Ship offshore operation heave compensation prediction method and system and storage medium
By employing an improved k-nearest neighbor selection method and a bidirectional gated recurrent neural network, the problem of missing heave data during ship operations at sea was solved, enabling accurate compensation and prediction, and improving the accuracy and reliability of data processing.
Patent Information
- Application Number
- CN202511752732.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-03-20
AI Technical Summary
In existing technologies, the lack of heave data during ship operations at sea leads to a decline in dataset quality, affecting the accuracy and reliability of prediction results. Furthermore, existing algorithms cannot effectively utilize contextual information, limiting the accuracy and reliability of heave prediction.
An improved k-nearest neighbor selection method is used to fill in missing data, and a bidirectional gated recurrent neural network is constructed. Combined with forward and backward propagation modules, the bidirectional recurrent neural network is used to predict ship heave and sag, making full use of time series information.
It improves the accuracy of compensation and prediction of missing heave values, realizes scientific monitoring and protection of ship operations at sea, and ensures the accuracy and reliability of prediction results.
Smart Images
Figure CN121705770A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of marine electronic information technology, specifically to a method, system, and storage medium for predicting heave compensation during marine operations. Background Technology
[0002] During maritime operations, ships are affected by wind, waves, currents, and their own motion, resulting in six degrees of freedom of motion, which significantly impacts the efficiency and safety of maritime operations. Among these, heave motion has the most significant impact, potentially causing collisions between loads and other offshore structures or personnel, and even leading to serious accidents such as sling breakage. Therefore, it is necessary to predict future heave based on historical and current measurements to improve the response speed of active heave compensation systems.
[0003] Under the influence of wind, waves, and currents, a ship's measurement system is susceptible to external interference, exhibiting problems such as poor output stability and intermittent transmission channel interruptions. This results in a certain degree of incompleteness in the collected ship heave data. The lack of ship heave data directly leads to a decline in the overall data quality of the dataset, causing distortion in prediction results based on the measurement data and affecting the reliability and accuracy of the compensation system.
[0004] The effects of ocean waves on ships exhibit strong continuity and temporal sequence. Therefore, clustering imputation is typically used to process and fill in missing data, followed by nonlinear prediction algorithms to predict ship heave and sag. Existing missing value imputation methods, when the data loss rate is high, are limited by the insufficient number of complete samples, restricting the upper limit of parameter k. Furthermore, they fail to deeply mine usable information from incomplete samples, leading to data instability and resource waste. Existing gated recurrent network algorithms can only achieve unidirectional data flow from past to future, resulting in a narrow model perspective, weak insight into the inherent relationships within the sequence, and an inability to accurately capture and utilize contextual information, thus limiting the accuracy and reliability of heave and sag prediction. Summary of the Invention
[0005] To address the aforementioned deficiencies in existing technologies, a method, system, and storage medium for predicting and compensating for heave during ship operations at sea are provided. This improves the accuracy of compensation and prediction for missing heave values during ship operations at sea, enabling scientific monitoring and protection of the ship's operations at sea.
[0006] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: a method for predicting the heave of ships at sea, comprising the following steps: S1. Collect heave data of the vessel during its operations at sea; S2. Determine if the measurement data is missing. If yes, continue to step S3; otherwise, proceed to step S4. S3. Based on the clustering algorithm, an improved k-nearest neighbor selection method is used to fill in the missing ship heave data. S4. Construct a time-sequential forward propagation gated recurrent unit module and a time-sequential backward propagation gated recurrent unit module to form a bidirectional recurrent neural network (RNN), so that the ship's heave and sag timing signal data can obtain the associated previous moment's state information and next moment's state information in the neural network. S5. A bidirectional gated cyclic network algorithm is used to predict the heave of ships.
[0007] According to the above technical solution, in step S3, the improved k-nearest neighbor selection method specifically includes the following steps: S3.1 Calculate the Euclidean distance between the sample to be filled and other samples, and sort them; S3.2 Randomly select the m samples closest to the sample to be filled; from the m samples, select the k samples that cover the missing columns of the sample to be filled the most; S3.3 Fill the missing numerical column with the mean of k samples, and determine whether the filling is complete. If yes, the data filling is complete and proceed to step S4. If not, continue to step S3.4. S3.4 Determine whether m reaches 20% of the total number of samples. If yes, fill the remaining missing columns with the mean of the numerical columns of the m samples and complete the data filling, then proceed to step S4. If not, continue to step S3.5. S3.5 Adjust the number of samples selected to 2m, and repeat step S3.2; S3.6 Iterate the above process until all missing values of the sample to be filled are filled.
[0008] According to the above technical solution, the initial value of m is taken as 5%-10% of the total number of samples.
[0009] According to the above technical solution, when the total number of samples is small, the initial value of m should be greater than 3.
[0010] According to the above technical solution, the specific method for updating the ship heave timing data in step S4 is as follows: ; ; ; In the formula: To output status information, It is a linear mapping relationship. For bidirectional gated cyclic networks in forward timing state The weight, For bidirectional gated cyclic networks in reverse timing state The weight, This is used to bias the state information.
[0011] According to the above technical solution, in step S5, the bidirectional gated recurrent network algorithm specifically includes the following steps: S5.1 Preprocess the ship heave data by dividing the data into training and testing sets and performing data normalization. S5.2 Initialize the neural network parameters, set the number of hidden layers, the number of neurons, and the learning rate, and construct a bidirectional gated recurrent neural network model; S5.3 Input the training set into the model and use the Adam optimizer to continuously update the network weights. and , S5.4 Determine if the training steps have been reached. If yes, continue to step S5.5. If no, return to step S5.3 to continue optimizing the network weights. S5.5. Traverse the training set multiple times until the set maximum number of iterations is reached to complete the training of the model; S5.6 Input the test set into the trained model for prediction, and perform inverse normalization on the prediction results to output the final predicted value of ship operation heave.
[0012] According to the above technical solution, the formula for data normalization is as follows: ; In the formula: This is the original data. For the normalized data, The maximum value in the dataset. It is the minimum value in the dataset.
[0013] According to the above technical solution, in step S5.2, the number of hidden layers and the number of neurons are determined based on computing power.
[0014] This invention also provides a ship heave compensation prediction system for offshore operations, comprising: An automated data acquisition module is used to collect heave data of ships during offshore operations. The heave compensation module is used to effectively fill in missing heave values for ships. The heave prediction module is used to predict the heave of a ship.
[0015] The present invention also provides a computer-readable storage medium, comprising: a determination program for a heave prediction method based on a bidirectional gated cyclic network stored on the computer-readable storage medium, wherein when the determination program for the heave prediction method based on a bidirectional gated cyclic network is executed by a processor, the heave compensation prediction method for ship maritime operations based on a bidirectional gated cyclic network described above is implemented.
[0016] The present invention has the following beneficial effects: 1. The present invention systematically constructs a ship heave compensation prediction model. In the case of partial missing measurement data, an improved k-nearest neighbor selection method is used to effectively fill in the missing ship heave values. Furthermore, a bidirectional gated cyclic network algorithm is used to predict the ship heave, thereby achieving accurate compensation and future prediction of ship heave.
[0017] 2. The present invention adopts an improved k-nearest neighbor selection method, which not only considers sample similarity more accurately, but also the integrity of the samples, ensuring good continuity between the generated filling data and the relevant neighborhood data, and improving the accuracy of numerical filling when there are many missing values.
[0018] 3. This invention utilizes multi-layer neural networks to improve the depth of the bidirectional gated recurrent network system model, enhance its self-learning ability, fully utilize the relevant information of the time sequence before and after the data degradation state, and effectively transmit it between neural nodes, thereby improving the accuracy of the operation process prediction model. Attached Figure Description
[0019] Figure 1 A flowchart of a method for predicting ship heave compensation during offshore operations provided by an embodiment of the present invention; Figure 2 A flowchart of a method for predicting ship heave compensation based on an optimized clustering algorithm provided in an embodiment of the present invention; Figure 3 The flowchart illustrates a method for predicting ship heave based on a bidirectional gated cyclic network, as provided in this embodiment of the invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0021] This invention provides a method for predicting ship heave compensation during offshore operations, such as... Figure 1 As shown, it includes the following steps: S1. Collect heave data of the vessel during its operations at sea; S2. Determine if the measurement data is missing. If yes, continue to step S3; otherwise, proceed to step S4. S3. Based on the clustering algorithm, an improved k-nearest neighbor selection method is used to fill in the missing ship heave data; specifically, step S3 is as follows: Figure 2 As shown, it includes the following sub-steps: S3.1 Calculate the Euclidean distance between the sample to be filled and other samples, and sort them; S3.2 Randomly select the m samples closest to the sample to be filled; from the m samples, select the k samples that cover the missing columns of the sample to be filled the most, where k is half of m.
[0022] S3.3 Fill the missing numerical column with the mean of k samples, and determine whether the filling is complete. If yes, the data filling is complete and proceed to step S4. If not, continue to step S3.4. S3.4 Determine whether m reaches 20% of the total number of samples. If yes, fill the remaining missing columns with the mean of the numerical columns of the m samples and complete the data filling, then proceed to step S4. If not, continue to step S3.5. S3.5 Adjust the number of samples selected to 2m, and repeat step S3.2; S3.6 Iterate the above process until all missing values of the sample to be filled are filled.
[0023] Specifically, the initial value of m is taken as 5%-10% of the total number of samples. When the total number of samples is small, the initial value of m can be greater than 3. The improved k-nearest neighbor selection method not only considers sample similarity more accurately, but also considers the integrity of the samples, thus effectively improving the accuracy of numerical imputation when there are many missing values.
[0024] S4. Construct a time-sequential forward propagation gated recurrent unit module and a time-sequential backward propagation gated recurrent unit module to form a bidirectional recurrent neural network (RNN), so that the ship's heave and sag timing signal data can obtain the associated previous moment's state information and next moment's state information in the neural network. Specifically, Recurrent Neural Networks (RNNs), as a special type of artificial neural network algorithm, are capable of processing time-series data and capturing its dynamic characteristics. During training, when an RNN is expanded into a multi-layer feedforward neural network, excessive memory layers due to parameter sharing and time step size lead to long training cycles and gradient vanishing or exploding problems. Gated Recurrent Units (RNNs), however, retain only update and reset gates, reducing the computational burden on the model, improving processing speed, and effectively mitigating gradient vanishing or exploding issues during backpropagation.
[0025] Specifically, in step S4, learning is performed in both clockwise and counterclockwise directions, so that the current state information is derived from the input. , forward state of the previous moment and the reverse state of the previous moment The following is a joint decision regarding the specific method for updating the ship heave sequence data: ; ; ; In the formula: To output status information, It is a linear mapping relationship. For bidirectional gated cyclic networks in forward timing state The weight, For bidirectional gated cyclic networks in reverse timing state The weight, This is used to bias the state information.
[0026] Specifically, by optimizing the forward and reverse bidirectional update algorithm, the algorithm model's scope and ability to analyze the intrinsic relationships in time series data can be improved, making the algorithm's prediction of ship heave more accurate.
[0027] S5. A bidirectional gated recurrent network algorithm is used to predict the ship's heave. Specifically, step S5 is as follows: Figure 3 As shown, it includes the following sub-steps: S5.1 Preprocess the ship heave data by dividing the data into training and testing sets and performing data normalization. S5.2 Initialize the neural network parameters, set the number of hidden layers, the number of neurons, and the learning rate, and construct a bidirectional gated recurrent neural network model; S5.3 Input the training set into the model and use the Adam optimizer to continuously update the network weights. and , S5.4 Determine if the training steps have been reached. If yes, continue to step S5.5. If no, return to step S5.3 to continue optimizing the network weights. S5.5. Traverse the training set multiple times until the set maximum number of iterations is reached to complete the training of the model; S5.6 Input the test set into the trained model for prediction, and perform inverse normalization on the prediction results to output the final predicted value of ship operation heave.
[0028] Specifically, the formula for data normalization is as follows: ; In the formula: This is the original data. For the normalized data, The maximum value in the dataset. It is the minimum value in the dataset.
[0029] Specifically, the number of hidden layers, the number of neurons, and the learning rate are determined based on computing power. Preferably, 3-4 hidden layers are selected, 50-200 neurons are selected, and the learning rate is selected as 0.8-0.9.
[0030] This invention also provides a ship heave compensation prediction system for offshore operations, comprising: An automated data acquisition module is used to collect heave data of ships during offshore operations. The heave compensation module is used to effectively fill in missing heave values for ships. The heave prediction module is used to predict the heave of a ship.
[0031] The present invention also provides a computer-readable storage medium, comprising: a determination program for a heave prediction method based on a bidirectional gated cyclic network stored on the computer-readable storage medium, wherein when the determination program for the heave prediction method based on a bidirectional gated cyclic network is executed by a processor, the heave compensation prediction method for ship maritime operations based on a bidirectional gated cyclic network described above is implemented.
[0032] In summary, this invention constructs a dynamic compensation and prediction model for ship heave during maritime operations. It dynamically collects data on the ship's maritime operations and, when some heave data is missing, uses an improved k-nearest neighbor selection method based on a clustering algorithm to fill in the missing data. A bidirectional gated cyclic network is employed to predict ship heave, making the prediction more accurate and enabling scientific monitoring and protection of the ship's maritime operations.
[0033] The embodiments described above are some, but not all, of the embodiments of this application. The detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
Claims
1. A method for predicting heave compensation during ship operations at sea, characterized in that, Includes the following steps: S1. Collect heave data of the vessel during its operations at sea; S2. Determine if the measurement data is missing. If yes, continue to step S3; otherwise, proceed to step S4. S3. Based on the clustering algorithm, an improved k-nearest neighbor selection method is used to fill in the missing ship heave data. S4. Construct a time-sequential forward propagation gated recurrent unit module and a time-sequential backward propagation gated recurrent unit module to form a bidirectional recurrent neural network (RNN), so that the ship's heave and sag timing signal data can obtain the associated previous moment's state information and next moment's state information in the neural network. S5. A bidirectional gated cyclic network algorithm is used to predict the heave of ships.
2. The method for predicting ship heave during offshore operations according to claim 1, characterized in that: In step S3, the improved k-nearest neighbor selection method specifically includes the following steps: S3.1 Calculate the Euclidean distance between the sample to be filled and other samples, and sort them; S3.2 Randomly select the m samples closest to the sample to be filled; from the m samples, select the k samples that cover the missing columns of the sample to be filled the most; S3.3 Fill the missing numerical column with the mean of k samples, and determine whether the filling is complete. If yes, the data filling is complete and proceed to step S4. If not, continue to step S3.
4. S3.4 Determine whether m reaches 20% of the total number of samples. If yes, fill the remaining missing columns with the mean of the numerical columns of the m samples and complete the data filling, then proceed to step S4. If not, continue to step S3.
5. S3.5 Adjust the number of samples selected to 2m, and repeat step S3.2; S3.6 Iterate the above process until all missing values of the sample to be filled are filled.
3. The method for predicting ship heave during offshore operations according to claim 2, characterized in that: The initial value of m is set to 5%-10% of the total sample size.
4. The method for predicting ship heave during offshore operations according to claim 3, characterized in that: When the total number of samples is small, the initial value of m should be greater than 3.
5. The method for predicting ship heave during offshore operations according to claim 1, characterized in that: In step S4, the specific method for updating the ship heave time series data is as follows: ； ； ； In the formula: To output status information, It is a linear mapping relationship. For bidirectional gated cyclic networks in forward timing state The weight, For bidirectional gated cyclic networks in reverse timing state The weight, This is used to bias the state information.
6. The method for predicting ship heave during offshore operations according to claim 1, characterized in that: In step S5, the bidirectional gated recurrent network algorithm specifically includes the following steps: S5.1 Preprocess the ship heave data by dividing the data into training and testing sets and performing data normalization. S5.2 Initialize the neural network parameters, set the number of hidden layers, the number of neurons, and the learning rate, and construct a bidirectional gated recurrent neural network model; S5.3 Input the training set into the model and use the Adam optimizer to continuously update the network weights. and , S5.4 Determine if the training steps have been reached. If yes, continue to step S5.
5. If no, return to step S5.3 to continue optimizing the network weights. S5.
5. Traverse the training set multiple times until the set maximum number of iterations is reached to complete the training of the model; S5.6 Input the test set into the trained model for prediction, and perform inverse normalization on the prediction results to output the final predicted value of ship operation heave.
7. The method for predicting ship heave during offshore operations according to claim 6, characterized in that: The formula for data normalization is as follows: ； In the formula: The original data, For the normalized data, The maximum value in the dataset. It is the minimum value in the dataset.
8. The method for predicting ship heave during offshore operations according to claim 6, characterized in that: In step S5.2, the number of hidden layers and neurons are determined based on computing power.
9. A ship's heave compensation prediction system for offshore operations, characterized in that, include: An automated data acquisition module is used to collect heave data of ships during offshore operations. The heave compensation module is used to effectively fill in missing heave values for ships. The heave prediction module is used to predict the heave of a ship.
10. A computer-readable storage medium, characterized in that, include: The computer-readable storage medium stores a determination program for a heave prediction method based on a bidirectional gated cyclic network. When the determination program for the heave prediction method based on a bidirectional gated cyclic network is executed by a processor, it implements the ship's heave compensation prediction method based on a bidirectional gated cyclic network as described in any one of claims 1-8.