Intelligent wharf bridge crane operation time prediction method and system and electronic equipment

Abstract

The invention relates to an intelligent wharf bridge crane operation time prediction method and system and electronic equipment, and the method comprises the steps: carrying out the data cleaning and feature coding of the obtained historical operation data and historical weather data of an intelligent wharf bridge crane, and constructing a bridge crane operation time prediction data set; based on a grey correlation analysis method, carrying out feature extraction on the bridge crane operation time prediction data set, and constructing a feature data set influencing the wharf bridge crane operation time; and constructing a bridge crane operation time multi-dimensional matrix based on the feature data set, and outputting a bridge crane operation time prediction value and a corresponding confidence interval by using a Monte Carlo prediction model according to an input bridge crane operation instruction and a current weather condition. Compared with the prior art, the method has the advantages of being sufficient in wharf bridge crane operation feature extraction, high in prediction accuracy, real-time, efficient and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent terminal operations, and particularly to a method, a system and an electronic device for predicting the operation time of a gantry crane at an intelligent terminal. Background Art

[0002] With the continuous development of global trade, container terminals, as key nodes in the flow of goods, are under huge economic and logistics pressure. In order to meet the growing import and export demands and significantly enhance the global competitiveness of Chinese ports, further reducing the labor costs at terminals and improving the terminal operation efficiency will be the inevitable path for the future development of ports. Therefore, the automation and intelligent transformation of traditional container terminals are extremely urgent.

[0003] The gantry crane is an important machine at the port, responsible for completing the loading and unloading operations of containers between the ship and the shore. Its loading and unloading efficiency will directly affect the time of the ship at the port. Therefore, the level of the gantry crane operation efficiency will directly represent the size of the import and export capacity of the terminal. Moreover, the loading and unloading efficiency of the gantry crane can be refined to the operation time for the gantry crane to complete the handling of each container. If the time taken for the gantry crane to complete each task instruction can be accurately predicted, it can not only quantitatively evaluate the instruction set specified by the operators, but also play a guiding role in further adjusting the scheduling decisions for the terminal operators. However, the port system often has characteristics such as complex and diverse constraint conditions, deep coupling of operations, and complex and changeable scenarios. The existing technologies often encounter difficulties in accurately predicting the gantry crane operation time.

[0004] Zeng Qingcheng et al., "Robust Optimization Model for Berth Planning in Container Terminals", Operations Research and Management Science, 2015, 24(2): 71-77, proposed a berth allocation optimization model considering the gantry crane loading and unloading efficiency. Before making the berth plan, the system uses the average loading and unloading volume per hour of the gantry crane as the loading and unloading efficiency, and divides the total number of containers loaded and unloaded by the ship by the loading and unloading efficiency to predict the operation duration of the container ship at the shore. When predicting the operation duration, this literature assumes that the gantry crane loading and unloading efficiency is a fixed value, without considering the influencing factors such as container differences, weather conditions, and ship sizes, thus unable to accurately predict the operation duration at the shore.

[0005] Jeffery Karafa et al., "The berth allocation problem with stochastic vessel handling times", International Journal of Advanced Manufacturing Technology, 2013, Vol. 65: 473 - 484, proposed a berth allocation model based on uncertain vessel handling times. When formulating the berth plan, the system does not set the quay crane handling efficiency as a fixed value, but assumes that the probability distribution of the vessel handling operation duration is determined, thereby establishing an exact algorithm and reducing the statistical error of the container ship terminal operation time. The method proposed in this literature requires a large amount of data support to accurately predict the terminal operation time. In addition, the probability distribution of the terminal operation duration is affected by many external factors and is often difficult to accurately model, resulting in a low prediction accuracy.

[0006] Han Zonglei et al., "A prediction model for container ship port operation time based on neural network", Computer Applications and Software, 2021, 38(2): 78 - 84, proposed a data-driven container ship operation time prediction system. The system considered the complex non-linear relationship between the operation duration and various influencing factors such as vessel type, number of quay cranes, and weather, and established a BP neural network to complete the prediction of the container ship operation time, thereby better formulating the berth plan. In the process of making the data set, this literature did not sufficiently clean the data to remove outliers. In addition, this literature did not conduct a detailed analysis of the impact of input features on the operation duration, which may introduce irrelevant variables and affect the prediction effect.

[0007] After retrieval, Chinese Invention Application No. CN202311737354.X discloses a quay crane operation efficiency prediction method and system. The system collects the tooling data, natural data, and human data that affect the quay crane operation efficiency, conducts feature analysis on the data through the mutual information method, and completes the prediction of the quay crane operation efficiency based on the BP neural network. The drawback of this method is that it cannot give a probability distribution and confidence interval for the predicted value, resulting in a low credibility of individual predicted values.

[0008] In summary, the existing technologies can, to a certain extent, complete the prediction of the quay crane operation time, but there are certain problems, such as being divorced from the terminal working conditions and external influencing factors, difficult to accurately model, insufficient consideration of features, imperfect feature analysis, and no guarantee for the credibility of the prediction results. These problems make it impossible for the existing technologies to accurately predict the quay crane operation time under the premise of fully considering various terminal working conditions and external environmental influencing factors, nor can they give a confidence interval for the prediction results to characterize their credibility.

[0009] How to achieve accurate prediction of the operation time of intelligent quay bridge cranes has become a technical problem to be solved. Summary of the Invention

[0010] The purpose of the present invention is to provide an intelligent quay bridge crane operation time prediction method, system and electronic device to overcome the defects of the above-mentioned existing technologies.

[0011] The purpose of the present invention can be achieved through the following technical solutions:

[0012] According to one aspect of the present invention, there is provided an intelligent quay bridge crane operation time prediction method, the method comprising:

[0013] Obtain the historical operation data and historical weather data of the intelligent quay bridge crane;

[0014] Clean the obtained data and perform feature encoding to construct a quay bridge crane operation time prediction data set;

[0015] Based on the grey relational analysis method, extract features from the quay bridge crane operation time prediction data set to construct a feature data set affecting the operation time of the quay bridge crane;

[0016] Based on the feature data set, construct a multi-dimensional matrix of the quay bridge crane operation time, and for the input quay bridge crane operation instruction and the current weather condition, use the Monte Carlo prediction model to output the quay bridge crane operation time prediction value and the corresponding confidence interval.

[0017] Preferably, the feature encoding is to digitally encode the quay bridge crane operation features corresponding to each quay bridge crane operation instruction, including:

[0018] Extract the starting point and destination information of the quay bridge crane operation and encode it to form a starting point code and a destination code;

[0019] Extract the wind speed magnitude and encode it to form a wind speed code;

[0020] Extract the container weight information of the quay bridge crane operation and encode it to form a container weight code;

[0021] Extract the quay bridge crane operation mode and encode it to form an operation mode code;

[0022] Extract the instruction priority and encode it to form an instruction priority code;

[0023] Extract the container type record and encode it to form a container type code;

[0024] Extract the ship information data and encode it to form a container ship width code and a container ship draft depth code;

[0025] And extract the operation date label and encode it to form an operation date code.

[0026] Preferably, the feature encoding process is specifically as follows: using the gantry crane operation instruction number as the unique index, when performing feature encoding, directly perform feature encoding on numerical data, and first convert non-numerical data into numerical data and then perform feature encoding to form a gantry crane operation time prediction data set.

[0027] Preferably, the construction of the feature data set affecting the gantry crane operation time at the terminal includes: using the gantry crane operation time as the reference sequence, using the remaining gantry crane operation information as the comparison sequence respectively, calculating the size of the grey correlation coefficient, and sorting according to the grey correlation coefficient from large to small, and selecting the n features with large grey correlation coefficients to construct the feature data set, where n is greater than 1.

[0028] Preferably, the dimension of the multi-dimensional matrix of the gantry crane operation time is the feature of the feature data set, and the element is the gantry crane operation time distribution that satisfies the current dimension.

[0029] Preferably, the method further includes inputting the feature data set of the gantry crane planned operation into the trained neural network prediction model to output the predicted value of the gantry crane operation time in real time.

[0030] More preferably, the neural network includes an input layer, a hidden layer and an output layer, each layer includes a number of neurons, and the connection between layers is completed through the full connection between neurons; the number of neurons in the input layer is the same as the number of features in the feature data set.

[0031] According to another aspect of the present invention, an intelligent terminal gantry crane operation time prediction system is provided, and the system includes:

[0032] Data preprocessing module: used to match the obtained intelligent terminal gantry crane operation historical data and historical weather data and then perform data cleaning, including removing missing values, outliers and extreme values; perform feature encoding on the cleaned data to construct a gantry crane operation time prediction data set;

[0033] Feature engineering module: perform feature extraction on the gantry crane operation time prediction data set after the data preprocessing module completes feature encoding to construct a feature data set affecting the gantry crane operation time at the terminal;

[0034] And the gantry crane operation time prediction module: input the gantry crane operation time feature data set into the prediction model to output the predicted value of the gantry crane operation time.

[0035] Preferably, the prediction model includes a Monte Carlo prediction model and a neural network prediction model, where the Monte Carlo prediction model outputs the predicted value of the gantry crane operation time and the corresponding confidence interval; the neural network prediction model outputs the predicted value of the gantry crane operation time in real time.

[0036] According to a third aspect of the present invention, there is provided an electronic device, including a memory and a processor. A computer program is stored on the memory, and when the processor executes the program, the method described above is implemented.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1) The present invention proposes an intelligent quay crane operation time prediction system. By matching and preprocessing the historical operation data of the quay crane and weather data, a quay crane operation time prediction data set is obtained; through the grey relational analysis method, the feature engineering of the quay crane operation time prediction data set is completed, and a quay crane operation feature data set is constructed; based on the feature data set, the Monte Carlo prediction model is used to output more accurate quay crane operation time prediction values and corresponding confidence intervals. The accurate and reliable prediction results play an important guiding role in the decision-making of quay operation personnel and have great market value.

[0039] 2) The present invention extracts multiple quay crane operation features in the quay crane operation instructions for feature coding, conducts grey relational analysis with the quay crane operation duration, and selects the features that have a greater impact on the quay crane operation time. The features are fully considered, and the features with small impacts are eliminated, resulting in higher prediction accuracy and efficiency.

[0040] 3) The present invention can also use the neural network prediction model to output prediction results in real time, which can meet the quay operation scenarios with high real-time requirements.

[0041] 4) The present invention also compares the prediction method with the benchmark value. The comparison results show that the prediction method of the present invention can complete the time prediction function of the quay crane for each task instruction based on various working conditions of the existing historical data, and the prediction accuracy is significantly improved compared with the benchmark value. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic flowchart of the intelligent quay crane operation time prediction method in the present invention;

[0043] Figure 2 It is a schematic diagram of the intelligent quay crane operation mode in the present invention;

[0044] Figure 3 It is a schematic structural diagram of the intelligent quay crane operation time prediction system in the present invention;

[0045] Figure 4 It is a schematic structural diagram of the neural network prediction model of the intelligent quay crane operation time prediction system in the present invention;

[0046] Figure 5 It is a comparison chart of the error percentile of the intelligent quay crane operation time prediction result and the benchmark value in the present invention;

[0047] Figure 6 This is a schematic structural diagram of the electronic device in the present invention. Specific embodiments

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] The present invention comprehensively considers each instruction in the gantry crane operation data and constructs a data set with the gantry crane operation instruction number as the unique index. Digital encoding is performed on the gantry crane operation characteristics corresponding to each instruction, including but not limited to: the starting point and destination of the gantry crane operation, wind speed, container weight, operation mode, instruction priority, container type, ship information, operation date, etc. The grey relational analysis is respectively performed on the above gantry crane operation characteristics and the gantry crane operation duration to obtain the characteristics that have a greater impact on the gantry crane operation time. Based on these characteristics and their corresponding gantry crane operation times, a Monte Carlo prediction model and a neural network prediction model are introduced. Among them, the Monte Carlo prediction model can give the predicted value of the gantry crane operation time under the corresponding working conditions and the confidence interval of the prediction result. The neural network prediction model can achieve a faster and more accurate prediction of the gantry crane operation time through GPU parallel acceleration.

[0050] This embodiment relates to an intelligent terminal gantry crane operation time prediction method, as Figure 1 shown in the following, the method includes the following steps:

[0051] Step 1: Obtain the historical data of the intelligent terminal gantry crane operation and the historical weather data, and preprocess the data, including: after matching the operation historical data and the historical weather data, removing the missing values, outliers, and extreme values in the data, and performing labeling processing (i.e., feature encoding) on the data to construct a gantry crane operation time prediction data set.

[0052] The method for removing outliers is specifically as follows: First, the gantry crane operation duration data set D is arranged in ascending order to form an ordered data set:

[0053] D sorted ={t (1) ,t (2) ,...,t (n)}

[0054] In the formula, t (1) , t (2) , t (n) are all gantry crane operation durations, t (1) ≤t (2) ≤...≤t(n) 。

[0055] Next, it is necessary to set the upper and lower bounds of the quay crane operation duration dataset. Based on the data distribution of the outliers, set the lower quantile q lower = 0.99, q upper = 0.01, and calculate the upper quantile t upper and the lower quantile t lower :

[0056] t lower = t (klower)

[0057] t upper = t (kupper)

[0058] In the formula, represents rounding up, and n is the number of quay crane operation durations in the dataset. After setting, the quay crane operation duration data greater than the upper bound or less than the lower bound will be excluded, and finally a new dataset D filtered is as follows:

[0059] D filtered = {t i ∈ D | t lower ≤ t i ≤ t upper}

[0060] Step 2: Based on the grey relational analysis method, extract the features of the quay crane operation time prediction dataset, construct a feature dataset that affects the quay crane operation time at the terminal, and use the corresponding quay crane operation time under each feature combination as the label dataset.

[0061] Step 3: Input the feature dataset of the quay crane operation time into the quay crane operation time prediction model, and output the quay crane operation time prediction value. The quay crane operation time prediction model is divided into a Monte Carlo prediction model and a neural network prediction model, which are obtained after training on the training samples. The Monte Carlo prediction model outputs the quay crane operation time prediction value and gives the confidence interval corresponding to the prediction value; the neural network prediction model outputs the quay crane operation time prediction value.

[0062] Verify this embodiment with a specific example. This example collects the quay crane operation instruction data, operation record data of a domestic container terminal from January 2024 to May 2024, and the wind speed data during this period. Figure 2 The verification steps for the intelligent terminal quay crane operation mode are as follows:

[0063] In Step 1, preprocess the collected data, including:

[0064] Take the difference between the timestamp of the completion instruction of the bridge crane and the timestamp of the start instruction, and record the time taken for the bridge crane to complete each instruction as the operation time of the bridge crane.

[0065] Remove missing values, outliers, and extreme values from the dataset.

[0066] And perform labeling processing (i.e., feature encoding) on the data. The bridge crane operation information processed includes but is not limited to:

[0067] Extract the starting point and destination point information of the bridge crane operation and encode it to form the starting point encoding and destination point encoding;

[0068] Extract the wind speed magnitude and encode it to form the wind speed encoding;

[0069] Extract the container weight information of the bridge crane operation and encode it to form the container weight encoding;

[0070] Extract the bridge crane operation mode and encode it to form the operation mode encoding;

[0071] Extract the instruction priority and encode it to form the instruction priority encoding;

[0072] Extract the container type record and encode it to form the container type encoding;

[0073] Extract the ship information data and encode it to form the container ship width encoding and container ship draft depth encoding;

[0074] Extract the operation date label and encode it to form the operation date encoding.

[0075] Form a new bridge crane operation time prediction dataset from the original data and store it in a storage medium.

[0076] Construct a dataset with the bridge crane operation instruction number as the unique index; when performing feature encoding, directly perform feature encoding on numerical data, and first convert non-numerical data to numerical data and then perform feature encoding to form the bridge crane operation time prediction dataset.

[0077] Feature encoding specifically includes:

[0078] Extraction of the starting point (ship, transfer platform) and destination point (transfer platform, ship) information of the bridge crane operation: For the transfer platform, extract the lane number information where it is located; for the ship, extract the container coordinate information (including row number, layer number), and directly use the extracted information as the starting point encoding and destination point encoding.

[0079] Extraction of wind speed magnitude: Label the wind speed magnitude in the environmental information according to the national standard GB / T28591-2012 of the China Meteorological Administration, and store the obtained wind speed label as the wind speed encoding.

[0080] Extraction of the container weight information in the gantry crane operation: The container weight is discretized into labels according to specific criteria, and the obtained labels are stored as the container weight codes. The criteria for container weight labeling are as follows:

[0081]

[0082] Extraction of the gantry crane operation mode information: The gantry crane operation is divided into single-container operation and double-container operation. Labels are added to these two modes, where the label for single-container operation is 1 and the label for double-container operation is 2.

[0083] Extraction of the instruction priority information: The gantry crane operation dataset contains priority information, and its numerical value is directly used as the instruction priority code.

[0084] Extraction of the container type record: The container type information is extracted from the gantry crane operation dataset. The container types are divided into general containers, refrigerated containers, dangerous goods containers, and oversize containers. The container type codes are as follows:

[0085]

[0086] Extraction of the ship information data: The ship width data and the draft data are extracted from the gantry crane operation dataset, and their corresponding numerical values are discretized into labels according to specific criteria. Among them, the container ship width labels are encoded according to the ship width value WIDE as follows:

[0087]

[0088] The container draft labels are encoded according to the ship draft size DEEP as follows:

[0089]

[0090] Extraction of the operation date label: A holiday label is added to the date recorded in the gantry crane operation dataset. Non-holiday dates are encoded as 1, and holiday dates are encoded as 2 to form the operation date code.

[0091] In step 2, the grey relational analysis method is a multi-factor statistical analysis method. The grey relational coefficients between the reference sequence and the comparison sequences are calculated, and the influence degree of the comparison sequences on the reference sequence is characterized by the magnitude of the grey relational coefficients. The larger the grey relational coefficient, the greater the influence degree of the comparison sequence on the reference sequence. The formula for calculating the grey relational coefficient is

[0092]

[0093] In the formula, x 0 (k) is the k-th element of the reference sequence, and x i (k) is the k-th element of the i-th comparison sequence. ρ is a coefficient with a value between (0,1). In this example, ρ takes 0.5.

[0094] Based on the grey relational analysis method, taking the quay crane operation time as the reference sequence and the other quay crane operation information as the comparison sequences respectively, calculate the magnitudes of the grey relational coefficients, and select the features with larger grey relational coefficients to construct the feature dataset.

[0095] Based on the magnitudes of the grey relational coefficients, select the following 8 features as the feature dataset: starting point code, destination point code, wind speed code, container weight code, operation mode code, container ship width code, container ship draft code, and container type code, and take the quay crane operation time as the label dataset.

[0096] In step 3, divide the dataset (including the feature dataset and the label dataset) that has completed feature engineering into training samples and test samples according to a certain ratio. In this example, the data volume of the training samples is 1,048,036, and the data volume of the test samples is 229,028.

[0097] Based on the training samples, complete the training of the Monte Carlo prediction model and the neural network prediction model. Based on the Monte Carlo algorithm, process the training samples to construct a multi-dimensional matrix of quay crane operation time. The dimension of the multi-dimensional matrix is the feature of the feature dataset in the training samples; the elements of the multi-dimensional matrix of quay crane operation time are the quay crane operation time distributions that satisfy the current dimension. The label distributions under each feature of the quay crane operation are input into the Monte Carlo prediction model; input the training samples into the neural network to complete the training to form the neural network prediction model;

[0098] Input the label samples into the prediction model to obtain the output results; among them, the Monte Carlo prediction model outputs the predicted value of the operation time and its corresponding confidence interval under the current quay crane operation feature; the neural network model outputs the predicted value of the operation time under the current quay crane operation feature;

[0099] The Monte Carlo prediction model is based on the Monte Carlo algorithm, which is a statistical method based on the law of large numbers and the central limit theorem. Its theory holds that a large number of independent and identically distributed random trials will present an approximate normal distribution, which is characterized as follows:

[0100]

[0101] Among them, μ is the sample mean, σ is the sample standard deviation, n is the total sample size, Φ(x) is the distribution function of the observed samples, and X i is any observed sample.

[0102] Based on the quay crane operation time distributions under each feature of the training samples, construct a probability model; conduct n independent and identically distributed random trials based on the probability model to generate a large number of random samples, obtain the prediction results based on the statistical analysis of the random samples, and give the confidence interval for the results based on the calculation of the prediction results and the variance.

[0103] The structure of the neural network prediction model is as Figure 4 shown, consisting of 1 input layer, 3 hidden layers and 1 output layer. The number of neurons in the input layer is the same as the number of features in the feature dataset, which is 9 in this example; the number of neurons in the 3 hidden layers is 64, 512, and 64 respectively; the number of neurons in the output layer is 1.

[0104] During network training, the predicted value of the quay crane operation time under the corresponding input quay crane operation features is obtained through the forward propagation of the neural network. The network calculates the loss through the loss function and trains the connection weights between neurons based on the error backpropagation, thus completing the training of the neural network prediction model.

[0105] The test samples are respectively input into the trained Monte Carlo prediction model and the neural network prediction model.

[0106] For the trained Monte Carlo prediction model, based on the quay crane operation time distribution under the quay crane operation conditions that meet the corresponding features, the distribution characteristics are used as the prediction object of the quay crane operation time in this example. That is, based on the probability distribution of the actual operation duration of the quay crane under specific working condition features, n independent random trials are carried out. In this example, n = 1024. The mathematical expectation and standard deviation of the random trials are statistically calculated. The predicted value of the quay crane operation time under the corresponding working condition features is given based on the mathematical expectation; the confidence interval of the predicted value of the quay crane operation time under the corresponding working condition features is given based on the mathematical expectation and standard deviation.

[0107] For the trained neural network prediction model, the predicted value of the quay crane operation time under the corresponding quay crane operation features of the test sample is obtained through the forward propagation of the neural network.

[0108] Taking the mean absolute percentage error as the evaluation index of the present invention. The formula for calculating the mean absolute percentage error is:

[0109]

[0110] In the formula, y i is the actual operation duration for the quay crane to complete the i-th instruction; is the predicted operation duration for the quay crane to complete the i-th instruction, and n is the total number of instructions.

[0111] In the above example, the prediction results of the quay crane operation time in the container terminal are evaluated. Among them, the benchmark value is completed by the method of expected prediction, and the average percentage error of the benchmark value is 22.77%. After testing, the mean absolute percentage error of the Monte Carlo prediction model of the present invention is 19.42%, and the error is reduced by 14.71% compared with the benchmark value; the mean absolute percentage error of the neural network prediction model is 19.52%, and the error is reduced by 14.27% compared with the benchmark value. The error percentile change curve of the mean absolute percentage error is as Figure 5As shown, the predicted results of the gantry crane operation time obtained by the present invention are superior to the reference value in terms of the error percentile.

[0112] From the example results, it can be seen that the intelligent quay crane operation time prediction method proposed by the present invention can complete the time prediction function of the quay crane for each task instruction based on various working conditions of the existing historical data, and the prediction accuracy is significantly improved compared with the reference value. The technical achievements of the present invention can not only provide data support for the dock workers to formulate various plans, but also quantitatively evaluate the existing different plans, and have great application value and market potential.

[0113] This embodiment also relates to an intelligent quay crane operation time prediction system, as Figure 3 , the system includes:

[0114] Data preprocessing module: used to obtain the historical data of the quay crane operation, introduce weather data, and after completing data matching, remove the missing values, outliers, and abnormal values in the data; perform feature encoding on the data to construct a quay crane operation time prediction data set.

[0115] Feature engineering module: perform feature engineering on the data set encoded by the above module based on feature engineering methods to construct a feature data set affecting the quay crane operation time;

[0116] And a quay crane operation time prediction module: input the quay crane operation time feature data set obtained by the above module into the prediction model, and output the predicted value of the quay crane operation time; the prediction model is divided into a Monte Carlo prediction model and a neural network prediction model;

[0117] Among them, the data preprocessing module constructs a data set with the quay crane operation instruction number as the unique index; when performing feature encoding, directly perform feature encoding on numerical data, and first convert non-numerical data into numerical data and then perform feature encoding to form a new quay crane operation time prediction data set.

[0118] Among them, the feature engineering module performs feature engineering on the cleaned data set to construct a quay crane operation feature data set, and uses the quay crane operation time as the label data set; the feature engineering method uses the grey relational analysis method. The grey relational analysis method is a multi-factor statistical analysis method, which calculates the grey relational coefficient between the reference sequence and the comparison sequence, and characterizes the influence degree of the comparison sequence on the reference sequence through the size of the grey relational coefficient. The larger the grey relational coefficient, the greater the influence degree of the comparison sequence on the reference sequence.

[0119] Taking the quay crane operation time as the reference sequence and the quay crane operation features as the comparison sequence, calculate the size of the grey relational coefficient of each feature, and select the features with larger grey relational coefficients to construct the feature data set.

[0120] The neural network structure includes an input layer, a hidden layer, and an output layer. Each layer is composed of several neurons, and the connections between layers are completed through full connections between neurons. The output h of a neuron is:

[0121]

[0122] In the formula, x i is the input of the i-th neuron, θ i is the connection weight between the i-th neuron and the current neuron, b is the threshold of the current neuron, and f is the activation function.

[0123] The input layer of the neural network inputs the dataset of quay crane operation characteristics, and the label of the output layer is the quay crane operation time. During training, error backpropagation is used to update the connection weights between neurons until the network converges. The predicted value of the quay crane operation time under the corresponding input quay crane operation characteristics is obtained through the forward propagation of the neural network. The ReLU activation function is used for the activation functions of the hidden layer and the output layer of the neural network.

[0124] The present invention can accurately characterize various working conditions of quay cranes. Among them, the Monte Carlo prediction model can give the real-time probability distribution of the predicted value of the quay crane operation time; the neural network prediction model can achieve faster real-time prediction results based on GPU acceleration, and has high application value in the field of automated container terminals.

[0125] Such as Figure 6 , the electronic device of the present invention includes a central processing unit (CPU), which can execute various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or computer program instructions loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM, and RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0126] Multiple components in the device are connected to the I / O interface, including: an input unit, such as a keyboard, a mouse, etc.; an output unit, such as various types of displays, speakers, etc.; a storage unit, such as a disk, an optical disc, etc.; and a communication unit, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit allows the device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0127] The processing unit executes the various methods and processes described above. For example, in some embodiments, the method may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed onto the device via the ROM and / or the communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute the method by any other suitable means (e.g., by means of firmware).

[0128] The functions described above herein can be performed at least in part by one or more hardware logic components. By way of example and not limitation, the types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0129] The program code for implementing the method of the present invention may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, a special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on the machine, partially on the machine, as a stand-alone software package partially on the machine and partially on a remote machine, or entirely on the remote machine or server.

[0130] In the context of the present invention, a machine-readable medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0131] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An intelligent method for predicting the operation time of a dock crane, characterized in that: The method comprises: Obtain historical operation data and historical weather data of intelligent terminal bridge cranes; Perform data cleaning and feature encoding on the acquired data to construct a data set for predicting the operation time of the bridge crane; Based on the grey correlation analysis method, the feature extraction of the crane operation time prediction data set is carried out to construct the feature data set that affects the crane operation time at the wharf. A multidimensional matrix of bridge crane operation time is constructed based on the feature data set, and the Monte Carlo prediction model is used to output the predicted value of bridge crane operation time and the corresponding confidence interval according to the input bridge crane operation instructions and current weather conditions.

2. The intelligent method for predicting the operation time of a dock crane according to claim 1 is characterized in that: The feature coding is to digitally code the bridge crane operation feature corresponding to each bridge crane operation instruction, including: Extract the starting point and destination point information of the bridge crane operation and encode them to form the starting point code and destination point code; Extract the wind speed and encode it to form a wind speed code; Extract the container weight information of the bridge crane operation and encode it to form a container weight code; Extract the bridge crane operation mode and encode it to form an operation mode code; Extracting and encoding the instruction priority to form an instruction priority code; Extract the container type record and encode it to form the container type code; Extract ship information data and encode it to form container ship width code and container ship draft depth code; And extract the operation date label and encode it to form the operation date code.

3. The intelligent method for predicting the operation time of a dock crane according to claim 1 is characterized in that: The feature coding process is specifically as follows: using the bridge crane operation instruction number as a unique index, when performing feature coding, the numerical data is directly feature coded, and the non-numerical data is first converted into numerical data and then feature coded, to form a bridge crane operation time prediction data set.

4. The intelligent method for predicting the operation time of a dock crane according to claim 1 is characterized in that: The method for constructing a feature data set that affects the operation time of the dock crane includes: taking the operation time of the crane as a reference sequence, taking the remaining crane operation information as a comparison sequence, calculating the size of the grey correlation coefficient, and sorting the grey correlation coefficients from large to small, and selecting n features with large grey correlation coefficients to construct a feature data set, where n is greater than 1.

5. The intelligent method for predicting the operation time of a dock crane according to claim 1 is characterized in that: The dimensions of the multidimensional matrix of bridge crane operation time are the characteristics of the feature data set, and the elements are the distribution of bridge crane operation time that meets the current dimensions.

6. The intelligent method for predicting the operation time of a dock crane according to claim 1, characterized in that: The method also includes inputting a feature data set of the planned operation of the bridge crane into a trained neural network prediction model, and outputting a predicted value of the bridge crane operation time in real time.

7. The intelligent method for predicting the operation time of a dock crane according to claim 6 is characterized in that: The neural network includes an input layer, a hidden layer and an output layer. Each layer includes a number of neurons. The connection between layers is completed through full connection between neurons. The number of neurons in the input layer is the same as the number of features in the feature data set.

8. A system for executing the intelligent dock crane operation time prediction method according to any one of claims 1 to 7, characterized in that: The system includes: Data preprocessing module: used to clean the data after matching the historical data of intelligent terminal crane operation and historical weather data, including removing missing values, abnormal values ​​and outliers; feature encoding the cleaned data to construct a crane operation time prediction data set; Feature engineering module: extracts features from the crane operation time prediction data set after feature encoding by the data preprocessing module, and constructs a feature data set that affects the crane operation time at the wharf; And the bridge crane operation time prediction module: the bridge crane operation time feature data set is input into the prediction model, and the bridge crane operation time prediction value is output.

9. A system according to claim 8, characterized in that: The prediction model includes a Monte Carlo prediction model and a neural network prediction model, wherein the Monte Carlo prediction model outputs a prediction value of the bridge crane operation time and a corresponding confidence interval; the neural network prediction model outputs a prediction value of the bridge crane operation time in real time.

10. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

Images