Railway container empty container allocation capacity parameter dynamic calibration method and system

Abstract

The present application provides a railway container empty container allocation capacity parameter dynamic calibration method and system, belongs to the railway container transportation technical field, obtains the historical empty container data of each handling station in the container use management information system, carries out the classical time sequence seasonal decomposition, identifies the seasonal cycle length, comprehensively considers the trend, seasonal term, error term, holiday term, uses the Prophet prediction model to predict the upper limit of the empty container quantity that can be discharged of the handling station in the planning period limited by the related capacity, calibrates the empty container return capacity, predicts the upper limit of the empty container quantity that can be adjusted into of the handling station in the planning period limited by the related capacity, and calibrates the container receiving capacity; according to the Prophet prediction of the historical empty container data in each period in the station, the handling station stacking capacity is calibrated. The present application can guide to formulate more scientific and reasonable railway container empty container allocation scheme, improve the accuracy and flexibility, reduce the work load of the dispatch personnel, and optimize the efficiency and benefit of the railway container transportation.

Description

Technical Field

[0001] This invention relates to the field of railway container allocation technology, specifically to a method and system for dynamically calibrating the empty container allocation capacity parameters of railway containers. This method is used to dynamically calibrate the empty container allocation capacity parameters of railway container handling stations, and to guide the development of more scientific, reasonable, and feasible railway empty container allocation plans. Background Technology

[0002] Determining the parameters of railway empty container allocation capacity is a crucial step in developing a railway empty container allocation plan. The purpose is to determine the operational capacity of container handling stations related to empty container allocation based on production data, and further integrate this capacity into the supply and demand calculation methods and empty container allocation models of railway container handling stations. This aims to obtain a more scientific and reasonable railway empty container allocation plan. Figure 1 As shown.

[0003] In railway transportation, the allocation of empty containers is a crucial step. Empty container allocation involves adjusting container flows that exhibit significant imbalances in space and time to prevent container shortages or backlogs at stations. It directly impacts railway efficiency and transportation costs, and empty container allocation plans are typically developed on a rolling five-day cycle. However, empty container allocation is actually constrained by many capacity factors, such as the storage capacity of container handling stations (the upper limit of container levels at the station within the planned cycle), container receiving capacity (the upper limit of the number of empty containers that can be received at the station within the planned cycle), and empty return capacity (the upper limit of the number of empty containers transferred from outside the station within the planned cycle). These capabilities guide the development of railway empty container allocation plans. Currently, railway production dispatchers, based on historical data and their own work experience, indirectly and implicitly incorporate these capacity parameters to some extent when manually developing empty container allocation plans after qualitative and quantitative analysis. However, due to differences in the work capabilities of dispatchers, and the influence of various uncertain factors such as road network conditions, empty vehicle allocation, and station operation capabilities (e.g., storage capacity is affected by the scale of the handling station's yard and loading and unloading capacity, container receiving capacity is affected by the station's operation capacity, and empty return capacity is affected by road network conditions and empty vehicle transport resources), the storage capacity, container receiving capacity, and empty return capacity obtained from rough analysis are not refined enough, which in turn leads to problems such as low fulfillment rate of the compiled empty container allocation plan.

[0004] In the daily operation and scheduling of railway container transportation, it is necessary to accurately calculate the relevant capacity parameters for empty container allocation. Determining the empty container allocation capacity parameters of each railway container handling station within the planned cycle—that is, the constraint and corrective effect of increasing empty container allocation capacity on the formulation of empty container allocation plans—can guide the optimization of railway container empty container allocation plans. Currently, my country uses methods including manual analysis and estimation, as well as quantitative methods that establish membership functions based on the fuzzy relationship between empty container transport volume and capacity limitations, to determine the empty return capacity and container receiving capacity of handling stations. Regarding the storage capacity of container handling stations, there are currently two main methods for determining the storage capacity of railway container handling stations in my country: one is the use of manual analysis and estimation, and the other is a design approach based on different perspectives to establish a fixed value.

[0005] (1) Manual analysis and calculation of the empty return capacity, container receiving capacity, and storage capacity of the handling station. At present, the daily transportation operation and empty container allocation of my country's railway containers mainly rely on manual work, which includes the container handling station manually analyzing and calculating to determine the empty container allocation capacity parameters. Specifically, regarding the empty return capacity, dispatchers calculate the upper limit of the number of empty containers that the handling station can transport and dispatch during the planned period based on the current empty car resources of the handling station and the estimated arrival of empty cars within the planned period, combined with the busyness and congestion of the connected railway lines (empty return capacity). Regarding the container receiving capacity, dispatchers consider the impact of factors such as the level of station equipment and the layout of facilities and equipment on the reception of arriving empty containers, and then determine the upper limit of the number of empty containers that the handling station can receive from other handling stations during the planned period (container receiving capacity). Finally, regarding the storage capacity, dispatchers ensure that the empty containers currently at the handling station do not exceed a certain warning line (storage capacity) based on historical data and their own work experience, combined with loading and unloading capacity. If it is foreseen that the number of empty containers at the station is approaching the warning line, the empty containers currently at the handling station will be promptly transferred to other stations. The empty container allocation capacity parameters obtained through this manual method are greatly affected by the varying abilities of dispatchers, resulting in low accuracy and susceptibility to personnel changes. Furthermore, it cannot handle the workload of continuously increasing railway container business volume.

[0006] (2) Fuzzy processing of empty container return capacity and container receiving capacity at handling stations. This implementation scheme assumes that the empty container allocation capacity parameter has a certain degree of fuzziness. Based on the fuzzy relationship between the empty container allocation volume and capacity limits, it proposes a membership function as a quantitative indicator for the empty container allocation capacity limit. However, this method is difficult to adapt to the constantly fluctuating international and domestic transportation market and has certain limitations.

[0007] (3) Set the storage capacity of the processing station to a fixed value through formula calculation. Most existing studies calculate the annual storage capacity of the processing station based on factors such as the total number of container yards on the ground, the average number of stacked layers in the yard, the number of working days per year, the average storage period of containers, and the container imbalance coefficient in the yard. However, since the calculated storage capacity is the annual storage capacity, it cannot meet the high spatiotemporal granularity and precision requirements of the current empty container allocation with a five-day planning cycle.

[0008] The current technology for determining the empty container allocation capacity parameters of railway container handling stations has the following main drawbacks: (1) When it comes to the empty container allocation problem in railway container transportation, accuracy and flexibility are key factors. Traditional manual analysis and calculation methods often require a lot of manpower and time, and are easily affected by subjective factors, resulting in inaccurate and unreliable results. (2) Although fuzzy processing methods can consider uncertain factors, their computational efficiency and accuracy are limited for complex transportation networks and large-scale datasets. (3) The method of setting the storage capacity of handling stations to a fixed value through formula calculation is relatively macroscopic and lacks sufficient spatiotemporal granular refinement. Summary of the Invention

[0009] The purpose of this invention is to provide a method and system for dynamically calibrating the empty container allocation capacity parameters of railway containers, so as to solve at least one of the technical problems existing in the background art.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] On the one hand, the present invention provides a method for dynamically calibrating parameters of railway container empty container allocation capacity, including:

[0012] Obtain historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle;

[0013] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be dispatched by the processing station during the planned cycle, which is subject to relevant capacity limitations. The return empty capacity in the empty container dispatching capacity parameter is dynamically calibrated.

[0014] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be transferred into the terminal during the planned cycle, which is subject to relevant capacity limitations. The container receiving capacity parameter in the empty container allocation capacity is dynamically calibrated.

[0015] Prophet predictions are made based on historical data of empty containers at the station for each period to determine the station's storage capacity.

[0016] Optionally, the classic time series seasonal decomposition of daily air-released data includes: initially setting the seasonal period; and decomposing the time series data using additive or multiplicative decomposition.

[0017] Optionally, additive factorization includes:

[0018] Calculating the trend term: Assuming that the periodic components exhibit the same characteristics in each period, the moving average method is used to calculate the trend term. That is, by periodically grouping the data, the average value of each group of data is calculated, thereby obtaining the trend of the data.

[0019] Calculate the detrended sequence: Perform detrending processing on the daily empty dataset of the original processing station. By eliminating the trend effect in the original dataset, a new column named "detrended" is generated.

[0020] Calculating the seasonal component: Calculate the average value based on the time granularity of the data's seasonality; subtract the average of the detrended values ​​from these average values; expand the number of results.

[0021] Calculate the residual term: Subtract the seasonal term from the detrended value.

[0022] Optionally, the formula for calculating the moving average is as follows:

[0023]

[0024] In the formula SMA t X represents the simple moving average at time point t. t-1 X represents the actual value in the previous period. t-2 X t-3 and X t-n These represent the actual values ​​for the previous two periods, the previous three periods, and up to the previous n periods, respectively, where n represents the number of periods in the moving average.

[0025] Optional, daily empty data seasonality period length identification, including:

[0026] In the classic time series seasonal decomposition, the seasonal period is initially set to weekly to decompose the daily outflow data. This yields the decomposition results of the time series data, including trend, seasonal, and residual terms, assuming a weekly seasonal period. Then, the decomposition evaluation index α, related to the residual term in this decomposition result, is calculated. w ;

[0027] Next, we set the seasonality period to monthly to decompose the daily data, obtaining time series data decomposition results containing trend, seasonal, and residual terms, assuming a monthly seasonality period. We then calculate the decomposition evaluation index α related to the residual terms in this decomposition result. m ;

[0028] min{α w ,α m} is compared with a set standard value, if min{α} w ,α m If the value is less than or equal to the standard value, then according to α w α m The size relationship determines the seasonal period length of daily air data: if α w If the α value is relatively smaller, then the seasonal period length (period) of the daily empty data extracted from the processing station is one week; if α m If the seasonal period length of the daily empty data from the processing station is relatively smaller, then the seasonal period length is set to one month; the extracted seasonal period length is used as the input cycle length parameter for the next Prophet prediction.

[0029] Conversely, if min{α w ,α m If the value is greater than the standard value, Prophet will be used to predict and automatically identify the cycle length parameter.

[0030] Optionally, calculate the evaluation index α. w and α m :

[0031]

[0032] In the formula, α is the evaluation index, in percentage (%). samples r is the number of data points in the time series data. t Let y be the t-th residual in the residual term of the decomposition result. t This represents the t-th data point in the original time series data.

[0033] Optional, daily empty data Prophet predictions, including:

[0034] Prophet considers trend, seasonality, holiday, and error terms to predict daily short-selling data for future periods.

[0035] y t =g(t)+s(t)+h(t)+∈ t

[0036] In the formula: g(t) represents the trend term, the non-periodic trend of the time series; s(t) represents the seasonal term, the periodic change in units of weeks, months, and years; h(t) represents the holiday term, indicating whether there is a holiday on that day; ∈ t This represents the error term, also known as the residual term; y t This represents the prediction result of the short-lived time series, which includes the predicted short-lived value, the upper bound of the prediction, and the lower bound of the prediction.

[0037] Secondly, the present invention provides a dynamic calibration system for railway container empty container allocation capacity parameters, comprising:

[0038] The acquisition module is used to acquire historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle.

[0039] The empty container return capacity prediction module is used to predict the upper limit of the number of empty containers that can be dispatched by the terminal during the planned cycle, based on the length of the seasonal cycle and taking into account trends, seasonal factors, error factors, and holiday factors. It uses the Prophet prediction model to dynamically calibrate the empty container return capacity in the empty container dispatch capacity parameters.

[0040] The container receiving capacity prediction module is used to predict the upper limit of the number of empty containers that the terminal can transfer in during the planned cycle, based on the length of the seasonal cycle, taking into account trends, seasonal factors, error factors, and holiday factors, using the Prophet prediction model. It dynamically calibrates the container receiving capacity in the empty container allocation capacity parameters.

[0041] The storage capacity prediction module is used to perform Prophet prediction based on historical empty container data for each period, and to calibrate the storage capacity of the processing station.

[0042] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the dynamic calibration method for railway container empty container allocation capacity parameters as described above.

[0043] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the dynamic calibration method for railway container empty container allocation capacity parameters as described above.

[0044] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the dynamic calibration method for railway container empty container allocation capacity parameters as described above.

[0045] The beneficial effects of this invention are as follows: Based on a large amount of historical data on container utilization and management, this invention proposes a series of technical methods for dynamically calibrating the empty container return capacity, container receiving capacity, and storage capacity of handling stations. These methods enable the calculation of empty container allocation capacity parameters that can adapt to the cyclical rolling development of empty container allocation plans, thus achieving dynamic calibration of railway container empty container allocation capacity parameters. The calibration results can guide the development of more scientific and reasonable railway container empty container allocation plans, improving their accuracy and flexibility, and reducing the workload of dispatchers, thereby optimizing the efficiency and effectiveness of railway container transportation.

[0046] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This diagram illustrates the role of empty container allocation capacity in the existing technology of empty container allocation process in railway containers.

[0049] Figure 2 This is a flowchart of the method for calibrating the return empty capacity parameters of the processing station according to an embodiment of the present invention.

[0050] Figure 3 This is a flowchart of the method for calibrating the capacity parameters of the container receiving station as described in an embodiment of the present invention.

[0051] Figure 4 This is a flowchart of the method for calibrating the storage capacity parameters of the processing station according to an embodiment of the present invention. Detailed Implementation

[0052] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0053] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0054] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0055] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0056] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0057] Example 1

[0058] In this embodiment 1, a dynamic calibration system for railway container empty container dispatching capacity parameters is first provided, including: an acquisition module, used to acquire historical empty container data of each handling station in the container utilization management information system, perform classical time series seasonal decomposition on the historical empty container data, and identify the seasonal cycle length; an empty return capacity prediction module, used to predict the upper limit of the number of empty containers that can be dispatched by the handling station under relevant capacity restrictions in the planned cycle based on the seasonal cycle length, comprehensively considering trends, seasonal items, error items, and holiday items, and using the Prophet prediction model, dynamically calibrating the empty return capacity in the container empty container dispatching capacity parameters; a container receiving capacity prediction module, used to predict the upper limit of the number of empty containers that can be transferred in by the handling station under relevant capacity restrictions in the planned cycle based on the seasonal cycle length, comprehensively considering trends, seasonal items, error items, and holiday items, and using the Prophet prediction model, dynamically calibrating the container receiving capacity in the container empty container dispatching capacity parameters; and a storage capacity prediction module, used to perform Prophet prediction based on historical empty container data at the station for each cycle, and calibrate the storage capacity of the handling station.

[0059] In this embodiment, the above-described system is used to dynamically calibrate the parameters of railway container empty container allocation capacity. This includes: acquiring historical empty container data from each processing station in the container utilization management information system; performing classical time-series seasonal decomposition on the historical empty container data and identifying the length of the seasonal cycle; based on the length of the seasonal cycle, comprehensively considering trends, seasonality, error factors, and holidays, using the Prophet prediction model to predict the upper limit of the number of empty containers that a processing station can dispatch within the planned cycle, subject to relevant capacity limitations, and dynamically calibrating the return empty container capacity parameter in the container empty container allocation capacity parameters; based on the length of the seasonal cycle, comprehensively considering trends, seasonality, error factors, and holidays, using the Prophet prediction model to predict the upper limit of the number of empty containers that a processing station can receive within the planned cycle, subject to relevant capacity limitations, and dynamically calibrating the receiving capacity parameter in the container empty container allocation capacity parameters; and performing Prophet prediction based on historical empty container data for each period to calibrate the storage capacity of the processing station.

[0060] Classic time series seasonality decomposition of daily air-release data includes: initially setting the seasonal period; and decomposing the time series data using additive or multiplicative decomposition. Additive decomposition includes:

[0061] Calculating the trend term: Assuming that the periodic components exhibit the same characteristics in each period, the moving average method is used to calculate the trend term. That is, by periodically grouping the data, the average value of each group of data is calculated, thereby obtaining the trend of the data.

[0062] Calculate the detrended sequence: Perform detrending processing on the daily empty dataset of the original processing station. By eliminating the trend effect in the original dataset, a new column named "detrended" is generated.

[0063] Calculating the seasonal component: Calculate the average value based on the time granularity of the data's seasonality; subtract the average of the detrended values ​​from these average values; expand the number of results.

[0064] Calculate the residual term: Subtract the seasonal term from the detrended value.

[0065] The formula for calculating the moving average is as follows:

[0066]

[0067] In the formula SMA t X represents the simple moving average at time point t. t-1 X represents the actual value in the previous period. t-2 X t-3 and X t-n These represent the actual values ​​for the previous two periods, the previous three periods, and up to the previous n periods, respectively, where n represents the number of periods in the moving average.

[0068] Identifying the seasonal cycle length of daily airborne data, including:

[0069] In the classic time series seasonality decomposition, the seasonality period is initially set to weekly to decompose the daily outflow data. This yields the decomposition results of the time series data, which includes trend, seasonal, and residual terms, assuming a weekly seasonality period. Then, the decomposition evaluation index α related to the residual term in the decomposition results is calculated. w ;

[0070] Next, we set the seasonality period to monthly to decompose the daily data, obtaining time series data decomposition results containing trend, seasonal, and residual terms, assuming a monthly seasonality period. We then calculate the decomposition evaluation index α related to the residual terms in this decomposition result. m ;

[0071] min{α w ,α m} is compared with a set standard value, if min{α} w ,α m If the value is less than or equal to the standard value, then according to α w α m The size relationship determines the seasonal period length of daily air data: if α w If the α value is relatively smaller, then the seasonal period length (period) of the daily empty data extracted from the processing station is one week; if α m If the seasonal period length of the daily empty data from the processing station is relatively smaller, then the seasonal period length is set to one month; the extracted seasonal period length is used as the input cycle length parameter for the next Prophet prediction.

[0072] Conversely, if min{α w ,α m If the value is greater than the standard value, Prophet will be used to predict and automatically identify the cycle length parameter.

[0073] Calculate the evaluation index α w and α m :

[0074]

[0075] In the formula, α is the evaluation index, in percentage (%). samples r is the number of data points in the time series data. t Let y be the t-th residual in the residual term of the decomposition result. t This represents the t-th data point in the original time series data.

[0076] Daily short-selling data Prophet predictions include:

[0077] Prophet considers trend, seasonality, holiday, and error terms to predict daily short-selling data for future periods.

[0078] y t =g(t)+s(t)+h(t)+∈ t

[0079] In the formula: g(t) represents the trend term, the non-periodic trend of the time series; s(t) represents the seasonal term, the periodic change in units of weeks, months, and years; h(t) represents the holiday term, indicating whether there is a holiday on that day; ε t This represents the error term, also known as the residual term; y t This represents the prediction result of the short-lived time series, which includes the predicted short-lived value, the upper bound of the prediction, and the lower bound of the prediction.

[0080] Since the upper limit of the forecast can accurately reflect the return capacity in the empty container allocation capacity parameter to a certain extent, the sum of the upper limit values ​​of the predicted air dispatch during the planned cycle is taken as the return capacity.

[0081] The following section details the steps for calculating and fitting each term.

[0082] (1) Calculate the trend term g(t)

[0083] The core component of Prophet is the trend term, which is used to analyze and fit non-periodic changes in time series. It offers two trend models: a saturated growth model and a piecewise linear model. These models accurately capture trend changes in time series, thereby improving forecast accuracy.

[0084] The saturated growth model does not exhibit an infinite upward trend, but rather reaches a saturation point after a certain threshold, with its saturation value dynamically changing over time. The piecewise linear model, on the other hand, cannot define a trend. However, both models incorporate varying degrees of assumptions and parameters to adjust smoothness, aiding in model optimization. Their calculation formulas are as follows:

[0085] Formula for saturated growth trend function:

[0086] Piecewise linear trend function formula: g lim (t)=(r+a(t)δ)t+(d+a(t)T γ )

[0087] In the formula: v represents the model carrying capacity; r represents the growth rate; δ and γ represent fitness; a(t) represents the number of times the mutation point changes before time t; and d represents the offset.

[0088] (2) Calculate the seasonal term s(t)

[0089] s(t) represents the periodic variation of the time series and can be used to simulate various periodic trends, such as weekly, monthly, and yearly changes. It is expressed using a Fourier series. The cycle length can be set here, such as the period after seasonal cycle length identification. If not set, it will be automatically identified by default. The specific calculation formula is as follows:

[0090]

[0091] In the formula: N represents the number of cycles used in the model; T represents the expected period length of the time series; 2n represents the number of parameters that need to be estimated to fit the seasonality; the setting of N needs to be considered in conjunction with T. For annual periodicity, T is set to 365.25 and N is 10. For weekly seasonality, T is set to 7 and N is set to 3. The larger N is, the better the fit to complex seasonality.

[0092] (3) Calculate the holiday term h(t)

[0093] Holidays and major events have a significant impact on time series forecasting. These impacts are usually predictable; for example, events like the Spring Festival travel rush and the National Day and May Day holidays, which see large increases in passenger traffic, will affect the capacity of railway container freight transport. However, these events are foreseeable. Incorporating these factors as prior knowledge into the model is crucial to improving its accuracy.

[0094] h(t) represents the impact of non-periodic, irregular holidays. By customizing the holiday list, the model can handle predictions under holiday or emergency scenarios, improving the accuracy of the return-to-empty capability parameter calibration. The principle is as follows:

[0095] h(t)=Z(t)k

[0096] In the formula: Z(t) is the indicator function; k represents the scope of influence of holidays.

[0097] (4) Calculate the error term ∈ t

[0098] Error term ∈ t This represents the noise component not reflected in the model and assumes that the noise factor follows a normal distribution.

[0099] (5) Obtain the upper bound of the prediction

[0100] The upper and lower bounds of the prediction are obtained by calculating the confidence interval, as follows: Based on the confidence level (usually 95%) and the standard deviation of the error term, the width of the confidence interval is calculated. The wider the confidence interval, the greater the uncertainty of the prediction. The upper and lower bounds of the prediction are obtained by adding or subtracting the width of the confidence interval from the predicted value.

[0101] Example 2

[0102] In this embodiment 2, a method for dynamically calibrating the empty container allocation capacity parameters of railway containers is provided. Specifically, this includes dynamic calibration of the empty container return capacity of the handling station, dynamic calibration of the container receiving capacity of the handling station, and dynamic calibration of the storage capacity of the handling station, ultimately achieving dynamic calibration of the empty container allocation capacity parameters of railway container handling stations. Wherein:

[0103] (1) Dynamic calibration technology for empty container return capacity of processing stations: After performing classical time series seasonal decomposition on the historical daily empty container dispatch data of each processing station obtained from statistics, and then obtaining the seasonal cycle length through seasonal cycle length identification technology, the Prophet prediction model is used to comprehensively consider trends, seasonal items, error items, holiday items, etc. to judge the upper limit of the number of empty containers that can be dispatched by the processing station under the relevant capacity restrictions in the planning cycle, and then dynamically calibrate the empty container return capacity in the empty container dispatch capacity parameter.

[0104] (2) Dynamic calibration technology for container receiving capacity of handling stations: After performing classical time series seasonal decomposition on the historical daily empty arrival data of each handling station obtained from statistics, and then obtaining the seasonal cycle length through seasonal cycle length identification technology, the Prophet prediction model is used to comprehensively consider trends, seasonal items, error items, holiday items, etc. to judge the upper limit of the number of empty containers that the handling station can transfer in the planned cycle due to relevant capacity restrictions, and then dynamically calibrate the container receiving capacity in the empty container allocation capacity parameter.

[0105] (3) Dynamic calibration technology for the storage capacity of the processing station. Considering that the data of empty boxes at the station is collected in the planned cycle, Prophet prediction is directly performed based on the historical data of empty boxes at the station in each cycle to calibrate the storage capacity of the processing station.

[0106] The three technologies mentioned above—dynamic calibration technology for the empty return capacity of processing stations, dynamic calibration technology for the container receiving capacity of processing stations, and dynamic calibration technology for the storage capacity of processing stations—are related and have certain operational processes.

[0107] (1) Data source: All three technologies require statistical analysis of the historical empty container data of each processing station in the container utilization management information system in order to obtain relevant parameters and indicators.

[0108] (2) Data Processing: Given the same data source, all three technologies require data processing and analysis to obtain the necessary prediction results and calibration parameters. Specifically, the dynamic calibration technology for the empty return capacity of the processing station and the dynamic calibration technology for the container receiving capacity of the processing station both require classical time series seasonal decomposition of historical data and identification of the seasonal cycle length in order to use the Prophet prediction model for prediction and analysis. The dynamic calibration technology for the storage capacity of the processing station, on the other hand, requires direct Prophet prediction of historical data to calibrate the storage capacity of the processing station.

[0109] (3) Parameter Influence: Although these three technologies calibrate different capacity parameters, the parameters among them will also have a certain influence. For example, in the dynamic calibration technology of the storage capacity of the processing station, the storage capacity will affect the calibration results of the empty return capacity and container receiving capacity of the processing station, because the storage capacity of the processing station directly determines whether the processing station can receive more empty containers or discharge more empty containers, thus affecting the calibration of the entire empty container allocation capacity parameter.

[0110] In this embodiment, the dynamic calibration technology for the return empty capacity of the processing station includes three parts, such as... Figure 2 As shown. The methods include: 1) Classical time-series seasonal decomposition technology for daily air dispatch data, which decomposes the components of the time-series data of air dispatches from processing stations to provide decomposition results for seasonal period length identification technology; 2) Seasonal period length identification technology for daily air dispatch data, which identifies the seasonal period length based on indicators related to the residual term in the decomposition results; and 3) Prophet prediction model technology for daily air dispatch data, which predicts daily air dispatches within future planning periods, so that the sum of the upper bounds of the predictions can be used to determine the dynamic return air dispatch capability of the processing station for that period.

[0111] For the classic time series seasonality decomposition technique for daily air release data, a time series is a sequence of data points arranged in chronological order and containing some information. Time series data typically includes characteristics such as data development trends (rising, falling, remaining unchanged) and patterns of change (seasonality). These characteristics usually have a certain regularity and predictability.

[0112] In this embodiment, time series data specifically has the following characteristics:

[0113] (1) Trend. Time series data usually exhibit a certain trend in data evolution. A certain variable, in the process of time or change of independent variable, shows a slow and long-term continuous upward, downward or unchanged trend, but the magnitude of the change may be different.

[0114] (2) Seasonality. A factor exhibits regular changes with peaks and troughs due to external influences, such as the changing of natural seasons. Specifically, seasonality in time series data refers to the periodic changes in the data, which are usually closely related to seasonal time nodes such as year, month, quarter, and week. Seasonal factors can generally be divided into additive seasonality and multiplicative seasonality.

[0115] Additive seasonality refers to the fact that the periodic variation in time series data does not change over time; in other words, the magnitude of data change remains constant regardless of time. Multiplicative seasonality, on the other hand, refers to the fact that the periodic variation in time series data changes over time; that is, the magnitude of data change exhibits a linear relationship with time. This phenomenon can be described as the magnitude of change in time series data varying with time.

[0116] Considering the characteristics of empty container allocation capacity at container handling stations, this invention sets the seasonality factor of the time series as additive seasonality in both the daily empty container dispatch data classical time series seasonal decomposition technology and the daily empty container arrival data classical time series seasonal decomposition technology.

[0117] (3) Randomness. At certain times, it exhibits random fluctuations, but overall it shows statistical regularity. Specifically, residuals refer to the part remaining after removing trend and seasonal characteristics from time series data. It is generally believed that the residuals of time series data with seasonal characteristics follow a normal distribution with a mean of 0. Residuals are regarded as a kind of white noise signal, so we can obtain residuals by gradually eliminating trend and seasonal characteristics in time series data.

[0118] Classical time series seasonality decomposition involves breaking down a time series into three parts: a trend term, a seasonal term, and a residual term. Assuming the time series is y, the trend is T, the seasonality is S, and the residual is R, then:

[0119] y = T + S + R

[0120] In this embodiment, the classic time series seasonal decomposition technique for daily empty data includes the following steps:

[0121] (1) Initially set the seasonal cycle. Time series data usually show a periodic trend, and this pattern of change is often caused by seasonal factors. If the time granularity of the time series data is time (such as hour), its seasonal cycle may be 24; if the time granularity is day, the seasonal cycle may be 7; if the time granularity is month, the seasonal cycle may be 12; if the time granularity is quarter, the seasonal cycle may be 4.

[0122] (2) Additive decomposition. Since the seasonality factors of time series can be divided into additive seasonality and multiplicative seasonality, two methods can be used to decompose time series data: additive decomposition and multiplicative decomposition.

[0123] However, since this embodiment has selected additive seasonality as the seasonality factor for the corresponding time series, only the details of the additive decomposition method will be introduced, including the following:

[0124] 1) Calculating the trend term. Assuming the periodic component exhibits the same characteristics within each period, the moving average method is used to calculate the trend term. This involves periodically grouping the data and calculating the average of each group to determine the data trend. A simple formula for calculating the moving average is as follows:

[0125]

[0126] In the formula SMA t X represents the simple moving average at time point t. t-1 X represents the actual value in the previous period. t-2 X t-3 and X t-n These represent the actual values ​​for the previous two periods, the previous three periods, and up to the previous n periods, respectively, where n represents the number of periods in the moving average.

[0127] 2) Calculate the detrended sequence. Detrend the daily empty dataset of the original processing station by eliminating the trend effect in the original dataset and generating a new column named "detrended".

[0128] 3) Calculate the seasonal component. The method for calculating the seasonal component is based on the following approach: ① Calculate the average value according to the time granularity of the data's seasonality (hour, day, month, quarter, year, etc.) (for example, if the data's seasonality is monthly, then calculate the average value of each month (January-December) across all years); ② Subtract the detrended average value from these average values; ③ Expand the number of results.

[0129] 4) Calculate the residual term. Calculating the residual term is relatively simple; just subtract the seasonal term from the detrended value.

[0130] For the technique of identifying the seasonality period length of daily air release data, the first step is to initially set the seasonality period to 7 in the classic time series seasonality decomposition to decompose the daily air release data, obtaining the time series data decomposition results containing trend, seasonal, and residual terms with the assumed seasonality period of 7. Then, the decomposition evaluation index α related to the residual term in this decomposition result is calculated. w Its definition is as follows:

[0131]

[0132] In the formula, α is the evaluation index, in percentage (%). samples r is the number of data points in the time series data. t Let y be the t-th residual in the residual term of the decomposition result. t This represents the t-th data point in the original time series data.

[0133] Next, we set the seasonality period to monthly to decompose the daily data, obtaining time series data decomposition results containing trend, seasonal, and residual terms, assuming a monthly seasonality period. Similarly, we calculate the decomposition evaluation index α related to the residual term in this decomposition result. m .

[0134] In obtaining the evaluation index α w and α m Then, min{α w ,α m} is compared with a set standard value (e.g., 10%). If min{α} w ,α m If the percentage is less than or equal to 10%, then according to α w α m The size relationship determines the seasonal period length of daily air data: if α w If it is smaller, then the seasonal period length of the daily empty data of the processing station is week (7); if α m If the value is smaller, then the seasonal cycle length (period) of the daily empty data from the processing station is extracted as month (30.5); the extracted seasonal cycle length (period) is used as the input cycle length parameter for the next Prophet prediction. Conversely, if min{α w ,α m If the percentage is greater than 10%, the seasonal cycle length (period) cannot be accurately determined at this time. Instead, Prophet will automatically identify the cycle length parameter.

[0135] For the Prophet forecasting model for daily air shipment data, as mentioned above, the time series decomposition method divides the time series into several parts: trend, seasonality, and residuals. Prophet has made necessary improvements and optimizations based on this approach to adapt to the fact that in real life and transportation operations, in addition to seasonality, trends, and randomness, holiday effects are also common influencing factors. Therefore, Prophet considers all four aspects simultaneously when predicting daily air shipment data for future periods, namely:

[0136] y t =g(t)+s(t)+h(t)+∈ t

[0137] In the formula: g(t) represents the trend term, the non-periodic trend of the time series; s(t) represents the seasonal term, generally the periodic change in units of weeks, months, and years; h(t) represents the holiday term, indicating whether there is a holiday on that day; ∈ tThis represents the error term, also known as the residual term. The Prophet forecasting method fits these terms and then sums them to obtain the forecast result of the air departure time series. The forecast result includes the predicted air departure value, the upper bound of the forecast, and the lower bound of the forecast. Since the upper bound of the forecast can accurately reflect the return air capacity in the empty container allocation capacity parameter to a certain extent, the sum of the upper bound values ​​of the predicted air departure for the planned period is taken as the return air capacity.

[0138] The following section details the steps for calculating and fitting each term.

[0139] (1) Calculate the trend term g(t)

[0140] The core component of Prophet is the trend term, which is used to analyze and fit non-periodic changes in time series. It offers two trend models: a saturated growth model and a piecewise linear model. These models accurately capture trend changes in time series, thereby improving forecast accuracy.

[0141] The saturated growth model does not exhibit an infinite upward trend, but rather reaches a saturation point after a certain threshold, with its saturation value dynamically changing over time. The piecewise linear model, on the other hand, cannot define a trend. However, both models incorporate varying degrees of assumptions and parameters to adjust smoothness, aiding in model optimization. Their calculation formulas are as follows:

[0142] Formula for saturated growth trend function:

[0143] Piecewise linear trend function formula: g lim (t)=(r+a(t)δ)t+(d+a(t)T γ )

[0144] In the formula: v represents the model carrying capacity; r represents the growth rate; δ and γ represent fitness; a(t) represents the number of times the mutation point changes before time t; and d represents the offset.

[0145] (2) Calculate the seasonal term s(t)

[0146] s(t) represents the periodic variation of the time series and can be used to simulate various periodic trends, such as weekly, monthly, and yearly changes. It is expressed using a Fourier series. The cycle length can be set here, such as the period after seasonal cycle length identification. If not set, it will be automatically identified by default. The specific calculation formula is as follows:

[0147]

[0148] In the formula: N represents the number of cycles used in the model; T represents the expected period length of the time series; 2n represents the number of parameters that need to be estimated to fit the seasonality; the setting of N needs to be considered in conjunction with T. For annual periodicity, T is set to 365.25 and N is 10. For weekly seasonality, T is set to 7 and N is set to 3. The larger N is, the better the fit to complex seasonality.

[0149] (3) Calculate the holiday term h(t)

[0150] Holidays and major events have a significant impact on time series forecasting. These impacts are usually predictable; for example, events like the Spring Festival travel rush and the National Day and May Day holidays, which see large increases in passenger traffic, will affect the capacity of railway container freight transport. However, these events are foreseeable. Incorporating these factors as prior knowledge into the model is crucial to improving its accuracy.

[0151] h(t) represents the impact of non-periodic, irregular holidays. By customizing the holiday list, the model can handle predictions under holiday or emergency scenarios, improving the accuracy of the return-to-empty capability parameter calibration. The principle is as follows:

[0152] h(t)=Z(t)k

[0153] In the formula: Z(t) is the indicator function; k represents the scope of influence of holidays.

[0154] (4) Calculate the error term ε t

[0155] Error term ε t This represents the noise component not reflected in the model and assumes that the noise factor follows a normal distribution.

[0156] (5) Obtain the upper bound of the prediction

[0157] The upper and lower bounds of the prediction are obtained by calculating the confidence interval, as follows: Based on the confidence level (usually 95%) and the standard deviation of the error term, the width of the confidence interval is calculated. The wider the confidence interval, the greater the uncertainty of the prediction. The upper and lower bounds of the prediction are obtained by adding or subtracting the width of the confidence interval from the predicted value.

[0158] In this embodiment, the dynamic calibration technology for the receiving capacity of the service station includes three parts, such as... Figure 3As shown. The methods include: 1) Classical time-series seasonal decomposition technology for daily arrival data, which decomposes the components of the arrival time-series data of the processing station to provide decomposition results for the seasonal period length identification technology; 2) Seasonal period length identification technology for daily arrival data, which identifies the seasonal period length based on indicators related to the residual term in the decomposition results; and 3) Prophet prediction model technology for daily arrival data, which predicts daily arrivals within the future planning period, so that the sum of the upper bounds of the predictions can be used to determine the dynamic receiving capacity of the processing station for that period.

[0159] The detailed steps of the dynamic calibration technology for the container receiving capacity of the processing station are similar to those of the dynamic calibration technology for the empty return capacity of the processing station. It is only necessary to replace the historical daily empty sending data entered at the beginning of the method with the historical daily empty arrival data, and finally determine the container receiving capacity of the processing station based on the sum of the predicted daily empty arrival upper limits for the future period.

[0160] In this embodiment, regarding the dynamic calibration technology for the storage capacity of processing stations, considering the application scenarios of storage capacity in the empty container allocation capacity parameters and the characteristics of the data collection time nodes for the current empty containers at the station in historical periods, the dynamic calibration technology for the storage capacity of processing stations of this invention is based on the current empty container data at the station in historical periods. After setting the data sampling time interval, a fitted Prophet model is created, and then the Prophet model is used to predict the current empty containers at the station in the next planning period. Specifically, as follows... Figure 4 As shown. Since the current empty container data is based on the empty container allocation cycle as the statistical interval, the obtained upper bound of the prediction is the processing station's storage capacity for the future planned cycle. The Prophet prediction principle and method in this dynamic calibration technology for processing station storage capacity are the same as the Prophet prediction model technology for daily empty container data, and will not be elaborated here.

[0161] Example 3

[0162] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, a method for dynamically calibrating parameters of railway container empty container allocation capacity is implemented. This method includes:

[0163] Obtain historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle;

[0164] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be dispatched by the processing station during the planned cycle, which is subject to relevant capacity limitations. The return empty capacity in the empty container dispatching capacity parameter is dynamically calibrated.

[0165] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be transferred into the terminal during the planned cycle, which is subject to relevant capacity limitations. The container receiving capacity parameter in the empty container allocation capacity is dynamically calibrated.

[0166] Prophet predictions are made based on historical data of empty containers at the station for each period to calibrate the station's storage capacity.

[0167] Example 4

[0168] This embodiment 4 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, and the memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute a method for dynamically calibrating railway container empty container allocation capacity parameters, the method including:

[0169] Obtain historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle;

[0170] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be dispatched by the processing station during the planned cycle, which is subject to relevant capacity limitations. The return empty capacity in the empty container dispatching capacity parameter is dynamically calibrated.

[0171] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be transferred into the terminal during the planned cycle, which is subject to relevant capacity limitations. The container receiving capacity parameter in the empty container allocation capacity is dynamically calibrated.

[0172] Prophet predictions are made based on historical data of empty containers at the station for each period to calibrate the station's storage capacity.

[0173] Example 5

[0174] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing a method for dynamically calibrating parameters of railway container empty container allocation capacity. This method includes:

[0175] Obtain historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle;

[0176] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be dispatched by the processing station during the planned cycle, which is subject to relevant capacity limitations. The return empty capacity in the empty container dispatching capacity parameter is dynamically calibrated.

[0177] Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be transferred into the terminal during the planned cycle, which is subject to relevant capacity limitations. The container receiving capacity parameter in the empty container allocation capacity is dynamically calibrated.

[0178] Prophet predictions are made based on historical data of empty containers at the station for each period to calibrate the station's storage capacity.

[0179] In summary, the calibration method provided in this invention is based on historical data from the container utilization management information system. After performing classical time series seasonal decomposition on the statistically obtained daily empty (arrival) data to obtain the seasonal cycle length, it uses the Prophet prediction model to dynamically calibrate the return empty capacity (receiving capacity) parameter in the empty container allocation capacity parameters, comprehensively considering trends, seasonal components, error components, and holiday components. Simultaneously, it directly performs Prophet prediction on the historical empty container data for each period to calibrate the storage capacity of the handling station. This invention includes key technologies such as dynamic calibration of the return empty capacity, dynamic calibration of the receiving capacity, and dynamic calibration of the storage capacity of the handling station, ultimately achieving accurate calculation of the empty container allocation capacity parameters of railway container handling stations. By dynamically and accurately measuring these fluctuating and complex capacity parameters for each planning cycle, this invention can improve the current situation of heavy workload and insufficient accuracy of manual methods, enabling the formulation of more scientific and feasible empty container scheduling schemes, and providing a data foundation for the future automated compilation of empty container allocation schemes.

[0180] This invention designs and integrates technologies such as time series decomposition, Prophet prediction for dynamic calibration of empty container return capacity of handling stations, dynamic calibration of container receiving capacity of handling stations, and dynamic calibration of storage capacity of handling stations. It forms a method for calibrating empty container allocation capacity parameters of railway container handling stations. This method can realize the dynamic calibration of empty container allocation capacity parameters of all container handling stations within the future empty container allocation plan cycle, and can be used to guide the formulation of scientific and reasonable railway container empty container allocation plans.

[0181] The key innovations of this invention are as follows:

[0182] (1) A method for identifying the seasonal cycle length of railway arrival-departure time series is proposed, which considers the trend, seasonality and randomness of the time series. The method for identifying the seasonal cycle length of the time series is based on the change pattern of container arrival-departure. The seasonal cycle is set to weeks and months to obtain different time series decomposition results. The relevant evaluation index is calculated for the residual terms in the decomposition results to identify the seasonal cycle length, which improves the prediction accuracy of the subsequent Prophet technology.

[0183] (2) The Prophet technique is used to obtain empty container allocation capacity parameters based on predicted upper bounds. First, classical time-series seasonal decomposition is performed on historical daily empty departure and arrival data from large-scale container operation management data. After identifying the seasonal cycle length, the obtained seasonal cycle length is input into the Prophet method as the cycle length for Prophet prediction. Then, the return empty capacity or container receiving capacity of the handling station is obtained based on the sum of the predicted upper bounds. Second, the Prophet method is used to predict the empty containers in the station for future planned cycles based on historical empty container data for each cycle. Then, the storage capacity of the handling station is determined based on the predicted upper bound. Using the robust and flexible Prophet technique to obtain predicted upper bounds for calibrating empty container allocation capacity parameters results in high accuracy.

[0184] (3) For the first time, dynamic calibration of empty container allocation capacity parameters of railway container handling stations has been achieved. This invention proposes a method for calibrating empty container allocation capacity parameters of railway container handling stations, including key technologies such as dynamic calibration of station return empty capacity, station receiving capacity, and station storage capacity. For the first time, dynamic calibration of empty container allocation capacity parameters of existing railway container handling stations has been achieved. This is of great significance for providing scientific basis for railway empty container allocation capacity and for providing basic data for the automated preparation of future empty container allocation plans.

[0185] This invention is the first to achieve dynamic calibration of empty container allocation capacity parameters at railway container handling stations. The core focus of this invention is the approach of dynamically mining and calibrating empty container allocation capacity based on historical container usage data. Previously, the empty container allocation capacity parameters considered when developing railway empty container allocation plans were all static, making it difficult to adapt to dynamic changes in the market and railway network capacity. This invention, however, employs a data-driven prediction-based dynamic calibration method, which can reflect dynamic changes under actual conditions and provide a more scientific and reasonable basis for railway empty container allocation.

[0186] Compared with existing technologies, this invention has the following two major advantages:

[0187] (1) Higher accuracy of calibration results. This invention integrates multi-dimensional historical and predictive data, such as historical daily outbound empty container data, historical daily arrival empty container data, and historical empty container data for each period at the container handling station. This enables comprehensive analysis and mining of the empty container allocation capacity parameters of the handling station. At the same time, for container handling stations with different seasonal cycle lengths, the corresponding seasonal cycle length after identification is used as the cycle length predicted by Prophet, so as to effectively improve the accuracy of calibration results.

[0188] (2) The calibration results are more consistent with actual railway transportation operations. This invention uses the Prophet method for prediction, which, in addition to incorporating the generally considered trend, seasonal, and error terms into the model, also considers the impact of holidays and calculates the holiday term. Holiday factors have a significant impact on railway empty container scheduling. For example, during holidays, increased railway passenger traffic necessitates additional passenger trains, making railway freight capacity on the line even more strained, and further leading to a certain degree of contraction in railway empty container allocation capacity. Therefore, Prophet prediction can more accurately reflect this impact compared to other prediction models, and is consistent with the daily railway empty container allocation approach.

[0189] (1) In this invention, the classic time series seasonal decomposition is used to decompose the historical daily empty or empty time series data of the processing station. In addition to the classic time series seasonal decomposition, there are also X11 decomposition method, SEATS decomposition method, STL decomposition method, fbprophet decomposition, etc., which can all be used for time series decomposition.

[0190] (2) In the present invention, the technologies for dynamic calibration of empty container return capacity, container receiving capacity, and storage capacity of processing stations use Prophet prediction to obtain the upper bound value of the prediction, and then calibrate the empty container allocation capacity parameters. In addition to Prophet prediction, the NeuralProphet combined prediction model can also be used to obtain the predicted value, the upper bound value of the prediction, and the lower bound value of the prediction, and then the empty container allocation capacity parameters are determined based on the upper bound value of the prediction.

[0191] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0192] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0193] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0194] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for dynamically calibrating parameters of railway container empty container allocation capacity, characterized in that, include: Obtain historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle; Identifying the seasonality period length of daily outbound data includes: initially setting the seasonality period to weekly in classic time series seasonality decomposition to decompose daily outbound data, obtaining time series data decomposition results containing trend, seasonal, and residual terms with the assumed seasonality period of weekly, and then calculating decomposition evaluation indicators related to the residual terms in the decomposition results. Next, the daily outflow data is decomposed by setting the seasonality period to monthly, obtaining time series data decomposition results containing trend, seasonal, and residual terms, assuming a monthly seasonality period. Decomposition evaluation indicators related to the residual terms in these results are then calculated. ; min{ , } is compared with a set standard value, if min{ , If the value is less than or equal to the standard value, then according to... , The size relationship determines the seasonal period length of daily air data: if If the seasonal period length (period) of the daily empty data extracted from the processing station is relatively smaller, then the seasonal period length is one week; if If the seasonal period length (period) of the daily empty data from the processing station is relatively smaller, then the extracted seasonal period length is set to months; this extracted seasonal period length is used as the input cycle length parameter for the next Prophet prediction; conversely, if min{ , If the value is greater than the standard value, then Prophet is used to predict and automatically identify the cycle length parameter. Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be dispatched by the processing station during the planned cycle, which is subject to relevant capacity limitations. The return empty capacity in the empty container dispatching capacity parameter is dynamically calibrated. Based on the length of the seasonal cycle, and taking into account trends, seasonality, error, and holidays, the Prophet prediction model is used to predict the upper limit of the number of empty containers that can be transferred into the terminal during the planned cycle, which is subject to relevant capacity limitations. The container receiving capacity parameter in the empty container allocation capacity is dynamically calibrated. Prophet predictions are performed based on historical empty container data for each period to determine the processing station's storage capacity; daily empty container data Prophet predictions include: Prophet considers trend, seasonality, holiday, and error terms to predict daily short-selling data for future periods. In the formula: This indicates the trend term, the non-periodic change trend of the time series; Indicates seasonal items, showing periodic changes in units of weeks, months, and years; The "Holiday" option indicates whether a public holiday applies on that day. This refers to the error term, also known as the residual term. This represents the prediction result of the short-lived time series, which includes the predicted short-lived value, the upper bound of the prediction, and the lower bound of the prediction.

2. The method for dynamically calibrating the parameters of railway container empty container allocation capacity according to claim 1, characterized in that, The classic time series seasonal decomposition of daily air-released data includes: initially setting the seasonal period; and decomposing the time series data using additive or multiplicative decomposition.

3. The method for dynamically calibrating the parameters of railway container empty container allocation capacity according to claim 2, characterized in that, Additive factorization includes: Calculating the trend term: Assuming that the periodic components exhibit the same characteristics in each period, the moving average method is used to calculate the trend term. That is, by periodically grouping the data, the average value of each group of data is calculated, thereby obtaining the trend of the data. Calculate the detrended sequence: Perform detrending processing on the daily empty dataset of the original processing station. By eliminating the trend effect in the original dataset, a new column named "detrended" is generated. Calculating the seasonal component: Calculate the average value based on the time granularity of the data's seasonality; subtract the average of the detrended values ​​from these average values; expand the number of results. Calculate the residual term: Subtract the seasonal term from the detrended value.

4. The method for dynamically calibrating the parameters of railway container empty container allocation capacity according to claim 3, characterized in that, The formula for calculating the moving average is as follows: In the formula This represents the simple moving average at time point t. This represents the actual value in the previous period. , and These represent the actual values ​​for the previous two periods, the previous three periods, and up to the previous n periods, respectively, where n represents the number of periods in the moving average.

5. The method for dynamically calibrating the parameters of railway container empty container allocation capacity according to claim 1, characterized in that, Calculate evaluation indicators and : In the formula The evaluation indicator is expressed as a percentage. The number of data points in the time series data. Let t be the t-th residual in the decomposition result. This represents the t-th data point in the original time series data.

6. A dynamic calibration system for railway container empty container allocation capacity parameters, characterized in that, include: The acquisition module is used to acquire historical empty container data from each processing station in the container utilization management information system, perform classic time series seasonal decomposition on the historical empty container data, and identify the length of the seasonal cycle. Identifying the seasonality period length of daily outbound data includes: initially setting the seasonality period to weekly in classic time series seasonality decomposition to decompose daily outbound data, obtaining time series data decomposition results containing trend, seasonal, and residual terms with the assumed seasonality period of weekly, and then calculating decomposition evaluation indicators related to the residual terms in the decomposition results. Next, the daily outflow data is decomposed by setting the seasonality period to monthly, obtaining time series data decomposition results containing trend, seasonal, and residual terms, assuming a monthly seasonality period. Decomposition evaluation indicators related to the residual terms in these results are then calculated. ; min{ , } is compared with a set standard value, if min{ , If the value is less than or equal to the standard value, then according to... , The size relationship determines the seasonal period length of daily air data: if If the seasonal period length (period) of the daily empty data extracted from the processing station is relatively smaller, then the seasonal period length is one week; if If the seasonal period length (period) of the daily empty data from the processing station is relatively smaller, then the extracted seasonal period length is set to months; this extracted seasonal period length is used as the input cycle length parameter for the next Prophet prediction; conversely, if min{ , If the value is greater than the standard value, then Prophet is used to predict and automatically identify the cycle length parameter. The empty container return capacity prediction module is used to predict the upper limit of the number of empty containers that can be dispatched by the terminal during the planned cycle, based on the length of the seasonal cycle and taking into account trends, seasonal factors, error factors, and holiday factors. It uses the Prophet prediction model to dynamically calibrate the empty container return capacity in the empty container dispatch capacity parameters. The container receiving capacity prediction module is used to predict the upper limit of the number of empty containers that the terminal can transfer in during the planned cycle, based on the length of the seasonal cycle, taking into account trends, seasonal factors, error factors, and holiday factors, using the Prophet prediction model. It dynamically calibrates the container receiving capacity in the empty container allocation capacity parameters. The storage capacity prediction module is used to perform prophet prediction based on historical empty container data for each period, and to calibrate the storage capacity of the processing station; daily empty container data prophet prediction includes: Prophet considers trend, seasonality, holiday, and error terms to predict daily short-selling data for future periods. In the formula: This indicates the trend term, the non-periodic change trend of the time series; Indicates seasonal items, showing periodic changes in units of weeks, months, and years; The "Holiday" option indicates whether a public holiday applies on that day. This refers to the error term, also known as the residual term. This represents the prediction result of the short-lived time series, which includes the predicted short-lived value, the upper bound of the prediction, and the lower bound of the prediction.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the dynamic calibration method for railway container empty container allocation capacity parameters as described in any one of claims 1-5.

8. An electronic device, characterized in that, include: The electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the dynamic calibration method for railway container empty container allocation capacity parameters as described in any one of claims 1-5.

Images