A bulk carrier main engine power prediction method and system

Abstract

The application provides a bulk carrier main engine power prediction method and system, and relates to the technical field of green ships and intelligent shipping. The method realizes high-precision and high-robustness prediction of the bulk carrier main engine power through the following steps: sister ship operation big data collection preprocessing, Pearson correlation feature screening, wave resistance explainable separation based on the two-factor Froude method and the iteration of resistance and propulsion efficiency, fusion of SVM to build a physically mechanism-constrained grey-box prediction model, and comparison and optimization of a black-box model. The method has physical explainability, can meet the consistency verification of the Minimum Propulsion Power Guidelines, can be quickly modeled and calculated online in real time relying on conventional operation data, can be reused for the same series of ship types, and has outstanding engineering practicability and maritime compliance.

Description

Technical Field

[0001] This invention relates to the field of green ship and intelligent shipping technology, specifically to a method and system for predicting the main engine power of bulk carriers. Background Technology

[0002] Existing methods for estimating the main engine power of bulk carriers are mainly divided into three categories: the extrapolation method based on similar ship types, the principle calculation method based on ship resistance performance, and the regression prediction method based on machine learning. All of these methods have obvious shortcomings in practical engineering applications.

[0003] First, the estimation method based on similar ship types relies on reference ship type data to indirectly estimate the power of the target ship. The calculation is simple, but the accuracy is highly dependent on the consistency between the reference ship and the target ship. The indirect estimation process is prone to introducing cumulative errors, and its applicability is limited in the scenario of ocean-going bulk carriers with large variations in sea state, loading, and speed.

[0004] Second, the performance calculation method based on hydrodynamic principles has a clear physical meaning and strong interpretability, but it requires a large number of high-precision ship parameters and environmental parameters, which are difficult to obtain completely during actual ship navigation. At the same time, the assumptions used to simplify the calculation will reduce the forecast accuracy, and the calculation process is complex and time-consuming, which cannot meet the needs of online and real-time power forecasting for actual ships.

[0005] Third, black-box prediction methods based on machine learning can fit nonlinear relationships using real ship operation data and have high accuracy within specific datasets. However, they rely entirely on data-driven approaches, lack support from the physical mechanisms of ship navigation, and the results are uninterpretable and difficult to verify through maritime regulations. Furthermore, the models are sensitive to data quality and have insufficient generalization ability and poor robustness in extrapolation scenarios such as unseen sea conditions and loading conditions.

[0006] In summary, existing technologies cannot simultaneously achieve a balance between forecast accuracy, environmental adaptability, real-time performance, and physical interpretability. They struggle to meet the demands for accurate, stable, and interpretable forecasts of bulk carrier main engine power under complex wind, wave, and current conditions, and also cannot effectively support the consistency verification of the "Minimum Propulsion Power Guidelines." Therefore, the industry urgently needs a high-precision, interpretable, and generalizable method for forecasting bulk carrier main engine power that combines actual ship operational data with ship physics. Summary of the Invention

[0007] This invention addresses the technical problems of existing bulk carrier main engine power forecasting methods, such as insufficient forecasting accuracy, poor physical interpretability, weak adaptability to complex sea conditions, and difficulty in meeting the consistency verification of the "Minimum Propulsion Power Guidelines". It provides a bulk carrier main engine power forecasting method and system, which achieves high-precision, high-stability, and interpretable forecasting of main engine power under complex wind, wave, and current environments by preprocessing real ship operation big data, interpretably separating wave drag, and integrating physical mechanisms with machine learning modeling.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for predicting the main engine power of a bulk carrier, the method comprising the following steps: Collect big data on the operation of sister bulk carriers, and perform preprocessing such as missing value removal, draft and speed filtering, and angle standardization on the collected data. The Pearson correlation coefficient method was used to screen the power and wave drag-in correlation features of the preprocessed data to obtain effective features. Based on the two-dimensional method and drag-propulsion efficiency iteration, the separation of wave drag on a real ship is completed by utilizing effective features to obtain the wave drag and resistance components. By integrating SVM machine learning with ship hydrodynamic mechanisms, a gray box prediction model for main engine power is constructed using the wave drag increase and drag components as inputs. The forecast model is evaluated using R², MSE, RMSE, and MAE indices, and the final forecast result is output by comparing it with the black box model.

[0009] On the other hand, the present invention also provides a bulk carrier main engine power forecasting system, which includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the system is triggered to execute the above-described bulk carrier main engine power forecasting method.

[0010] Compared with the prior art, the beneficial effects of the present invention are: 1. Forecast accuracy and robustness are significantly improved, and adaptability to complex sea states is stronger: By standardizing and preprocessing the big data from sister ship operations and screening relevant features, noise and redundant information interference are effectively reduced. A gray-box model constrained by physical mechanisms is employed to accurately capture the nonlinear relationships between multiple factors such as wind, waves, current, draft, and speed. Compared to traditional empirical methods, pure mechanistic methods, and pure machine learning black-box methods, this approach results in smaller prediction errors and more stable output. It maintains high forecast accuracy and strong generalization ability even under extrapolated sea states and varying loading conditions, effectively adapting to nonlinear disturbances caused by complex sea states.

[0011] 2. Possesses physical interpretability and meets maritime regulatory verification requirements: Based on the two-dimensional Froude method and the drag-propulsion efficiency iteration, the wave drag is explained and separated in an interpretable manner. The physical relationship between power and each drag component is clarified, making the forecast process consistent with the hydrodynamic mechanism of ships and the results traceable and verifiable. It solves the defects of pure black box models that are not interpretable and pure mechanism methods that are difficult to obtain parameters. It can meet the consistency verification requirements of the "Minimum Propulsion Power Guidelines" and is more suitable for actual ship engineering and maritime audit scenarios.

[0012] 3. The project boasts outstanding practicality and can achieve efficient real-time forecasting: It eliminates the need for a large number of high-precision parameters that are difficult to collect on-site, and can complete modeling and forecasting based on conventional operational data of actual ships. The overall process is simplified and efficient, and can support online and real-time calculation of bulk carrier main engine power. At the same time, the model based on sister ship data can be quickly reused to the same series of ship types, which greatly reduces the cost of testing and calculation, and is more suitable for promotion and application in actual shipping engineering.

[0013] 4. Verification using actual ship data shows that the forecast accuracy of this invention is significantly better than that of a purely data-driven model: The comparison results of the gray-box model and the pure machine learning black-box model on the same test set show that the determination coefficient R² of the gray-box model is 0.9988, significantly higher than that of the black-box model (0.9854); the root mean square error (RMSE) is 0.0333, and the mean absolute error (MAE) is 0.0220, which are only 27.3% and 24.3% of those of the black-box model, respectively. These data fully demonstrate that by introducing physical mechanism constraints, this invention significantly improves the accuracy and consistency of host power prediction while maintaining interpretability, exhibiting a clear technical advantage over purely data-driven methods.

[0014] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0015] Figure 1 This is a flowchart of the prediction of ship main engine power based on black box and gray box models provided by the present invention; Figure 2 This is a comparison chart of the actual ship power prediction results and big data using the gray box model based on the SVM algorithm provided by this invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0017] Unless the context requires otherwise, throughout the specification and claims, the word “comprising” and its variations, such as “including” and “having”, shall be understood to have an open, inclusive meaning, that is, to be interpreted as “including, but not limited to”.

[0018] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0019] The singular forms “a” and “the” used in this specification and the appended claims include plural references unless otherwise expressly stated herein. It should be noted that the term “or” is generally used to mean “and / or” unless otherwise expressly stated herein.

[0020] In the following description, in order to clearly demonstrate the structure and working method of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0021] This invention provides a method and system for predicting the main engine power of bulk carriers. The implementation details of embodiments of this invention are described below with reference to the accompanying drawings. The following content is only for ease of understanding and is not essential for implementing this solution. This method is based on measured wave environment and navigation data. Through iterative analysis of drag increase and propulsion efficiency caused by wave drag, a physically interpretable path is established. Furthermore, it combines supervised learning from machine learning with black-box model predictions to improve power prediction accuracy. For example... Figure 1 As shown, this method proceeds in five progressive steps, encompassing the entire process of establishing, analyzing, and predicting the ship's main engine power database. The specific steps are as follows: S1. Big data collection and preprocessing; We collected and analyzed real-world operational data from a series of sister bulk carriers of the same design, model, and shipyard on multiple major global shipping routes. We compiled statistics on the operating duration, service life, and filtered valid data for each vessel, providing comprehensive data support for model building. Simultaneously, we created track maps for each vessel, visually presenting their navigation paths and operational status; we used frequency maps to statistically analyze and display the top three key data points; and we summarized the statistical results of various influencing factors for each vessel, providing detailed numerical evidence.

[0022] Data filtering: 1. Remove missing data Missing values ​​can be efficiently identified using the built-in functions of data processing software (such as Python's pandas library, R language, or specialized statistical software). This is achieved by setting specific filtering rules (completeness, reasonableness, and relevance) to ensure that only complete and accurate records are retained. While removing missing data for core signals, it's also necessary to consider other related signals. If the missing core signal is closely related to the missing data of other signals (e.g., both caused by the same instrument malfunction), it may be necessary to remove the missing data of these related signals simultaneously to ensure the overall consistency and reliability of the data. After removing missing data, a final integrity verification is performed on the data.

[0023] 2. Data filtering based on draft and speed Considering the need to incorporate data from model tests and potential flow calculations into the analysis, the draft and speed data used for analysis must not exceed the range of model tests and potential flow calculations. The speed range within which static resistance can be calculated is determined through rapid response reports. Combining the range of operational data with the scope of ship performance research, the selection and classification rules for draft and speed are determined as follows: 1) Algorithm for classifying draft conditions in, n To divide the quantity, For maximum and minimum draft, a To select a range, For the first k Water level conditions.

[0024] 2) Speed ​​range division algorithm in, m Divide the number according to speed, , The maximum and minimum speeds allowed for model testing / speed reports. Select a range for a single speed. For the first j Each speed condition.

[0025] By using the above two-way constraints of draft and speed, we ensure that the operational data falls within the effective range for calculating hydrostatic resistance and wave drag, thus guaranteeing the rationality and accuracy of subsequent physical mechanism calculations.

[0026] 3. Handling angles with periodic properties Due to the heading angle ( θ h ) and heading angle ( θ dSince angle-related data such as angles change periodically, in order to ensure that the input angle data has periodic properties, the scheme inputs angles in radians.

[0027] 4. Wave angle ( θ e Data conversion The actual ship operation data only contains the ship's wave direction angle and the absolute wave direction angle relative to the geodetic coordinate system. Therefore, a relative angle conversion of the ship's wave direction angle is required based on the actual ship operation data. In the wave drag prediction of the main engine power gray box model, the wave drag increase is equal in magnitude and consistent in direction along the ship's length, and equal in magnitude but opposite in direction along the ship's beam. To further conduct effective analysis of wave drag increase, it is necessary to convert the wave direction angle relative to the ship to 0. -180 between.

[0028] relative to the ship's wave angle The wave angle is relative to the Earth coordinate system. Let be the heading angle relative to the Earth coordinate system, when Then it is right-hand drive. Then it is left-hand drive.

[0029] Actual wave angle for: S2. Correlation Feature Analysis; The Pearson correlation coefficient measures the presence and strength of a linear correlation between two features. Therefore, it is used to measure the correlation between power and other features. The formula is as follows: In the formula: The value indicates the strength of the correlation; a positive value indicates a positive correlation between the increase in the feature and the target feature, while a negative value indicates a negative correlation. Here, the black-box power forecast model directly considers the relationship between the host power and each feature, while the gray-box power forecast model considers the relationship between wave drag and each feature.

[0030] S3, Wave resistance enhancement and separation; Given the existing hydrostatic resistance test data of the ship model, the hydrostatic resistance data of the actual ship can be calculated using two dimensions. Then, based on the difference in main engine speed between windy and waveless conditions, the wind and wave resistance increase of the actual ship under windy and waveless conditions can be estimated. Subtracting the wind resistance from the wind resistance can yield the estimated wave resistance increase (in the following calculation formula, the coefficient of the model ship is suffixed with m, and the coefficient of the actual ship is suffixed with s).

[0031] 1. Estimation of static water resistance of a real ship based on the two-dimensional method The two-dimensional method divides the total resistance of a ship into frictional resistance. and residual resistance The friction coefficient can be calculated. and residual drag coefficient The friction resistance coefficient was then calculated. The residual resistance coefficient of a real ship is calculated using the ITTC 1957 formula. Friction resistance coefficient of ship model The static water resistance coefficient of the model ship After conversion, the following are the various resistance coefficients and resistance calculation formulas.

[0032] coefficient of friction resistance as follows: in The Reynolds number is given by the following formula: For the ship's speed, Because of the ship's long waterline, is the viscosity coefficient of water.

[0033] Actual ship residual coefficient The calculation is as follows: The total still water resistance coefficient of the model boat The calculation is as follows: For the water density in the ship model test, This represents the wetted surface area of ​​the ship model.

[0034] Total still water resistance coefficient of the actual ship for: The wetted surface area of ​​the bilge keel. This represents the wetted surface area of ​​the actual ship. This is the roughness subsidy coefficient. This is the air drag coefficient.

[0035] air drag coefficient as follows: It is the projected area of ​​the hull and superstructure above the waterline on the mid-section.

[0036] Total static resistance and still water effective power The calculation is as follows: 2. Calculation of wind resistance of actual ship The actual wind resistance of a ship is calculated using the following resistance formula: relative wind direction angle The drag coefficient at the bottom, Let the mass density of air be 1.226 kg / m³. 3 , It represents the projected area of ​​the hull and superstructure above the waterline on the mid-section. Relative wind speed, For ground speed.

[0037] 3. Real-world wave drag separation based on drag-propulsion efficiency iteration Assuming a certain ship propulsion efficiency Changes and increased drag from wind and waves and hydrostatic resistance The ratio satisfies the following formula: Ship wind and wave propulsion efficiency for: The power of the ship's main shaft in wind and waves is: The total resistance in the wind and waves, For the ship's speed relative to the water, For shaft system efficiency ( (0.99).

[0038] If we know the relationship between propulsion efficiency and the drag ratio of wind and waves... and the ship's hydrostatic resistance Then, the total increase in drag of the ship due to wind and waves can be obtained by solving the above equation using the iterative method. The static resistance of a ship The relationship between propulsion efficiency and the drag ratio caused by wind and waves can be obtained using still water model tests. It can be obtained using variable load model tests or CFD calculations.

[0039] At the model scale, self-propulsion calculations are performed at a fixed speed and different propeller speeds to simulate the increase in propeller speed under different wind and wave conditions and increased drag. The corresponding forces at different speeds are compared to those in the model test. Calculated using the following formula: in, and These represent the resistance of a self-propelled vessel with a propeller and the resistance of a vessel without a propeller, respectively. The corresponding increase in actual ship resistance is calculated using the following formula: in, This is the correction value for hydrostatic friction resistance (i.e., the forced force corresponding to the hydrostatic self-propulsion point). For the model's scaling ratio, and The water densities are measured at the scales of the actual ship and the model, respectively.

[0040] Calculate propeller thrust and torque at different speeds, and use the ITTC real-ship performance prediction method (ITTC7.5–02–03–01.4) to calculate the actual ship propulsion efficiency at different speeds and the total drag coefficient in wind and waves. as follows: in, This represents the actual wetted surface area of ​​the ship. This coefficient is used in the ITTC standard procedure to calculate propulsion efficiency and is a necessary intermediate parameter for fitting propulsion efficiency.

[0041] Based on the actual ship performance prediction method, we can obtain the drag increase at different speeds (i.e., different drag increases). The propulsion efficiency of the actual ship (below) Then calculate its relationship with the self-point ( The difference in propulsion efficiency (obtainable through linear interpolation) Then, function fitting is used. and The relationship between propulsion efficiency and drag ratio at this speed was obtained. By performing similar experiments at other speeds, the relationship between propulsion efficiency and drag increase at different speeds can be obtained.

[0042] effective power The calculation formula is as follows: in, This formula represents the actual power received by the propeller; it also specifies the main shaft power. With effective power The conversion relationship provides a physical basis for the subsequent calculation of the host power from the wave drag amplification in the gray box model.

[0043] Wave resistance The calculation is as follows: The iterative method uses CFD calculations to simulate the changes in propeller speed of a ship under different wind and wave conditions, and after obtaining different data points, a function is used to fit the relationship, thus obtaining the relationship between propulsion efficiency and wind and wave drag ratio. Then, an iterative method is used to solve the problem. The interpolation between the total wind and wave drag increase in the current (n+1)th iteration and the total wind and wave drag increase in the nth iteration is less than 0.00001, at which point the iteration ends. Compared with the propeller formula correction method that uses interpolation data constructed based on propeller speed, this method better reflects the changes in various drag coefficients during ship navigation and can better predict wave drag increase.

[0044] S4. Construction of gray box model for host power prediction; Support Vector Machines (SVMs) are fundamentally binary classification models. Their basic model constructs a linear classifier with the largest margin in the feature space; margin maximization is their core distinguishing feature from the perceptron. The learning strategy of SVM aims to maximize the margin, which can be formalized as a convex quadratic programming problem. This problem is equivalent to minimizing a regularized hinge loss function. The essence of the SVM learning algorithm is the optimization solution of this convex quadratic programming problem.

[0045] Suppose it contains The training set of sample pairs for each real-world ship operation data sample is as follows: ;in, For the first The input column vector consists of real-ship operational data from a training sample. ; This represents the host power and wave resistance output values ​​corresponding to this sample.

[0046] Let the linear regression function established in the high-dimensional feature space be: in, It is a nonlinear mapping function. Let be the weight vector to be solved. This is the bias term to be optimized.

[0047] definition The linearly insensitive loss function is: in, These are the predicted values ​​of main engine power and wave drag returned by the linear regression function. For the corresponding host power and wave resistance values, This is the regression error threshold, which means that if and The difference between them is less than or equal to If so, the loss is 0.

[0048] Following the approach used in SVM classification models, a slack variable ξ is introduced. i ξ i The above solution for the weight vector With bias term The optimization problem can be expressed in the following mathematical form: in, As a penalty factor, The larger the value, the greater the training error. The stronger the sample penalty, The error requirements for the regression function are specified. A smaller value indicates a smaller error in the regression function and a smaller error in predicting the main engine power and wave increase. To solve the above equation, the Lagrangian function is introduced, and the solution is transformed into a dual form, where... and It is a Lagrange multiplier in the duality problem.

[0049] in, This is the kernel function.

[0050] Suppose that the optimal solution obtained by solving the dual form is Then there is in, This is the optimal weight vector obtained through optimization. The optimal bias term obtained through optimization. This represents the number of support vectors.

[0051] The prediction function (regression function) for main engine power and wave drag is: The power of the ship's main shaft in wind and waves is: in, The total resistance in the wind and waves, For the ship's speed relative to the water, For shaft system efficiency ( (0.99). To increase drag due to predicted wind and waves, For hydrostatic resistance, To improve efficiency.

[0052] Improve efficiency In The values ​​can be interpolated from the data in the speed report. It can be calculated using the following formula: Different ship types have different coefficients. (Still water resistance) Interpolation calculations can be performed based on the data table; the predicted total wave resistance is... ,in To predict increased wave resistance, For wind resistance.

[0053] First, CFD calculations were used to obtain the relationship between ship propulsion efficiency and wind and wave drag under different propulsion conditions. By numerically simulating different propeller speeds and corresponding resistance conditions, a functional relationship between propulsion efficiency and drag ratio was established. Then, wave drag was separated using an iterative method, which served as the data source for prediction by the black-box model SVM algorithm. Next, the wave drag predicted by the black-box model SVM algorithm, along with still water resistance and wind resistance, were substituted into the propulsion power calculation formula to obtain the main engine power prediction result, thereby constructing a gray-box model of the main engine power of a bulk carrier.

[0054] S5. Model Evaluation and Comparison.

[0055] The parameters for model evaluation and comparative analysis are defined as follows: R² (Determination coefficient) MSE (mean square error) RMSE (Root Mean Square Error) and MAE (Mean absolute error).

[0056] Coefficient of determination ( R ²) is used to determine how close the data is to the fitted regression line, mean squared error (MSE). MSE The root mean square error (RMSE) is the average of the squared differences between predicted and actual values. RMSE ) is the square root of the mean square error, and the mean absolute error ( MAE The average of the absolute values ​​of the differences between the predicted and actual values ​​is given by the following formula: in, It is the actual value. It is the average of the true values. It is a predicted value. This refers to the sample size, and the same applies below.

[0057] The prediction accuracy, stability, and generalization ability of the gray box model are quantitatively evaluated using four indicators: R², MSE, RMSE, and MAE. The error between the model's power predictions on the test set and the actual ship measurements is calculated. Based on the processed wave direction, the power prediction results of the gray box model are visualized in polar coordinates and then plotted as a radar chart, as shown below. Figure 2 As shown in the figure, the actual power values ​​in the operational data and the power values ​​predicted by the gray box model are displayed. The fitting curve is fitted with a cubic polynomial. Due to the large number of data points, plotting them all on the radar chart would result in too dense a point distribution, making it difficult to display them effectively. Therefore, one data point is selected in increments of 20, resulting in approximately 300 points in the radar chart. This facilitates observation of the fitting effect in conjunction with the evaluation data in the table above. The gray box model is compared and validated with a pure machine learning black box model under the same parameters and dataset. The prediction errors, robustness, and extrapolation adaptability of the two models under different sea states and loading conditions are compared and analyzed. Based on the evaluation results, the model feature weights and hyperparameters are optimized to output the final main engine power prediction results that meet the requirements of maritime regulations and engineering applications.

[0058] Table 1 Evaluation results of gray box model and black box model Comparing the evaluation results of the gray-box model and the black-box model in Table 1, we can see that the R² of the gray-box model is 0.9988, significantly higher than that of the black-box model (0.9854). The RMSE and MAE of the gray-box model are 0.0333 and 0.0220, respectively, far superior to those of the black-box model (0.1219 and 0.0906). These results verify that the gray-box modeling method of this invention, which integrates physical mechanisms and SVM, can effectively improve the accuracy and robustness of bulk carrier main engine power prediction compared to purely data-driven methods, and also possesses good physical interpretability, making it suitable for practical engineering applications.

[0059] The present invention also provides a bulk carrier main engine power forecasting system, which includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the system is triggered to execute the above-described bulk carrier main engine power forecasting method.

[0060] This invention discloses a method and system for predicting the main engine power of bulk carriers, belonging to the field of green shipping and intelligent shipping technology. It aims to solve the problems of low accuracy, poor physical interpretability, weak adaptability to complex sea conditions, and inability to meet the verification requirements of the "Minimum Propulsion Power Guidelines" in existing power prediction methods. Based on big data from the operation of sister bulk carriers, this invention employs standardized preprocessing and Pearson correlation feature screening. It uses a two-dimensional Froude method combined with drag increase-propulsion efficiency iteration to achieve interpretable separation of wave drag increase. It integrates SVM machine learning to construct a gray-box prediction model constrained by physical mechanisms, and completes the entire prediction process—from data preprocessing and feature analysis to drag increase separation, model construction, and evaluation comparison—through black-box model comparison and optimization.

[0061] Verification using real-ship data shows that the gray-box model proposed in this invention significantly outperforms the purely data-driven black-box model in several quantitative indicators. The aforementioned comparative data fully demonstrates that, while maintaining physical interpretability, this invention significantly improves the prediction accuracy and stability of main engine power compared to existing machine learning methods, exhibiting higher consistency and smaller prediction bias.

[0062] This invention enables high-precision, robust, and interpretable real-time forecasting of main engine power for bulk carriers under complex wind, wave, and current conditions, significantly improving generalization capabilities under extrapolated sea states and variable loading conditions. The forecasting process conforms to hydrodynamic mechanisms, and the results are traceable and verifiable, meeting the consistency verification requirements of maritime regulations. Modeling can be performed using only routine operational data, making the process efficient and low-cost. It can be quickly reused for the same series of ship types, demonstrating outstanding engineering practicality and maritime compliance, effectively overcoming the inherent defects of traditional empirical methods, pure mechanism methods, and pure machine learning black-box methods.

[0063] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention. Such modifications or substitutions should all fall within the scope of the invention, or any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A bulk carrier main engine power prediction method, characterized by, The method includes the following steps: Collect big data on the operation of sister bulk carriers, and perform preprocessing such as missing value removal, draft and speed filtering, and angle standardization on the collected data. The Pearson correlation coefficient method was used to screen the power and wave drag-in correlation features of the preprocessed data to obtain effective features. Based on the two-dimensional method and drag-propulsion efficiency iteration, the separation of wave drag on a real ship is completed by utilizing effective features to obtain the wave drag and resistance components. By integrating SVM machine learning with ship hydrodynamic mechanisms, a gray box prediction model for main engine power is constructed using the wave drag increase and drag components as inputs. The gray box forecasting model is evaluated using R², MSE, RMSE, and MAE indices, and the final forecast results are output by comparing it with the black box model.

2. The method of claim 1, wherein, The specific aspects of the draft speed selection include: Considering the need to incorporate data from model tests and potential flow calculations into the analysis, the draft and speed data used for analysis must not exceed the range of model tests and potential flow calculations. The speed range for calculating still water resistance is determined through rapid response reports. Combining the range of operational data with the scope of ship performance research, the selection and classification rules for draft and speed are determined as follows: 1) Algorithm for classifying draft conditions ； wherein, n is the number of divisions, is the maximum minimum draft, a is the selected range, is the first k draft condition; 2) Speed ​​range division algorithm ； wherein, m number of divisions for speed, , maximum, minimum speed allowed for model test / rapidity report, single speed selection range, first j speed condition.

3. The method of claim 1, wherein, The angle standardization specifically includes: The actual ship operation data only contains the ship's wave direction angle and the absolute wave direction angle of the waves relative to the geodetic coordinate system. Therefore, it is necessary to perform a relative angle conversion for the ship's wave direction angle based on the actual ship operation data. In the wave drag prediction of the main engine power gray box model, the wave drag is equal in magnitude and direction along the ship's length, and equal in magnitude and opposite in direction along the ship's beam. In order to analyze the wave drag, the wave direction angle relative to the ship needs to be converted to... between, relative to the ship's wave angle ; in, The wave angle is relative to the Earth coordinate system. Let be the heading angle relative to the Earth coordinate system, when Then it is right-hand drive. Then it is left-hand drive; Actual wave angle for: 。 4. The method according to claim 1, characterized in that, The Pearson correlation coefficient method is used to screen the power and wave drag correlation features of the preprocessed data to obtain effective features, specifically including: The Pearson correlation coefficient measures whether there is a linear correlation between two features and the strength of that correlation. Therefore, the Pearson correlation coefficient is used to measure the correlation between power and other features. The formula is as follows: ； In the formula: The value indicates the strength of the correlation. A positive value indicates a positive correlation between the increase of the feature and the target feature, while a negative value indicates a negative correlation between the increase of the feature and the target feature.

5. The method according to claim 1, characterized in that, The method, based on the two-dimensional method and drag-propulsion efficiency iteration, utilizes effective features to separate wave drag on a real ship, obtaining wave drag and resistance components. This includes the estimation of still water resistance on a real ship based on the two-dimensional method. The specific steps are as follows: The two-dimensional method divides the total resistance of a ship into frictional resistance. and residual resistance By calculating the friction resistance coefficient and residual drag coefficient The friction resistance coefficient was then calculated. The residual resistance coefficient of a real ship is calculated using the ITTC 1957 formula. Friction resistance coefficient of ship model The static water resistance coefficient of the model ship After conversion, the following are the various resistance coefficients and resistance calculation formulas: Friction resistance coefficient as follows: ； ； in The Reynolds number is given by the following formula: ； ； For the ship's speed, Because of the ship's long waterline, The viscosity coefficient of water; Actual ship residual coefficient The calculation is as follows: ； The total still water resistance coefficient of the model boat The calculation is as follows: ； For the water density in the ship model test, The wetted surface area of ​​the ship model; Total still water resistance coefficient of the actual ship for: ； The wetted surface area of ​​the bilge keel. This represents the wetted surface area of ​​the actual ship. This is the roughness subsidy coefficient. This refers to the air drag coefficient; air drag coefficient as follows: ； It is the projected area of ​​the hull and superstructure above the waterline on the mid-section. Total static resistance and still water effective power The calculation is as follows: ； 。 6. The method according to claim 5, characterized in that, The method based on the two-dimensional method and drag-propulsion efficiency iteration utilizes effective features to complete the separation of wave drag on a real ship, obtaining wave drag and resistance components. It also includes the calculation of wind resistance on a real ship. The specific steps are as follows: The actual wind resistance of a ship is calculated using the following formula: ； relative wind direction angle The drag coefficient at the bottom, Let the mass density of air be 1.226 kg / m³. 3 , It represents the projected area of ​​the hull and superstructure above the waterline on the mid-section. Relative wind speed, For ground speed.

7. The method according to claim 6, characterized in that, The method based on the two-dimensional method and drag-propulsion efficiency iteration, which utilizes effective features to complete the separation of wave drag and resistance in actual ships, and obtains the wave drag and resistance components, also includes the separation of wave drag and resistance in actual ships based on drag-propulsion efficiency iteration. The specific steps are as follows: Assuming a certain ship propulsion efficiency Changes and increased resistance from wind and waves and hydrostatic resistance The ratio satisfies the following formula: ； Ship wind and wave propulsion efficiency for: ； The power of the ship's main shaft in wind and waves is: ； The total resistance in the wind and waves, For the ship's speed relative to the water, For shaft system efficiency; If the relationship between propulsion efficiency and the drag ratio of wind and waves is obtained... and the ship's hydrostatic resistance Then, by using the iterative method to solve the above equations, we can obtain the total increase in drag of the ship due to wind and waves. The static resistance of a ship The relationship between propulsion efficiency and wind and wave drag ratio was obtained using still water model tests. Obtained through variable load model tests or CFD calculations; At the model scale, self-propulsion calculations are performed at a fixed speed and different propeller speeds to simulate the increase in propeller speed under different wind and wave conditions and drag-inducing effects; the forces at different speeds are compared with those in the model test. Calculated using the following formula: ； in, and These represent the resistance of a self-propelled vessel with a propeller and the resistance of a vessel without a propeller, respectively; the corresponding increase in actual ship resistance is calculated using the following formula: ； in, This is the correction value for hydrostatic friction resistance. For the model's scaling ratio, and The water density at the scales of the actual ship and the model are respectively. The propulsion efficiency of a real ship at different speeds and the total drag coefficient in wind and waves were calculated using the ITTC real-ship performance prediction method. as follows: ； in, This represents the actual wetted surface area of ​​the ship. The propulsion efficiency of a real ship at different speeds was obtained based on the actual ship performance prediction method. Then calculate the difference in propulsion efficiency between its propulsion point and the self-center point. Then, use function fitting. and The relationship between propulsion efficiency and drag ratio at this speed was obtained. ; effective power The calculation formula is as follows: ； in, This represents the actual power received by the propeller. Wave resistance The calculation is as follows: 。 8. The method according to claim 7, characterized in that, The aforementioned integration of SVM machine learning and ship hydrodynamic mechanisms, using the wave drag amplification and drag components as inputs, constructs a gray box prediction model for main engine power, specifically including: The learning strategy of Support Vector Machine (SVM) aims to maximize the margin, which is essentially an optimization solution of convex quadratic programming. Suppose it contains The training set of sample pairs for each real-world ship operation data sample is as follows: ;in, For the first The input column vector consists of real-ship operational data from a training sample. ; The corresponding host power and wave impedance output values ​​are for this sample. Let the linear regression function established in the high-dimensional feature space be: ； in, It is a nonlinear mapping function. Let be the weight vector to be solved. The bias term to be optimized; definition The linearly insensitive loss function is: ； in, These are the predicted values ​​of main engine power and wave drag returned by the linear regression function. For the corresponding host power and wave resistance values, This is the regression error threshold, indicating that if... and The difference between them is less than or equal to If , then the loss is equal to 0; Introducing slack variable ξ i ξ i The weight vector will be solved. With bias term The optimization problem can be expressed in the following mathematical form: ； in, As a penalty factor, The larger the value, the greater the training error. The stronger the sample penalty, The error requirements for the regression function are specified. The smaller the value, the smaller the error of the regression function and the smaller the error in predicting the host power and wave increase.

9. The method according to claim 8, characterized in that, The method for constructing a gray-box prediction model for main engine power using the wave drag increase and drag components as input, which integrates SVM machine learning and ship hydrodynamic mechanisms, also includes: Solving for the weight vector With bias term When solving optimization problems, the Lagrangian function is introduced, and the problem is transformed into a dual form for solution, where... and For Lagrange multipliers in the duality problem; ； in, For kernel functions; Suppose that the optimal solution obtained by solving the dual form is Then we have: ； ； in, This is the optimal weight vector obtained through optimization. The optimal bias term obtained through optimization. This represents the number of support vectors. The prediction functions for main engine power and wave resistance are: ； The power of the ship's main shaft in wind and waves is: ； in, The total resistance in the wind and waves, For the ship's speed relative to the water, For shaft system efficiency; To increase drag due to predicted wind and waves, For hydrostatic resistance, To improve efficiency; to improve efficiency In The values ​​were interpolated from the data in the speed report. The calculation is performed using the following formula: ； Coefficients for different ship types; static resistance Based on interpolation calculations using the data table, the predicted total wave resistance is: ,in To predict increased wave resistance, For wind resistance.

10. A bulk carrier main engine power prediction system, the system comprising a memory for storing computer program instructions and a processor for executing the program instructions, characterized in that, When the computer program instructions are executed by the processor, the system is triggered to execute the bulk carrier main engine power prediction method as described in any one of claims 1 to 9.

Images