CN-122022026-A - Estuary channel freight volume prediction method based on system dynamics

CN122022026ACN 122022026 ACN122022026 ACN 122022026ACN-122022026-A

Abstract

The application provides a estuary channel freight volume prediction method based on system dynamics, which relates to the technical field of water freight volume prediction, and comprises the steps of defining boundary data of a target estuary channel, and retrieving freight volume historical data for data cleaning; the method comprises the steps of carrying out causal influence analysis on a estuary channel based on a plurality of historical freight volume data, carrying out variable identification according to causal relations, drawing a causal relation graph, constructing an estuary channel freight volume prediction model, verifying the estuary channel freight volume prediction model by using the plurality of historical freight volume data, and carrying out multi-scene simulation through the estuary channel freight volume prediction model when verification is passed, so as to generate a multi-scene estuary channel freight volume prediction result. The application solves the technical problem that in the prior art, the prediction model only depends on the time sequence rule of the historical data, neglects the transportation demand, causes inaccurate freight volume prediction, and improves the accuracy of the estuary channel freight volume prediction through multi-scene simulation prediction.

Inventors

GAO TIANHANG
Mu Changze
LIU CHANGJIAN
YUAN ZIWEN
REN JING
WU HONGYU
HUANG CHUAN
SHEN YIHUA
Yu Xunran
LIANG XIAO

Assignees

交通运输部规划研究院

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (8)

1. A estuary channel freight volume prediction method based on system dynamics is characterized by comprising the following steps: Defining boundary data for a target estuary channel, and calling historical freight traffic data according to the boundary data to perform data cleaning to obtain a plurality of historical freight traffic data; Performing causal influence analysis on the estuary channel based on the plurality of historical freight traffic data to obtain causal relations, performing variable identification according to the causal relations, and drawing a causal relation graph, wherein the causal relation graph comprises a main feedback loop; Constructing a system dynamics model based on the causal relationship graph, drawing a system flow graph through the system dynamics model to define variables, and constructing a estuary channel freight volume prediction model; and verifying the estuary channel freight volume prediction model by utilizing the plurality of historical freight volume data, and performing multi-scene simulation through the estuary channel freight volume prediction model when verification is passed, so as to generate a multi-scene estuary channel freight volume prediction result.
2. The estuary channel freight volume prediction method based on system dynamics of claim 1, wherein boundary data is defined for the target estuary channel, and the freight volume history data is called according to the boundary data for data cleaning to obtain a plurality of history freight volume data, the method comprising: defining space boundary data based on the starting and stopping points of the estuary channel and the main channel range; defining time boundary data based on the initial year of the historical statistical data as a historical period starting point and the final year of the historical statistical data as a historical period ending point; constructing boundary data of a target estuary channel according to the space boundary data and the time boundary data; Searching a historical freight database of the estuary channel according to the boundary data to obtain freight volume historical data; carrying out abnormal value identification on the freight volume historical data, and eliminating the freight volume historical data according to an abnormal identification result to obtain a first cleaning result; Carrying out missing identification on the freight volume historical data, and carrying out missing value alignment on the freight volume historical data according to a missing identification result to obtain a second cleaning result; and integrating the first cleaning result and the second cleaning result to construct the plurality of historical freight volume data.
3. The estuary channel freight rate prediction method based on system dynamics of claim 1, wherein causal influence analysis is performed on the estuary channel based on the plurality of historical freight rate data to obtain causal relationships, variable identification is performed according to the causal relationships, and a causal relationship graph is drawn, wherein the causal relationship graph comprises a main feedback loop, and the method comprises: Carrying out freight volume influence identification on the estuary channel based on the plurality of historical freight volume data, and generating a plurality of influence factors, wherein the plurality of influence factors comprise causal relationships; Dividing a plurality of cargo transportation subsystems according to the causal relationship, traversing the plurality of cargo transportation subsystems to identify internal and external association influences of the estuary channels, and constructing a variable set to be analyzed; Matching and relevant analysis is carried out on the plurality of historical freight traffic data and the variable set to be analyzed, and a data relevant coefficient is obtained; Performing mutual analysis based on the data correlation coefficient, determining an interaction direction, and constructing a plurality of variable influence paths according to the interaction direction; Traversing the variable influence paths to perform feedback analysis, determining a main feedback loop, dynamically associating the variable influence paths with the main feedback loop, and constructing a dynamic feedback structure; the dynamic feedback structure is added to the causal relationship graph.
4. A estuary channel freight method based on system dynamics as set forth in claim 3, wherein the interaction direction is determined based on the mutual analysis of the data correlation coefficients, the method comprising: Setting a first preset positive correlation threshold, a second preset positive correlation threshold, a third preset negative correlation threshold and a fourth preset negative correlation threshold; Wherein the first preset positive correlation threshold is greater than the second preset positive correlation threshold, and the third preset negative correlation threshold is greater than the fourth preset negative correlation threshold; when the data correlation coefficient is positive, and the absolute value of the data correlation coefficient is larger than a first preset positive correlation threshold value, generating a strong positive influence factor; Generating a weak positive influence factor when the data correlation coefficient is a positive number and the absolute value of the data correlation coefficient is smaller than or equal to a second preset positive correlation threshold; when the data correlation coefficient is a negative number and the absolute value of the data correlation coefficient is larger than a third preset negative correlation threshold value, generating a strong negative influence factor; When the data correlation coefficient is a negative number and the absolute value of the data correlation coefficient is smaller than or equal to a fourth preset negative correlation threshold value, generating a weak negative influence factor; and defining the action direction of the strong positive influence factor and the weak positive influence factor as positive promotion, defining the action direction of the strong negative influence factor and the weak negative influence factor as negative inhibition, and constructing the interaction direction.
5. A method of estuary channel freight rate prediction based on system dynamics as defined in claim 3, wherein the dynamic feedback structure is added to the causal graph, the method comprising: Setting a dominant adjustment frame based on the primary feedback loop; Synchronizing other variable influence paths as input or output to the dynamic feedback structure for analysis, and constructing a plurality of link parameters, wherein the links comprise parallel link parameters and single link parameters; and constructing a causal relationship graph based on the parallel link parameters and the mapping of the single link parameters to key nodes of the main feedback loop and the main regulation framework.
6. The estuary channel freight volume prediction method based on system dynamics according to claim 1, wherein a system dynamics model is constructed based on the causal relationship graph, a system flow graph is drawn through the system dynamics model to define variables, and an estuary channel freight volume prediction model is constructed, the method comprises: carrying out variable identification on the system dynamics model based on the causal relationship graph, and determining a variable type; Recognizing variable connection relations by combining the variable types with the causal relation graph, and performing graphical language processing according to the variable connection relations to construct a system flow graph; dynamically analyzing the variable based on the system flow graph to obtain a variable dynamic structure; Integrating the system flow graph according to the variable dynamic structure to construct a estuary channel freight volume prediction model.
7. The estuary channel freight volume prediction method based on system dynamics of claim 1, wherein the estuary channel freight volume prediction model is validated by utilizing the plurality of historical freight volume data, and when the validation is passed, multi-scene simulation is performed through the estuary channel freight volume prediction model to generate a multi-scene estuary channel freight volume prediction result, the method comprises: Synchronizing the plurality of historical freight volume data to the estuary channel freight volume prediction model for fitting calculation to obtain fitting freight volume predicted values of a plurality of historical time points; Introducing actual freight magnitude of a estuary channel according to a plurality of historical time points, and performing error calculation based on the fitted freight magnitude predicted value and the actual freight magnitude to obtain a fitting error rate; verifying the estuary channel freight volume prediction model based on the fitting error rate, and setting a plurality of freight scenes when verification is passed; Mapping the plurality of freight scenes to the estuary channel freight volume prediction model for simulation, and generating a multi-scene estuary channel freight volume prediction result.
8. The estuary channel freight rate prediction method based on system dynamics of claim 7, wherein the validating the estuary channel freight rate prediction model based on the fitting error rate, the method comprises: setting an error threshold based on the fitting error rate, and judging whether the fitting error rates of a plurality of historical time points are smaller than the error threshold; If the fitting error rates of the historical time points are smaller than the error threshold value, verifying the estuary channel freight traffic prediction model; And if any value of the fitting error rates of the historical time points is larger than the error threshold, the estuary channel freight volume prediction model is not verified, the estuary channel freight volume prediction model is adjusted, and the estuary channel freight volume prediction model is verified in an iteration mode according to an adjustment result until the fitting error rates of the historical time points are smaller than the error threshold.

Description

Estuary channel freight volume prediction method based on system dynamics Technical Field The application relates to the technical field of water freight traffic prediction, in particular to a estuary channel freight traffic prediction method based on system dynamics. Background Today, inland shipping has become an important component of a number of integrated transportation systems, with a large amount of shipping being carried by a number of inland rivers each year. From the perspective of transport organization of inland navigation, inland navigation mainly comprises two parts, namely, section transport in inland and river intermodal transport connecting coastal and inland ports. In order to greatly promote river and sea intermodal transportation, the management department continuously optimizes the navigation rules of the estuary channel on one hand, improves the navigation efficiency of the channel, inputs a large amount of funds to carry out the estuary channel treatment engineering on the other hand, and improves the estuary channel passing capacity and the channel stability by improving the estuary channel conditions. From the viewpoint of the freight demand in the service coastal area, what extent the estuary channel management engineering needs to be carried out, what standard the channel water depth and channel width are, and what ton-level ship navigation is guaranteed depend on the prediction result of the estuary channel water freight traffic. In the current research of the estuary channel freight traffic or water freight traffic prediction, the main stream method adopts a time series prediction model with strong universality, and mainly two technical routes exist, namely a non-machine learning model represented by a gray prediction method, a regression prediction method and the like, and a machine learning model represented by a neural network model. In any case, the time series prediction model predicts only based on the trend change of the freight volume data, and neglects the effect of factors such as the deepest transportation demand change in the freight volume forming mechanism, so that a large deviation occurs in the prediction result. In summary, in the prior art, the prediction model only depends on the time sequence rule of the historical data, so that the transportation requirement is ignored, and the technical problem of inaccurate freight volume prediction exists. Disclosure of Invention The application aims to provide a estuary channel freight volume prediction method based on system dynamics, which is used for solving the technical problem that in the prior art, the freight volume prediction is inaccurate because a prediction model only depends on a historical data time sequence rule and the transportation demand is ignored. The application provides a estuary channel freight volume prediction method based on system dynamics, which comprises the steps of defining boundary data of a target estuary channel, calling freight volume historical data according to the boundary data to conduct data cleaning to obtain a plurality of historical freight volume data, conducting causal effect analysis on the estuary channel based on the plurality of historical freight volume data to obtain causal relations, conducting variable identification according to the causal relations, drawing a causal relation graph, wherein the causal relation graph comprises a main feedback loop, constructing a system dynamics model based on the causal relation graph, defining variables through a system flow graph drawn by the system dynamics model, constructing an estuary channel freight volume prediction model, verifying the estuary channel freight volume prediction model by utilizing the plurality of historical freight volume data, and conducting multi-scene simulation through the estuary channel freight volume prediction model when verification is passed, and generating a multi-scene estuary channel freight volume prediction result. The method comprises the steps of defining space boundary data based on a start point and a stop point of a estuary channel and a main channel range, defining time boundary data based on a start year of historical statistical data as a historical period starting point and an end year of the historical statistical data as a historical period ending point, constructing boundary data of a target estuary channel according to the space boundary data and the time boundary data, searching a historical freight database of the estuary channel according to the boundary data to obtain freight volume historical data, carrying out outlier identification on the freight volume historical data, removing the freight volume historical data according to an outlier identification result to obtain a first cleaning result, carrying out missing identification on the freight volume historical data, carrying out missing value supplement on the freight volume historical data according to a missing identificat