold function, so that the noise reduction capability is more flexible while the feature learning capability is kept. According to the scheme of the application, the request of real-time prediction of the quay crane operation duration in the berth planning process can be well met with relatively high average detection precision and robustness.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a method and system for lengthening operation time of a container ship. Background Art

[0002] In today's increasingly prosperous globalized world, ocean transportation plays a crucial role. Ports, as comprehensive transportation hubs connecting land and sea, not only provide loading and unloading services for import and export cargo, but also offer a variety of shipping services, making them key nodes in the logistics network. Container transportation, with its standardized processes for loading, unloading, transportation, and storage, has become the most important mode of transportation for international trade.

[0003] Vessel Turnaround Time (VTT) is the total time a vessel spends in port, including time spent at berth and waiting at anchor. It is a key indicator for measuring a port's overall turnover efficiency and assessing its service capabilities. Accurately predicting VTT helps reduce unnecessary ship detentions and improve overall port profitability. To develop optimal berthing plans before a vessel arrives, ports need to predict VTT. This primarily includes cargo loading and unloading time (VHT), as well as minor delays caused by factors such as weather. Because VHT is a primary component of VTT and is easier to estimate, many ports use VHT instead of VTT as the basis for berth allocation.

[0004] Traditional VHT estimation involves multiplying the number of pre-assigned quay cranes by a constant representing operational efficiency. This product is then divided by the total number of containers on board a container ship to arrive at a rough VHT. However, this rough estimate only considers the number of containers loaded and the number of quay cranes, ignoring other factors and their nonlinear effects. Therefore, the result is often inaccurate and requires constant updating and adjustment based on the vessel's berthing operations.

[0005] Therefore, many researchers have been working to identify more factors influencing VHT or to develop more complex models to improve prediction accuracy. Mokhtar and Shah, using data from the Port of Klang, Malaysia, from August 2005, proposed a multivariate linear regression model to predict VTT, factoring in factors such as the number of quay cranes, the number of trailers, and the volume of containers loaded and unloaded. K. Gayathma et al. analyzed three container ports using the Analytical Hierarchical Process (AHP) and found that transport-related delays and berthing-related delays significantly impact VTT, but the factors influencing each port vary. Jeffery Karafa et al. calculated delay probabilities from vessel operation data and used these probabilities to estimate VHT. MD Gracia et al. explored the impact of port loading and unloading strategies on VHT. Liming Guo et al. pointed out that VHT can be affected by weather uncertainties.

[0006] Many scholars have applied machine learning to navigation-related issues. For example, Meng Guan et al. established a "multiple MMSIs on a single ship" detection model based on AIS data. Similarly, some scholars have sought to solve the VTT and VHT prediction problems through machine learning methods: Vibhuti Dhingra et al. proposed a two-layer stochastic Markov model under the premise of optimizing quay crane allocation; Dejan Štepec et al. compiled FAL table data from the Port of Bordeaux, France, from 2008 to 2018 and established a VTT prediction model using the CatBoost method; Han Zonglei established a three-layer neural network VHT prediction model based on ship attributes (such as length and width) and weather, number of quay cranes, and specific container source composition; Dong Shan established a VHT prediction model for the Tianjin container port based on an extreme learning machine (ELM) model with a single hidden layer optimized by the particle swarm algorithm, and combined it with the principal component analysis method to reduce the dimensionality of the input features; Zhu Qianwen used a two-layer stacking to integrate multiple machine learning algorithms to establish a VHT prediction model for the Guangzhou Port. In recent years, some researchers have tried to use more complex deep network models to solve the problem: He Yuqing established the deep learning model core computing architecture DLM-CCA (Deep Learning Model Core Compound Architecture) based on the operation log of Guangzhou Port, which includes a four-layer VTT prediction model with stacked LSTM, Gaussian noise, bidirectional GRU and fully connected layers, and a VHT prediction model with stacked RNN, CNN and attention mechanism, but the model interpretability is poor; Hyeonsoo Shin et al. based on 22 years of data from Busan Port in South Korea (2001.01-2022-08), established a monthly LSTM time series prediction model for the average monthly waiting time of ships at anchorages and the average monthly loading and unloading time of containers, respectively. However, this model is suitable for making future predictions based on the long-term operation of the port, which is completely different from the VHT and VTT prediction problems for a single ship.

[0007] In summary, existing prediction models have several shortcomings in scenarios where accurate VHT prediction is required for port berthing plans for container ships. Furthermore, in real-world loading and unloading operations, VHT is affected by complex, difficult-to-quantify factors such as worker proficiency, and the data collection process itself can be subject to errors, which can lead to noisy data. To address these issues, this paper, based on the Deep Residual Shrinkage Network (DRSN-CW) model, which has both feature extraction and denoising capabilities, adopts a more flexible threshold function to reduce potential errors in practical applications. Summary of the Invention

[0008] The present invention provides a method and system for predicting the operation duration of container ships, which is used to solve the problems of incomplete feature analysis, insufficient feature-to-label interpretation, and failure to effectively eliminate noise interference in the prior art when predicting the operation duration of container ships. The solution of the present application can well meet the requirements of real-time prediction of quay crane operation duration during berth planning with high average detection accuracy and robustness. The present invention provides a method for predicting the operation duration of container ships, comprising: Collect target data that affects quay crane operation efficiency, including container source composition, ship static attributes, port allocation resources and weather data; The historical data of container ship operations are preprocessed and used as a training set; Use ADP-DRSN for regression fitting training to obtain the prediction model; The target data is cleaned by 3-sigma and boxplot method; Feature classification is performed based on the feature data source and input into the model through different channels; Build the ADP-DRSN deep learning model based on the ResNet18 architecture; Use an improved threshold function; The present invention also provides a case bridge operation efficiency prediction system, comprising: The data acquisition module is used to collect target data that affects the operation time of container ships, including container source composition, ship static attributes, port allocation resources and weather data; Feature analysis module, used to perform feature analysis on target data to obtain a prediction data set; The efficiency prediction module is used to input the prediction data set into the pre-trained prediction model to predict the operation time of container ships.

[0009] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, any of the above-mentioned methods for predicting the operation duration of a container ship is implemented.

[0010] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for predicting the operation duration of a container ship.

[0011] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for predicting the operation duration of a container ship.

[0012] The solution provided in this application can meet the requirements of a method for predicting the operation duration of container ships with high speed, light weight and high accuracy with high average detection accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1 This is one of the data preprocessing flow diagrams of the container ship operation duration prediction method provided by an embodiment of the present invention; Figure 2 This is the second data preprocessing flow diagram of the container ship operation duration prediction method provided by an embodiment of the present invention; Figure 3 is a schematic diagram of the structure of the prediction model provided by an embodiment of the present invention; Figure 4 2 is a schematic diagram of the structure of a container ship operation duration prediction system provided by an embodiment of the present invention; Figure 5 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0015] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0016] Figure 1 This is one of the data preprocessing flow diagrams of the container ship operation duration prediction method provided by an embodiment of the present invention.

[0017] Figure 2 This is the second data preprocessing flow diagram of the container ship operation duration prediction method provided by an embodiment of the present invention.

[0018] like Figure 1 and Figure 2 As shown, this embodiment provides a method for processing container ship operation duration prediction data, including: Step 101: Collect target data that affects the operation time of container ships; Step 102: performing integrated calculation on relevant data; Step 103: The target data is classified and processed to obtain a prediction data set.

[0019] This includes resources needed for container loading and unloading operations, such as container source composition, vessel static attributes, port allocation resources, and weather data. Container source composition can be directly obtained from port entry declaration forms, vessel static attributes can be read from a database, port allocation resources are compiled from historical operation records, and weather data is obtained through web crawlers.

[0020] A box plot, also known as a box plot or box diagram, is a statistical chart used to display the distribution of data. A box plot includes five statistical quantities for a set of data: minimum value, first quartile (Q1), median, third quartile (Q3), and maximum value.

[0021] General steps to remove outliers using a boxplot: Draw a box plot: First, draw a box plot of the data, including the box, median line, whiskers, etc.; Determine the range of outliers: Based on the box plot, calculate the range of the whiskers. Usually, the length of the whiskers is 1.5 times the interquartile range, which can be adjusted according to the specific situation; Identify outliers: Identify data points that are outside the whisker range on the box plot; these points are outliers; Delete outliers: Delete the data points identified as outliers from the data set; Redraw the box plot: After deleting the outliers, redraw the box plot and check if there are still outliers; Repeat the steps: If outliers still exist after deleting the outliers, you can iterate the above steps multiple times until there are no more outliers in the data.

[0022] In this embodiment, the boxplot method is used for data cleaning, which can effectively eliminate the interference of outliers.

[0023] In an exemplary embodiment, feature analysis of target data is performed using a mutual information method to obtain a prediction data set, including organizing historical operation records and performing one-hot encoding on non-numerical features.

[0024] One-hot encoding is a common method for converting categorical variables into binary vectors, often used in machine learning and deep learning tasks. The basic idea is to convert a categorical variable with a finite number of possible values ​​into a binary vector where only one element is 1 and the rest are 0. This encoding method preserves the mutual exclusivity between categorical variables, making it easier to calculate and process during model training.

[0025] In an exemplary embodiment, the prediction model is trained in the following manner: The historical operation time of container ships is pre-processed and used as the training set; ADP-DRSN was used to perform regression fitting on the training set to obtain the prediction model.

[0026] In an exemplary embodiment, the historical data of the quay crane operation is preprocessed, including: Noise was removed by boxplot and 3sigma methods; Relevant features are obtained by integrating container operation records.

[0027] In an exemplary embodiment, ADP-DRSN is used to perform regression fitting on the training set, including: Build a deep learning model based on the resnet-18 architecture; Add drsn-cw module; Modify the soft threshold function to a more flexible adaptive threshold function; Figure 3 It is a schematic diagram of the structure of the prediction model provided by an embodiment of the present invention.

[0028] like Figure 3 As shown in the figure, this model adopts the resnet-18 deep learning network architecture and adds a threshold noise reduction mechanism.

[0029] Traditional VHT estimation involves calculating the product of the number of pre-assigned quay cranes and a constant representing operational efficiency. This product is then divided by the total number of containers carried by the container ship to produce a rough VHT. However, this rough estimate only considers the number of loaded containers and the number of quay cranes, ignoring other factors and their nonlinear effects. Therefore, the result is often inaccurate and requires constant updating and adjustment based on the vessel's berthing status. For scenarios where ports require accurate VHT predictions when planning berthing for container ships, existing prediction models also have shortcomings. Furthermore, in real-world loading and unloading operations, VHT is affected by complex, difficult-to-quantify factors such as worker proficiency, and the data collection process itself can be subject to errors, often resulting in a certain amount of noise. The ADP-DRSN, while retaining the excellent feature extraction and denoising capabilities of the Deep Residual Shrinkage Network (DRSN-CW), adopts a more flexible threshold function, further minimizing potential errors in practical applications.

[0030] After testing, the ADP-DRSN model achieved an r² of 0.745 and a Pearson R² of 0.879 on the test set; and an r² of 0.919 and a Pearson R² of 0.960 on all data sets, including the training and test sets. These results are the highest compared with other deep learning prediction models under the ResNet-18 architecture, indicating that ADP-DRSN has a high accuracy in fitting data and can better meet the requirements of container ship operation duration prediction methods in real production processes.

[0031] The following is an example of a specific embodiment to illustrate the specific implementation process of the method for predicting the operation duration of a container ship provided by this application.

[0032] like Figure 1-4 As shown in Figure 2, the container ship operation time prediction method includes: 1. Feature extraction: First, more than 1.7 million container operation samples of a port in Liaoning were merged and processed according to English ship names, import voyages, and export voyages. The relevant features were integrated and extracted, and the container operation time of each ship was calculated as the data label. After processing, there were 2188 data items in total, including the amount of 20-foot loaded containers loaded, 40-foot loaded containers loaded, 45-foot loaded containers loaded, 20-foot empty containers loaded, 40-foot empty containers loaded, 20-foot loaded containers unloaded, 40-foot loaded containers unloaded, 45-foot loaded containers unloaded, and 20-foot loaded containers unloaded. Container composition data, including the number of empty 20-foot containers, the number of unloaded 40-foot containers, the number of loaded refrigerated containers, the number of loaded hazardous goods containers, the number of unloaded refrigerated containers, the number of unloaded hazardous goods containers, and the total number of containers; port resource allocation data, including the number of quay cranes, the number of yard equipment, the utilization rate of double cranes, the number of trailers, the berthing area (East Zone / West Zone), the total haul distance; vessel attribute data, including length, breadth, gross tonnage, gross deadweight, route number (trunk / feeder, domestic / foreign trade), and berthing time; weather data, including maximum and minimum temperatures, weather type, and high temperature flags; 2. Outlier removal: Since the original data comes from actual container port operation data, some outliers will inevitably appear. Using a reasonable outlier detection method to remove these outliers can enhance the credibility of the data and improve the goodness of model fit. Use 3-sigma and boxplot methods to remove outliers; 3. Model construction: Build the ADP-DRSN deep learning network, add the drsn-cw module based on the resnet-18 architecture and modify the threshold function, modify the threshold function, and introduce the small sub-network required for parameter calculation; 4. Predict the next container ship operation duration based on the established ADP-DRSN deep learning model.

[0033] The following describes a container ship operation duration prediction system provided by the present invention. The container ship operation duration prediction system described below and the container ship operation duration prediction method described above can be referenced to each other.

[0034] Figure 4 It is a structural diagram of a shore container ship operation duration prediction system provided by an embodiment of the present invention.

[0035] like Figure 4 As shown, the quay crane operation efficiency prediction system provided in this embodiment includes: The data collection module 401 is used to collect target data that affects the operation time of the container ship, and the target data includes tooling data, natural data and human data; Feature analysis module 402, used to perform feature analysis on target data using mutual information method to obtain a prediction data set; The duration prediction module 403 is used to input the prediction data set into the pre-trained prediction model to predict the operation duration of the container ship.

[0036] The specific implementation method of the container ship operation duration prediction system provided in this embodiment can be implemented with reference to the above embodiment and will not be repeated here.

[0037] Figure 5 An example of a physical structure diagram of an electronic device is shown below. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the container ship operation duration prediction method, which includes: Collect target data that affects container ship operation time; Integrate and classify the collected data to obtain the prediction data set; The prediction dataset is input into the pre-trained prediction model to predict the operation duration of container ships.

[0038] Furthermore, the logic instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0039] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the container ship operation duration prediction method provided by the above methods, which includes: Collect target data that affects container ship operation time; Integrate and classify the collected data to obtain the prediction data set; The prediction dataset is input into the pre-trained prediction model to predict the operation duration of container ships.

[0040] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting the operation duration of a container ship provided by the above methods is implemented, and the method includes: Collect target data that affects container ship operation time; Integrate and classify the collected data to obtain the prediction data set; The prediction dataset is input into the pre-trained prediction model to predict the operation duration of container ships.

[0041] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. That is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0042] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods of each embodiment or certain portions of the embodiments. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting the operation time of a container ship, characterized in that: include: Collecting target data that affects container ship operation time, including container source composition, ship static attributes, port allocation resources and weather data; Performing feature analysis on the target data to obtain a prediction data set; The prediction data set is input into a pre-trained prediction model to predict the operation duration of the container ship.

2. The method for predicting the operation duration of a container ship according to claim 1, characterized in that: After collecting target data that affects the operation time of container ships, the following steps are also included: Processing the target data by means of the boxplot method and the like to remove abnormal values; By changing the fixed soft threshold function to a more flexible adaptive threshold function based on the residual shrinkage network of the ResNet-18 architecture, we obtain the deep residual shrinkage network model ADP-DRSN with adaptive threshold function.

3. The method for predicting the operation duration of a container ship according to claim 1, characterized in that: Perform feature classification analysis on the target-related data to obtain a prediction data set, including: Container source information is directly obtained from the ship declaration form; combined with the ship's static data and route table, information about the ship's own attributes such as length, breadth, gross tonnage, total deadweight, and route (one-hot coded according to domestic trade / foreign trade, trunk line / branch line) can be obtained; combined with the ship's schedule data, data on the number of quay cranes, number of trailers, berthing time, and other related port resources are obtained; and using a web crawler to obtain weather-related information during operations to obtain the aforementioned forecast data set.

4. The method for predicting quay crane operation efficiency according to claim 1, characterized in that: The prediction model is trained in the following way: The historical data of container ship operations are preprocessed and used as a training set; ADP-DRSN was used for regression fitting training to obtain the prediction model.

5. The method for predicting the operation duration of a container ship according to claim 5, characterized in that: The data preprocessing of the historical data of the quay crane operation includes: The target data is cleaned by 3-sigma and boxplot method; Feature classification is performed based on the feature data source and input into the model through different channels.

6. The method for predicting the operation duration of a container ship according to claim 5, characterized in that: The ADP-DRSN deep learning model is used to perform regression fitting on the training set, including: Build the ADP-DRSN deep learning model based on the ResNet18 architecture; Use an improved threshold function.

7. The container ship operation time system is characterized by: include: A data acquisition module is used to collect target data that affects the operation time of container ships, including container source composition, ship static properties, port allocation resources and weather data; A feature analysis module, configured to perform feature analysis on the target data to obtain a prediction data set; The efficiency prediction module is used to input the prediction data set into the pre-trained prediction model to predict the operation time of the container ship.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for predicting quay crane operation efficiency according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for predicting quay crane operation efficiency according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting quay crane operation efficiency according to any one of claims 1 to 6 is implemented.

Images