Ship fusion trajectory prediction method based on deep sea wind power plant coordinate driving

Abstract

The invention belongs to the field of ship trajectory prediction, and discloses a ship fusion trajectory prediction method based on deep sea wind power plant coordinate driving, and the method comprises the steps: obtaining standard AIS data; respectively calculating a wind power plant repulsive force and a wind power plant angle factor, a position code of the ship at a corresponding moment, and feature vectors of the ship and the wind power plant; training a preset ship trajectory prediction model by using the position code of the ship at the corresponding moment, the feature vectors of the ship and the wind power plant and the hidden state output data at the previous moment to obtain a trained ship trajectory prediction model; and performing ship trajectory prediction by using the trained ship trajectory prediction model. According to the method, the deep and far sea wind power plant coordinates are taken into consideration factors of ship trajectory prediction, more targeted navigation suggestions and path planning are provided for the ship, and the safety and efficiency of ship navigation are improved.

Description

Technical Field

[0001] The present invention belongs to the field of ship trajectory prediction, and in particular relates to a ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive. Background Art

[0002] Ship trajectory prediction technology refers to the use of historical ship trajectory data and other related information to predict the navigation trajectory and behavior of the ship in the future through modeling and algorithm analysis. With the rise of deep learning, models such as recurrent neural networks (RNN), long short-term memory networks (LSTM) and attention mechanisms are widely used in ship trajectory prediction, which can better capture the complex relationship and time series information between data, thereby improving the prediction accuracy. Such methods are relatively simple in data processing and feature extraction, and often only consider basic information such as the position and speed of the ship. However, in reality, ship trajectory data may contain more complex information and features, such as the ship's navigation intention, navigation mode, ship type, etc., which are crucial for more accurate prediction of ship trajectories. Traditional methods often fail to fully explore and utilize this information, resulting in low prediction accuracy. In addition, these methods have limitations in dealing with long-term dependencies and time series relationships. Ship trajectory data has obvious time series, and the current position and speed of the ship are often affected by the previous moment or even earlier. The statistical models in traditional methods are difficult to effectively capture this time series relationship, resulting in inaccurate and incoherent predictions of the future development of the ship's trajectory.

[0003] LSTM (Long Short-Term Memory) is a variant of recurrent neural network (RNN), which aims to solve the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data. Although the design of LSTM enables it to better capture long-term dependencies, the structure of LSTM network is more complex than traditional RNN, including key components such as input gate, forget gate and output gate. This complexity often leads to more computing resources and training time in some cases, increasing the complexity of the model and the challenge of training. On the other hand, since LSTM contains more gated units, it means that the network needs to store and update more information in its internal memory units. This may lead to the risk of overfitting of LSTM in some short sequence tasks or when the data correlation is weak, because its complexity may make it easier to overfit a small amount of data.

[0004] In deep-sea environments, the increase in wind farms has brought new challenges to ship navigation, because ships need to avoid these wind farms in time to ensure the safety of navigation. In this case, it becomes crucial to consider the location and layout of deep-sea wind farms and incorporate them into the trajectory prediction model. However, the current trajectory prediction methods have some shortcomings in dealing with wind farm factors. Existing methods usually tend to regard trajectory prediction as a simple spatiotemporal sequence problem, ignoring the impact of special features such as wind farms in deep-sea environments on the trajectory of ship motion. This simplified model approach may result in the inability of ships to effectively cope with the presence of wind farms in actual navigation, thereby increasing the potential risk of collision. In addition, existing methods often lack sufficient refinement and real-time performance when considering wind farm factors. The layout of deep-sea wind farms may change with time and environmental conditions, and current models often find it difficult to update this information in a timely manner, resulting in prediction results that may not be accurate or timely. Summary of the invention

[0005] 1. Purpose of the invention The purpose of the present invention is to provide a ship fusion trajectory prediction method driven by deep-sea wind farm coordinates. By using a gated recurrent unit (GRU) to replace a long short-term memory network (LSTM), and incorporating the deep-sea wind farm coordinates into the considerations of ship trajectory prediction, this design not only reduces the amount of network parameters and reduces the risk of overfitting, but also can more comprehensively consider the situation of the ship's surrounding environment, provide ships with more targeted navigation suggestions and path planning, and improve the safety and efficiency of ship navigation.

[0006] (II) Technical solution To solve the above problems, the present invention provides a ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive, the method comprising: Acquire standard AIS data, wherein the standard AIS data includes the latitude and longitude coordinates and heading data of the ship; the heading data includes the heading of the ship; Calculating the wind farm repulsive force according to the latitude and longitude coordinates of the ship and the wind farm coordinates; Calculating the wind farm angle factor according to the heading data, the angle formed by the heading of the ship and the direction of the wind farm; Using a first fully connected network to extract features from the latitude and longitude coordinates of the ship to obtain a position code of the ship at a corresponding time; Using a second fully connected network to extract features of the wind farm repulsive force and the wind farm angle factor, to obtain feature vectors of the ship and the wind farm; The preset ship trajectory prediction model is trained using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain a trained ship trajectory prediction model; the preset ship trajectory prediction model includes a multi-layer GRU network; The trained ship trajectory prediction model is used to predict the ship trajectory.

[0007] Furthermore, before calculating the wind farm repulsion force according to the ship's longitude and latitude coordinates and the wind farm coordinates, it is also included to determine whether to introduce the wind farm coordinates based on the actual distance between the ship and the surrounding wind farms. The wind farm repulsion force calculation formula is as follows: ; In the formula, is the wind farm repulsion force, is an exponential function, and are two normal numbers, represents the distance between the ship and the wind farm, It represents the minimum distance between the ship and the wind farm. The distance D between the ship and the wind farm is calculated by the longitude and latitude coordinates of the ship and the coordinates of the wind farm.

[0008] Furthermore, it is characterized in that the wind farm angle factor calculation formula is as follows: ; In the formula, is the wind farm angle factor, , is the heading vector of the ship in longitude and latitude, , are the longitude vector and latitude vector from the ship to the wind farm, It is the angle formed by the heading of the ship and the direction of the wind farm.

[0009] Furthermore, the calculation formula for the position code of the ship at the corresponding time is as follows: ; In the formula, is the position code of the ship at time t, is the ReLU nonlinear activation function, is the first network training weight, , Respectively expressed in The longitude and latitude of the ship at the moment.

[0010] Furthermore, it is characterized in that the calculation formula of the characteristic vector of the ship and the wind farm is as follows: ; in, is the characteristic vector of the ship and the wind farm, Train the weights for the second network.

[0011] Furthermore, the preset ship trajectory prediction model is trained using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain the trained ship trajectory prediction model, including: Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm And the hidden state output data of the previous moment All are input into the first layer of GRU network to obtain the first hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm , the hidden state output data of the previous moment The first hidden state at time t is input into the second layer of GRU network to obtain the second hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of ships and wind farms , the hidden state output data of the previous moment The second hidden state at time t is input into the third layer of GRU network to obtain the third hidden state at time t. The final trajectory of the ship is predicted according to the third hidden state at time t; wherein the multi-layer GRU network includes a first layer GRU network, a second layer GRU network and a third layer GRU network.

[0012] Furthermore, the processing process of each layer of the GRU network includes: Receive the ship's position code at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment ; Encode the position of the ship at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment Both are input into the update gate, and the update gate vector is calculated ; Encode the position of the ship at time t And the hidden state output data of the previous moment Both are input into the reset gate, and the reset gate vector is calculated , Combined with the ship's position code at time t , the hidden state output data of the previous moment and reset gate vector , calculate the candidate hidden state at time t ; Based on the candidate hidden state , update gate vector And the hidden state output data of the previous moment , calculate the hidden state at time t.

[0013] Furthermore, the update gate vector The calculation formula is as follows: ; In the formula, is the weight matrix of the update gate, is the bias term of the update gate, is the Sigmoid activation function.

[0014] Furthermore, the reset gate vector The calculation formula is as follows: ; In the formula, To reset the gate weight matrix, is the bias term for resetting the gate, is the Sigmoid activation function.

[0015] Furthermore, the candidate hidden state at time t The calculation formula is as follows: ; In the formula represents element-wise multiplication, is the weight matrix of the candidate hidden states, is the bias term of the candidate hidden state, () is the hyperbolic tangent function.

[0016] (III) Beneficial effects The present invention proposes a ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive, aiming to optimize the navigation path planning in combination with the specific navigation characteristics of the ship, thereby improving the safety and efficiency of ship navigation, and also has a positive effect on maritime traffic management and monitoring. First, the ship trajectory prediction model preset by the present invention includes a multi-layer GRU network, and the gated recurrent unit (GRU) is selected to replace the long short-term memory network (LSTM) in order to simplify the model structure and reduce complexity. Compared with LSTM, GRU has a simpler structure, which only contains update gates and reset gates. The concise performance of GRU reduces the number of parameters in the network, reduces the computational cost of the model, improves the training efficiency, and also helps the model to better generalize to new data, avoiding overfitting of training data, thereby improving the robustness and generalization ability of the model. In addition, the present invention incorporates the coordinates of deep-sea wind farms into the considerations of ship trajectory prediction, and performs trajectory prediction based on the distance and angle between the ship position and the wind farm, demonstrating a deeper concern for maritime navigation safety and efficiency. The principle is: based on the actual distance between the ship and the surrounding wind farms, determine whether to introduce the wind farm coordinates. When the judgment result is that the wind farm coordinates need to be introduced, calculate the wind farm repulsion force according to the ship's longitude and latitude coordinates and the wind farm coordinates; calculate the wind farm angle factor according to the heading data, the angle formed by the ship's heading and the direction of the wind farm, and then perform feature extraction on the above data through a fully connected network, input the corresponding results into a multi-layer GRU network for training, generate a trained ship trajectory prediction model, and finally perform ship trajectory prediction through the trained ship trajectory prediction model. The present invention innovatively utilizes the location information of the wind farm, combined with the relative position relationship between the current position of the ship and the wind farm, to further improve the accuracy and reliability of ship trajectory prediction, so that ships can better avoid obstacles such as wind farms in complex marine environments and avoid potential collision risks. By combining factors such as the ship's position and the location, distance and angle of the wind farm, the ship's surrounding environment can be considered more comprehensively, providing the ship with more targeted navigation advice and path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive of the present invention; Figure 2 is a flow chart of ship trajectory prediction according to a specific embodiment of the present invention; Figure 3 Schematic diagram of the GRU structure of a specific embodiment of the present invention; Figure 4 Is the Sigmoid activation function of a specific embodiment of the present invention; Figure 5 for Figure 2Schematic diagram of the horizontal propagation (based on the application of time series) of the three-layer GRU network structure. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0019] like Figure 1 As shown, the present invention provides a ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive, the method comprising the following steps: S101, acquiring standard AIS data, wherein the standard AIS data includes the latitude and longitude coordinates and heading data of the ship; the heading data includes the heading of the ship; S102, calculating the wind farm repulsive force according to the latitude and longitude coordinates of the ship and the wind farm coordinates; S103, calculating the wind farm angle factor according to the heading data, the angle formed by the heading of the ship and the direction of the wind farm; S104, using a first fully connected network to extract features from the longitude and latitude coordinates of the ship to obtain a position code of the ship at a corresponding time; S105, using a second fully connected network to perform feature extraction on the wind farm repulsive force and the wind farm angle factor to obtain feature vectors of the ship and the wind farm; S106, training a preset ship trajectory prediction model using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain a trained ship trajectory prediction model; the preset ship trajectory prediction model includes a multi-layer GRU network; S107, using the trained ship trajectory prediction model to perform ship trajectory prediction.

[0020] Further, in S101, standard AIS data is obtained, and this step specifically includes: first obtaining ship trajectory data through the AIS system, and then preprocessing the ship trajectory data to obtain standard AIS data. The standard AIS data includes the latitude and longitude coordinates and heading data of the ship; the heading data includes the heading of the ship; in this step: the AIS system is a system for automatic information exchange between ships and between ships and shore-based facilities. In trajectory prediction technology, ship trajectory data is usually provided by the automatic identification system (AIS), which is installed on ships. The AIS system aims to improve the efficiency and safety of maritime traffic management. By transmitting the position, navigation speed, heading and other information of the ship in real time, it enables ships to identify, track and avoid collisions with each other, and also provides important data support for maritime monitoring and emergency rescue. In this step, the ship trajectory data (i.e., ship AIS data) is obtained by the AIS system; due to some problems in the actual application of ship AIS data, such as the lack of a complete information verification mechanism, the data may contain a large amount of error information. These error messages may be caused by problems in positioning equipment, communication facilities, software design, etc. These erroneous data may interfere with the analysis of ship trajectory behavior, thereby affecting the accuracy and reliability of ship trajectory prediction. Therefore, when preprocessing AIS data, it is necessary to clean and verify the data to remove abnormal and erroneous data, thereby improving the credibility and accuracy of the data and ensuring the effectiveness of subsequent analysis and modeling. For data preprocessing, we focus on the collection of ship AIS data in a specific longitude and latitude area, and then perform preliminary cleaning of the data to remove non-core columns such as length, width, ship status and date to obtain preprocessed data (standard AIS data). Standard AIS data only retains the data tuples of the five key elements of the ship's unique identity, timestamp, longitude, latitude (latitude and longitude coordinates), and heading. The timestamp is a dynamic information of the ship, which can be updated in real time to reflect the current status and movement trajectory of the ship. Next, in order to conduct an in-depth analysis of the navigation trajectory of a specific MMSI (Maritime Mobile Service Identity, which is the abbreviation of the maritime mobile service identification code, used to identify ships. This code must be kept consistent with the actual situation of the ship at all times), we perform a data deduplication step to traverse the entire AIS data set and accurately screen and extract all data records that match the MMSI. Subsequently, these records are sorted according to the order of timestamps, and the redundant data with the same time and MMSI due to repeated reception by the system is identified and eliminated. This is to ensure the purity, consistency and accuracy of the data, and lay a solid and accurate data foundation for subsequent trajectory analysis and modeling. Next, data conversion is performed, which usually involves converting the data into a format suitable for analysis.This includes operations such as converting date and time fields into standard formats, converting geographic coordinates into numerical formats, and converting ship types into codes for subsequent analysis.

[0021] Ship AIS data The latitude and longitude information is standardized, that is, the longitude and latitude data in the historical records are normalized. This process aims to unify the measurement scales between different features so that all features are at the same numerical level, thereby accelerating the convergence speed of the neural network and helping to improve the generalization ability of the model when processing new data. The calculation formula is as follows: ; (1) In formula (1), , Indicates the maximum and minimum longitudes in the AIS data of each ship. , Indicates the maximum and minimum latitudes in the AIS data of each ship, ( , ) are respectively expressed in The longitude and latitude of the ship at the time, ( , )、( , ) is part of the standard AIS data, , are the heading vector of the ship in longitude and latitude respectively.

[0022] Furthermore, before step S102, it also includes: based on the actual distance between the ship and the surrounding wind farms, judging whether to introduce the wind farm coordinates, specifically including: before performing trajectory prediction, we need to screen the geographical location of the ship. According to the provisions of the relevant safe driving distance of the ship, we use 15 nautical miles as the threshold of the distance between the ship and the surrounding wind farms in this article. 15 nautical miles is equivalent to a threshold of a safe distance. If the actual distance between the ship and the wind farm is less than 15 nautical miles, the coordinates of the wind farm are taken into account in this method; if the distance is greater than 15 nautical miles, the future ship heading will not coincide with the location of the wind farm, so the wind farm coordinates are excluded. The present invention mainly solves the problem of ship trajectory prediction within the range of less than 15 nautical miles between the actual distance between the ship and the wind farm.

[0023] Further, S102, calculating the wind farm repulsive force according to the latitude and longitude coordinates of the ship and the wind farm coordinates, specifically includes: For the wind farm coordinates, the distance between the ship and the wind farm is quantified by calculating the distance to the ship, which is calculated as: ; (2) In formula (2), The distance between the ship and the wind farm, ( , ) are the wind farm coordinates; After calculating the distance between the ship and the wind farm Afterwards, the wind farm repulsion force can be calculated : ; (3) In formula (3), is the wind farm repulsion force, is an exponential function, and are two normal numbers, represents the distance between the ship and the wind farm, Indicates the minimum distance between a ship and a wind farm. , and The optimal choice is usually determined by considering the characteristics of the ship and the navigation environment in a specific water area. The purpose of this construction is that when the distance between the ship and the wind farm is greater than When the ship is within a safe distance from the wind farm, there is no need to pay too much attention to the impact of the wind farm on the ship's trajectory, so an exponential function is used to reduce the impact of the wind farm repulsion. On the contrary, if the distance between the ship and the wind farm is less than When , it indicates that the ship and the wind farm are in a dangerous and easy-to-collision state, it is necessary to pay close attention to the positional relationship between the two and significantly increase the repulsive force of the wind farm.

[0024] Further, in S103, the angle factor of the wind farm is calculated according to the heading data, the angle formed by the heading of the ship and the direction of the wind farm. The design principle of this step is: in addition to calculating the repulsive force between the ship and the wind farm, it is also necessary to note the positional relationship between the ship's movement direction and the wind farm, that is, the wind farm in front of the ship is more important than the wind farm behind the ship. This is a special positional relationship determined by the movement direction of the ship. Therefore, the present invention introduces the angle factor k between the ship and the wind farm, and its calculation formula is as follows: ; (4) In formula (4), is the angle factor of the wind farm (the angle factor between the ship and the wind farm), ( , ) is the ship heading vector, the ship heading vector is denoted as , , are the longitude vector and latitude vector formed from the ship to the wind farm respectively. The vector formed from the ship to the wind farm is , It is the angle formed by the heading of the ship and the direction of the wind farm.

[0025] Further, in S104, the first fully connected network is used to extract features from the latitude and longitude coordinates of the ship to obtain the position code of the ship at the corresponding time, specifically including: embedding the standard AIS data (pre-processed data), and encoding the pre-processed ship data using the first fully connected network layer. The calculation formula of the position code of the ship at the corresponding time is as follows; ; (5) In formula (5), is the position code of the ship at time t, is the ReLU nonlinear activation function, is the first network training weight, which corresponds to the weight of the first fully connected layer and is used to calculate , , Respectively expressed in The longitude and latitude of the ship at the moment.

[0026] Further, in S105, the second fully connected network is used to perform feature extraction on the wind farm repulsive force and the wind farm angle factor to obtain feature vectors of the ship and the wind farm, specifically including: After calculating the wind farm repulsion force F and the wind farm angle factor k, the two are subjected to feature extraction through the second fully connected network operation. The calculation method is: ; (6) In formula (6), is the ReLU nonlinear activation function, The training weights for the second network are the weights corresponding to the second fully connected layer, which are used to calculate is the vector of ship and wind farm characteristics (wind farm impact factor).

[0027] Further, in S106, the preset ship trajectory prediction model is trained using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain a trained ship trajectory prediction model, which specifically includes: Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm And the hidden state output data of the previous moment All are input into the first layer of GRU network to obtain the first hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm , the hidden state output data of the previous moment The first hidden state at time t is input into the second layer of GRU network to obtain the second hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of ships and wind farms , the hidden state output data of the previous moment The second hidden state at time t is input into the third layer of GRU network to obtain the third hidden state at time t; The final trajectory of the ship is predicted according to the third hidden state at time t; wherein the preset ship trajectory prediction model includes a multi-layer GRU network, and the multi-layer GRU network includes a first-layer GRU network, a second-layer GRU network, and a third-layer GRU network. The preset trajectory prediction model is composed of three layers of GRU. The three-layer GRU is set because the data of ship trajectories in reality will contain a lot of complex information and features. If the structure is too small, it may not be able to accurately capture these features. However, if a structure with more than three layers is set, it is more likely to make the model more complex, and there may be overfitting problems, which cannot be generalized in practical applications. The three-layer GRU structure is as follows: Figure 2 and Figure 5 As shown, Figure 2 In the AIS information part (MMSI corresponds to the unique identity of the ship, time is the timestamp, longitude is the longitude, latitude is the latitude, and COG (Course Over Ground)) is the heading. After the ship coordinates and wind farm coordinates are calculated and processed, they are input into the preset trajectory prediction model. The three vertical layers are the first layer GRU network, the second layer GRU network, and the third layer GRU network from top to bottom. The third layer GRU network is close to the prediction result output part. Figure 5 The vertical 3 layers correspond to Figure 1 The three-layer structure of Figure 2 The lower part, and the three horizontal columns on the right are expanded in chronological order to describe the specific working principle of the model. The third hidden state at time t (current time) is output in the last layer of GRU network (third layer), which is the predicted value of the final trajectory point of the ship at time t. The role of the first hidden state at time t and the second hidden state at time t in the longitudinal GRU input and output is to facilitate the inspection of the parameters in the middle of the training model. Figure 5 for Figure 2 The expanded diagram of the lower middle part (GRU three-layer module), the GRU module in the center of the figure and Figure 3 The internal structures of the GRU modules in the figure are the same.

[0028] Furthermore, the processing process of each layer of the GRU network, the GRU network structure includes a reset gate and an update gate. The reset gate determines how to combine the new input information with the previous memory, and the update gate defines the amount of the previous memory saved to the current time step. The processing process is as follows Figure 3 As shown, specifically including: 1) Receive the ship's position code at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment In this step, for each time step t, the GRU receives the input ship position , and the hidden state at the previous time step ; 2) Encode the position of the ship at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment Both are input into the update gate, and the update gate vector is calculated The main function of the update gate is to control the hidden state passed down from the previous time step and the candidate hidden state at the current time step How to combine to generate the hidden state of the current time step .when When it is close to 0, it means that it relies more on the current candidate hidden state; when When it is close to 1, it means that it relies more on the hidden state of the previous moment, and the update gate vector The calculation formula is as follows: ; (7) In formula (7), is the weight matrix of the update gate, is the bias term of the update gate, is the Sigmoid activation function, such as Figure 4 As shown in the figure, the horizontal axis of the activation function represents the input value, and the vertical axis represents the output value. The sigmoid activation function can map the output of the neuron to between 0 and 1.

[0029] 3) Encode the position of the ship at time t And the hidden state output data of the previous moment Both are input into the reset gate, and the reset gate vector is calculated , the reset gate is used to control how to combine the hidden state of the previous time step with the current input to calculate the candidate hidden state. The reset gate vector The calculation formula is as follows: ; (8) In formula (8), To reset the gate weight matrix, is the bias term for resetting the gate, Sigmoid activation function, the reset gate determines the hidden state of the previous time step by controlling the size of its output value Candidate hidden state at the current time step The degree of participation in the calculation process. When the reset gate is close to 0, it means that the hidden state of the previous time step is almost completely ignored, that is, the historical information is "forgotten". The reset gate adjusts the current input through its output value and historical information In generating candidate hidden states When the reset gate is close to 1, it means that more historical information is retained, which affects the calculation of candidate hidden states together with the current input. It helps the model more effectively handle long-term and short-term dependencies in sequence data by controlling the forgetting of historical information and the combination of current input and historical information.

[0030] 4) Combined with the ship's position code at time t , the hidden state output data of the previous moment and reset gate vector , calculate the candidate hidden state at time t , candidate hidden states Represents the candidate hidden state at the current moment, which will update the hidden state of the previous time step according to the current input and reset gate; the candidate hidden state at time t The calculation formula is as follows: ; (9) In the formula represents element-wise multiplication, is the weight matrix of the candidate hidden states, is the bias term of the candidate hidden state, () is the hyperbolic tangent function. It should be noted that the parameters of each GRU module , , , , , The value needs to be adjusted according to the actual situation. The calculation of the candidate hidden state directly involves the input of the current time step , which enables the model to capture important information in the current input. Although the calculation of the candidate hidden state also depends on the hidden state of the previous time step , but this dependency is achieved by resetting the gate When the reset gate is close to 0, the model will ignore most of the historical information, and when the reset gate is close to 1, more historical information will be retained.

[0031] 5) Finally, according to the candidate hidden state , update gate vector And the hidden state output data of the previous moment , calculate the hidden state at time t, the calculation formula is: ; (10) Further, in S107, the trained ship trajectory prediction model is used to predict the ship trajectory, and real-time data is obtained and input into the trained ship trajectory prediction model to obtain the ship trajectory prediction data for each time step. , and predict the final trajectory point of the ship.

[0032] The present invention provides a ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive, which aims to optimize the navigation path planning in combination with the specific navigation characteristics of the ship, thereby improving the safety and efficiency of ship navigation, and also has a positive effect on maritime traffic management and monitoring. The method proposes to incorporate the location of the wind farm, the distance and angle between the ship and the wind farm into the ship trajectory prediction, and introduces the wind farm repulsion force F and the wind farm angle factor k. This innovative method aims to use the location information of the wind farm, combined with the relative position relationship between the current position of the ship and the wind farm, to more accurately predict the future trajectory and action of the ship. By considering the influencing factors of the wind farm, such as location, direction and distance, the performance of the trajectory prediction model can be effectively improved. The prediction model can more accurately predict the navigation path of the ship, avoid potential collision risks, and improve navigation safety. Incorporating factors such as the site selection location of the deep-sea wind farm, the distance and angle between the ship and the wind farm into the ship trajectory prediction model can achieve more comprehensive environmental perception and intelligent decision-making. The present invention also proposes to use a fully connected network to extract and encode the features of the ship's longitude and latitude position, the repulsive force of the wind farm, and the angle factor of the wind farm, and finally input the above information into a network composed of multiple layers of GRU to predict the ship's trajectory. The present invention adopts a GRU network because the GRU network has an update gate and a reset gate. These gating mechanisms help the model better capture the long-term dependencies in the time series data. The network structure composed of GRU is simpler, reducing the number of parameters and computational complexity. In ship trajectory prediction, because the navigation of a ship is affected by many factors, it is necessary to consider the state information over a long period of time in the past. This simplicity helps to reduce the risk of overfitting, speed up training, and perform better when the amount of data is small.

Claims

1. A ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive, characterized in that: include: Acquire standard AIS data, wherein the standard AIS data includes the latitude and longitude coordinates and heading data of the ship; the heading data includes the heading of the ship; Calculating the wind farm repulsive force according to the latitude and longitude coordinates of the ship and the wind farm coordinates; Calculating the wind farm angle factor according to the heading data, the angle formed by the heading of the ship and the direction of the wind farm; Using a first fully connected network to extract features from the latitude and longitude coordinates of the ship to obtain a position code of the ship at a corresponding time; Using a second fully connected network to extract features of the wind farm repulsive force and the wind farm angle factor, to obtain feature vectors of the ship and the wind farm; The preset ship trajectory prediction model is trained using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain a trained ship trajectory prediction model; the preset ship trajectory prediction model includes a multi-layer GRU network; The trained ship trajectory prediction model is used to predict the ship trajectory.

2. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 1 is characterized in that: Before calculating the wind farm repulsion force according to the ship's latitude and longitude coordinates and the wind farm coordinates, it is also included to determine whether to introduce the wind farm coordinates based on the actual distance between the ship and the surrounding wind farms. The wind farm repulsion force calculation formula is as follows: ； In the formula, is the wind farm repulsion force, is an exponential function, and are two normal numbers, represents the distance between the ship and the wind farm, It represents the minimum distance between the ship and the wind farm. The distance D between the ship and the wind farm is calculated by the longitude and latitude coordinates of the ship and the coordinates of the wind farm.

3. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 2 is characterized in that: The calculation formula of the wind farm angle factor is as follows: ； In the formula, is the wind farm angle factor, , are the heading vectors of the ship in longitude and latitude respectively, , are the longitude vector and latitude vector from the ship to the wind farm, It is the angle formed by the heading of the ship and the direction of the wind farm.

4. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 3 is characterized in that: The calculation formula for the position code of the ship at the corresponding time is as follows: ； In the formula, is the position code of the ship at time t, is the ReLU nonlinear activation function, is the first network training weight, , Respectively expressed in The longitude and latitude of the ship at the moment.

5. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 4 is characterized in that: The calculation formula of the characteristic vector of the ship and the wind farm is as follows: ； in, is the characteristic vector of the ship and the wind farm, Train the weights for the second network.

6. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 5 is characterized in that: The method of training a preset ship trajectory prediction model by using the position code of the ship at the corresponding time, the feature vector of the ship and the wind farm, and the hidden state output data at the previous time to obtain a trained ship trajectory prediction model includes: Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm And the hidden state output data of the previous moment All are input into the first layer of GRU network to obtain the first hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of the ship and the wind farm , the hidden state output data of the previous moment The first hidden state at time t is input into the second layer of GRU network to obtain the second hidden state at time t; Encode the position of the ship at time t , the characteristic vectors of ships and wind farms , the hidden state output data of the previous moment The second hidden state at time t is input into the third layer of GRU network to obtain the third hidden state at time t; The final trajectory of the ship is predicted according to the third hidden state at time t; wherein the multi-layer GRU network includes a first layer GRU network, a second layer GRU network and a third layer GRU network.

7. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 6 is characterized in that: The processing of each layer of the GRU network includes: Receive the ship's position code at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment ; Encode the position of the ship at time t , the characteristic vectors of ships and wind farms And the hidden state output data of the previous moment Both are input into the update gate, and the update gate vector is calculated ; Encode the position of the ship at time t And the hidden state output data of the previous moment Both are input into the reset gate, and the reset gate vector is calculated , Combined with the ship's position code at time t , the hidden state output data of the previous moment and reset gate vector , calculate the candidate hidden state at time t ; Based on the candidate hidden state , update gate vector And the hidden state output data of the previous moment , calculate the hidden state at time t.

8. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 7 is characterized in that: The update gate vector The calculation formula is as follows: ； In the formula, is the weight matrix of the update gate, is the bias term of the update gate, is the Sigmoid activation function.

9. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 7 is characterized in that: The reset gate vector The calculation formula is as follows: ； In the formula, To reset the gate weight matrix, is the bias term for the reset gate.

10. The ship fusion trajectory prediction method based on deep-sea wind farm coordinate drive according to claim 7 is characterized in that: The candidate hidden state at time t The calculation formula is as follows: ； In the formula represents element-wise multiplication, is the weight matrix of the candidate hidden states, is the bias term of the candidate hidden state, () is the hyperbolic tangent function.

Images