A port planning method and system based on big data

Abstract

The invention discloses a port planning method and system based on big data, comprising: continuously acquiring port AIS data as port ship behavior characteristic data, establishing a time series model based on ship traffic; statistically analyzing the time pattern of ship occupation of berths, acquiring the time interval of ship entry and navigation, using Poisson distribution to analyze the probability density of port ships for the time interval, and acquiring water area data of port operation area; performing principal component analysis on the water area data of the port operation area, acquiring the characteristic value of ship behavior in the port area, and constructing a port throughput prediction model according to the change of time series; establishing a port planning investment optimization model, acquiring a port planning decision that adapts to the actual needs of the port according to the optimization value of the port planning investment optimization model, constructing a ship traffic flow combination prediction model, acquiring a high-precision ship traffic flow prediction value, being able to optimize the planning of port ship navigation information, and providing decision support for port channel design.

Description

Technical Field

[0001] The present invention relates to the technical field of port planning, and in particular to a port planning method and system based on big data. Background Art

[0002] Maritime transport plays an important role in global cargo transportation. As an important hub of maritime transport, the service efficiency of ports is related to whether the entire supply chain system can operate efficiently, which has aroused great concern from all parties about the efficiency of port ship services. In the evaluation of world-class ports, it is difficult to compare the ship service efficiency of various ports horizontally due to the lack of unified evaluation indicators for ship service efficiency and differences in data statistical caliber.

[0003] At present, port service efficiency data mainly comes from statistics from port enterprises, shipping companies, and research and consulting institutions. However, these data are statistically delayed and have inconsistent statistical calibers, making it difficult to conduct horizontal comparisons of the service efficiency of various ports. Selecting data that can simultaneously reflect the service efficiency of global ports and ships and proposing effective measurement methods has become a common concern in the industry.

[0004] The existing port planning decision-making methods are mainly based on AIS data to measure the port ship service efficiency. They can estimate six key performance indicators in real time, including ship flow, berth usage and allocation, loading and unloading efficiency, throughput, and ship collision risk in the port area. However, there are still the following problems in the analysis of a large amount of historical AIS data and the relationship between ship navigation planning and port investment optimization:

[0005] (1) In the process of AIS data extraction, due to the large amount of historical data, it is difficult to remove data noise, and the accuracy of the obtained ship traffic flow characteristics is relatively low, resulting in relatively low accuracy of ship navigation prediction;

[0006] (2) At present, the definition and classification of ship behavior are mostly limited to ship navigation behavior, using the characteristics of ship motion to distinguish ship behavior patterns, without considering the perspective of the port ship movement process. Although it can effectively improve the project efficiency in the implementation stage of waterway construction, it is unable to provide information such as ship waiting time, number of ships waiting to berth, port service level, etc. in the early planning and design stage, and cannot provide corresponding decision-making support for the optimization of waterway design schemes. Summary of the invention

[0007] The purpose of the present invention is to provide a port planning method and system based on big data, so as to solve the technical problems in the prior art that the prediction accuracy of ship navigation in the prior art is relatively low, and the information such as ship waiting time, number of ships waiting to berth, port service level, etc. cannot be provided in the early planning and design stage of the port, and the corresponding decision-making cannot be provided for the optimization of the port channel design scheme.

[0008] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0009] A first aspect of the present invention provides a port planning method based on big data, comprising the following steps:

[0010] Continuously acquiring port AIS data as port ship behavior characteristic data, establishing a time series model based on ship traffic volume according to the time sequence of ships arriving at the port, and acquiring a ship dynamic data table;

[0011] Analyze the time interval of ships arriving at the port in the ship dynamic data table, count the time pattern of ships occupying berths, obtain the time interval of ships entering the port, use Poisson distribution to analyze the probability density of ships in the port for the time interval, input the probability density of ships in the port into the background server through the identification field, call the data screening function to complete the extraction of specific spatial features of the port, and obtain the water area data of the port operation area;

[0012] By performing principal component regression analysis on the water area data of the port operation area, the ship behavior characteristic value of the port operation area is obtained, the gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series to obtain the port throughput demand prediction value;

[0013] An improved genetic algorithm is used to establish a port planning investment optimization model for the port throughput demand forecast value, and a port planning decision that adapts to the actual needs of the port is obtained based on the optimization value of the port planning investment optimization model.

[0014] As a preferred solution of the present invention, a time series is extracted from the ship behavior characteristic data according to the time of arrival at the port, the traffic volume of the ship in the port area is obtained, and a time series model based on the ship traffic volume is constructed according to the traffic volume, including:

[0015] Classify the ship behavior characteristic data according to ship type, arrival time and destination to establish an AIS dynamic data table structure, match the ship behavior characteristic data according to the MMSI identifier in the AIS dynamic data table structure to obtain an AIS data table with both dynamic and static attributes;

[0016] The AIS data table is grouped and counted by a counter in different time periods every day, and the distribution law of the time interval between two ships before and after the ship in the same operation area of ​​the port is analyzed to obtain the time interval between the ships arriving at the port;

[0017] Establishing a time series based on the ship behavior characteristic data according to the ship arrival time within the time interval, and using the automatic correlation function ACF on the time series to obtain the correlation of the ship arrivals at equal intervals;

[0018] A time series model based on ship traffic is established according to the correlation of the ship arrival at the port, the time series model adopts a SARIMA time series structure to obtain a non-stationary time series, and a ship dynamic data table is obtained according to the non-stationary time series;

[0019] The SARIMA time series structure obtains the correlation through the automatic correlation function ACF of the ship arrival time, and uses the correlation to establish the test statistic to obtain the non-stationary time series Q L :

[0020]

[0021] Where m represents the time series of the ship arrival time, n represents the number of ships arriving at the port, i (i≤n) represents the i-th ship arrival, p 2 (i) represents the autocorrelation function of the residual sequence of the i-th ship, is the autocorrelation function of the time observation series, χ 2 (n) represents the autocorrelation function of all ships.

[0022] As a preferred solution of the present invention, the time interval of ship arrival at the port is obtained according to the ship dynamic data table, the time pattern of ship occupation of berths is counted, and the port ship traffic organization model of ships entering the port and navigating in the same time interval is obtained, including:

[0023] Extracting the starting time of the ship's arrival at the port from the ship dynamic data table, and queuing the arriving ships in sequence according to the starting time;

[0024] Determine the time that a ship occupies a berth in a port area according to the type of ship, and analyze the berth occupation time to obtain the dispatch time of different types of ships in the port;

[0025] Determine whether the ship within the scheduling time has received the scheduling instruction. If the scheduling instruction is received, select a time node within the scheduling time to schedule the ship, and the ship scheduling instruction is determined according to the actual speed of the corresponding ship and the navigation direction of the port area; if the scheduling instruction is not received, the corresponding ship continues to wait;

[0026] The arrival time nodes of the ships receiving the dispatching instructions within the dispatching time are counted, the number of navigable ships within the dispatching time is obtained, and the ship traffic organization model is planned in the port area according to the number of navigable ships.

[0027] As a preferred solution of the present invention, the navigation time interval of all navigable ships in the port area is obtained according to the ship traffic organization model, and the probability density of port ships is analyzed using Poisson distribution for the navigation time interval, including:

[0028] The navigation time of all ships in the ship traffic organization model is taken as the unit time of ships entering the port, the number of ships passing through is counted, the time interval of ships passing through is analyzed to obtain the law of ships arriving at the port, and the probability density Q of ships in the port is analyzed using the Poisson distribution function j , the expression is:

[0029]

[0030] Among them, t represents the unit time of the ship entering the port, Q j represents the probability that the number of ships arriving at time point t is j, and λ represents the average number of ships arriving at time point t.

[0031] As a preferred solution of the present invention, AnyLogic software is used to simulate the probability density of the ships in the port to obtain the navigation distribution characteristics of the ships in the port area. The navigation distribution characteristics are input into the background server through the identification field to call the data screening function to complete the extraction of the port's specific spatial characteristics and obtain the water area data of the port operation area.

[0032] As a preferred solution of the present invention, the gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function for the ship behavior characteristic value, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series, including:

[0033] The ship behavior characteristic value is collected as sample data, and the ship behavior characteristic value is expressed as (x ij ) m*n Matrix, m represents the number of rows of the ship behavior characteristic value, n represents the number of columns of the ship behavior characteristic value, for the (x ij ) m*n The matrix is ​​subjected to principal component analysis PCA to obtain the standard deviation of the characteristic values ​​of the ship behavior;

[0034] According to the standard deviation, Cov((x ij ) function obtains the covariance matrix of the principal component index, calculates the eigenvalues ​​of the covariance matrix based on the principal component index, and sorts the eigenvalues;

[0035] The sorted eigenvalues ​​are input into the GM (1,1) model to obtain port throughput forecast data.

[0036] As a preferred solution of the present invention, a residual optimization combination model is established for the port throughput forecast data to predict the port throughput demand value, including:

[0037] Sorting the port ship traffic volume data according to the eigenvalues ​​of the principal component index, obtaining a ship traffic flow data sequence, determining the stationarity of the ship traffic flow data sequence, calculating the difference order of the ship traffic flow data sequence using the SARIMA time series structure, and determining the order of the residual optimization model according to the difference order;

[0038] Establishing a ship traffic flow prediction model based on the SARIMA model in the SARIMA time series structure, and calculating the output value of the ship traffic flow prediction model to obtain ship traffic flow prediction residual data;

[0039] Inputting the ship traffic flow prediction residual data into the BLSTM neural network model, training the ship traffic flow prediction residual data, obtaining a maximum number of training times according to a loss function in the BLSTM neural network model, and obtaining test data of the BLSTM neural network model after the maximum number of training times is met;

[0040] The test data is used as a throughput demand prediction value for ship traffic flow.

[0041] As a preferred solution of the present invention, a weight allocation model is established for the port planning area according to the throughput demand forecast value, and an improved genetic algorithm is used to establish a port planning investment optimization model to obtain a port planning decision that adapts to the actual needs of the port, including:

[0042] The throughput demand forecast value is quantitatively analyzed for the number of ship traffic accidents, and the ship accident forecast value y is obtained by assigning appropriate weights to obtain the weight allocation model t , the expression is:

[0043] y t =β1y1+β2y2

[0044] Among them, β1 and β2 represent the weight value of the linear weight distribution model and the weight value of the nonlinear weight distribution model respectively, y1 and y2 represent the prediction results of the linear model and the prediction results of the nonlinear model respectively;

[0045] The ship accident prediction value y t The weight value of the combined model is calculated with the minimum RMSE as the indicator, and the ship traffic flow prediction result based on the optimal weighted combined model is obtained.

[0046] As a preferred solution of the present invention, an improved genetic algorithm is used to obtain the ship traffic type and operating cost according to the ship traffic flow prediction result, and a linear model of the service efficiency of the port planning area and the scale of the port is established according to the ship throughput, and the expression is:

[0047]

[0048] in, represents the service efficiency of port k in year t, represents the initial service efficiency of port k, β represents the mean of the weight values ​​of the linear weight allocation model and the nonlinear weight allocation model, represents the unit time cost of transportation to port k in year t, It represents the total cost of cargo at port k in year t.

[0049] A second aspect of the present invention is an analysis system for the port planning decision-making method based on big data according to any one of claims 1 to 9, comprising:

[0050] A data acquisition unit is used to continuously acquire port AIS data to obtain a set of data characterizing the behavior characteristics of port ships, and to collect the arrival time of port ships to obtain a ship traffic time series based on the ship traffic volume;

[0051] A model management unit, which obtains the ship arrival time according to the time series model of the ship traffic volume, manages the ship traffic organization model in sequence according to the ship arrival time, and obtains the port throughput prediction value to realize the planning of ship navigation in the port area;

[0052] The investment optimization unit obtains the service efficiency of the port under the condition of adapting to the ship throughput based on the planning of ship navigation in the port area and the linear model of the service efficiency of the port planning area and the scale of the port, so as to evaluate the investment efficiency of the future planning of the port

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention adopts the SARIMA model to predict and analyze the port ship traffic flow, and combines the residual optimization combination prediction method to construct a ship traffic flow combination prediction model, which can effectively process massive AIS data, obtain high-precision ship traffic flow data, and obtain high-precision ship traffic flow prediction values.

[0055] The construction of a simulation model of the port ship traffic organization system using AnyLogic software improves the intuitiveness of the model, effectively obtains comprehensive water data of the port operation area, optimizes the planning of port ship navigation information, and provides appropriate decision support for the design of port ship passage channels. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0057] Figure 1 A flow chart of a port planning decision method provided by an embodiment of the present invention;

[0058] Figure 2 A system structure diagram of a port planning decision-making method provided by an embodiment of the present invention;

[0059] 1. Data collection unit; 2- Model management unit; 3. Investment optimization unit. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] like Figure 1 As shown, the present invention provides a port planning method based on big data, comprising the following steps:

[0062] Continuously acquiring port AIS data as port ship behavior characteristic data, establishing a time series model based on ship traffic volume according to the time sequence of ships arriving at the port, and acquiring a ship dynamic data table;

[0063] In this embodiment, the port AIS data is the real value of uninterrupted monthly ship traffic flow. The port AIS data is plotted into a line graph based on the ship traffic flow, which is convenient for observing and analyzing the fluctuations and changing trends of the data. The ship behavior characteristic data based on the ship traffic flow is extracted according to the line graph, so that the acquisition of ship behavior characteristic data is more comprehensive and can effectively provide ship port data.

[0064] In this embodiment, the time series model uses the navigation time of adjacent ships and the irreversible time sequence as the time attribute to obtain the correlation in the ship navigation process. It can effectively obtain the correlation between adjacent ships in the port at equal intervals and can effectively measure the smoothness of the ship navigation.

[0065] Analyze the time interval of ships arriving at the port in the ship dynamic data table, count the time pattern of ships occupying berths, obtain the time interval of ships entering the port, use Poisson distribution to analyze the probability density of ships in the port for the time interval, input the probability density of ships in the port into the background server through the identification field, call the data screening function to complete the extraction of specific spatial features of the port, and obtain the water area data of the port operation area;

[0066] In this embodiment, a Poisson probability distribution model of port ships is established according to the time interval of ships entering the port for navigation, so as to effectively obtain the distribution law of port ships.

[0067] By performing principal component regression analysis on the water area data of the port operation area, the ship behavior characteristic value of the port operation area is obtained, the gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series to obtain the port throughput demand prediction value;

[0068] In this embodiment, principal component regression analysis is performed on the water area data of the port operation area to mainly obtain the eigenvalues ​​of the ship data, and the eigenvalues ​​are processed using a gray number differential function to obtain the gray amount of the ship behavior eigenvalues. This allows key indicators to be screened out from the massive water area data of the port operation area for modeling, making the port throughput prediction model more accurate.

[0069] An improved genetic algorithm is used to establish a port planning investment optimization model for the port throughput demand forecast value, and a port planning decision that adapts to the actual needs of the port is obtained based on the optimization value of the port planning investment optimization model.

[0070] In this embodiment, the port planning investment optimization model mainly analyzes the port's attractiveness to cargo and the port's ship navigation capacity based on the data characteristics of the port's ships. In addition to the port's ship transportation costs, it can also comprehensively consider the time cost of land cargo and the hinterland shippers' accessibility to land and sea areas, and obtain the connection between the port's service efficiency and the port's scale.

[0071] Extract the time series of the ship behavior characteristic data according to the time of arrival at the port, obtain the traffic volume of the ship in the port area, and build a time series model based on the ship traffic volume according to the traffic volume, including:

[0072] Classify the ship behavior characteristic data according to ship type, arrival time and destination to establish an AIS dynamic data table structure, match the ship behavior characteristic data according to the MMSI identifier in the AIS dynamic data table structure to obtain an AIS data table with both dynamic and static attributes;

[0073] In this embodiment, the AIS data table is constructed and designed based on the relational database management system MySQL, and all three types of data tables related to the system application are stored in a database file named "aisdata", including 3 initial data, 5 data stored in the data processing module, and 17 data stored in other modules.

[0074] The AIS data table is grouped and counted by a counter in different time periods every day, and the distribution law of the time interval between two ships before and after the ship in the same operation area of ​​the port is analyzed to obtain the time interval between the ships arriving at the port;

[0075] In this embodiment, the time interval for ships to arrive at the port is obtained by performing first-order difference on the time series of ship traffic flow to obtain a differential sequence, and the corresponding time interval for ships to arrive at the port is obtained in the differential sequence.

[0076] Establishing a time series based on the ship behavior characteristic data according to the ship arrival time within the time interval, and using the automatic correlation function ACF on the time series to obtain the correlation of the ship arrivals at equal intervals;

[0077] A time series model based on ship traffic is established according to the correlation of the ship arrival at the port, the time series model adopts a SARIMA time series structure to obtain a non-stationary time series, and a ship dynamic data table is obtained according to the non-stationary time series;

[0078] The SARIMA time series structure obtains the correlation through the automatic correlation function ACF of the ship arrival time, and uses the correlation to establish the test statistic to obtain the non-stationary time series Q L :

[0079]

[0080] Where m represents the time series of the ship arrival time, n represents the number of ships arriving at the port, i (i≤n) represents the i-th ship arrival, p 2 (i) represents the autocorrelation function of the residual sequence of the i-th ship, is the autocorrelation function of the time observation series, χ 2 (n) represents the autocorrelation function of all ships.

[0081] In this embodiment, the model value of the SARIMA time series structure is set to SARIMA (2, 1, 1) (1, 1, 1) 13 The AIC value is the smallest, and SARIMA (2,1,1) (1,1,1) is finally selected. 13The model is the optimal model to predict ship traffic flow data. A white noise check is performed on the established SARIMA model sequence. From the ACF and PACF diagrams of the residual structure sequence of the SARIMA model, it can be seen that the residual sequence value of the ship travel movement flow fluctuates around 0. All the nodes of the ACF and PACF diagrams fall within the 97% confidence interval, which proves that they are unrelated.

[0082] According to the ship dynamic data table, the time interval of ship arrival at the port is obtained, the time pattern of ship occupation of berth is counted, and the port ship traffic organization model of ships entering the port and navigating at the same time interval is obtained, including:

[0083] Extracting the starting time of the ship's arrival at the port from the ship dynamic data table, and queuing the arriving ships in sequence according to the starting time;

[0084] Determine the time that a ship occupies a berth in a port area according to the type of ship, and analyze the berth occupation time to obtain the dispatch time of different types of ships in the port;

[0085] Determine whether the ship within the scheduling time has received the scheduling instruction. If the scheduling instruction is received, select a time node within the scheduling time to schedule the ship, and the ship scheduling instruction is determined according to the actual speed of the corresponding ship and the navigation direction of the port area; if the scheduling instruction is not received, the corresponding ship continues to wait;

[0086] The arrival time nodes of the ships receiving the dispatching instructions within the dispatching time are counted, the number of navigable ships within the dispatching time is obtained, and the ship traffic organization model is planned in the port area according to the number of navigable ships.

[0087] In this embodiment, the port vessel traffic organization system is regarded as a queuing system for research, and the port vessel traffic organization system after transformation also constitutes a port vessel traffic organization queuing system model.

[0088] In this embodiment, when the port ship traffic organization system is regarded as a queuing system, the ships entering and leaving the port can be converted into customers in the queuing system; the ships stranded in the anchorages and berths outside the port during the process of entering and leaving the port queue up to form a queue; and the berths, waterways, and converging waters in the port can all be regarded as service agencies.

[0089] The navigation time interval of all navigable ships in the port area is obtained according to the ship traffic organization model, and the probability density of port ships is analyzed by Poisson distribution for the navigation time interval, including:

[0090] The navigation time of all ships in the ship traffic organization model is taken as the unit time of ships entering the port, the number of ships passing through is counted, the time interval of ships passing through is analyzed to obtain the law of ships arriving at the port, and the probability density Q of ships in the port is analyzed using the Poisson distribution function j , the expression is:

[0091]

[0092] Among them, t represents the unit time of the ship entering the port, Q j represents the probability that the number of ships arriving at time point t is j, and λ represents the average number of ships arriving at time point t.

[0093] In this embodiment, when the ships in the ship traffic organization model are regarded as customers of the queuing system, when the ships enter the port, the arrival pattern of the ships is the arrival pattern of the customers. According to the probability density function of the Poisson distribution of the ship arrival pattern, the Poisson distribution of the ship arrival is obtained, and the time intervals between two adjacent arriving ships are made independent of each other. The time that the ship occupies the terminal berth is regarded as the service time of the terminal berth, which is mainly the sum of the ship's berthing time, operating time and unberthing time. Common ship types in specific ports are selected as representative ship types to calculate the time required for the terminal berth.

[0094] The probability density of ships in the port is simulated by AnyLogic software to obtain the navigation distribution characteristics of ships in the port area. The navigation distribution characteristics are input into the background server through the identification field to call the data screening function to complete the extraction of port-specific spatial characteristics and obtain the water area data of the port operation area.

[0095] In this embodiment, in order to build a simulation model, the relevant modules of the process modeling library in AnyLogic are mainly used to realize the construction of a simulation model of the port ship traffic organization system, which improves the intuitiveness of the model; and can associate controls and parameters, and adjust them in real time during the simulation process, with good interactivity.

[0096] The gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series, including:

[0097] The ship behavior characteristic value is collected as sample data, and the ship behavior characteristic value is expressed as (x ij ) m*n Matrix, m represents the number of rows of the ship behavior characteristic value, n represents the number of columns of the ship behavior characteristic value, for the (x ij ) m*n The matrix is ​​subjected to principal component analysis PCA to obtain the standard deviation of the characteristic values ​​of the ship behavior;

[0098] According to the standard deviation, Cov((x ij ) function obtains the covariance matrix of the principal component index, calculates the eigenvalues ​​of the covariance matrix based on the principal component index, and sorts the eigenvalues;

[0099] The sorted eigenvalues ​​are input into the GM (1,1) model to obtain port throughput forecast data.

[0100] In this embodiment, principal component analysis PCA is used to reduce the dimension of multiple variables into relatively few independent variables, which can largely ensure and change the problems caused by multicollinearity. Relatively independent variables can be obtained through the dimension reduction operation method, and the information of the original variables can be retained to a large extent.

[0101] In this embodiment, the GM (1,1) model is used to obtain the port throughput forecast data, which can provide confidence and confidence intervals. At the same time, in the process of regression analysis, the correlation between various influencing factors and the forecast target is fully considered, which is of great significance for improving the credibility of the forecast.

[0102] A residual optimization combination model is established for the port throughput forecast data to forecast the port throughput demand value, including:

[0103] Sorting the port ship traffic volume data according to the eigenvalues ​​of the principal component index, obtaining a ship traffic flow data sequence, determining the stationarity of the ship traffic flow data sequence, calculating the difference order of the ship traffic flow data sequence using the SARIMA time series structure, and determining the order of the residual optimization model according to the difference order;

[0104] In this embodiment, the stationarity test and differential processing of the SARIMA time series are performed to determine the number of differences to be 1, the order of the residual optimization model is determined by observing the autocorrelation function graph (ACF) and the partial autocorrelation function graph (PACF) of the SARIMA stationary structure order, and Python modeling is used to calculate the AIC values ​​of all parameter combinations and find the minimum value to determine the optimal model combination.

[0105] Establishing a ship traffic flow prediction model based on the SARIMA model in the SARIMA time series structure, and calculating the output value of the ship traffic flow prediction model to obtain ship traffic flow prediction residual data;

[0106] Inputting the ship traffic flow prediction residual data into the BLSTM neural network model, training the ship traffic flow prediction residual data, obtaining a maximum number of training times according to a loss function in the BLSTM neural network model, and obtaining test data of the BLSTM neural network model after the maximum number of training times is met;

[0107] The test data is used as a throughput demand prediction value for ship traffic flow.

[0108] In this embodiment, the ship traffic flow prediction model uses residual data as the basic data for model prediction, and adopts the residual optimization combined prediction method to construct a ship traffic flow combined prediction model. In this combined prediction method, the linear model plays a leading role and the nonlinear model plays an auxiliary role, which effectively avoids the influence of the influencing factors of ship traffic flow on the prediction accuracy and gives full play to their respective advantages.

[0109] A weight allocation model is established for the port planning area according to the throughput demand forecast value, and an improved genetic algorithm is used to establish a port planning investment optimization model to obtain a port planning decision that adapts to the actual needs of the port, including:

[0110] The throughput demand forecast value is quantitatively analyzed for the number of ship traffic accidents, and the ship accident forecast value y is obtained by assigning appropriate weights to obtain the weight allocation model t , the expression is:

[0111] y t =β1y1+β2y2

[0112] Among them, β1 and β2 represent the weight value of the linear weight distribution model and the weight value of the nonlinear weight distribution model respectively, y1 and y2 represent the prediction results of the linear model and the prediction results of the nonlinear model respectively;

[0113] The ship accident prediction value y t The weight value of the combined model is calculated with the minimum RMSE as the indicator, and the ship traffic flow prediction result based on the optimal weighted combined model is obtained.

[0114] In this embodiment, a weight distribution model is used to quantitatively analyze the number of ship traffic accidents, and the estimated values ​​of each model in the combined estimation model are obtained by assigning appropriate weights, thereby obtaining the estimated value of the accident, realizing the ship traffic flow prediction of the SARIMA model and the ship traffic flow prediction of the BLSTM neural network model, and effectively verifying the effectiveness of the residual optimization combination model. The residual part of the SARIMA prediction result is used as the initial data of the model for simulation training, and the prediction result of the combination of BLSTM neural network and random forest feature selection is combined with the prediction result of SARIMA as the final combined model prediction result, which effectively improves the prediction accuracy.

[0115] According to the ship traffic flow prediction results, the improved genetic algorithm is used to obtain the ship traffic type and operating cost, and the service efficiency of the port planning area and the scale of the port are established according to the ship throughput. The expression is:

[0116]

[0117] in, represents the service efficiency of port k in year t, represents the initial service efficiency of port k, β represents the mean of the weight values ​​of the linear weight allocation model and the nonlinear weight allocation model, represents the unit time cost of transportation to port k in year t, It represents the total cost of cargo at port k in year t.

[0118] In this embodiment, a linear model of the service efficiency of the port planning area and the scale of the port is established through the ship throughput, the port planning and investment costs are clarified, and the port revenue, port construction cost, inland transportation cost of the shipper's cargo and port time cost are combined into one model. When the port group is at the optimal scale, the port revenue and congestion cost are offset. From the perspective of the highest port ship service efficiency and the larger the port scale, when the ship transportation demand is effectively predicted, a corresponding regional port planning investment model is provided to obtain a port planning investment plan.

[0119] like Figure 2 As shown, the present invention provides an analysis system for the port planning decision-making method based on big data according to any one of claims 1 to 9, comprising:

[0120] The data acquisition unit 1 is used to continuously acquire the port AIS data to obtain a set of data representing the behavior characteristics of the port ships, and to collect the arrival time of the port ships to obtain the ship traffic time series based on the ship traffic volume;

[0121] Model management unit 2, obtaining the ship arrival time according to the time series model of the ship traffic volume, managing the ship traffic organization model in sequence according to the ship arrival time, and obtaining the port throughput prediction value to realize the planning of ship navigation in the port area;

[0122] The investment optimization unit 3 obtains the service efficiency of the port under the condition of adapting to the ship throughput based on the planning of ship navigation in the port area and the linear model of the service efficiency of the port planning area and the scale of the port, so as to evaluate the investment efficiency of the future planning of the port.

[0123] The present invention adopts the SARIMA model to predict and analyze the port ship traffic flow, and combines the residual optimization combination prediction method to construct a ship traffic flow combination prediction model, which can effectively process massive AIS data, obtain high-precision ship traffic flow data, and obtain high-precision ship traffic flow prediction values.

[0124] The construction of a simulation model of the port ship traffic organization system using AnyLogic software improves the intuitiveness of the model, effectively obtains comprehensive water data of the port operation area, optimizes the planning of port ship navigation information, and provides appropriate decision support for the design of port ship passage channels.

[0125] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A port planning method based on big data, characterized in that: The following steps are involved: Continuously acquiring port AIS data as port ship behavior characteristic data, establishing a time series model based on ship traffic volume according to the time sequence of ships arriving at the port, and acquiring a ship dynamic data table; Analyze the time interval of ships arriving at the port in the ship dynamic data table, count the time pattern of ships occupying berths, obtain the time interval of ships entering the port, use Poisson distribution to analyze the probability density of ships in the port for the time interval, input the probability density of ships in the port into the background server through the identification field, call the data screening function to complete the extraction of specific spatial features of the port, and obtain the water area data of the port operation area; By performing principal component regression analysis on the water area data of the port operation area, the ship behavior characteristic value of the port operation area is obtained, the gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series to obtain the port throughput demand prediction value; Using an improved genetic algorithm to establish a port planning investment optimization model for the port throughput demand forecast value, and obtaining a port planning decision that adapts to the actual needs of the port based on the optimization value of the port planning investment optimization model; The gray quantity of the ship behavior characteristic value is obtained based on the ship time series by using the gray number differential function, and the port throughput prediction model is constructed according to the change of the gray quantity in the time series, including: The ship behavior characteristic value is collected as sample data, and the ship behavior characteristic value is expressed as (x ij ) m*n Matrix, m represents the number of rows of the ship behavior characteristic value, n represents the number of columns of the ship behavior characteristic value, for the (x ij ) m*n The matrix is ​​subjected to principal component analysis PCA to obtain the standard deviation of the characteristic values ​​of the ship behavior; According to the standard deviation, Cov((x ij ) function obtains the covariance matrix of the principal component index, calculates the eigenvalues ​​of the covariance matrix based on the principal component index, and sorts the eigenvalues; Inputting the sorted characteristic values ​​into the GM (1,1) model to obtain port throughput forecast data; A residual optimization combination model is established for the port throughput forecast data to forecast the port throughput demand value, including: Sorting the port ship traffic volume data according to the eigenvalues ​​of the principal component index, obtaining a ship traffic flow data sequence, determining the stationarity of the ship traffic flow data sequence, calculating the difference order of the ship traffic flow data sequence using a SARIMA time series structure, and determining the order of the residual optimization model according to the difference order; Establishing a ship traffic flow prediction model based on the SARIMA model in the SARIMA time series structure, and calculating the output value of the ship traffic flow prediction model to obtain ship traffic flow prediction residual data; Inputting the ship traffic flow prediction residual data into the BLSTM neural network model, training the ship traffic flow prediction residual data, obtaining a maximum number of training times according to a loss function in the BLSTM neural network model, and obtaining test data of the BLSTM neural network model after the maximum number of training times is met; Using the test data as a throughput demand forecast value for ship traffic flow; A weight allocation model is established for the port planning area according to the throughput demand forecast value, and an improved genetic algorithm is used to establish a port planning investment optimization model to obtain a port planning decision that adapts to the actual needs of the port, including: The throughput demand forecast value is quantitatively analyzed for the number of ship traffic accidents, and the ship accident forecast value y is obtained by assigning appropriate weights to obtain the weight distribution model t , the expression is: y t =β1y1+β2y2 Among them, β1 and β2 represent the weight value of the linear weight distribution model and the weight value of the nonlinear weight distribution model respectively, y1 and y2 represent the prediction results of the linear model and the prediction results of the nonlinear model respectively; The ship accident prediction value y t The weight value of the combined model is calculated with the minimum RMSE as the indicator, and the ship traffic flow prediction result based on the optimal weighted combined model is obtained.

2. According to the big data-based port planning method of claim 1, it is characterized in that: Extract the time series of the ship behavior characteristic data according to the time of arrival at the port, obtain the traffic volume of the ship in the port area, and build a time series model based on the ship traffic volume according to the traffic volume, including: Classify the ship behavior characteristic data according to ship type, arrival time and destination to establish an AIS dynamic data table structure, match the ship behavior characteristic data according to the MMSI identifier in the AIS dynamic data table structure to obtain an AIS data table with both dynamic and static attributes; The AIS data table is grouped and counted by a counter in different time periods every day, and the distribution law of the time interval between two ships before and after the ship in the same operation area of ​​the port is analyzed to obtain the time interval between the ships arriving at the port; Establishing a time series based on the ship behavior characteristic data according to the ship arrival time within the time interval, and using the automatic correlation function ACF on the time series to obtain the correlation of the ship arrivals at equal intervals; A time series model based on ship traffic is established according to the correlation of the ship arrival at the port, the time series model adopts a SARIMA time series structure to obtain a non-stationary time series, and a ship dynamic data table is obtained according to the non-stationary time series; The SARIMA time series structure obtains the correlation through the automatic correlation function ACF of the ship arrival time, and uses the correlation to establish the test statistic to obtain the non-stationary time series Q L : Where m represents the time series of the ship arrival time, n represents the number of ships arriving at the port, i (i≤n) represents the i-th ship arrival, p 2 (i) represents the autocorrelation function of the residual sequence of the i-th ship, is the autocorrelation function of the time observation series, χ 2 (n) represents the autocorrelation function of all ships.

3. The port planning method based on big data according to claim 2 is characterized in that: According to the ship dynamic data table, the time interval of ship arrival at the port is obtained, the time pattern of ship occupation of berth is counted, and the port ship traffic organization model of ships entering the port and navigating at the same time interval is obtained, including: Extracting the starting time of the ship's arrival at the port from the ship dynamic data table, and queuing the arriving ships in sequence according to the starting time; Determine the time that a ship occupies a berth in a port area according to the type of ship, and analyze the berth occupation time to obtain the dispatch time of different types of ships in the port; Determine whether the ship within the scheduling time has received the scheduling instruction. If the scheduling instruction is received, select a time node within the scheduling time to schedule the ship, and the ship scheduling instruction is determined according to the actual speed of the corresponding ship and the navigation direction of the port area; if the scheduling instruction is not received, the corresponding ship continues to wait; The arrival time nodes of the ships receiving the dispatching instructions within the dispatching time are counted, the number of navigable ships within the dispatching time is obtained, and the ship traffic organization model is planned in the port area according to the number of navigable ships.

4. The port planning method based on big data according to claim 3 is characterized in that: The navigation time interval of all navigable ships in the port area is obtained according to the ship traffic organization model, and the probability density of port ships is analyzed by Poisson distribution for the navigation time interval, including: The navigation time of all ships in the ship traffic organization model is taken as the unit time of ships entering the port, the number of ships passing through is counted, the time interval of ships passing through is analyzed to obtain the law of ships arriving at the port, and the probability density Q of ships in the port is analyzed using the Poisson distribution function j , the expression is: Among them, t represents the unit time of the ship entering the port, Q j represents the probability that the number of ships arriving at time point t is j, and λ represents the average number of ships arriving at time point t.

5. The port planning method based on big data according to claim 3 is characterized in that: The probability density of ships in the port is simulated by AnyLogic software to obtain the navigation distribution characteristics of ships in the port area. The navigation distribution characteristics are input into the background server through the identification field to call the data screening function to complete the extraction of port-specific spatial characteristics and obtain the water area data of the port operation area.

6. The port planning method based on big data according to claim 5 is characterized in that: According to the ship traffic flow prediction results, the improved genetic algorithm is used to obtain the ship traffic type and operating cost, and the service efficiency of the port planning area and the scale of the port are established according to the ship throughput. The expression is: in, represents the service efficiency of port k in year t, represents the initial service efficiency of port k, β represents the mean of the weight values ​​of the linear weight allocation model and the nonlinear weight allocation model, represents the unit time cost of transportation to port k in year t, It represents the total cost of cargo at port k in year t.

7. An analysis system for the port planning decision method based on big data according to any one of claims 1 to 6, characterized in that: include: A data acquisition unit (1) is used to continuously acquire port AIS data to obtain a set of data representing the behavior characteristics of port ships, and to acquire the arrival time of port ships to obtain a ship traffic time series based on the ship traffic volume; A model management unit (2) is used to obtain the ship arrival time according to the time series model of the ship traffic volume, manage the ship traffic organization model in sequence according to the ship arrival time, and obtain the port throughput prediction value to realize the planning of ship navigation in the port area; The investment optimization unit (3) obtains the service efficiency of the port under the condition of adapting to the ship throughput based on the planning of ship navigation in the port area and the linear model of the service efficiency of the port planning area and the scale of the port, so as to evaluate the investment efficiency of the future planning of the port.

Images