A data-driven ship stall estimation method and system

Abstract

The present invention discloses a data-driven ship stall estimation method and system. The method includes: collecting historical ocean environment information, historical ship navigation information, and historical ship main engine information; performing a first data preprocessing on the historical ocean environment information and historical ship navigation information, and performing a second data preprocessing on the historical ship main engine information; calculating a first impact speed based on the ground speed and current velocity; establishing a mapping relationship between the main engine speed and the ground speed based on wind speed and wave height screening samples to obtain a still water speed calculation model, and calculating a second impact speed; obtaining the ship stall value by the difference between the first impact speed and the second impact speed; and constructing a ship stall prediction model that inputs real-time ship navigation information and outputs the current ship stall estimation result. This method achieves rapid and accurate calculation of ship stall through data-driven means, avoids the complex calculation process of traditional mechanism models, and significantly enhances its engineering application value.

Description

Technical Field

[0001] The present invention relates to the field of ship management, and in particular to a data-driven ship stall estimation method and system. Background Art

[0002] During the navigation process, especially in the ocean, ships are inevitably affected by the wind, waves, and currents of the navigation environment. Compared with the calm water navigation environment, the ship speed is reduced, and the ship may stall. How to measure the stall of the ship under the influence of wind, waves and currents is of great significance for the implementation of route optimization and speed optimization. The existing ship stall is mainly based on the working characteristics of the ship, main engine, and propeller. The logical relationship between the various components in the system is analyzed, the ship resistance under wind, waves and currents is deduced, and then based on the force balance principle, the ship stall under different navigation environments is finally obtained. However, many parameters such as ship performance, main engine operating conditions, propellers, etc. are required. These parameter values ​​will change with the sailing time. The calculation process is cumbersome, which greatly limits the practical application and promotion. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to propose a data-driven ship stall estimation method and system thereof to solve the problem that ship stall analysis relies on mechanism models and has insufficient engineering practicality.

[0004] To achieve the above technical objectives, in a first aspect, the present application provides a data-driven ship stall estimation method, comprising:

[0005] Collect multi-source ship navigation information, including historical ocean environment information, historical ship navigation information, and historical ship main engine information. Historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, current speed, and current direction. Historical ship navigation information includes speed over ground and course over ground. Historical ship main engine information includes main engine speed and ship power.

[0006] Performing a first data preprocessing on the historical ocean environment information and the historical ship navigation information, the first data preprocessing including deleting abnormal data marks and interpolating and filling null value data, to obtain preprocessed historical ocean environment information and preprocessed historical ship navigation information;

[0007] Performing a second data preprocessing on the historical ship main engine information, the second data preprocessing including speed mutation point detection and correction, to obtain preprocessed historical ship main engine information;

[0008] The speed of the ship after the influence of wind and waves is calculated based on the speed over ground and still water resistance, wind resistance, wave resistance increase, and propeller efficiency, and recorded as the first impact speed;

[0009] The wind speed and wave height that meet the preset conditions are screened out from the pre-processed historical ocean environment information to generate sample information. Based on the sample information, a mapping relationship between the ship's main engine speed and the ground speed is established to obtain a still water speed calculation model.

[0010] The still water speed is predicted based on the still water speed calculation model and the historical ship main engine speed, which is recorded as the second influencing speed;

[0011] Calculating a ship stall value based on the first influencing speed and the second influencing speed, where the ship stall value is the difference between the second influencing speed and the first influencing speed;

[0012] A ship stall prediction model is constructed, using preprocessed historical ocean environment information and preprocessed historical ship engine information as input features and ship stall values ​​as output labels for model training, resulting in a trained ship stall prediction model.

[0013] In addition, real-time ship navigation information is collected and input into the trained ship stall prediction model, and the ship stall calculation result corresponding to the current ship is output.

[0014] In some embodiments, a first data preprocessing is performed on the historical ocean environment information and the historical ship navigation information. The first data preprocessing includes deleting abnormal data marks and interpolating and filling null value data to obtain the preprocessed historical ocean environment information and the preprocessed historical ship navigation information, including:

[0015] For wind speed and wave height, a sliding window statistical method is used to identify the first abnormal data point that exceeds the preset fluctuation range and mark it for deletion;

[0016] The ground speed and the ship's main engine speed are checked for rationality, and the second abnormal data point that does not conform to the ship's power performance is deleted;

[0017] For missing data on wind speed and direction, the synchronous observation data of adjacent ships on the same route are used to perform spatiotemporal correlation interpolation;

[0018] For missing data on wave height and direction, spatial correlation interpolation is performed using wave propagation models combined with data from observation stations in adjacent sea areas.

[0019] The missing data of ground speed are interpolated by model estimation through the mapping relationship between it and the ship's main engine speed;

[0020] The pre-processed historical ocean environment information and the pre-processed historical ship navigation information are obtained.

[0021] In some embodiments, a second data preprocessing is performed on the historical ship main engine information, and the second data preprocessing includes speed mutation point detection and correction, including:

[0022] Calculate the first-order difference of the ship's main engine speed in adjacent time windows and identify abnormal mutation points that exceed the preset speed change threshold;

[0023] Generate a ship power change curve based on the ship power, and perform a secondary check on the identified abnormal mutation points based on the ship power change curve to exclude valid mutations caused by normal operating conditions, and obtain a final set of abnormal mutation points. The final set of abnormal mutation points includes short-term pulse mutation points and continuous abnormal mutation points.

[0024] The short-time pulse mutation point is smoothed by using the median filter algorithm to obtain the first corrected data set;

[0025] The ship's main engine speed-power characteristic curve is used to reconstruct data for the continuous abnormal mutation point to obtain the second corrected data set;

[0026] Verifying the first anomaly correction data set and the second anomaly correction data set according to the physical feasibility of the ship power system to obtain a final anomaly correction data set;

[0027] Generate preprocessed historical ship host information based on the final anomaly correction dataset.

[0028] In some embodiments, the speed of the ship after being affected by wind and waves is calculated based on the ground speed and the current speed, and recorded as the first affected speed, including:

[0029] Establish a ship motion coordinate system and decompose the ground speed in the pre-processed historical ship navigation information into the longitudinal ground speed component and the transverse ground speed component;

[0030] The current velocity in the pre-processed historical ocean environment information is converted into the longitudinal current velocity component and the transverse current velocity component in the same coordinate system;

[0031] Calculate the resultant velocity vector of the ship relative to the water flow, including:

[0032] Longitudinal composite velocity component = longitudinal ground speed component - longitudinal flow velocity component;

[0033] Transverse composite velocity component = transverse ground speed component - transverse flow velocity component;

[0034] Generate a composite velocity vector according to the longitudinal composite velocity component and the transverse composite velocity component;

[0035] The first impact velocity is calculated based on the magnitude of the resultant velocity vector.

[0036] In some embodiments, wind speeds and wave heights that meet preset conditions are screened from historical ocean environment information to generate sample information, and a mapping relationship between the ship's main engine speed and the ground speed is established based on the sample information to obtain a still water speed calculation model, including:

[0037] Select wind speed and wave height that meet preset conditions from the pre-processed historical ocean environment information as sample information;

[0038] The sample information is verified for ship status, invalid samples with abnormal main engine or excessive rudder angle are excluded, and a valid still water sample set is obtained;

[0039] Based on the effective still water sample set, a nonlinear regression method is used to establish a mapping relationship model between the ship's main engine speed and the ground speed.

[0040] During the training process of the mapping relationship model, the ship's draft is introduced as an auxiliary variable to dynamically verify the established mapping relationship model. When the prediction deviation exceeds the preset allowable range, the model update mechanism is triggered to update the mapping relationship model.

[0041] Repeat the iterative training until the prediction deviation of the mapping relationship model is within the preset allowable range, which is recorded as the still water speed calculation model.

[0042] In some embodiments, the still water speed is calculated and predicted based on the still water speed calculation model and the historical ship main engine speed, and is recorded as the second influencing speed, including:

[0043] Based on the still water speed calculation model, the historical ship main engine speed is used as the model input parameter, and the initial still water speed prediction value is calculated through the mapping relationship;

[0044] According to the real-time ship draft, the initial still water speed prediction value is dynamically compensated and corrected to obtain the corrected still water speed value;

[0045] Smoothing the corrected still water speed value to eliminate the influence of random measurement errors and generate the second impact speed;

[0046] The second impact speed is associated with the historical ship main engine speed and the real-time ship draft and stored in the ship navigation status database.

[0047] In some embodiments, the first impact speed is expressed by formula (1), which is as follows:

[0048] ;

[0049] In formula (1), is the first impact speed after the impact of wind and waves, is the speed over ground in still water, is the hydrostatic resistance that first affects the velocity, is the hydrostatic resistance of the ship's speed over the ground in still water, is wind resistance, To increase the resistance of waves, is the propeller efficiency under wind and wave conditions, is the propeller efficiency in still water conditions.

[0050] In some embodiments, a ship stall prediction model is constructed, using historical ocean environment information and historical ship main engine information as input features and ship stall values ​​as output labels for model training, to obtain a trained ship stall prediction model, including:

[0051] Extracting environmental characteristic parameters from pre-processed historical navigation data, and extracting operation characteristic parameters from pre-processed historical ship main engine information to form a model input feature set;

[0052] The ship's stall value is calculated based on the second impact speed and the actual ground speed as the model output label;

[0053] Build a ship stall prediction model, screen environmental characteristic parameters and operational characteristic parameters through feature importance analysis, and obtain key input features;

[0054] A sample dataset is extracted from the sample database and divided into a training set and a validation set. The ship stall prediction model is trained based on the training set and validated based on the validation set. A cross-validation model is used to tune the model's hyperparameters during the training and validation process. An early stopping mechanism is introduced to prevent overfitting of the ship stall prediction model.

[0055] The ship stall prediction model is iteratively trained until the training is completed to obtain a trained ship stall prediction model.

[0056] In some embodiments, real-time ship navigation information is collected and input into a trained ship stall prediction model to output a ship stall estimation result corresponding to the current ship, including:

[0057] Real-time ship navigation information includes environmental monitoring information and real-time operating parameters. Environmental monitoring information includes real-time wind speed, real-time wave height, and real-time wave direction. Real-time operating parameters include real-time ship main engine speed, real-time draft, and real-time rudder angle.

[0058] Extract features from real-time ship navigation information and generate real-time feature vectors;

[0059] The real-time feature vector is input into the trained ship stall prediction model to obtain the predicted ship stall value at the current moment;

[0060] The predicted ship stall value is compared and verified with the current actual speed change curve, and the verified predicted ship stall value is dynamically smoothed to generate the ship stall calculation result corresponding to the current ship.

[0061] In the second aspect, the present invention also provides a ship stall estimation system based on data driving, which is applicable to the estimation method described in the first aspect. The system includes a data acquisition unit, a stall calculation unit and a stall prediction model. The data acquisition unit is used to collect multi-source information of ship navigation, and the multi-source information of ship navigation includes historical ocean environment information, historical ship navigation information and historical ship main engine information. The historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, flow speed and flow direction. The historical ship navigation information includes speed over ground and course over ground. The historical ship main engine information includes ship main engine speed and ship power. The stall calculation unit is used to perform a first data preprocessing on the historical ocean environment information and the historical ship navigation information. The first data preprocessing includes abnormal data mark deletion and null value data interpolation filling. The historical ship main engine information is subjected to a second data preprocessing, and the second data preprocessing includes speed mutation point detection and correction. According to the speed over ground and still water resistance, wind speed and wind speed, the stall calculation unit performs a second data preprocessing on the historical ship main engine information. The second data preprocessing includes speed mutation point detection and correction. The speed of the ship after the influence of wind and waves is calculated by taking into account the resistance, wave resistance increase, and propeller efficiency, which is recorded as the first influence speed; the wind speed and wave height that meet the preset conditions are screened out from the historical ocean environment information to generate sample information, and a mapping relationship between the ship's main engine speed and the ground speed is established based on the sample information to obtain a still water speed calculation model; the real-time still water speed is calculated according to the still water speed calculation model and the real-time ship main engine speed, which is recorded as the second influence speed; the ship's stall value is calculated based on the first influence speed and the second influence speed, and the ship's stall value is the difference between the second influence speed and the first influence speed; the stall prediction model is used to construct a ship stall prediction model, with historical ocean environment information and historical ship main engine information as input features and the ship stall value as the output label for model training to obtain a trained ship stall prediction model; and real-time ship navigation information is collected and input into the trained ship stall prediction model to output the ship stall estimation result corresponding to the current ship.

[0062] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0063] The present invention provides a data-driven method and system for estimating ship stall. The method comprises: collecting historical ocean environment information, historical ship navigation information, and historical ship main engine information; performing a first data preprocessing on the historical ocean environment information and historical ship navigation information, and performing a second data preprocessing on the historical ship main engine information; calculating a first influencing speed based on ground speed and still water resistance, wind resistance, wave added resistance, and propeller efficiency; establishing a mapping relationship between main engine speed and ground speed based on wind speed and wave height screening samples to obtain a still water speed calculation model, and calculating a second influencing speed; obtaining a ship stall value by taking the difference between the first influencing speed and the second influencing speed; and constructing a ship stall prediction model that inputs real-time ship navigation information and outputs a current ship stall estimation result. This method achieves rapid and accurate calculation of ship stall through data-driven means, avoids the complex calculation process of traditional mechanism models, and significantly enhances its engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 is a flow chart of the calculation method described in the specific implementation method;

[0066] Figure 2 is a data preprocessing flow chart described in a specific embodiment;

[0067] Figure 3 2. It is a schematic diagram of the calculation model of the main engine speed and still water speed described in the specific implementation manner;

[0068] Figure 4 2. It is a performance diagram of the ship stall prediction model described in the specific implementation manner;

[0069] Figure 5 2 is a flow chart of the calculation system according to the specific implementation method. DETAILED DESCRIPTION

[0070] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It is particularly noted that the following examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. Similarly, the following examples are only some embodiments of the present invention and are not intended to be exhaustive. All other embodiments obtained by those of ordinary skill in the art without creative effort are intended to fall within the scope of protection of the present invention.

[0071] See also Figures 1 to 4In a first aspect, this embodiment provides a data-driven ship stall estimation method, comprising:

[0072] Collect multi-source information of ship navigation, which includes historical ocean environment information, historical ship navigation information and historical ship host information. Historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, current speed and current direction, which can be expressed as ,in, For historical marine environmental information, is the wind speed, For wind direction, For the high waves, For the wave direction, is the wave cycle, is the flow rate, is the flow direction; the historical ship navigation information includes the speed over the ground and the course over the ground, which is expressed as ,in, For historical ship navigation information, is the ground speed, is the course to the ground; the historical ship main engine information includes the ship main engine speed and ship power, which is expressed as ,in, For historical ship host information, is the main engine speed of the ship, is the ship power; finally the matrix ,in, Provide multi-source information for ship navigation.

[0073] Performing a first data preprocessing on the historical ocean environment information and the historical ship navigation information, the first data preprocessing including deleting abnormal data marks and interpolating and filling null value data, to obtain preprocessed historical ocean environment information and preprocessed historical ship navigation information;

[0074] Performing a second data preprocessing on the historical ship main engine information, the second data preprocessing including speed mutation point detection and correction, to obtain preprocessed historical ship main engine information;

[0075] The speed of the ship after the influence of wind and waves is calculated based on the speed over ground and still water resistance, wind resistance, wave resistance increase, and propeller efficiency, and recorded as the first impact speed;

[0076] The wind speed and wave height that meet the preset conditions are screened out from the pre-processed historical ocean environment information to generate sample information. Based on the sample information, a mapping relationship between the ship's main engine speed and the ground speed is established to obtain a still water speed calculation model.

[0077] The still water speed is predicted based on the still water speed calculation model and the historical ship main engine speed, which is recorded as the second influencing speed;

[0078] Calculating a ship stall value based on the first influencing speed and the second influencing speed, where the ship stall value is the difference between the second influencing speed and the first influencing speed;

[0079] A ship stall prediction model is constructed, using preprocessed historical ocean environment information and preprocessed historical ship engine information as input features and ship stall values ​​as output labels for model training, resulting in a trained ship stall prediction model.

[0080] In addition, real-time ship navigation information is collected and input into the trained ship stall prediction model, and the ship stall calculation result corresponding to the current ship is output.

[0081] In this embodiment, the collection of multi-source information of ship navigation can be achieved through various sensor equipment on the ship. Preferably, the wind speed and direction data are measured in real time by the ultrasonic anemometer on the top of the mast, which can capture the real wind field conditions 10 meters above the sea surface; the wave height and wave direction are inversely calculated by the hull motion response analysis system, combined with the pitch and roll data of the inertial measurement unit, which can accurately reflect the actual effect of the waves on the ship; the speed and direction of the current are measured by the acoustic Doppler profiler installed on the bottom of the ship, and the measurement depth covers all layers of water flow within the draft range of the ship. The speed and heading relative to the ground in the ship's navigation data are obtained by fusing GPS signals with compass data, and the main engine speed and power data are directly read from the ship's automation system. These multi-source information are transmitted to the central server through the ship's local area network, and a timestamp alignment algorithm is used to ensure the time consistency of the data.

[0082] During the first data preprocessing of historical marine environmental information and historical ship navigation information, a dynamic threshold algorithm is used to mark and delete abnormal data, and a reasonable fluctuation range is set according to the climatic characteristics of the ship's navigation area. For example, the wave height fluctuation range allowed on the North Atlantic winter route is 20% wider than that on the summer route. The interpolation and filling of null value data takes into account the spatial correlation of marine environmental elements and adopts the radial basis function interpolation method centered on the ship route to ensure that the filled data conforms to the spatial distribution pattern of the actual marine environment. The second data preprocessing of historical ship main engine information focuses on speed mutations. By analyzing the main engine load characteristic curve, it distinguishes between normal operating condition switching and equipment abnormalities, and uses a reconstruction algorithm based on the power-speed relationship to correct the speed data confirmed to be abnormal.

[0083] The process of establishing the still water speed calculation model fully considers the differences in the hydrodynamic performance of ships under different loading conditions. When screening sample information, in addition to the basic wind speed and wave height conditions, the ship's draft and trim data are also introduced as auxiliary screening indicators to ensure the representativeness of the training samples. Optionally, the model uses a segmented regression method to divide the main engine speed into three intervals: low speed, medium speed and high speed, and establish mapping relationships respectively to more accurately reflect the nonlinear characteristics of the ship's propulsion efficiency as the speed changes. The model update mechanism is set to automatic trigger mode. When the actual speed of five consecutive voyages deviates from the predicted value by more than 3%, the system automatically starts the incremental learning process.

[0084] The training process of the ship stall prediction model pays special attention to the interaction between environmental factors. Preferably, a sliding time window is used to process the continuous data stream in the real-time calculation link. The window length is automatically adjusted according to the current sea conditions. When the wind and waves are calm, a longer time window is used to improve stability. When encountering severe sea conditions, the time window is shortened to enhance response speed. Furthermore, the calculation results are output in the form of a probability distribution, providing a risk warning reference for navigation decisions. This embodiment achieves precise quantification of the complex physical process of ship stall through a data-driven approach.

[0085] This embodiment ensures the accuracy and reliability of input data by collecting multi-source information on ship navigation and undergoing a strict data preprocessing process. The dynamic threshold algorithm and radial basis function interpolation method are used to process historical marine environmental information and historical ship navigation information, effectively eliminating the influence of abnormal data and null data; the integrity of the main engine data is guaranteed by detecting and correcting the speed mutation points of historical ship main engine information. The established still water speed calculation model takes into account the differences in the hydrodynamic performance of the ship under different loading conditions and accurately reflects the nonlinear characteristics of the ship's propulsion efficiency. The ship stall prediction model analyzes the interaction between environmental factors, combines sliding time window technology and probability distribution output, and realizes the accurate calculation of ship stall, providing reliable data support for navigation decisions.

[0086] In some embodiments, a first data preprocessing is performed on the historical ocean environment information and the historical ship navigation information. The first data preprocessing includes deleting abnormal data marks and interpolating and filling null value data to obtain the preprocessed historical ocean environment information and the preprocessed historical ship navigation information, including:

[0087] For wind speed and wave height, a sliding window statistical method is used to identify the first abnormal data point that exceeds a preset fluctuation range and mark it for deletion. The method for determining the preset fluctuation range includes: dynamically adjusting the fluctuation threshold parameters of each sea area based on the statistical distribution characteristics of historical meteorological data in the ship's navigation area;

[0088] The ground speed and the ship's main engine speed are checked for rationality, and the second abnormal data point that does not conform to the ship's power performance is deleted;

[0089] For missing data on wind speed and direction, we use synchronous observation data from adjacent ships on the same route to perform spatiotemporal interpolation. This interpolation includes: establishing a spatial grid with the longitude and latitude of the ships as coordinates, and selecting observation data from ships on the same route within adjacent time windows in the time dimension as the interpolation basis;

[0090] For missing data on wave height and direction, spatial correlation interpolation is performed using wave propagation models combined with data from observation stations in adjacent sea areas.

[0091] The missing data of ground speed are interpolated by model estimation through the mapping relationship between it and the ship's main engine speed;

[0092] The pre-processed historical ocean environment information and the pre-processed historical ship navigation information are obtained. The pre-processed historical ocean environment information must meet the data integrity requirements and ensure that each timestamp contains a minimum data set of three basic parameters: wind speed, wave height, and current speed.

[0093] In this embodiment, when performing the first data preprocessing on historical marine environmental information and historical ship navigation information, a sliding window statistical method uses a fixed-length data window (typically one hour) to calculate the moving average and standard deviation of wind speed and wave height within the window, thereby identifying the first abnormal data point that exceeds a preset fluctuation range. The method for determining the preset fluctuation range specifically involves establishing an extreme value distribution model based on 10 years of historical meteorological data for each typical ship navigation area. A generalized Pareto distribution is then used to fit the extreme wind and wave characteristics of each sea area in different seasons, thereby dynamically adjusting the fluctuation threshold parameters for each navigation area.

[0094] When verifying the rationality of the ground speed and main engine speed, the system uses the main engine characteristic curve and hull resistance curve from the ship design manual to establish a speed-speed interval matrix that includes state parameters such as the ship's draft and trim. If the correspondence between the ground speed and main engine speed is detected to be outside the reasonable range defined by this matrix, it is identified as a second abnormal data point that needs to be deleted. This verification method fully considers the differences in the ship's power performance under different loading conditions and avoids misjudgments caused by single-threshold verification.

[0095] When using spatiotemporal correlation interpolation to address missing wind speed and direction data, the route is first divided into a 0.1°×0.1° geographic grid. A one-hour time window is then established, centered on the missing point. Using a spatial inverse distance weighted algorithm and a time decay factor, a weighted average of observations from adjacent ships on the same route is calculated to ensure that the interpolation result conforms to the spatiotemporal evolution of the actual wind field. Spatial correlation interpolation of the wave propagation model uses a simplified version of the third-generation wave model, SWAN, combined with spectral data from nearby ocean observation stations. Wave height and direction parameters in the missing area are reconstructed by solving the wave action balance equation.

[0096] When modeling and interpolating missing data for speed over ground, a preprocessed database of historical ship engine information is used to construct a non-parametric probabilistic model of engine speed and speed over ground based on kernel density estimation. This model generates a confidence interval for speed over ground under current speed conditions, with the median value used as the optimal interpolation result. The resulting preprocessed historical ocean environment information must ensure that each timestamp contains at least three basic parameters: wind speed, wave height, and current velocity. These parameters constitute the minimum data dimension requirement for ship stall analysis.

[0097] This embodiment uses a hierarchical data quality control strategy to preserve the physical characteristics of the original data while ensuring its integrity and consistency. This is particularly true when processing data with strong spatiotemporal correlations, such as marine environmental data. By integrating the ship's own observation data with external reference data, the reliability of data preprocessing is effectively improved.

[0098] In some embodiments, a second data preprocessing is performed on the historical ship main engine information, and the second data preprocessing includes speed mutation point detection and correction, including:

[0099] Calculate the first-order difference of the ship's main engine speed in adjacent time windows to identify abnormal mutation points that exceed a preset speed change threshold. The method for determining the preset speed change threshold includes:

[0100] Obtain the rated speed change rate parameter according to the ship's main engine model;

[0101] Combined with historical navigation data, the speed fluctuation range under various working conditions is obtained;

[0102] Dynamically adjust the threshold to adapt to the speed change characteristics under different navigation conditions;

[0103] Generate a ship power change curve based on the ship power, and perform a secondary check on the identified abnormal mutation points based on the ship power change curve to exclude valid mutations caused by normal operating conditions, and obtain a final set of abnormal mutation points. The final set of abnormal mutation points includes short-term pulse mutation points and continuous abnormal mutation points.

[0104] The short-time pulse mutation point is smoothed by using the median filter algorithm to obtain the first corrected data set;

[0105] The ship's main engine speed-power characteristic curve is used to reconstruct data for the continuous abnormal mutation point to obtain the second corrected data set;

[0106] Verifying the first anomaly correction data set and the second anomaly correction data set according to the physical feasibility of the ship power system to obtain a final anomaly correction data set;

[0107] Generate preprocessed historical ship host information based on the final anomaly correction dataset.

[0108] In this embodiment, the principle behind using first-order differences to detect speed mutation points is that changes in a ship's main engine speed are essentially an inertial response process of the mechanical system, and its rate of change is strictly constrained by the main engine's dynamic characteristics. By calculating the speed difference between adjacent time windows, the instantaneous rate of speed change can be quantified. Under real-world operating conditions, this rate is limited by the main engine's maximum acceleration capability. When the difference exceeds a threshold determined by main engine performance parameters and historical operating condition statistics, it indicates an abnormal jump that violates physical laws, potentially due to sensor failure, data transmission errors, or recording anomalies.

[0109] The secondary verification of the power curve leverages the energy conservation principle of the ship's propulsion system: there is a definite coupling relationship between main engine power and speed changes. During normal operating conditions, the power and speed curves exhibit synchronous and gradual changes; however, abnormal sudden changes often manifest as speed-power decoupling. This cross-validation, based on physical laws, effectively eliminates anomalies caused by acceleration and deceleration operations.

[0110] The median filter algorithm is suitable for pulse-type mutations and can preserve the trend characteristics of the true signal. For persistent anomalies, the speed-power characteristic curve is used for reconstruction. This is because the long-term anomaly has destroyed the physical plausibility of the original data, and data that conforms to the power characteristics must be reconstructed based on the system's energy conversion relationship. The final physical verification phase introduces auxiliary parameters such as fuel consumption and establishes multi-dimensional constraints to ensure that the corrected data meets the complete operational logic of the ship's power system.

[0111] This embodiment uses differential operations to capture anomalies that violate the law of inertia, utilizes the power-speed coupling relationship to identify unreasonable jumps, and ultimately ensures the physical consistency of the data based on multi-parameter system constraints. Through a processing method based on domain knowledge, it can identify deep data anomalies that violate physical principles better than traditional statistical filtering, providing high-quality data that is both mathematically reasonable and physically realistic for subsequent analysis.

[0112] In some embodiments, the speed of the ship after being affected by wind and waves is calculated based on the ground speed and the current speed, and recorded as the first affected speed, including:

[0113] Establish a ship motion coordinate system and decompose the ground speed in the pre-processed historical ship navigation information into the longitudinal ground speed component and the transverse ground speed component;

[0114] The current velocity in the pre-processed historical ocean environment information is converted into the longitudinal current velocity component and the transverse current velocity component in the same coordinate system;

[0115] Calculate the resultant velocity vector of the ship relative to the water flow, including:

[0116] Longitudinal composite velocity component = longitudinal ground speed component - longitudinal flow velocity component;

[0117] Transverse composite velocity component = transverse ground speed component - transverse flow velocity component;

[0118] Generate a composite velocity vector according to the longitudinal composite velocity component and the transverse composite velocity component;

[0119] The first impact velocity is calculated according to the modulus of the composite velocity vector. The calculation formula is:

[0120] ;

[0121] In the above formula, is the first impact velocity, is the longitudinal composite velocity component, is the lateral composite velocity component.

[0122] In this embodiment, vector decomposition is performed by establishing a motion coordinate system. The actual movement of a ship under complex sea conditions is a multi-factor coupling process of fluid dynamics. Decomposing the speed and flow rate into longitudinal / transverse components for vector calculation is essentially applying the principle of relative motion in fluid mechanics to eliminate the interference of water flow on the ship's positioning data. This processing method can accurately remove the influence of environmental factors on the ship's motion and provide benchmark data that conforms to the laws of fluid dynamics for subsequent analysis. This embodiment realizes the mathematical decoupling of the ship's actual propulsion performance and environmental interference through coordinate system transformation. This method based on motion decomposition can better reflect the actual dynamic characteristics of the ship than the traditional direct calculation of speed.

[0123] In some embodiments, wind speeds and wave heights that meet preset conditions are screened from historical ocean environment information to generate sample information, and a mapping relationship between the ship's main engine speed and the ground speed is established based on the sample information to obtain a still water speed calculation model, including:

[0124] Select wind speed and wave height that meet preset conditions from the pre-processed historical ocean environment information as sample information;

[0125] The sample information is verified for ship status, invalid samples with abnormal main engine or excessive rudder angle are excluded, and a valid still water sample set is obtained;

[0126] Based on the effective still water sample set, a nonlinear regression method is used to establish a mapping relationship model between the ship's main engine speed and the ground speed.

[0127] During the training process of the mapping relationship model, the ship's draft is introduced as an auxiliary variable to dynamically verify the established mapping relationship model. When the prediction deviation exceeds the preset allowable range, the model update mechanism is triggered to update the mapping relationship model.

[0128] Repeat the iterative training until the prediction deviation of the mapping relationship model is within the preset allowable range, which is recorded as the still water speed calculation model.

[0129] In this embodiment, considering that the vessel's stall rate is minimal when the wind speed is less than force 3 and the wave height is less than 0.5m, it can be assumed that the ground speed under these wind and wave conditions is equal to the still water speed. Sample information that meets these pre-set conditions (wind speed < force 3 and wave height < 0.5m) is selected from the pre-processed historical ocean environment information. Invalid samples due to engine anomalies or excessive rudder angles are further excluded to form a valid still water sample set.

[0130] The ship's main engine speed in the effective still water sample set and the ground speed under the influence of wind and waves are combined into a speed-speed mapping matrix ,in, is the mapping relationship matrix between the main engine speed and the ship speed under still water conditions, which serves as the input data carrier of the mapping relationship model. is the main engine speed parameter in the pre-processed historical ship main engine information, is the speed over ground after filtering by wind and wave conditions (i.e., regarded as still water speed), based on the matrix , with the main engine speed as the horizontal axis and the ground speed as the vertical axis, the data visualization is carried out, and the linear regression method is used to establish the still water speed calculation model. Assume that the model equation is , using the least squares method to find the slope and intercept ,like Figure 3 As shown. Among them, is the predicted still water speed. The resulting still water speed calculation model is used to calculate the second impact speed based on the historical ship's main engine speed. During model training, the ship's draft should be introduced as an auxiliary variable for dynamic verification. Preferably, a model update mechanism is triggered when the prediction deviation exceeds 5%.

[0131] This embodiment ensures that the established stillwater speed calculation model accurately reflects the ship's performance characteristics under stillwater conditions by screening sample information that meets preset conditions and excluding invalid samples due to engine anomalies or excessive rudder angles. A nonlinear regression method is used to establish a mapping relationship model between the ship's engine speed and speed over ground, and the ship's draft is introduced as an auxiliary variable for dynamic verification, effectively improving the model's prediction accuracy. A model update mechanism is triggered when the prediction deviation exceeds the preset allowable range, ensuring continuous optimization of the stillwater speed calculation model and providing a reliable basis for the calculation of the second-influence speed.

[0132] In some embodiments, the still water speed is calculated and predicted based on the still water speed calculation model and the historical ship main engine speed, and is recorded as the second influencing speed, including:

[0133] Based on the still water speed calculation model, the historical ship main engine speed is used as the model input parameter, and the initial still water speed prediction value is calculated through the mapping relationship;

[0134] According to the real-time ship draft, the initial still water speed prediction value is dynamically compensated and corrected to obtain the corrected still water speed value;

[0135] Smoothing the corrected still water speed value to eliminate the influence of random measurement errors and generate the second impact speed;

[0136] The second impact speed is associated with the historical ship main engine speed and the real-time ship draft and stored in the ship navigation status database.

[0137] In this embodiment, the still water speed calculation model is a deterministic mapping relationship between the ship's main engine speed and ground speed, established using a valid still water sample set in the previous embodiment. This model is trained through nonlinear regression, and its input parameters are strictly limited to historical main engine speeds that have undergone secondary data preprocessing. The initial still water speed prediction is generated by substituting the preprocessed speed data into the model equation for a single-step forward calculation. Its physical meaning represents the theoretical ship speed under ideal still water conditions.

[0138] Dynamic compensation is achieved based on the nonlinear relationship between a ship's draft and hull resistance. This is achieved by establishing a draft-resistance correction coefficient lookup table, using the ship's real-time draft as an index parameter. Speed ​​compensation is then applied to the initial prediction to match the current draft. This lookup table is constructed by discretizing the draft-resistance curves obtained from tank tests, ensuring that the corrected stillwater speed value conforms to the ship's actual hydrodynamic characteristics.

[0139] Smoothing uses an exponentially weighted filtering algorithm based on time series. Its filter coefficients are adaptively adjusted based on the ship's inertial characteristics, primarily to eliminate speed fluctuations caused by speed sensor noise. The second-influence speed, the steady-state output value after smoothing, essentially reflects the theoretical still water speed that the ship can achieve at its current draft. By calculating the difference between this and the first-influence speed, it accurately quantifies speed losses due to wind and wave conditions.

[0140] The ship navigation status database is implemented using a time-series database architecture. Its storage relationships are established through a unified timestamp index, ensuring strict synchronization of the second impact speed with the historical ship engine speed and real-time ship draft at the corresponding moment. This database provides a benchmark data source for ship performance analysis and supports time range queries to obtain still water speed characteristics for any segment.

[0141] This embodiment accurately calculates the second impact speed using a still water speed calculation model. Combined with dynamic compensation and smoothing, this effectively eliminates the effects of draft variations and measurement noise, ensuring the second impact speed accurately reflects the ship's still water performance. The second impact speed is associated and stored with historical engine speeds and real-time draft depth, providing a reliable benchmark for stall analysis and improving stall estimation accuracy.

[0142] In some embodiments, the first impact speed is expressed by formula (1), which is as follows:

[0143] ;

[0144] In formula (1), V_s is the first impact speed after the influence of wind and waves, V_0 is the speed over the ground in still water, R_1 is the still water resistance of the first impact speed, R_2 is the still water resistance of the speed over the ground in still water, R_a is the wind resistance, R_aW is the wave resistance increase, η_1 is the propeller efficiency under wind and wave conditions, and η_2 is the propeller efficiency under still water conditions.

[0145] In this embodiment, the first impact speed calculation model established by formula (1) fully considers the force characteristics of the ship in the wind and wave environment from the perspective of fluid mechanics. and hydrostatic resistance Accurately characterizes the still water resistance characteristics of the hull at different speeds; wind resistance and wave resistance The effect of wind and wave environment on the additional resistance of ships was quantified; and The ratio of reflects the difference in propeller efficiency under wind and wave conditions and still water conditions. Formula (1) realizes the refined modeling of wind and wave interference factors, making the calculated first impact speed This model accurately reflects a ship's actual speed performance in a real ocean environment. Compared to traditional methods, this calculation model has clear physical meaning and theoretical basis, providing scientific and reliable benchmark data for subsequent calculations of ship stall values, significantly improving the accuracy and credibility of stall estimation results.

[0146] In some embodiments, a ship stall prediction model is constructed, using historical ocean environment information and historical ship main engine information as input features and ship stall values ​​as output labels for model training, to obtain a trained ship stall prediction model, including:

[0147] Extracting environmental characteristic parameters from pre-processed historical navigation data, and extracting operation characteristic parameters from pre-processed historical ship main engine information to form a model input feature set;

[0148] The ship's stall value is calculated based on the second impact speed and the actual ground speed as the model output label;

[0149] Build a ship stall prediction model, screen environmental characteristic parameters and operational characteristic parameters through feature importance analysis, and obtain key input features;

[0150] A sample dataset is extracted from the sample database and divided into a training set and a validation set. The ship stall prediction model is trained based on the training set and validated based on the validation set. A cross-validation model is used to tune the model's hyperparameters during the training and validation process. An early stopping mechanism is introduced to prevent overfitting of the ship stall prediction model.

[0151] The ship stall prediction model is iteratively trained until the training is completed to obtain a trained ship stall prediction model.

[0152] In this embodiment, in the process of constructing a ship stall prediction model, environmental characteristic parameters (including wind speed, wind direction, wave height, wave direction, wave period) and operating characteristic parameters (including ship main engine speed) are first extracted from the pre-processed historical navigation data to form a model input feature set; at the same time, the ship stall value is calculated as the output label based on the difference between the second impact speed and the speed over the ground. After the sample data set is randomly divided into a training set and a validation set in a ratio of 0.8:0.2, the BP neural network, random forest, gradient boosting decision tree and LASSO algorithm are used to establish prediction models respectively, and the key input features are screened by feature importance analysis. In the model training stage, the hyperparameters of various algorithms are optimized using Bayesian search combined with the ten-fold cross-validation method. The model is trained using the training set and the performance is evaluated using the validation set. The main evaluation indicators include the determination coefficient R², mean square error MSE, root mean square error RMSE and mean absolute error MAE (such as Figure 4An early stopping mechanism is introduced during the training process to prevent overfitting, and the trained ship stall prediction model is finally obtained after iterative optimization.

[0153] This example achieves high-precision ship stall prediction through a systematic model-building process. The model input feature set is constructed based on environmental characteristic parameters (such as wind speed and wave height) and operational characteristic parameters (such as main engine speed), combined with stall labels that influence speed calculation, ensuring data integrity. Multiple machine learning algorithms (such as BP neural networks and random forests) are used in parallel for modeling. Feature importance analysis and Bayesian hyperparameter optimization (10-fold cross-validation) are used to improve model generalization. An early stopping mechanism is introduced to effectively prevent overfitting. The final model, validated by multiple metrics such as R² and MSE, significantly improves the accuracy and reliability of stall prediction, providing intelligent decision-making support for ship energy efficiency management.

[0154] In some embodiments, real-time ship navigation information is collected and input into a trained ship stall prediction model to output a ship stall estimation result corresponding to the current ship, including:

[0155] Real-time ship navigation information includes environmental monitoring information and real-time operating parameters. Environmental monitoring information includes real-time wind speed, real-time wave height, and real-time wave direction. Real-time operating parameters include real-time ship main engine speed, real-time draft, and real-time rudder angle.

[0156] Extract features from real-time ship navigation information and generate real-time feature vectors;

[0157] The real-time feature vector is input into the trained ship stall prediction model to obtain the predicted ship stall value at the current moment;

[0158] The predicted ship stall value is compared and verified with the current actual speed change curve, and the verified predicted ship stall value is dynamically smoothed to generate the ship stall calculation result corresponding to the current ship.

[0159] In this embodiment, real-time ship navigation information refers to environmental monitoring information and operating status parameters collected in real time by a sensor system during navigation. Environmental monitoring information specifically includes real-time wind speed, wave height, and wave direction. These parameters can be obtained through equipment such as weather stations and wave radars and are used to characterize the current sea conditions. Real-time operating parameters include real-time main engine speed, draft, and rudder angle. These parameters are collected through equipment such as engine room monitoring systems and draft gauges and are used to reflect the ship's real-time operating status.

[0160] Real-time feature vectors are structured feature data generated by preprocessing raw real-time ship navigation information through standardization and normalization. Their dimensions match the input layer of the trained ship stall prediction model. Real-time feature vectors are generated through feature engineering methods to ensure that the input data format is consistent with the feature set used during model training.

[0161] The predicted stall value is the output of the trained stall prediction model based on the input real-time feature vector. The predicted stall value is verified by comparing it with the current actual speed curve to ensure consistency with actual navigation conditions. The verified predicted value is then processed through dynamic smoothing methods such as Kalman filtering to generate a stable and reliable stall estimation result, providing a basis for real-time decision-making in ship energy efficiency management.

[0162] This embodiment collects real-time environmental monitoring information and operational parameters, constructs a trained ship stall prediction model using real-time feature vector input, and implements online dynamic estimation of ship stall. The prediction results are verified against actual speed curves and then dynamically smoothed, significantly improving the real-time and accuracy of stall predictions. This provides a reliable, real-time decision-making basis for ship speed optimization and energy efficiency management.

[0163] See also Figure 5In the second aspect, the present embodiment further provides a ship stall estimation system based on data driving, which is applicable to the estimation method described in the first aspect. The system includes a data acquisition unit, a stall calculation unit and a stall prediction model. The data acquisition unit is used to collect multi-source ship navigation information, and the multi-source ship navigation information includes historical ocean environment information, historical ship navigation information and historical ship main engine information. The historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, flow speed and flow direction. The historical ship navigation information includes speed over ground and course over ground. The historical ship main engine information includes ship main engine speed and ship power. The stall calculation unit is used to perform a first data preprocessing on the historical ocean environment information and the historical ship navigation information. The first data preprocessing includes abnormal data mark deletion and null value data interpolation filling. The historical ship main engine information is subjected to a second data preprocessing. The second data preprocessing includes speed mutation point detection and correction. According to the speed over ground and hydrostatic resistance, The speed of the ship after the influence of wind and waves is calculated based on wind resistance, wave added resistance, and propeller efficiency, which is recorded as the first influencing speed; the wind speed and wave height that meet the preset conditions are screened out from the historical ocean environment information to generate sample information, and the mapping relationship between the ship's main engine speed and the ground speed is established based on the sample information to obtain the still water speed calculation model; the real-time still water speed is calculated according to the still water speed calculation model and the real-time ship main engine speed, which is recorded as the second influencing speed; the ship's stall value is calculated based on the first influencing speed and the second influencing speed, and the ship's stall value is the difference between the second influencing speed and the first influencing speed; the stall prediction model is used to construct a ship stall prediction model, with historical ocean environment information and historical ship main engine information as input features, and the ship stall value as the output label for model training to obtain a trained ship stall prediction model; and real-time ship navigation information is collected and input into the trained ship stall prediction model to output the ship stall estimation result corresponding to the current ship.

[0164] In this embodiment, the data acquisition unit refers to a hardware sensor network and data acquisition module used to collect various multi-source information during a vessel's voyage. It uses the ship's weather station, wave radar, GPS positioning system, and engine room monitoring equipment to obtain historical ocean environmental information (such as wind speed and wave height), historical ship navigation information (such as speed over ground), and historical ship engine information (such as engine speed). The stall calculation unit, a computing module with data processing capabilities, performs a first data preprocessing step to clean outliers in the ocean environmental data and a second data preprocessing step to correct sudden changes in the engine speed data to ensure data quality.

[0165] The data-driven ship stall estimation system proposed in this embodiment is used to implement the estimation method described in the first aspect, and its description is not repeated here.

[0166] This system uses a data acquisition unit to collect multi-source navigation data in real time. After data cleaning and feature calculation by the stall calculation unit, a trained stall prediction model outputs highly accurate stall estimation results. This data-driven processing flow fully automates the entire process from raw data acquisition to stall prediction, effectively improving the real-time and accuracy of ship energy efficiency analysis. The coordinated operation of various system units provides reliable technical support for ship navigation optimization.

[0167] By adopting the above technical solution, the present invention is different from the existing technology and has the following beneficial effects:

[0168] The present invention collects multi-source ship navigation information, including historical ocean environment information, historical ship navigation information, and historical ship main engine information, and establishes a data-driven ship stall estimation method through a rigorous data preprocessing process. This method ensures the accuracy and integrity of the input data through a first data preprocessing process that removes abnormal data labels and interpolates and fills null data in the historical ocean environment and historical ship navigation information, and a second data preprocessing process that detects and corrects speed mutation points in the historical ship main engine information. Based on the preprocessed data, the first impact speed of the ship after the influence of wind and waves is calculated, and a still water speed calculation model is established by screening sample information that meets preset conditions to obtain the second impact speed. By constructing a ship stall prediction model with historical ocean environment information and historical ship main engine information as input features and ship stall values ​​as output labels, accurate prediction of ship stall is achieved. Finally, by collecting ship navigation information in real time and inputting it into the trained prediction model, the stall estimation result of the current ship is output, providing a real-time decision-making basis for ship navigation optimization and energy efficiency management. The present invention realizes the full-process automation from data acquisition and preprocessing to model training and real-time prediction, which significantly improves the accuracy and reliability of ship stall prediction.

[0169] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0170] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0171] The above descriptions are only some embodiments of the present invention and do not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A data-driven ship stall estimation method, characterized in that: include: Collecting multi-source ship navigation information, including historical ocean environment information, historical ship navigation information, and historical ship main engine information. The historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, current speed, and current direction. The historical ship navigation information includes speed over ground and course over ground. The historical ship main engine information includes main engine speed and ship power. Performing a first data preprocessing on the historical ocean environment information and the historical ship navigation information, wherein the first data preprocessing includes deleting abnormal data marks and interpolating and filling null value data to obtain preprocessed historical ocean environment information and preprocessed historical ship navigation information; Performing a second data preprocessing on the historical ship main engine information, the second data preprocessing including speed mutation point detection and correction, to obtain preprocessed historical ship main engine information; The speed of the ship after the influence of wind and waves is calculated based on the speed over ground and still water resistance, wind resistance, wave resistance increase, and propeller efficiency, and recorded as the first impact speed; The first impact speed is expressed by formula (1), which is as follows: ； In formula (1), is the first impact speed after the impact of wind and waves, is the speed over ground in still water, is the hydrostatic resistance that first affects the velocity, is the hydrostatic resistance of the ship's speed over the ground in still water, is wind resistance, To increase the resistance of waves, is the propeller efficiency under wind and wave conditions, is the propeller efficiency under still water conditions; Select wind speed and wave height that meet preset conditions from the pre-processed historical ocean environment information as sample information; Verify the ship status of the sample information, exclude invalid samples with abnormal main engine or excessive rudder angle, and obtain a valid still water sample set; Based on the effective still water sample set, a nonlinear regression method is used to establish a mapping relationship model between the ship's main engine speed and the ground speed. A segmented regression method is used to divide the main engine speed into three intervals: low speed, medium speed and high speed, and mapping relationship models are established respectively. During the training process of the mapping relationship model, the ship's draft is introduced as an auxiliary variable. In the real-time calculation link, a sliding time window is used to process the continuous data stream. The window length is automatically adjusted according to the current sea conditions. When the wind and waves are calm, a longer time window is used to improve stability. When encountering severe sea conditions, the time window is shortened to enhance the response speed. The calculation results are output in the form of probability distribution. The established mapping relationship model is dynamically verified. When the prediction deviation exceeds the preset allowable range, the model update mechanism is triggered to update the mapping relationship model. Repeat the iterative training until the prediction deviation of the mapping relationship model is within a preset allowable range, which is recorded as the still water speed calculation model; Based on the still water speed calculation model, the historical ship main engine speed is used as the model input parameter, and the initial still water speed prediction value is calculated through the mapping relationship; According to the real-time draft of the ship, the initial still water speed prediction value is dynamically compensated and corrected to obtain a corrected still water speed value; Smoothing the corrected still water speed value to eliminate the influence of random measurement errors and generate the second impact speed; The second impact speed is associated with the historical ship main engine speed and the real-time ship draft and stored in the ship navigation status database; Calculating a ship stall value based on the first influencing speed and the second influencing speed, wherein the ship stall value is a difference between the second influencing speed and the first influencing speed; A ship stall prediction model is constructed, using preprocessed historical ocean environment information and preprocessed historical ship engine information as input features and ship stall values ​​as output labels for model training, resulting in a trained ship stall prediction model. In addition, real-time ship navigation information is collected and input into the trained ship stall prediction model, and the ship stall calculation result corresponding to the current ship is output.

2. The data-driven ship stall estimation method according to claim 1, characterized in that: A ship stall prediction model is constructed. The historical ocean environment information and historical ship main engine information are used as input features, and the ship stall value is used as the output label for model training. The trained ship stall prediction model is obtained, which includes: Extracting environmental characteristic parameters from pre-processed historical navigation data, and extracting operation characteristic parameters from pre-processed historical ship main engine information to form a model input feature set; Calculate the ship's stall value according to the second impact speed and the actual ground speed as a model output label; Build a ship stall prediction model, screen environmental characteristic parameters and operational characteristic parameters through feature importance analysis, and obtain key input features; A sample data set is extracted from a sample database, and the sample data set is divided into a training set and a validation set. The ship stall prediction model is trained according to the training set, and the ship stall prediction model is validated according to the validation set. A cross-validation mode is used to tune the hyperparameters of the model during the training and validation process, and an early stopping mechanism is introduced to prevent overfitting of the ship stall prediction model. The ship stall prediction model is iteratively trained until the training is completed to obtain a trained ship stall prediction model.

3. The data-driven ship stall estimation method according to claim 1, characterized in that: Collect real-time ship navigation information and input it into the trained ship stall prediction model, outputting the ship stall prediction results corresponding to the current ship, including: The real-time ship navigation information includes environmental monitoring information and real-time operating parameters. The environmental monitoring information includes real-time wind speed, real-time wave height, and real-time wave direction. The real-time operating parameters include real-time ship main engine speed, real-time draft, and real-time rudder angle. Extract features from real-time ship navigation information and generate real-time feature vectors; Inputting the real-time feature vector into the trained ship stall prediction model to obtain the predicted ship stall value at the current moment; The predicted ship stall value is compared and verified with the current actual speed change curve, and the verified predicted ship stall value is dynamically smoothed to generate the ship stall calculation result corresponding to the current ship.

4. The data-driven ship stall estimation method according to claim 1, characterized in that: Performing a first data preprocessing on the historical ocean environment information and the historical ship navigation information, wherein the first data preprocessing includes deleting abnormal data marks and interpolating and filling null value data, to obtain the preprocessed historical ocean environment information and the preprocessed historical ship navigation information, including: For wind speed and wave height, a sliding window statistical method is used to identify the first abnormal data point that exceeds the preset fluctuation range and mark it for deletion; The ground speed and the ship's main engine speed are checked for rationality, and the second abnormal data point that does not conform to the ship's power performance is deleted; For missing data on wind speed and direction, the synchronous observation data of adjacent ships on the same route are used to perform spatiotemporal correlation interpolation; For missing data on wave height and direction, spatial correlation interpolation is performed using wave propagation models combined with data from observation stations in adjacent sea areas. The missing data of ground speed are interpolated by model estimation through the mapping relationship between it and the ship's main engine speed; The pre-processed historical ocean environment information and the pre-processed historical ship navigation information are obtained.

5. The data-driven ship stall estimation method according to claim 1, characterized in that: Performing a second data preprocessing on the historical ship main engine information, the second data preprocessing including speed mutation point detection and correction, including: Calculate the first-order difference of the ship's main engine speed in adjacent time windows and identify abnormal mutation points that exceed the preset speed change threshold; Generate a ship power change curve based on the ship power, and perform a secondary check on the identified abnormal mutation points based on the ship power change curve to exclude valid mutations caused by normal operating condition switching, thereby obtaining a final abnormal mutation point set, which includes short-term pulse mutation points and continuous abnormal mutation points; The short-time pulse mutation point is smoothed by using the median filter algorithm to obtain the first corrected data set; The ship's main engine speed-power characteristic curve is used to reconstruct data for the continuous abnormal mutation point to obtain the second corrected data set; Verifying the first anomaly correction data set and the second anomaly correction data set according to the physical feasibility of the ship power system to obtain a final anomaly correction data set; Preprocessed historical ship host information is generated according to the final abnormal correction data set.

6. A data-driven ship stall estimation system, characterized in that: The method for inferring data according to any one of claims 1 to 5, wherein the system comprises: a data acquisition unit for collecting multi-source ship navigation information, wherein the multi-source ship navigation information includes historical ocean environment information, historical ship navigation information, and historical ship main engine information; the historical ocean environment information includes wind speed, wind direction, wave height, wave direction, wave period, current speed, and current direction; the historical ship navigation information includes speed over ground and course over ground; and the historical ship main engine information includes main engine speed and ship power; a stall calculation unit for performing a first data preprocessing on historical ocean environment information and historical ship navigation information, the first data preprocessing including deletion of abnormal data markers and interpolation and filling of null value data; performing a second data preprocessing on historical ship main engine information, the second data preprocessing including detection and correction of speed mutation points; calculating the speed of the ship after the influence of wind and waves based on the ground speed and still water resistance, wind resistance, wave resistance increase, and propeller efficiency, and recording it as a first influence speed; screening wind speeds and wave heights that meet preset conditions from the historical ocean environment information to generate sample information, establishing a mapping relationship between the ship main engine speed and the ground speed based on the sample information, and obtaining a still water speed calculation model; calculating the real-time still water speed based on the still water speed calculation model and the real-time ship main engine speed, and recording it as a second influence speed; calculating the ship stall value based on the first influence speed and the second influence speed, and the ship stall value is the difference between the second influence speed and the first influence speed; The stall prediction model is used to construct a ship stall prediction model, using historical ocean environment information and historical ship main engine information as input features and ship stall values ​​as output labels for model training to obtain a trained ship stall prediction model; and real-time ship navigation information is collected and input into the trained ship stall prediction model, outputting the ship stall estimation result corresponding to the current ship.

Images