Ship destination port and cargo flow direction prediction method based on shipping network and medium
By preprocessing and semantic information extraction of ship AIS data, combining long-term and short-term memory neural network model and shipping network data, we identify key nodes in the shipping network, and solve the accuracy and real-time prediction of ship destination ports and cargo flow directions in the existing technology. The problem of insufficient achieves in-depth understanding and accurate prediction of ship motion patterns.
Patent Information
- Application Number
- CN202411236092.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-12-20
AI Technical Summary
The existing technology lacks an in-depth understanding of ship behavior and sea traffic flow in the prediction of ship destination ports and cargo flow directions, resulting in poor prediction accuracy and real-time performance, and the inability to effectively identify complex maritime traffic patterns and navigation behaviors, making it difficult to achieve accuracy. forecast of ship destination port and cargo flow direction.
By obtaining ship AIS data for preprocessing and semantic information extraction, the ship shipping network data is generated, and inputting it into the long-term and short-term memory neural network multi-factor model, the key nodes of the shipping network are identified using trajectory morphology analysis method and geographic information system data, and the construction of the first The first and second prediction models are used to output the prediction results of the ship's destination port and cargo flow direction.
It has achieved an in-depth understanding of the ship's movement mode, improved the accuracy and real-time prediction of ship's destination port and cargo flow direction, and can scientifically dispatch the flow direction and destination port of the ship, which is suitable for scientific dispatch of maritime transportation.
Smart Images
Figure CN119168141A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of ship destination port and cargo flow prediction, and in particular relates to a method and medium for predicting ship destination port and cargo flow based on a shipping network. Background Art
[0002] The shipping network is a complex network system involving multiple elements such as ports, ships, cargo and routes. With the development of global trade and the increasing frequency of ocean transportation, transportation network prediction has become an important issue in the field of shipping. The existing methods for predicting ship destination ports and cargo flow directions have the following main shortcomings: First, most of the existing studies are based on ports, predicting the port berthing of all ships or a certain type of ships at a certain port, and lack research on the next destination port for a certain ship during navigation. Second, the prediction accuracy and real-time performance are not ideal. At present, the research basis of shipping is mainly to analyze the point trajectory of ships using longitude and latitude coordinates. Although this method can provide basic information on the location of ships, it lacks an in-depth understanding of ship behavior and maritime traffic flow. This analysis method only stays on the surface of geographic space, ignoring the semantic level or higher logical level information behind the ship trajectory, resulting in the inability to effectively identify and express complex maritime traffic patterns, navigation behaviors and their changes, and it is even more difficult to analyze the spatial organization relationship of maritime transportation activities from a semantic height to achieve ship destination and flow prediction.
[0003] Chinese invention patent CN116805172B discloses a method for predicting crude oil ship destination port driven by AIS data. Through AIS data, the position of the ship is tracked after the ship departs from the oil loading port, and its destination port is predicted, which solves the problem of low accuracy of prediction results caused by the lack of spatial information and ship transportation preference information in the prediction process of crude oil ship transportation destination port; at the same time, by constructing a ship spatial information extraction network and a ship preference information extraction network, and fusing them with the ship dynamic information extraction network to construct an information fusion network for predicting the destination port, the prediction accuracy is improved. However, the above method cannot realize the prediction of cargo flow direction, and requires complex algorithms and a large amount of computing resources, which may be difficult to deploy in some resource-constrained environments. Therefore, it is necessary to design a prediction method for ship destination port and cargo flow direction, deeply understand the movement mode of the ship, and accurately predict the ship destination port and cargo flow direction. Summary of the invention
[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method and medium for predicting the ship's destination port and cargo flow direction based on the shipping network, deeply understand the movement pattern of the ship, and accurately predict the ship's destination port and cargo flow direction.
[0005] The purpose of the present invention can be achieved by the following technical solutions:
[0006] The present invention provides a method for predicting a ship's destination port and cargo flow direction based on a shipping network, comprising the following steps:
[0007] Obtaining and preprocessing ship AIS data, wherein the ship AIS data includes the port information of the ship, the time of arrival and departure of the ship, the ship draft and the ship's water mobile service identification code, and the port information of the ship includes the port code and the longitude and latitude of the port;
[0008] Extracting semantic information from the preprocessed ship AIS data to obtain ship shipping network data, wherein the ship shipping network data includes ship track data and shipping network key node data;
[0009] The ship shipping network data is input into a long short-term memory neural network multi-factor model to obtain prediction results of the ship's destination port and cargo flow direction.
[0010] Furthermore, the specific process of the pretreatment is as follows:
[0011] Delete the data with wrong number of digits of ship's maritime mobile service identification code;
[0012] Delete the data with wrong number of port codes;
[0013] Symbolize the port's longitude and latitude.
[0014] Furthermore, the long short-term memory neural network multi-factor model is constructed based on the long short-term memory neural network, including a first prediction model and a second prediction model. The first prediction model takes the ship shipping network data as input and outputs a prediction result of the ship's destination port.
[0015] Furthermore, the second prediction model takes the prediction result of the ship's destination port and the change in the ship's draft as input, and outputs the prediction result of the cargo flow direction, wherein the change in the ship's draft is calculated based on the ship's draft in the AIS data.
[0016] Furthermore, when training the long short-term memory neural network multi-factor model, shipping network data of ships whose proportion of frequent port calls exceeds a set threshold are selected as training samples.
[0017] Furthermore, the calculation formula for the proportion of frequently called ports is as follows:
[0018]
[0019] Among them, p is the proportion of frequent ports, n is the total number of ports called by the ship during the study period, and n oThe number of ports whose call times exceed the set threshold.
[0020] Furthermore, the generation process of the ship shipping network data is specifically as follows:
[0021] Obtain the pre-processed AIS data, draw the ship's trajectory using the trajectory morphology analysis method, and obtain the ship's trajectory data;
[0022] Obtain geographic information system data, segment the drawn ship trajectory according to the bifurcation points, make a route node marking line segment on each ship trajectory, identify the key nodes on the ship trajectory in combination with the geographic information system data, and obtain the key node data of the shipping network.
[0023] Furthermore, the route node marking line segment on each route segment forms an angle of 90 degrees with the ship track.
[0024] Furthermore, the key nodes include stop points, entry and exit points, and route turning points.
[0025] The present invention also provides a computer-readable storage medium on which a computer program is stored, and the program implements the above method when executed by a processor.
[0026] Compared with the prior art, the present invention has the following beneficial effects:
[0027] 1. The present invention proposes a prediction method for a ship's destination port and cargo flow direction based on a shipping network. First, the ship's AIS data is obtained and preprocessed. The ship's AIS data has the advantages of strong real-time performance and comprehensive information, including the ship's port information, the ship's arrival and departure time, the ship's draft and the ship's water mobile service identification code. The ship's port information includes the port code and the port's longitude and latitude; then, semantic information is extracted from the preprocessed ship AIS data to obtain the ship's shipping network data. The ship's shipping network data includes the ship's trajectory data and the shipping network's key node data, which helps to deeply understand the ship's behavior; finally, the ship's shipping network data is input into a long short-term memory neural network multi-factor model. The long short-term memory neural network multi-factor model is constructed based on the long short-term memory neural network, which can effectively capture the long-term dependency in the input information and improve the accuracy of the prediction results. According to the prediction results of the ship's destination port and the cargo flow direction, the flow direction of the cargo carried by the ship and the destination port can be judged in advance to achieve scientific scheduling of maritime transportation.
[0028] 2. In the present invention, the specific process of obtaining ship shipping network data is as follows: obtaining pre-processed AIS data, drawing the trajectory of the ship using the trajectory morphology analysis method, and obtaining the ship trajectory data, which can intuitively display the movement of the ship and realize visual analysis; segmenting the drawn ship trajectory according to the bifurcation point, and making a route node marking line segment on each segment of the ship trajectory. The route node marking line segment on each route is at a 90-degree angle to the ship trajectory, which can prevent the direction of the ship's travel at the route node from coinciding with the marking line, resulting in an inability to determine which route node the ship is currently passing through; combining the geographic information system data to identify the key nodes on the ship trajectory, the key nodes include stop points, entry and exit points, and route turning points. The key node data of the shipping network finally obtained is helpful to accurately identify the position of the ship and fully reflect and understand the behavior of the ship.
[0029] 3. In the present invention, the long short-term memory neural network multi-factor model includes a first prediction model and a second prediction model. The first prediction model takes the ship shipping network data as input and outputs the prediction result of the ship's destination port. The second prediction model takes the prediction result of the ship's destination port and the ship's draft change calculated according to the ship's draft in the AIS data as input, and outputs the prediction result of the cargo flow direction. This hierarchical structure helps to gradually refine the prediction results and improve the overall prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 is a flow chart of the method of the present invention;
[0031] Figure 2 This is a schematic diagram of ship shipping network data, where L represents the route node marking line segment, S represents the ship trajectory segment, and D represents the key node. DETAILED DESCRIPTION
[0032] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.
[0033] Example:
[0034] This embodiment provides a method for predicting the destination port and cargo flow of ships based on the shipping network. Figure 1 As shown, the following steps are included:
[0035] S1. Obtain ship AIS data and perform preprocessing.
[0036] Ship AIS data has the advantages of strong real-time performance and comprehensive information, including the port information of the ship, the time of arrival and departure of the ship, the ship's draft and the ship's marine mobile service identification code MMSI. The port information of the ship includes the port name, port code and port longitude and latitude. Due to the long data duration, the large number of ships and the number of ports docked, the AIS data is relatively complex and may contain errors, so the data needs to be pre-processed to improve the data quality. The processing steps are as follows:
[0037] First, delete the data with incorrect MMSI digits (normally it should be 9 digits), and delete the data with obvious errors in the port code or port name (for example, the port code digits are incorrect, or the port name is not in the port name list). Then, the longitude and latitude of the port are symbolized in the direction. Specifically, the east longitude and north latitude are represented by positive numbers, and the west longitude and south latitude are represented by negative numbers. Then delete the data with the absolute value of longitude greater than 180 and the data with the absolute value of latitude greater than 90. At the same time, since the ship will send AIS information to the shore station every few seconds when it docks, it is necessary to determine that the ship docks at the port in sequence, in order to ensure that the same ship only retains one docking record in the same port for a continuous time. When there are multiple ships with the same MMSI records, compare whether the name of the port where the ship docks and the docking time are the same. If there is a difference in the port name or the docking time, its data is retained. Otherwise, the latest record is retained and the rest of the data is deleted.
[0038] S2. Extract semantic information from the preprocessed ship AIS data to obtain ship shipping network data.
[0039] The ship shipping network data includes ship trajectory data and shipping network key node data. The generation process is as follows:
[0040] The MMSI of the ship is used as the ship's identity to extract the ship's trajectory information. The real-time position of the ship in the AIS data and the timetable corresponding to the position are used to extract the information of the ship entering and leaving the departure fence and the arrival fence, and the arrival and departure information of the same port and key network nodes are integrated. Finally, the departure and arrival information of the same ship that is too similar and abnormal is cleaned.
[0041] S201, drawing the track of the ship according to the pre-processed AIS data, marking the main routes and turning points, identifying the ship's usual routes by using the track morphology analysis method, and obtaining the ship's track data.
[0042] S202, segmenting the drawn ship track according to the bifurcation points, such as Figure 2As shown, a route node is marked on the main route before the bifurcation point, and each route is marked on each branch route after the bifurcation point. Make a route node marking line segment on each ship trajectory. In order to prevent the direction of the ship's travel at the route node from coinciding with the marking line, making it impossible to determine which route node the ship is currently passing through, the route node marking line segment is at a certain angle to the ship's trajectory, usually set at 90 degrees. With the route node marking line segment as the hypotenuse, a water area is determined according to the longitude and latitude coordinates of the two end points of the line segment. When the current longitude and latitude coordinates of the ship are within the range of the water area, it can be considered that the ship is currently passing through this route node. If the current longitude and latitude coordinates of the ship are not within the longitude and latitude range of the two end points of any route node marking line segment, it can be determined that the ship is not currently passing through the route node.
[0043] S203. Obtain geographic information system (GIS) data, identify key nodes on the ship's trajectory based on it, and obtain key node data of the shipping network. Key nodes include stopover points, entry and exit points, and route turning points, which help to accurately identify the ship's position and fully reflect and understand the ship's behavior.
[0044] S3. Input the ship shipping network data into the long short-term memory neural network multi-factor model to obtain the prediction results of the ship's destination port and cargo flow.
[0045] The long short-term memory neural network multi-factor model is built on the long short-term memory neural network (LSTM). LSTM is a time recursive neural network that uses a unique gating unit to learn and memorize sequence data. It is suitable for processing and predicting problems that are highly related to time series and can achieve high-precision predictions. LSTM controls the long-term state through three control switches. The first switch is responsible for controlling the continued preservation of the long-term state, the second switch is responsible for controlling the input of the immediate state into the long-term state, and the third switch is responsible for controlling whether the long-term state is used as the current output.
[0046] The long short-term memory neural network multi-factor model includes a first prediction model and a second prediction model. The first prediction model uses multiple data (including ship shipping network data and marine meteorological data) of a ship passing through a certain node as input to predict the next destination port, and further corrects the ship's destination port as the nodes passed by the ship increase in sequence. After the prediction of the ship's destination port is completed, the cargo flow direction of the ship can be known by judging whether the ship has loaded or unloaded at the port. The prediction result of the ship's destination port and the change of the ship's draft are input into the second prediction model to predict the cargo flow direction. The change of the ship's draft is calculated according to the ship's draft in the AIS data. The change of the ship's draft = the draft before the ship arrives at the port - the draft before the ship leaves the port, and the thresholds of the two are set according to the normal loading and unloading rules of the ship.
[0047] Specifically, the first prediction model for predicting the destination port is designed as follows:
[0048] Input layer: Ship shipping network data, including historical ship routes, port stay time and other information, as well as marine meteorological information such as wind speed, waves, and ocean currents that affect ship navigation.
[0049] Data processing: Normalize the input data to make it suitable for neural network processing. Construct time series data, where each time step contains the data of the ship at a certain node.
[0050] LSTM layer: Design a multi-layer LSTM network to handle long-term dependencies in time series data. Each LSTM unit contains an input gate, a forget gate, an output gate, and a cell state.
[0051] Output layer: used to predict the next destination port of the ship, and the output result is the probability distribution of each possible destination port.
[0052] Training and optimization: The cross entropy loss function is used for training, and an optimization algorithm such as Adam is adopted. As the ship passes through new nodes, the input data of the model is updated and the destination port prediction results are further revised.
[0053] The second prediction model for predicting cargo flow is designed as follows:
[0054] Input layer: the output result of the first prediction model (i.e. the predicted destination port) and the ship draft change data.
[0055] Data processing: Normalize the input data. It may be necessary to encode the destination data into a format that can be used by the neural network.
[0056] Feature merging: Merge the destination port prediction results and the ship draft change data into a comprehensive feature vector.
[0057] Output layer: The output result is the probability distribution of cargo loading and unloading (inflow / outflow) or direct classification result.
[0058] Threshold setting: According to the normal loading and unloading rules of the ship, set the threshold of the ship's draft change to assist in determining the direction of cargo flow.
[0059] Training and Optimization: Use appropriate loss functions for training and choose a suitable optimizer.
[0060] When training the long short-term memory neural network multi-factor model, the shipping network data of ships whose frequent port calls exceed the set threshold are selected as training samples. In this embodiment, ships whose frequent port calls are greater than 70% are selected as sample data for research. The calculation formula for the frequent port calls is as follows:
[0061]
[0062] Among them, p is the proportion of frequent ports, n is the total number of ports called by the ship during the study period, and n o The number of ports whose call times exceed the set threshold.
[0063] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0064] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.
Claims
1. A method for predicting the destination port and cargo flow of a ship based on a shipping network, characterized in that: The following steps are involved: Obtaining and preprocessing ship AIS data, wherein the ship AIS data includes the port information of the ship, the time of arrival and departure of the ship, the ship draft and the ship's water mobile service identification code, and the port information of the ship includes the port code and the longitude and latitude of the port; Extracting semantic information from the preprocessed ship AIS data to obtain ship shipping network data, wherein the ship shipping network data includes ship track data and shipping network key node data; The ship shipping network data is input into a long short-term memory neural network multi-factor model to obtain prediction results of the ship's destination port and cargo flow direction.
2. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 1, characterized in that: The specific process of the pretreatment is as follows: Delete the data with wrong number of digits of ship's maritime mobile service identification code; Delete the data with wrong number of port codes; Symbolize the port's longitude and latitude.
3. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 1, characterized in that: The long short-term memory neural network multi-factor model is constructed based on the long short-term memory neural network, and includes a first prediction model and a second prediction model. The first prediction model takes the ship shipping network data as input and outputs a prediction result of the ship's destination port.
4. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 3, characterized in that: The second prediction model takes the prediction result of the ship's destination port and the change in the ship's draft as input, and outputs the prediction result of the cargo flow direction. The change in the ship's draft is calculated based on the ship's draft in the AIS data.
5. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 1, characterized in that: When training the long short-term memory neural network multi-factor model, shipping network data of ships whose proportion of frequent port calls exceeds a set threshold are selected as training samples.
6. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 5, characterized in that: The calculation formula for the proportion of frequent port calls is as follows: Among them, p is the proportion of frequent ports, n is the total number of ports called by the ship during the study period, and n o The number of ports whose call times exceed the set threshold.
7. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 1, characterized in that: The generation process of the ship shipping network data is as follows: Obtain the pre-processed AIS data, draw the ship's trajectory using the trajectory morphology analysis method, and obtain the ship's trajectory data; Obtain geographic information system data, segment the drawn ship trajectory according to the bifurcation points, make a route node marking line segment on each ship trajectory, identify the key nodes on the ship trajectory in combination with the geographic information system data, and obtain the key node data of the shipping network.
8. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 7, characterized in that: The route node marking line segments on each route segment are at a 90-degree angle to the ship's trajectory.
9. The method for predicting the destination port and cargo flow direction of a ship based on a shipping network according to claim 7, characterized in that: The key nodes include stop points, entry and exit points, and route turning points.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.