A ship heave prediction system and method based on Conv-Bi-LSTM model

Abstract

This invention proposes a ship heave prediction system and method based on a Conv-Bi-LSTM model. The method involves acquiring historical ship motion data, establishing a motion information dataset, preprocessing the dataset, randomly splitting it into training and test datasets, constructing a ship heave prediction model using a Conv-Bi-LSTM model, training the Conv-Bi-LSTM network with the training dataset, and inputting the test dataset into the ship heave prediction model for prediction. The prediction accuracy of the test data is then obtained, and the model's performance is evaluated using mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE). This invention uses historical motion information with multiple degrees of freedom as input to the model for comprehensive prediction of ship heave, fully utilizing the forward and reverse time state information of the motion time series, thus improving the prediction accuracy of heave motion.

Description

Technical Field

[0001] This invention belongs to the field of ship motion state prediction technology, specifically, it relates to a ship heave motion prediction system and prediction method based on the Conv-Bi-LSTM model. Background Technology

[0002] When ships navigate at sea, they are subjected to environmental disturbances such as sea winds and waves, resulting in complex rolling motions that affect aircraft takeoffs and landings, cargo handling, and other operations, causing six degrees of freedom of rolling motion. Among these, heave motion, as the primary form of passive motion, significantly impacts ship navigation, helicopter takeoffs and landings, and weapon launches, requiring effective prediction algorithms for ship control and compensation. However, in complex sea conditions, the motions of each degree of freedom are interconnected, making accurate prediction of heave motion information difficult using traditional methods. Therefore, developing new ship motion prediction methods is crucial for improving the safety of ship navigation and operations.

[0003] In recent years, research on ship heave prediction has attracted increasing attention. Machine learning, due to its advantages in nonlinear fitting, is often applied to the prediction of nonlinear ship heave. With the rapid development of computer technology, deep learning theory has been widely used in time series prediction. Recurrent Neural Networks (RNNs) are an effective method for processing time series, but due to the problems of vanishing and exploding gradients, RNNs cannot effectively establish long-term dependency models of ship motion sequences.

[0004] To address this shortcoming, a Long Short-Term Memory (LSTM) network model has been developed. Due to its recurrent structure and gating mechanism, LSTM can retain information for a longer period. It is considered a state-of-the-art method for solving time series correlation problems. LSTM possesses the advantages of long-term input memory and explicit memory, improving upon the long-term dependency problem present in RNNs. It has been widely applied in many fields, particularly in time series modeling, such as the prediction of ocean waves and ship motion.

[0005] A genetic algorithm (GA) is used to optimize the LSTM model, and the GA-LSTM model is used to predict ship trajectories. The dimension of the LSTM network input vector is studied, and an input vector space optimization technique based on the impulse response function (IRF) and autocorrelation function (ACF) is proposed. However, due to the long training and prediction time of the LSTM model for time series data with long time spans, and the fact that it does not fully utilize other degrees of freedom and inverse feature information of the ship, the gradient vanishing problem is not completely solved, and the model needs further improvement. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a ship heave prediction system and method based on the Conv (Convolutional Neural Networks)-Bi-LSTM model. The prediction results of this model are compared with those of the Long Short-Term Memory (LSTM) network model, the ordinary backpropagation model, and the Kalman filter model. Based on this comparison, a heave prediction model based on multi-feature information is designed. By using historical motion information from multiple degrees of freedom as input to the model, a comprehensive prediction of ship heave is made, fully utilizing the forward and reverse time state information of the motion time series, thus improving the prediction accuracy of heave motion.

[0007] This invention is achieved through the following technical solution:

[0008] A method for predicting ship heave motion based on the Conv-Bi-LSTM model:

[0009] The control method specifically includes the following steps:

[0010] Step 1: Obtain historical ship motion data, establish a motion information dataset, preprocess the dataset, and randomly split the dataset at a ratio of 19:1 to obtain a training dataset and a test dataset.

[0011] Step 2: Construct a ship heave prediction model using the Conv-Bi-LSTM model. Train the Conv-Bi-LSTM network using the training dataset described in Step 1 to obtain the ship heave prediction model.

[0012] Step 3: Input the test dataset into the ship heave motion prediction model obtained in Step 2 for prediction, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators.

[0013] Furthermore, in step one, the ship's historical motion data includes motion information for six degrees of freedom: time, roll, pitch, bow roll, sway, sway, and heave.

[0014] Furthermore, in step one, the preprocessing includes repairing, denoising, and normalizing the abnormal data in the dataset.

[0015] Furthermore, in step two, the Conv-Bi-LSTM model consists of stacked convolutional layers and bidirectional LSTM layers, and the Dropout method is used to prevent overfitting of the model.

[0016] The Conv-Bi-LSTM model consists of two convolutional layers, one bidirectional LSTM layer, and one full output layer, and uses the Dropout method to prevent overfitting.

[0017] The kernel size is 10, the stride is 1, the number of output channels in the first layer is 2, and the number of output channels in the second layer is 4; the batch size is 64, the learning rate is 0.001, and the dropout probability is 0.2.

[0018] A more accurate ship heave prediction model is obtained by training the Conv-Bi-LSTM model.

[0019] Furthermore, in step three, the single-input prediction model based on Conv-Bi-LSTM from step two is used to predict the future heave motion information of the ship. The input data of the single-input prediction model is the dataset of the ship's heave motion. Multiple convolutional kernels are used to extract features from the input data, and then the data is input into the Bi-LSTM layer. Finally, a fully connected layer is used to output the ship's heave motion state for multiple future time intervals.

[0020] Then, based on the Conv-Bi-LSTM multi-input prediction model, the future heave motion information of the ship is predicted. The input data of the multi-input prediction model are the six degrees of freedom motion information of the ship: roll, pitch, bow, sway, swell and heave. The comprehensive prediction of the ship's heave is achieved through more feature information.

[0021] Furthermore, in step three, the performance of the model is evaluated using evaluation indicators. The four error evaluation indicators selected in the experiment are mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE).

[0022] The formulas for the evaluation indicators are shown below:

[0023]

[0024]

[0025]

[0026]

[0027] Where N is the total number of predicted samples in the test set, i is the predicted sample index, and y(i) is the true value of the i-th predicted sample. Let be the predicted value of the i-th predicted sample.

[0028] A ship heave prediction system based on the Conv-Bi-LSTM model:

[0029] The prediction system includes: a ship motion data establishment and processing module, a ship heave prediction model establishment module, and a prediction performance evaluation module;

[0030] The ship motion data establishment and processing module is used to acquire historical ship motion data, establish a dataset of motion information, preprocess the dataset, randomly split the dataset according to a ratio, and obtain a training dataset and a test dataset.

[0031] The ship heave prediction model building module is used to construct a ship heave prediction model using a Conv-Bi-LSTM model. The Conv-Bi-LSTM network is trained using the training dataset to obtain the ship heave prediction model.

[0032] The prediction performance evaluation module is used to input the test dataset into the ship heave motion prediction model, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators.

[0033] A marine vehicle equipped with the aforementioned ship heave prediction system.

[0034] An electronic device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the above method.

[0035] A computer-readable storage medium for storing computer instructions that, when executed by a processor, implement the steps of the above-described method.

[0036] Beneficial effects of the invention

[0037] (1) The addition of convolutional layers can utilize the periodic and local features of ship historical motion information. The bidirectional network structure makes full use of the forward and reverse time state information of the motion time series, improving the prediction accuracy of heave motion.

[0038] (2) The multi-input prediction model utilizes the relationship between the six degrees of freedom information of roll, pitch, yaw, sway, heave and heave, and takes the historical motion information of multiple degrees of freedom as the input of the model to make a comprehensive prediction of the ship's heave.

[0039] (3) Experiments show that the prediction model based on Conv-Bi-LSTM has higher prediction accuracy and better adaptability to sea states of all levels. It can also achieve higher prediction accuracy under high sea state conditions. Attached Figure Description

[0040] Figure 1 This is the Conv-Bi-LSTM network structure of the present invention;

[0041] Figure 2 This invention provides information on the six degrees of freedom motion of a ship.

[0042] Figure 3 This is a comparison chart of the errors of the single-input prediction model for sea states 3 to 6 of this invention and other prediction models, wherein...

[0043] (a) Prediction error of sea state level 3 using a single-input model;

[0044] (b) Prediction error of sea state level 4 using a single-input model;

[0045] (c) Prediction error of sea state level 5 using a single-input model;

[0046] (d) Prediction error of sea state level 6 using a single-input model;

[0047] Figure 4 This is a comparison chart of the errors of the multi-input prediction model and the single-input prediction model for sea states three to six of this invention.

[0048] (a) Prediction error of the third-level sea state in the multi-input model;

[0049] (b) Prediction error of sea state level 4 using multi-input model;

[0050] (c) Prediction error of sea state level 5 using multi-input model;

[0051] (d) Prediction error of sea state level 6 in multi-input model. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Combination Figures 1 to 4 .

[0054] A method for predicting ship heave motion based on the Conv-Bi-LSTM model:

[0055] The control method specifically includes the following steps:

[0056] Step 1: Obtain historical ship motion data, establish a motion information dataset, preprocess the dataset, randomly split the dataset proportionally, and obtain training and test datasets.

[0057] Step 2: Construct a ship heave prediction model using the Conv-Bi-LSTM model. Train the Conv-Bi-LSTM network using the training dataset described in Step 1 to obtain the ship heave prediction model.

[0058] Step 3: Input the test dataset into the ship heave motion prediction model obtained in Step 2 for prediction, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators.

[0059] like Figure 2 As shown, in step one, the ship's historical motion data includes motion information for six degrees of freedom: time, roll, pitch, bow roll, sway, sway, and heave.

[0060] In step one, the preprocessing includes repairing, denoising, and normalizing outlier data in the dataset. The data is then passed to the convolutional layer. A key feature of convolutional layers is weight sharing; that is, a convolutional kernel uses the same set of weights to extract features in the same way when performing convolution operations on each part of the data, ultimately resulting in a single feature value.

[0061] Weight sharing can effectively extract features and reduce model complexity by decreasing the number of weights. In this invention, a combination of convolutional and pooling layers is used to extract temporal, periodic, and local features from ship motion data. Finally, the extracted feature values ​​are passed to a Bi-LSTM network for multi-step prediction of ship motion.

[0062] In step two, the Conv-Bi-LSTM model consists of stacked convolutional layers and bidirectional LSTM layers, and the Dropout method is used to prevent overfitting of the model.

[0063] In the process of predicting ship heave motion, the output at the current moment is related not only to the forward time state but also to the reverse time state. In order to extract time series features in both directions, the prediction model uses forward input data in the first layer and reverse input data in the second layer. The output is determined by the two layers of neural networks.

[0064] The Conv-Bi-LSTM model consists of two convolutional layers, one bidirectional LSTM layer, and one full output layer, and uses the Dropout method to prevent overfitting.

[0065] Bi-LSTM can fully utilize the bidirectional temporal features of ship history information. To prevent overfitting as the network size increases, Dropout is used during model training, where neurons in the LSTM hidden layers are randomly and temporarily deleted according to probability. Finally, the output is calculated through fully connected layers. The specific network structure of the single-input prediction model is as follows:

[0066] As shown in Table 1:

[0067]

[0068] Table 1 Parameters of Single-Input Conv-Bi-LSTM Network Model

[0069] The kernel size is 10, the stride is 1, the number of output channels in the first layer is 2, and the number of output channels in the second layer is 4; the batch size is 64, the learning rate is 0.001, and the dropout probability is 0.2.

[0070] A more accurate ship heave prediction model is obtained by training the Conv-Bi-LSTM model.

[0071] The trained single-input prediction model is used to predict ship motion at various sea states. The prediction error for ship heave in sea states three to six is ​​as follows: Figure 3 As shown.

[0072] The higher the sea state level, the greater the heave prediction error. Compared with other prediction methods, the Conv-Bi-LSTM prediction model has better prediction performance under all sea state levels. As the sea state level increases, the prediction error of all four prediction models continues to increase, but the prediction error of the CNN-Bi-LSTM prediction model increases the slowest. The model has higher prediction accuracy under high sea state conditions.

[0073] In step three, the single-input prediction model based on Conv-Bi-LSTM from step two is used to predict the future heave motion information of the ship. The input data of the single-input prediction model is the dataset of the ship's heave motion. Multiple convolutional kernels are used to accurately extract the features from the input data. Then the data is input into the Bi-LSTM layer, and finally the fully connected layer is used to output the ship's heave motion state for multiple future time intervals.

[0074] Then, based on the Conv-Bi-LSTM multi-input prediction model, the future heave motion information of the ship is predicted. The input data of the multi-input prediction model are the six degrees of freedom motion information of the ship: roll, pitch, bow, sway, swell and heave. The comprehensive prediction of the ship's heave is achieved through more feature information.

[0075] In step three, the performance of the model is evaluated using evaluation metrics. The four error evaluation metrics selected for the experiment are mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE).

[0076] The formulas for the evaluation indicators are shown below:

[0077]

[0078]

[0079]

[0080]

[0081] Where N is the total number of predicted samples in the test set, i is the predicted sample index, and y(i) is the true value of the i-th predicted sample. Let be the predicted value of the i-th predicted sample.

[0082] A ship heave prediction system based on the Conv-Bi-LSTM model:

[0083] The prediction system includes: a ship motion data establishment and processing module, a ship heave prediction model establishment module, and a prediction performance evaluation module;

[0084] The ship motion data establishment and processing module is used to acquire historical ship motion data, establish a dataset of motion information, preprocess the dataset, randomly split the dataset according to a ratio, and obtain a training dataset and a test dataset.

[0085] The ship heave prediction model building module is used to construct a ship heave prediction model using a Conv-Bi-LSTM model. The Conv-Bi-LSTM network is trained using the training dataset to obtain the ship heave prediction model.

[0086] The prediction performance evaluation module is used to input the test dataset into the ship heave motion prediction model, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators.

[0087] The model performance was evaluated using four metrics: Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). Prediction errors are shown in Table 2.

[0088]

[0089] Table 2 Prediction errors of different models

[0090] It can be seen that the Conv-Bi-LSTM prediction model has a significantly better prediction effect than other models. The percentage improvement index is used to describe the difference between the two models on the same error index.

[0091] At sea states 3 to 6, in terms of mean average error (MAE), the Conv-Bi-LSTM prediction model reduced the prediction error by 35.39%, 27.77%, 50.94%, and 51.88% compared to the LSTM model; by 44.03%, 53.57%, 61.08%, and 76.69% compared to the Kalman model; and by 62.32%, 74.82%, 82.92%, and 80.66% compared to the BP prediction model.

[0092] Regarding the root mean square error (RMSE), the Conv-Bi-LSTM prediction model reduced the prediction error by 32.04%, 26.35%, 61.16%, and 57.15% compared to the LSTM model; by 37.99%, 56.47%, 62.03%, and 74.69% compared to the Kalman model; and by 60.17%, 74.23%, 84.64%, and 81.35% compared to the BP prediction model.

[0093] By analyzing the prediction error and percentage improvement of the four models, it can be concluded that the prediction performance of the Conv-Bi-LSTM prediction model is significantly better than that of the other three models.

[0094] There is a certain mapping relationship between ship heave and its historical roll, pitch, bow roll, sway, and pitch. By utilizing information from multiple degrees of freedom of ship motion, a comprehensive prediction of ship heave can be made. The prediction error for ship heave in sea states three to six is ​​as follows: Figure 4 As shown in Table 3, the prediction errors of the model are as follows.

[0095]

[0096]

[0097] Table 3. Prediction errors of single-input and multi-input prediction models.

[0098] Multi-input prediction models make better use of various motion information for model training and prediction, resulting in higher prediction accuracy and reliability.

[0099] Compared to the Conv-Bi-LSTM prediction model, the mul-Conv-Bi-LSTM prediction model reduced the mean error (MAE) by 47.78%, 24.57%, 26.98%, and 22.25% for sea states 3 to 6, respectively; and the root mean square error (RMSE) by 45.38%, 15.86%, 27.20%, and 14.09%, respectively.

[0100] It can be seen that the multi-input prediction model has further improved the prediction performance compared with the single-input prediction model, and is suitable for comprehensive prediction of ship heave motion by collecting a variety of external information and ship motion information.

[0101] A marine vehicle equipped with the aforementioned ship heave prediction system.

[0102] An electronic device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the above method.

[0103] A computer-readable storage medium for storing computer instructions that, when executed by a processor, implement the steps of the above-described method.

[0104] The memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DRRAM). It should be noted that the memory of the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0105] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means such as coaxial cable, optical fiber, digital subscriber line (DSL), or wireless means such as infrared, wireless, or microwave. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media such as floppy disks, hard disks, magnetic tapes, optical media such as high-density digital video discs (DVDs), or semiconductor media such as solid-state drives (SSDs).

[0106] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0107] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as execution by a hardware decoding processor, or as execution by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0108] The foregoing has provided a detailed description of the ship heave prediction system and method based on the Conv-Bi-LSTM model proposed in this invention, and has elucidated the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for predicting ship heave motion based on a Conv-Bi-LSTM model, characterized in that: The prediction method specifically includes the following steps: Step 1: Obtain historical ship motion data, establish a motion information dataset, preprocess the dataset, and randomly split the dataset at a ratio of 19:1 to obtain a training dataset and a test dataset. Step 2: Construct a ship heave prediction model using the Conv-Bi-LSTM model. Train the Conv-Bi-LSTM network using the training dataset described in Step 1 to obtain the ship heave prediction model. Step 3: Input the test dataset into the ship heave motion prediction model obtained in Step 2 for prediction, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators. In step three, the single-input prediction model based on Conv-Bi-LSTM from step two is used to predict the future heave motion information of the ship. The input data of the single-input prediction model is the dataset of the ship's heave motion. Multiple convolutional kernels are used to extract features from the input data, and then the data is input into the Bi-LSTM layer. Finally, a fully connected layer is used to output the ship's heave motion state for multiple future time intervals. Then, based on the Conv-Bi-LSTM multi-input prediction model, the future heave motion information of the ship is predicted. The input data of the multi-input prediction model are the six degrees of freedom motion information of the ship: roll, pitch, bow, sway, swell and heave. The comprehensive prediction of the ship's heave is achieved through more feature information.

2. The prediction method according to claim 1, characterized in that: In step one, the ship's historical motion data includes motion information for six degrees of freedom: time, roll, pitch, bow roll, sway, sway, and heave.

3. The prediction method according to claim 2, characterized in that: In step one, the preprocessing includes repairing, denoising, and normalizing the abnormal data in the dataset.

4. The prediction method according to claim 3, characterized in that: In step two, the Conv-Bi-LSTM model consists of stacked convolutional layers and bidirectional LSTM layers, and the Dropout method is used to prevent overfitting of the model. The Conv-Bi-LSTM model consists of two convolutional layers, one bidirectional LSTM layer, and one full output layer, and uses the Dropout method to prevent overfitting. The kernel size is 10, the stride is 1, the number of output channels in the first layer is 2, and the number of output channels in the second layer is 4; the batch size is 64, the learning rate is 0.001, and the dropout probability is 0.

2. A more accurate ship heave prediction model is obtained by training the Conv-Bi-LSTM model.

5. The prediction method according to claim 4, characterized in that: In step three, the performance of the model is evaluated using evaluation metrics. The four error evaluation metrics selected for the experiment are mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE). The formulas for the evaluation indicators are shown below: Where N is the total number of predicted samples in the test set, and i is the predicted sample index. Let i be the true value of the i-th predicted sample. Let be the predicted value of the i-th predicted sample.

6. A prediction system for a ship heave motion prediction method based on the Conv-Bi-LSTM model according to any one of claims 1 to 5, characterized in that: The prediction system includes: a ship motion data establishment and processing module, a ship heave prediction model establishment module, and a prediction performance evaluation module; The ship motion data establishment and processing module is used to acquire historical ship motion data, establish a dataset of motion information, preprocess the dataset, randomly split the dataset according to a ratio, and obtain a training dataset and a test dataset. The ship heave prediction model building module is used to construct a ship heave prediction model using a Conv-Bi-LSTM model. The Conv-Bi-LSTM network is trained using the training dataset to obtain the ship heave prediction model. The prediction performance evaluation module is used to input the test dataset into the ship heave motion prediction model, obtain the prediction accuracy of the test data, and evaluate the performance of the model using the mean error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation indicators.

7. A maritime vehicle, characterized in that: The marine vehicle is equipped with the ship heave prediction system as described in claim 6.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images