Ship air sail navigation aid energy efficiency dynamic optimization method and system and ship

Abstract

The invention discloses a ship air sail navigation aiding energy efficiency dynamic optimization method and system and a ship, and relates to the technical field of ship energy conservation. Dynamic parameters in the ship navigation process are collected in real time; establishing a dynamic coupling model; the coupling model is utilized to quantify the dynamic interference of the hull attitude parameters on the air sail state parameters into attack angle offset; a reinforcement learning algorithm is adopted to simulate the attack angle offset, and the optimal attack angle of the air sail is obtained; according to the optimal attack angle and the environmental parameters, target adjustment parameters of the air sail are calculated; and driving the air sail to adjust the sail body rotation angle and the sail surface inclination angle in real time according to the target adjustment parameters. According to the method, the air sails always work in the efficient interval with the maximum lift coefficient and the minimum resistance coefficient, the pneumatic interference model is introduced, the total auxiliary thrust of the multiple sails is improved, the problem that the energy-saving effect is reduced due to dynamic deviation of the attack angle is effectively solved, and the situation that the fuel consumption of a main engine is reduced is stably achieved.

Description

Technical Field

[0001] This invention relates to the field of marine energy conservation technology, specifically to a method, system, and vessel for dynamic optimization of the energy efficiency of marine air sail-assisted navigation. Background Technology

[0002] With the global shipping industry's push towards carbon neutrality, ship energy efficiency improvement technologies have become a core direction for green shipping development. Air sail systems, as a low-cost, high-return energy-saving solution, capture wind energy to provide auxiliary thrust to ships, reducing main engine fuel consumption by 10%-30% and significantly reducing carbon emissions. In recent years, new air sail technologies such as rigid wing sails and rotary sails have been increasingly used on ocean-going vessels such as bulk carriers and container ships.

[0003] However, the navigation conditions of ships in the actual marine environment are far more complex and dynamic than ideal conditions. Ocean-going vessels are constantly exposed to the coupled effects of various marine environmental loads such as wind, waves, and currents, inevitably resulting in six degrees of freedom of hull motion, including roll, pitch, and bow. This complex and dynamic navigation environment places higher demands on the efficient and stable operation of airsail navigation systems. Traditional airsail control strategies, whether early manual operations or later automatic control systems, are mostly designed and parameter-tuned based on static or quasi-static conditions (such as assuming stable hull attitude, constant wind speed and direction, and uniform straight-line navigation). These strategies typically only consider the average or slow changes in environmental parameters such as wind speed and direction, and adjust the sail's angle of attack accordingly in hopes of obtaining ideal lift.

[0004] The core limitation of traditional control strategies is that the deviation of the aerosail angle of attack caused by the ship's motion is not quantified and compensated in real time, resulting in a decrease in energy efficiency and safety. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, and vessel for dynamic optimization of the energy efficiency of ship air sails for navigation, in order to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic optimization of the energy efficiency of ship air sails for navigation, comprising:

[0007] Real-time acquisition of dynamic parameters during ship navigation, including at least environmental parameters, ship attitude parameters, and airsail status parameters;

[0008] By processing the dynamic parameters, a dynamic coupling model is established;

[0009] The dynamic disturbance of the ship's attitude parameters to the air sail state parameters is quantified into an angle of attack offset using the coupling model.

[0010] A reinforcement learning algorithm is used to simulate the angle of attack offset, and the optimal angle of attack of the air sail under the current working conditions is obtained.

[0011] Based on the optimal angle of attack and the environmental parameters, the target adjustment parameters of the air sail are calculated through a preset geometric change model. The target adjustment parameters include at least the target angle of the sail body and the target tilt angle of the sail surface relative to the hull.

[0012] The target adjustment parameters are output to drive the air sail to adjust the sail rotation angle and sail tilt angle in real time according to the target adjustment parameters.

[0013] In a preferred embodiment, the real-time acquisition of dynamic parameters during ship navigation specifically includes:

[0014] The real-time acquisition of dynamic parameters during ship navigation specifically includes:

[0015] The environmental parameters include real-time wind speed, real-time wind direction, and sea state level;

[0016] The ship's attitude parameters include roll angle, pitch angle, roll rate, pitch rate, actual speed, and actual heading;

[0017] The airsail state parameters include the current angle of attack, the current sail rotation angle, real-time aerodynamic forces, and sail structural stress.

[0018] In a preferred embodiment, establishing the dynamic coupling model specifically refers to establishing a wind-wave-ship-sail multibody dynamic coupling model;

[0019] The coupling model also incorporates the ship's cargo status and the real-time power of the main engine as input variables;

[0020] The coupled model is trained using historical navigation data.

[0021] In a preferred embodiment, the step of using a reinforcement learning algorithm to simulate the angle-of-attack offset and derive the optimal angle of attack for the airsail under the current operating conditions specifically includes:

[0022] The reinforcement learning algorithm takes maximizing the aerodynamic efficiency of the air sail and satisfying the ship's navigation safety constraints as its dual objective functions, and uses the angle of attack offset as a disturbance term in the algorithm input.

[0023] The safety constraints include at least: critical angle of attack for airsail stall, maximum stress threshold for sail structure, hull stability threshold, and main engine load constraint.

[0024] The weights of the biobjective function are dynamically adjusted according to the sea state level.

[0025] In a preferred embodiment, the calculation of the target adjustment parameters of the airsail using a preset geometric change model specifically includes:

[0026] The geometric transformation model is constructed based on the mechanical structural characteristics of the air sail, including the transmission ratio of the sail rotation joints and the geometric constraints of the connection mechanism between the sail and the hull.

[0027] In a preferred embodiment, the method further includes a feedback correction step:

[0028] The actual angle of attack, actual aerodynamic force, and hull attitude parameters of the air sail after adjustment according to the target adjustment parameters are collected in real time and compared with the optimal angle of attack and safety constraint threshold.

[0029] This invention also provides a dynamic optimization system for the energy efficiency of ship air sails for navigation, comprising:

[0030] Multi-dimensional perception module: used to collect the dynamic parameters and transmit them to the data processing module;

[0031] Data processing module: used to filter and reduce noise of the collected dynamic parameters, and output clean data to the model building module;

[0032] Model building module: used to build and train a wind-wave-ship-sail multibody dynamics coupling model and quantify the angle of attack offset;

[0033] Algorithm solution module: Built-in reinforcement learning algorithm, used to input angle of attack offset and safety constraints, to solve for the optimal angle of attack;

[0034] Parameter calculation module: Built-in geometric change model and aerodynamic disturbance model, used to calculate target adjustment parameters;

[0035] Execution drive module: containing hydraulic actuators or electric servo mechanisms, used to receive target adjustment parameters and drive the air sail to adjust its attitude;

[0036] Feedback and early warning module: Used to compare actual parameters with thresholds and trigger correction, early warning or emergency actions.

[0037] In a preferred embodiment, a ship includes a central control system for implementing a dynamic optimization method for ship airsail-assisted navigation energy efficiency as described in any of the preceding embodiments.

[0038] In a preferred embodiment, the vessel is connected to a base, a motor is fixedly installed inside the base, a rotating main shaft is fixedly installed on the top of the motor, an air-pressurized mainsail is installed outside the rotating main shaft, a supporting inner cylinder is installed inside the air-pressurized mainsail, an upper transmission disk is rotatably installed on the top of the rotating main shaft, a lower transmission disk is rotatably installed on the bottom of the rotating main shaft, a control sub-shaft is rotatably installed inside the lower transmission disk, a tail fin auxiliary sail is installed outside the control sub-shaft, an upper connecting disk is rotatably installed on the top of the control sub-shaft, another control sub-shaft is rotatably connected inside the upper connecting disk, and the control sub-shaft located on the left side is rotatably installed inside the upper transmission disk.

[0039] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0040] This invention dynamically captures environmental, hull attitude, and airsail state parameters through a multi-dimensional real-time sensing module. It then quantifies the angle-of-attack shift caused by hull roll and pitch using a wind-wave-ship-sail multibody dynamics coupling model. Finally, a reinforcement learning algorithm is used to solve for the optimal angle of attack, ensuring the airsail always operates within the efficient range of maximum lift and minimum drag. Simultaneously, an aerodynamic interference model is introduced in the multi-sail configuration to increase the total auxiliary thrust of the sails, effectively avoiding the energy-saving effect reduction caused by dynamic angle-of-attack shifts in traditional methods, and stably achieving reduced main engine fuel consumption.

[0041] This invention incorporates multi-dimensional safety constraints through a reinforcement learning algorithm and dynamically adjusts the weights of the two objectives based on sea conditions. A feedback correction module compares actual parameters with thresholds in real time, immediately triggering parameter recalculation when deviations reach the safety warning line. Under abnormal operating conditions, it triggers audible and visual warnings and an emergency retraction procedure, effectively avoiding the risks of sail stall, structural damage, or capsizing, thus ensuring safe navigation. It enhances adaptability to complex operating conditions and covers navigation needs across multiple scenarios.

[0042] This invention incorporates the ship's cargo state and the real-time power of the main engine into a multibody dynamics coupling model, which can adapt to the angle of attack offset quantification requirements under different cargo loads and main engine operating states. Through a sea state classification adjustment mechanism, it can prioritize energy efficiency in calm sea states and prioritize safety in harsh sea states. Through a periodic self-calibration function, it can compensate for the decrease in accuracy caused by equipment aging and parameter drift, ensuring stable optimization results during long-term navigation. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0044] Figure 1This is a flowchart of the method of the present invention.

[0045] Figure 2 This is a system block diagram of the present invention.

[0046] Figure 3 This is a schematic diagram of the connection between the control sub-shaft and the lower transmission disk of the present invention.

[0047] Figure 4 This is a schematic diagram of the connection between the upper connecting plate and the control sub-shaft of the present invention.

[0048] Legend:

[0049] 1. Air-pressurized mainsail; 2. Tail fin auxiliary sail; 3. Upper connecting plate; 4. Base; 5. Rotating main shaft; 6. Control sub-shaft; 7. Motor; 8. Lower transmission plate; 9. Support inner cylinder; 10. Upper transmission plate. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1, please refer to Figure 1 As shown in this embodiment, a method for dynamic optimization of the energy efficiency of ship air sails for navigation includes:

[0052] S1. Real-time acquisition of dynamic parameters during ship navigation, including at least environmental parameters, ship attitude parameters, and airsail status parameters;

[0053] S2. By processing the dynamic parameters, a dynamic coupling model is established;

[0054] S3. Using the coupling model, the dynamic disturbance of the hull attitude parameters to the air sail state parameters is quantified as the angle of attack offset.

[0055] S4. Use reinforcement learning algorithm to simulate angle of attack offset and obtain the optimal angle of attack of air sail under the current working conditions;

[0056] S5. Based on the optimal angle of attack and the environmental parameters, calculate the target adjustment parameters of the air sail using a preset geometric change model. The target adjustment parameters include at least the target angle of the sail body and the target tilt angle of the sail surface relative to the hull.

[0057] S6. Output the target adjustment parameters to drive the air sail to adjust the sail rotation angle and sail tilt angle in real time according to the target adjustment parameters.

[0058] As described in S1-S6 above, the navigation conditions of ships in the actual marine environment are far more complex and dynamic than ideal conditions. Ocean-going vessels are constantly exposed to the coupled effects of various marine environmental loads such as wind, waves, and currents, inevitably resulting in six degrees of freedom of hull motion, including roll, pitch, and bow. This complex and dynamic navigation environment places higher demands on the efficient and stable operation of airsail navigation systems. Traditional airsail control strategies, whether early manual operations or later automatic control systems, are mostly designed and parameter-tuned based on static or quasi-static conditions (such as assuming stable hull attitude, constant wind speed and direction, and uniform straight-line navigation). These strategies typically only consider the average or slow changes in environmental parameters such as wind speed and direction, and adjust the sail's angle of attack accordingly to achieve ideal lift. The core limitation of traditional control strategies is that the airsail angle of attack deviation caused by hull motion is not quantified and compensated in real time, leading to decreased energy efficiency and safety.

[0059] This invention dynamically captures environmental, hull attitude, and airsail state parameters through a multi-dimensional real-time sensing module. It then quantifies the angle-of-attack shift caused by hull roll and pitch using a wind-wave-ship-sail multibody dynamics coupling model. Finally, a reinforcement learning algorithm is used to solve for the optimal angle of attack, ensuring the airsail always operates within the efficient range of maximum lift and minimum drag. Simultaneously, an aerodynamic interference model is introduced in the multi-sail configuration to increase the total auxiliary thrust of the sails, effectively avoiding the energy-saving effect reduction caused by dynamic angle-of-attack shifts in traditional methods, and stably achieving reduced main engine fuel consumption.

[0060] This invention incorporates multi-dimensional safety constraints through a reinforcement learning algorithm and dynamically adjusts the weights of the two objectives based on sea conditions. A feedback correction module compares actual parameters with thresholds in real time, immediately triggering parameter recalculation when deviations reach the safety warning line. Under abnormal operating conditions, it triggers audible and visual warnings and an emergency retraction procedure, effectively avoiding the risks of sail stall, structural damage, or capsizing, thus ensuring safe navigation. It enhances adaptability to complex operating conditions and covers navigation needs across multiple scenarios.

[0061] This invention incorporates the ship's cargo state and the real-time power of the main engine into a multibody dynamics coupling model, which can adapt to the angle of attack offset quantification requirements under different cargo loads and main engine operating states. Through a sea state classification adjustment mechanism, it can prioritize energy efficiency in calm sea states and prioritize safety in harsh sea states. Through a periodic self-calibration function, it can compensate for the decrease in accuracy caused by equipment aging and parameter drift, ensuring stable optimization results during long-term navigation.

[0062] In one embodiment, step S1, specifically including the real-time acquisition of dynamic parameters during ship navigation, includes:

[0063] S11. The environmental parameters include real-time wind speed, real-time wind direction, and sea state level. Real-time wind speed and direction are collected by a three-dimensional ultrasonic anemometer installed on the top of the ship's mast, and wave height data is collected by a wave height sensor on the bottom of the ship and matched with the sea state level.

[0064] (1) The wave height sensor is installed in the mid-section of the hull below the waterline of the ship. This position allows direct contact with the waves during navigation, capturing the vertical fluctuation amplitude of the waves in real time. The wave height sensor uses the ultrasonic ranging principle or the pressure sensing principle: when the waves act on the hull, the sensor emits ultrasonic waves and receives the reflected signals from the sea surface, calculates the vertical distance between the sensor and the instantaneous sea surface, and then combines it with the real-time draft of the ship to calculate the actual wave height data; or it senses the pressure change of the waves impacting the hull through the pressure sensor, and combines it with the fluid dynamics model to infer the wave height, ensuring that the wave height data acquisition error is ≤0.1m, meeting the accuracy requirements of sea state level matching.

[0065] (2) The real-time wave height data collected by the wave height sensor is transmitted to the data processing unit of the multi-dimensional sensing module via the ship's CAN bus, and then matched to the sea state according to the internationally accepted sea state classification standard (simplified version of the correspondence between Beaufort scale and sea state):

[0066] When the real-time wave height is ≤1.25m, it is determined to be sea state ≤3 (corresponding to calm to light wave conditions): under this condition, the waves have less interference with the roll and pitch of the ship, the ship's attitude is relatively stable, and the weight of maximizing the aerodynamic efficiency of the air sail in the subsequent reinforcement learning algorithm can be set to 70%-80%, and the critical angle of attack for air sail stall is executed according to the baseline value of 15°-20°.

[0067] When the real-time wave height is greater than 1.25m, it is determined to be sea state > 3 (corresponding to medium to very large waves): Under this condition, the impact of the waves on the hull is enhanced, the roll and pitch amplitudes increase, which can easily lead to a sharp increase in the airsail angle of attack deviation. At this time, the system automatically increases the weight of the ship's navigation safety constraints to 60%-70%, and at the same time lowers the critical angle of attack for airsail stall by 2°-3° to avoid the risk of stall caused by the angle of attack exceeding the critical value due to sea wave interference.

[0068] The core value of this matching logic lies in achieving dynamic adaptation between operating conditions and control strategies: Traditional airsail control strategies do not distinguish between sea state levels and use fixed angle of attack control thresholds, which can easily lead to decreased energy efficiency or safety risks when sea state changes abruptly; however, by accurately matching sea state levels through wave height sensors, subsequent model calculations (angle of attack offset correction), algorithm optimization (dual target weight adjustment), and safety constraints (stall angle adjustment) can be more in line with the actual navigation environment. For example, when the sea state suddenly increases from level 3 to level 4 (wave height increases from 1.2m to 1.5m), the system can complete the sea state level determination and weight adjustment within 100ms, ensuring that the airsail control strategy shifts from energy efficiency priority to safety priority in a timely manner, avoiding the risks caused by wave interference.

[0069] S12. The ship attitude parameters include roll angle, pitch angle, roll angular velocity, pitch angular velocity, actual ship speed, and actual heading. The roll angle, pitch angle, and corresponding angular velocity are collected by the shipborne inertial measurement unit (IMU), and the actual ship speed and heading are collected by the GPS navigation system.

[0070] S13. The air sail state parameters include the current angle of attack, the current sail rotation angle, the real-time aerodynamic force, and the sail structure stress. The current angle of attack and the sail rotation angle are collected by an angle sensor installed at the root of the sail, the real-time aerodynamic force is collected by a pressure sensor on the surface of the sail, and the sail structure stress is collected by a strain gauge attached to the sail structure.

[0071] S14. The acquisition frequency of the dynamic parameters shall not be less than 10Hz to ensure real-time response to changes in hull attitude and environment.

[0072] As described in S11-S14 above, by clearly defining the acquisition objects, hardware carriers, and frequency requirements of dynamic parameters, accurate and real-time data support is provided for the entire optimization method. On the one hand, specialized equipment such as 3D ultrasonic anemometers, IMUs, and pressure sensors ensure the accuracy of environmental, ship attitude, and airsail state parameter acquisition, avoiding errors from traditional manual methods or single sensors. On the other hand, an acquisition frequency of no less than 10Hz can capture ship attitude fluctuations and environmental changes in real time, laying the foundation for subsequent wind-wave-ship-sail model quantification of angle of attack offset and reinforcement learning to solve for the optimal angle of attack. At the same time, accurate matching of sea state levels can provide a basis for adjusting the weights of the dual objective function and adapting the safety constraint threshold, ensuring the achievement of the dual objectives of airsail energy efficiency optimization and navigation safety from the data source.

[0073] In one embodiment, step S2, establishing the dynamic coupling model specifically includes:

[0074] S21. Establish a multibody dynamic coupling model of wind-wave-ship-sail, and the model also introduces the ship's cargo state (draft, center of gravity position) and the real-time power of the main engine as input variables.

[0075] The wind-wave-ship-sail multibody dynamics coupling model is not a simple superposition of the four elements, but rather establishes the dynamic interaction relationship between them through physical equations. The specific connection logic is as follows:

[0076] The coupling of wind and sail: Wind is the power source of the air sail, and its speed and direction directly determine the initial aerodynamic force (lift and drag) of the air sail. The model calculates the pressure distribution on the sail surface under different wind speeds through fluid dynamics equations, and then obtains the magnitude of the initial aerodynamic force. At the same time, it converts the transient changes of wind (gusts and turbulence) into the fluctuation coefficient of aerodynamic force to simulate the dynamic influence of real wind conditions on the sail.

[0077] Wave-ship coupling: Ocean waves generate roll and pitch moments through the buoyancy difference on the ship. The model is based on ship hydrodynamics theory. It calculates the force of the waves on the ship based on wave height and wave frequency. Then, combined with the ship's own stability parameters, it obtains the real-time variation law of roll angle, pitch angle and corresponding angular velocity, avoiding the limitations of traditional models that assume uniform wave action.

[0078] Coupled between ship and sail: The roll and pitch of the ship will drive the synchronous movement of the air sail, resulting in the angle of attack of the sail relative to the actual incoming flow. The model uses rigid body kinematics equations to transform the attitude parameters of the ship into the spatial attitude changes of the sail, and then quantifies the angle of attack offset to achieve a precise mapping of ship dynamics → sail attitude → angle of attack change.

[0079] Traditional multihull models only consider the empty ship state and ignore the impact of cargo on attitude. This step introduces these two parameters, which are primarily aimed at solving the model adaptability problem under different cargo conditions. Their specific functions are as follows:

[0080] The impact of draft: Draft is determined by the ship's cargo capacity (the larger the cargo capacity, the deeper the draft), which directly affects the hull's immersion depth in the waves. The deeper the draft, the smaller the equivalent area of ​​the hull impacted by the waves, and the amplitude of roll and pitch will be reduced (for example, a 50,000-ton bulk carrier with a full load and a draft of 12m will have a roll angle that is 20%-30% lower than when it is unloaded (with a draft of 5m)). The model uses draft as an input parameter for the ship's hydrodynamic equations to correct the calculation results of wave forces, ensuring that the simulation of roll and pitch is more consistent with the actual cargo-carrying state.

[0081] The influence of the center of gravity position: The distribution of cargo will change the position of the ship's center of gravity (e.g., cargo concentrated at the bow will shift the center of gravity forward, while cargo concentrated at the side will shift the center of gravity). The position of the center of gravity directly determines the ship's roll and pitch characteristics—the higher the center of gravity, the worse the roll stability (the roll angle is more likely to increase); the more off-center the center of gravity, the more likely the ship will tilt during pitch. The model adjusts the ship's stability parameters by using the center of gravity coordinates (x, y, z axis coordinates relative to the ship's baseline) so that the calculated roll and pitch angles can match different cargo distribution scenarios, avoiding inaccurate quantification of angle of attack offset due to changes in the center of gravity.

[0082] The real-time power of the main engine reflects the current power demand of the ship. This parameter is introduced to achieve coordinated matching between the air sail auxiliary thrust and the main engine power.

[0083] When the main engine power is high in real time, the ship is in a state of high power demand. At this time, the air sail needs to provide greater auxiliary thrust. The model will correct the target aerodynamic coefficient of the air sail by the main engine power to ensure that the quantification of the angle of attack offset is more in line with the thrust demand.

[0084] When the main engine power is low in real time, the ship is in a state of low power demand. Excessive airsail thrust may cause fluctuations in the main engine load. The model will limit the maximum aerodynamic force of the airsail based on the main engine power to avoid the problem of excessive thrust after the angle of attack offset is corrected.

[0085] S22. The model is trained using historical navigation data (including data related to sea state, ship attitude parameters, airsail angle of attack, and fuel consumption rate) to ensure that the calculation error of the angle of attack offset does not exceed 3%.

[0086] The historical navigation data mentioned in the steps (including data related to sea state, ship attitude parameters, airsail angle of attack, and fuel consumption rate) must be a structured dataset that meets the requirements of time synchronization, scene coverage, and accuracy. Its specific composition is as follows:

[0087] Data dimensions and collection frequency: The data must include timestamps, sea state levels, ship attitude parameters, airsail angle of attack, and fuel consumption rate. All parameters must be collected synchronously at a frequency of 10Hz to ensure the correlation of multiple parameters at the same point in time.

[0088] Scene coverage requirements: To ensure the model's versatility under different operating conditions, historical data must cover at least three core scenarios:

[0089] ① Different sea state scenarios (sea states 2-6, covering calm to moderate wave conditions);

[0090] ② Different cargo loading scenarios (empty, half-loaded, fully loaded, corresponding to draft depths of 5-12m);

[0091] ③ For different navigation scenarios (tailwind and sea, headwind and sea, crosswind and sea), the effective data volume for each scenario should be no less than 1,000 sets to avoid data bias leading to poor model generalization ability.

[0092] The model is trained using supervised learning and iterative optimization. The core idea is to teach the model the mapping relationship between input parameters (wind, waves, ship, sails, cargo, main engine power) and output parameters (angle of attack offset). Specific steps include:

[0093] Data preprocessing: First, historical data is cleaned to remove outliers, and then missing data is supplemented by interpolation. Then, the data is normalized to avoid affecting the training effect due to differences in parameter magnitude.

[0094] Model training and parameter iteration: A random forest regression algorithm is used to construct a mapping model between input and output. Historical data is divided into a training set (70%) and a test set (30%) in a 7:3 ratio.

[0095] Training phase: Using sea state level, roll angle, pitch angle, draft, and main engine power as input features, and the actual measured angle of attack offset as the label, the model's decision tree number, node splitting threshold, and other parameters are adjusted by gradient descent to minimize the mean square error between the model's predicted angle of attack offset and the actual measured value.

[0096] Validation and optimization phase: Use the test set to validate the model accuracy. If the error exceeds 3%, increase the amount of training data or adjust the algorithm parameters. Repeat the iteration until the error is ≤3% to ensure that the model's prediction accuracy meets the standard.

[0097] Error control does not rely solely on algorithms, but is achieved through a triple guarantee of data accuracy, model structure, and verification mechanisms:

[0098] Data accuracy assurance: The accuracy of the acquisition equipment directly determines the data quality—for example, the wind speed measurement error of the 3D ultrasonic anemometer is ±0.1m / s, and the roll angle measurement error of the IMU is ±0.1°, ensuring that the original data error of the input model is ≤1%, reducing errors from the source;

[0099] Model structure optimization: Add an error compensation factor to the model. For example, if historical data shows that the model prediction error tends to increase when the roll angle is >10°, then add a special compensation equation in this range (such as error compensation factor = 0.95 + 0.005 × roll angle) to correct the prediction results.

[0100] Real-time verification mechanism: After the model training is completed, real-time verification is required in actual ship testing. The model is deployed to the ship system, and the angle of attack offset calculated by the model is compared with the actual angle of attack offset back-inferred by the sail pressure sensor. If the mean error of 100 consecutive sets of data is >3%, the historical data is called again for incremental training to ensure that the error is stable within 3% in the long term.

[0101] S23. The angle of attack offset is coupled with the roll angle, pitch angle, roll velocity, and pitch velocity, that is, for every 5° increase in roll angle or every 3° increase in pitch angle, the angle of attack offset increases by 1°-1.5°.

[0102] S231. The physical principle that the angle of attack offset is coupled with the roll angle, pitch angle, roll rate, and pitch rate is based on the physical logic of sail attitude change → relative change in the direction of the incoming flow → angle of attack offset, which is specifically broken down as follows:

[0103] The relationship between roll angle and angle of attack offset: The air sail is fixed on the hull deck. When the hull rolls to the right (roll angle increases), the sail tilts to the right in sync with the hull, causing a change in the spatial attitude of the sail surface. Assuming that in the initial state, the sail surface is perpendicular to the horizontal plane (angle of attack = 12°, ideal value), when the roll angle increases by 5°, the sail surface tilts to the right by 5°. At this time, the angle of the actual incoming flow (assuming the wind direction remains unchanged) relative to the tilted sail surface will increase by 1°-1.5°.

[0104] The greater the roll rate, the greater the acceleration of the sail tilt, and the faster the rate of change of the angle of attack deviation (e.g., when the roll rate is 2° / s, the rate of change of the angle of attack deviation is 1.2 times that when the roll rate is 1° / s), making it easier for the angle of attack to deviate from the optimal value.

[0105] The relationship between pitch angle and angle of attack offset: When the hull pitches (bow pitches up or down), the sail pitches back and forth with the hull, changing the vertical angle between the sail and the incoming current. When the pitch angle increases by 3° (bow pitches up), the front of the sail rises and the rear sinks, and the angle of the actual incoming current relative to the sail increases by 1°-1.5° (angle of attack offset = 1°-1.5°).

[0106] The greater the pitch rate, the more drastic the dynamic changes in the sail's pitch, and the larger the fluctuation range of the angle of attack offset (e.g., when the pitch rate = 1.5° / s, the fluctuation range of the angle of attack offset is 1.5 times that when the pitch rate = 0.5° / s), requiring the model to track and correct in real time.

[0107] S232. The numerical value that the angle of attack deviation increases by 1°-1.5° for every 5° increase in roll angle or 3° increase in pitch angle is based on statistical laws derived from actual ship tests and simulation verification. The specific derivation process is as follows:

[0108] Statistical analysis of actual ship test data: A 50,000-ton bulk carrier (equipped with rigid wing air sails, sail height 20m, sail width 5m) was selected. Under sea state 3 (wave height 1.0m) and full load (draft 12m) conditions, the roll angle was controlled to increase from 0° to 15° (in increments of 5°), and the pitch angle was controlled to increase from 0° to 9° (in increments of 3°). The angle of attack deviation was measured simultaneously.

[0109] Roll angle 0°→5°: angle of attack offset increases from 0° to 1.1°;

[0110] Roll angle 5°→10°: angle of attack offset increased from 1.1° to 2.3°;

[0111] Roll angle 10°→15°: angle of attack offset increased from 2.3° to 3.5°;

[0112] Pitch angle 0°→3°: angle of attack offset increases from 0° to 1.2°;

[0113] The pitch angle increased from 3° to 6°, and the angle of attack offset increased from 1.2° to 2.5°.

[0114] The pitch angle increased from 6° to 9°, and the angle of attack offset increased from 2.5° to 3.8°.

[0115] Statistical analysis revealed that for every 5° increase in roll angle and 3° increase in pitch angle, the increase in angle of attack deviation remained stable in the range of 1°-1.5°. This pattern showed a deviation of ≤0.2° under different sea states (levels 2-4), demonstrating its universality.

[0116] Simulation verification supplement: A virtual model of the hull-air sail was constructed using ship dynamics simulation software to simulate the changes in angle of attack under different roll and pitch angles. The simulation results showed a consistency of ≥95% with the actual ship test data, further verifying the accuracy of the numerical relationship.

[0117] S233, Correction rules for specific values ​​that are dynamically adjusted according to the ship's cargo status;

[0118] The cargo loading status affects the specific value of the angle of attack offset through two parameters: draft and center of gravity position. The correction rules are formulated based on the logic of cargo loading status → hull attitude characteristics → angle of attack offset sensitivity, as follows:

[0119] Correction based on draft: The greater the draft, the better the stability of the hull in waves, the lower the sensitivity of roll and pitch to angle of attack deviation, and the smaller the correction factor.

[0120] For drafts ≤ 6m (unloaded): for every 5° increase in roll angle, the angle of attack offset increases by 1.3°-1.5°; for every 3° increase in pitch angle, the angle of attack offset increases by 1.3°-1.5° (correction factor = 1.3).

[0121] For a draft of 6-10m (half load): for every 5° increase in roll angle, the angle of attack offset increases by 1.1°-1.3°; for every 3° increase in pitch angle, the angle of attack offset increases by 1.1°-1.3° (correction factor = 1.1).

[0122] For a draft ≥10m (full load): for every 5° increase in roll angle, the angle of attack offset increases by 1.0°-1.2°; for every 3° increase in pitch angle, the angle of attack offset increases by 1.0°-1.2° (correction factor = 1.0); for example: when fully loaded (draft 12m), a 5° increase in roll angle results in an angle of attack offset of 1.0°-1.2°; when unloaded (draft 5m), the offset increases by 1.3°-1.5° with the same roll increment, which conforms to the actual law of good stability and low offset sensitivity when fully loaded.

[0123] Correction based on center of gravity position: A shift in the center of gravity position will exacerbate attitude changes in a certain direction of the hull, requiring targeted adjustments to the offset value:

[0124] Center of gravity deflection to the left > 0.5m (relative to the hull centerline): The angle of attack shift when the hull rolls to the left increases by 0.2°-0.3° compared to when it rolls to the right;

[0125] When the center of gravity shifts forward by more than 1m (relative to the midpoint of the hull): the angle of attack offset during hull pitching (bow pitching up) increases by 0.2°-0.3° compared to when the bow pitches down; the corrected values ​​ensure that the model can accurately reflect the impact of attitude deviations caused by uneven cargo distribution on the angle of attack, and avoid the distortion of the angle of attack offset due to the shift of the center of gravity.

[0126] As described in S21-S23 above, by introducing the ship's cargo status and real-time main engine power into the wind-wave-ship-sail multibody dynamics coupling model, the limitations of traditional models that rely solely on the environment and ship attitude are overcome. This allows for precise adaptation to the angle-of-attack offset quantification requirements under different cargo loads and main engine operating conditions, avoiding model calculation deviations caused by changes in operating conditions. By training model parameters based on historical navigation data, the angle-of-attack offset calculation error is controlled within 3%, providing accurate data support for subsequent reinforcement learning algorithms to solve for the optimal angle of attack and avoiding the problem of reduced airsail energy efficiency caused by inaccurate quantification in traditional methods. By clarifying the coupling relationship between the angle-of-attack offset and the roll, pitch, and corresponding angular velocities, and combining this with dynamic correction values ​​based on the cargo status, the interference of ship attitude changes on the airsail angle of attack can be captured in real time and accurately. This provides a clear basis for dynamically adjusting airsail parameters, ensuring that the airsail always operates within a high-efficiency and safe range.

[0127] In one embodiment, step S4 involves using a reinforcement learning algorithm to simulate the angle-of-attack offset and derive the optimal angle of attack for the airsail under the current operating conditions. Specifically, this includes:

[0128] S41. The reinforcement learning algorithm takes maximizing the aerodynamic efficiency of the air sail and satisfying the ship's navigation safety constraints as its dual objective function, and uses the angle of attack offset as the interference term in the algorithm input.

[0129] S411. The reinforcement learning algorithm here adopts the Deep Deterministic Policy Gradient (DDPG) framework. Its core logic is to achieve dual-objective optimization through the interaction of agent-environment-reward function. The specific adaptation method is as follows:

[0130] Intelligent agent: The control and decision-making unit corresponding to the air sail. Its core output is the candidate angle of attack of the air sail. It needs to optimize aerodynamic efficiency while meeting safety constraints.

[0131] Environment: Simulates a real-time ship navigation scenario, including state parameters such as wind (wind speed, wind direction), waves (sea state level), ship attitude (roll angle, pitch angle), and air sail status. The angle of attack offset is integrated into the state space as a dynamic disturbance of the environment. That is, during each interaction, the environment will input the angle of attack offset generated by the wind-wave-ship coupling into the state matrix in real time, so that the agent can perceive the reference offset trend of the angle of attack under the current working condition and avoid decision-making from deviating from the actual dynamic disturbance.

[0132] State space: The dimensions include real-time wind speed, wind direction, sea state level, roll angle, pitch angle, angle of attack offset, real-time main engine power, and sail structural stress, totaling 8 dimensions, to ensure that the intelligent agent fully perceives the operating conditions.

[0133] Action space: The target angle of attack output is the air sail, which needs to be continuously adjusted within the safety constraints.

[0134] S412, Reward function implementation of the dual objective function;

[0135] Reinforcement learning cannot directly handle dual objectives. It is necessary to transform the maximization of aerodynamic efficiency and the satisfaction of safety constraints into a comprehensive reward function, and balance the dual objectives through positive rewards and negative penalties. The specific design is as follows:

[0136] Positive bonus for aerodynamic efficiency (R1): The core indicator is the lift-to-drag ratio of the air sail (L / D, lift coefficient L1 / drag coefficient D1), and the formula is: R1=k1×(L / D-(L / D)) min ) / ((L / D) max -(L / D) min );

[0137] Among them, (L / D) max This represents the optimal lift-to-drag ratio for airfoils (e.g., rigid sails are approximately 12-15), (L / D) min It is the minimum effective lift-to-drag ratio (e.g., 4; below this value there is no energy saving effect), and k1 is the efficiency bonus coefficient (valued at 1.0-1.5, with 1.5 for sea state ≤3, prioritizing efficiency).

[0138] Example: When the lift-to-drag ratio is 12 (optimal value), R1 = 1.5 × (12-4) / (15-4) = 1.09, the reward is the maximum; when the lift-to-drag ratio is 4 (minimum value), R1 = 0, there is no efficiency reward.

[0139] Negative penalty for safety constraints (R2): If the candidate angle of attack output by the agent triggers the safety constraint boundary, a penalty is applied, with the formula: R2 = -k2 × Σ(ΔC / ΔC) max );

[0140] Where ΔC is the deviation of a certain constraint parameter from the threshold (e.g., when the sail stress exceeds 80% of the yield strength, ΔC = stress value - 0.8 × yield strength; when the roll angle exceeds 15°, ΔC = roll angle - 15°), ΔC max This is the maximum permissible deviation of the parameter (e.g., stress ΔC). max = 0.2 × yield strength, roll angle ΔC max =5°), k2 is the safety penalty coefficient (value 2.0-3.0, 3.0 when sea state > 3, to strengthen safety constraints);

[0141] Example: If the sail stress = 0.9 × yield strength (0.1 × yield strength for exceeding 80% threshold), then R2 = -3.0 × (0.1 × yield strength / (0.2 × yield strength)) = -1.5, the penalty offsets the efficiency reward, forcing the agent to adjust the angle of attack.

[0142] The comprehensive reward function (R) is: R = R1 + R2. The agent learns to make the angle of attack decision that maximizes R by continuously interacting with the environment, thus achieving a balance between efficiency and safety.

[0143] S413, The specific input method for angle of attack offset as an interference term;

[0144] The angle of attack offset, as a dynamic perturbation parameter of the environmental state, affects the agent's perception and decision-making regarding the actual angle of attack. The specific logic is as follows:

[0145] The candidate angle of attack output by the agent is the ideal angle of attack (the target value without offset), while in actual navigation, the angle of attack offset caused by the ship's attitude will make the actual angle of attack = candidate angle of attack + offset.

[0146] The algorithm pre-defines the mapping relationship between angle of attack offset and hull attitude in the environment model, calculates the offset in real time based on the hull attitude parameters, and incorporates the actual angle of attack into the state space to feed back to the agent.

[0147] If the actual angle of attack touches the safety constraint due to the offset, the environment will report the state that the actual angle of attack exceeds the threshold to the agent, triggering the R2 penalty, which forces the agent to reduce the candidate angle of attack and avoid safety risks in actual operation.

[0148] S42. The safety constraints include at least: the critical angle of attack for airsail stall, the maximum stress threshold of the sail structure, the hull stability threshold, and the main engine load constraint.

[0149] S421. The critical angle of attack for stall of the airsail is determined based on the airfoil characteristics and engineering safety margin. The specific logic is as follows:

[0150] Airfoil characteristics basics: Marine air sails commonly use high lift-to-drag ratio airfoils such as NACA63-415. The stall critical angle of attack for this type of airfoil in wind tunnel testing is approximately 18°-22°.

[0151] Engineering safety margin: Considering the combined effects of gusts and angle of attack deviations during actual navigation, a safety margin of 2°-3° needs to be reserved. Therefore, the stall critical angle of attack is set to 15°-20°.

[0152] Adaptability to different airfoils: If a rotary sail is used (the critical stall angle of attack is about 20°-25°), the upper limit of the threshold can be adjusted to 25°, but the core logic is that the airfoil's limit stall angle is -2°-3° safety margin.

[0153] S422. The maximum stress threshold of the sail structure is an engineering design criterion based on material fatigue life and structural safety redundancy, as detailed below:

[0154] Basic material properties: The main sail of the air sail is made of carbon fiber composite material, with a yield strength of approximately 1200 MPa and a breaking strength of approximately 1500 MPa; if a steel inner support cylinder (Q345 steel) is used, the yield strength is approximately 345 MPa.

[0155] Fatigue life guarantee: Ship air sails need to withstand alternating wind loads for extended periods. If the stress consistently exceeds 90% of the yield strength, the material is prone to fatigue cracking (fatigue life shortened to less than 50% of the design life); while 80% yield strength is achieved through 10 7 The safety threshold verified by the alternating load test is that under this stress, the material fatigue life can meet the design service life of the ship of 15-20 years.

[0156] Emergency redundancy: A 20% stress margin is reserved to cope with stress peaks caused by extreme gusts and avoid instantaneous structural failure.

[0157] S423, the hull stability thresholds (roll angle not exceeding 15°, pitch angle not exceeding 8°) are formulated in accordance with the International Maritime Organization (IMO) Convention for the Safety of Life at Sea (SOLAS) and the Ship Stability Code, as follows:

[0158] The basis for a roll angle of 15°: The SOLAS Convention requires that the roll angle of a ship in sea state 5 should be ≤15° and the roll period should be ≥6s. If the roll angle exceeds 15°, the ship is prone to roll resonance, which can lead to waves on the deck, cargo displacement, or even loss of stability.

[0159] The basis for the 8° pitch angle: an excessive pitch angle will cause the propeller to spill water when the bow pitches up and the bow to hit the waves when the bow pitches down. The IMO’s “Structural Strength Specification” stipulates that the pitch angle of ocean-going cargo ships should be ≤8° to avoid the dynamic load caused by pitching exceeding the structural design limit.

[0160] Synergy with airsails: The auxiliary thrust of airsails changes the hull's rolling moment, thus incorporating the hull stability threshold into the constraints to ensure that airsail adjustments do not exceed the ship's inherent stability safety boundaries.

[0161] S43. The main engine load constraint is to prevent excessive airsail thrust from causing the main engine load to fluctuate beyond the acceptable range. The specific basis is as follows:

[0162] Main engine power range characteristics: Rated power (P) of marine main engines n The load characteristic curve shows that when the load is <40% P n When the main engine is in a low load range, its speed stability is poor. If the auxiliary thrust provided by the air sail increases suddenly, the ship's sailing resistance will drop sharply, the main engine load will drop instantaneously, triggering the main engine idling protection and causing it to shut down.

[0163] The test basis for the lift reduction ratio: Through actual ship testing, when the main engine power = 40%P n hour:

[0164] Air sail lift reduced by 30%; main engine load fluctuation range reduced from ±15% to ±8%;

[0165] Air sail lift reduced by 50%: Main engine load fluctuation range can be controlled within ±5%, completely avoiding the risk of engine shutdown;

[0166] Dynamic adaptability: The lift reduction ratio is not a fixed value, but is dynamically adjusted according to the real-time power of the host unit. Precise control is achieved through a mapping table of host power and lift reduction ratio.

[0167] S44. The weights of the dual objective function are dynamically adjusted according to the sea state level: when the sea state level is ≤3, the weight of maximizing aerodynamic efficiency accounts for 70%-80%; when the sea state level is >3, the weight of satisfying navigation safety constraints is increased to 60%-70%, and at this time the critical angle of attack for airsail stall is reduced by 2°-3°.

[0168] S441. Sea state directly determines the amplitude of hull attitude fluctuations and the risk of angle-of-attack deviation. Therefore, the core logic of weight adjustment is: the more stable the sea state (≤3), the smaller the angle-of-attack deviation and the lower the safety risk, with priority given to aerodynamic efficiency; the more severe the sea state (>3), the larger the attitude fluctuations and the easier it is for the angle of attack to exceed the safety threshold, with priority given to ensuring navigation safety. The specific correspondence is as follows:

[0169]

[0170]

[0171] S442. The weighting ratio is the optimal range obtained through energy efficiency-safety trade-off tests under multiple sea states. The specific testing process is as follows:

[0172] Test conditions: 50,000-ton bulk carrier, equipped with 2 rigid sails, sea state 2-5, record fuel saving rate and safety incident rate under different weights;

[0173] Test results when sea state ≤ 3:

[0174] Efficiency weighting 70%: fuel saving rate 18%, safety incident rate 0.5%;

[0175] Efficiency weighting 80%: fuel saving rate 20%, safety incident rate 1.2%;

[0176] Efficiency weight > 80%: Fuel saving rate only increases to 20.5%, but the safety incident rate rises sharply to 3.8%. Therefore, an efficiency weight of 70%-80% is determined to be the optimal range.

[0177] Test results when sea state > 3:

[0178] Safety weighting 60%: fuel saving rate 12%, safety incident rate 0.8%;

[0179] Safety weight 70%: fuel saving rate 10%, safety incident rate 0.1%;

[0180] Safety weight > 70%: Fuel saving rate drops to 8%, and safety benefits are not significantly improved. Therefore, a safety weight of 60%-70% is determined to be the optimal range.

[0181] S443. When sea state > 3, the critical stall angle of attack is lowered by 2°-3° to address the risk that the angle of attack may momentarily exceed the threshold in rough sea states. The specific logic is as follows:

[0182] Angle of attack fluctuation characteristics in sea state >3: When the wave height is >1.25m, the instantaneous peak values ​​of the ship's roll and pitch will increase the angle of attack deviation from ≤1° to 1°-3°, and gusts will further exacerbate the angle of attack fluctuation.

[0183] The safety buffer effect of lowering the threshold: Lowering the stall critical angle of attack from 15°-20° to 13°-17° can reserve a fluctuation buffer space of 2°-3° - even if the angle of attack deviates by a maximum of 3°, the actual angle of attack will only reach 17°+3°=20° (not exceeding the original threshold), thus avoiding stall;

[0184] Impact on energy efficiency control: After lowering the threshold, the upper limit of the candidate angle of attack is reduced from 20° to 17°. However, due to the high resistance of the ship itself when the sea state is bad, the small reduction in auxiliary thrust has only a 1%-2% impact on the overall fuel saving rate, which is far less than the benefit of reducing safety risks.

[0185] S444. The complete process of using reinforcement learning algorithms to solve for the optimal angle of attack;

[0186] Initialization: Load the environment model trained with historical navigation data, preset safety constraint thresholds, and initial values ​​for the weights of the two objectives (determined according to the current sea state level);

[0187] Status awareness: The multi-dimensional perception module collects real-time status data, calculates the angle of attack offset, and inputs it into the environment model;

[0188] Action decision: Based on the current state, the agent outputs candidate angles of attack and checks whether they meet safety constraints;

[0189] Reward calculation: Calculate the lift-to-drag ratio (e.g., 13) to get R1 = 1.2. If there is no safety constraint trigger, then R2 = 0. The overall reward is R = 1.2.

[0190] Strategy update: Adjust the algorithm strategy according to the reward R. If the actual angle of attack exceeds the lowered threshold, then R2 = -2.5 and R = -1.3, forcing the strategy to adjust the candidate angle of attack to 15°.

[0191] Iterative optimization: Repeat steps 2-5 every 100ms to continuously optimize candidate angles of attack, and finally output the optimal angle of attack (e.g., 15.5°) that maximizes the overall reward R, ensuring a balance between efficiency and safety.

[0192] As described in S41-S44 above, the dual objective function is to maximize aerodynamic efficiency and satisfy navigation safety constraints. This ensures that the airsail operates within the high-efficiency range of maximum lift and minimum drag, while mitigating risks through multi-dimensional safety constraints (stall angle, structural stress, hull stability, and main engine load), avoiding the drawbacks of traditional algorithms that prioritize efficiency over safety or safety over failure rate. The angle-of-attack offset is used as an input disturbance term, enabling the optimal angle-of-attack solution to respond in real-time to changes in angle of attack caused by hull roll and pitch, solving the problem of reduced energy-saving effect caused by neglecting dynamic offsets in traditional static algorithms, and ensuring the accuracy of angle-of-attack control under complex sea conditions. Safety constraints cover the airsail, sail structure, hull stability, and main engine operation, blocking safety risks at multiple levels and preventing sail stall, structural damage, or main engine failure. The weights of the dual objectives are dynamically adjusted according to sea state levels, and the stall critical angle under severe sea conditions is lowered, achieving adaptability to improve energy efficiency in calm sea conditions and ensure safety in severe sea conditions, thus enhancing the system's adaptability to complex scenarios.

[0193] In one embodiment, step S5, calculating the target adjustment parameters of the airsail using a preset geometric change model, specifically includes:

[0194] S51. The geometric transformation model is constructed based on the mechanical structural characteristics of the air sail, including the transmission ratio of the sail rotation joint and the geometric constraints of the connection mechanism between the sail surface and the hull.

[0195] S511, Sail Angle: It is necessary to offset the angle between the wind direction and the longitudinal axis of the hull. For example, when the wind direction is 30° to the longitudinal axis of the hull, the sail angle needs to be adjusted to 30° (so that the sail is facing the direction of the incoming flow). Then, it should be combined with the angle of attack offset correction (if the roll causes the angle of attack to shift by 1°, the angle should be adjusted by an additional 1°).

[0196] S512, Sail Inclination Angle: It is necessary to compensate for the vertical shift in angle of attack caused by the ship's roll. For example, if the ship rolls 5° to the right, the sail needs to tilt 5° to the left (to counteract the effect of the roll on the horizontal attitude of the sail) to ensure that the angle between the sail chord and the actual incoming flow is equal to the optimal angle of attack.

[0197] S52. If the ship is equipped with two or more air sails, the model also introduces the aerodynamic interference model of each sail: by calculating the wake effect of the front sail on the rear sail (the wake wind speed attenuation coefficient is 0.6-0.8, and the wind direction deflection angle is 3°-5°), the target adjustment parameters of the rear sail are compensated and corrected - that is, the target turning angle of the rear sail is increased by 5°-8° compared with the front sail, and the sail tilt angle is decreased by 2°-3° compared with the front sail, to ensure that the total auxiliary thrust of multiple sails is increased by 5%-10%, and to avoid the additional rolling moment of the hull caused by the mutual interference of the aerodynamic forces of multiple sails exceeding 5kN·m;

[0198] S521. The interference of the fore sail on the aft sail originates from the wake region formed by the separation of airflow over the sail surface. Relevant parameters were obtained through wind tunnel testing and actual ship testing.

[0199] S522. Wake area range: The wake area of ​​the front sail is fan-shaped, with a length along the direction of the incoming flow that is 1.5-2 times the sail height and a width that is 2-3 times the sail width.

[0200] S523, Wind speed attenuation coefficient: In the core region of the wake (within 1-1.5 times the sail height from the front sail), the wind speed attenuates to 0.6-0.7 times the wind speed of the incoming wind from the front sail (because the kinetic energy of the airflow is absorbed by the front sail); at the edge of the wake (within 1.5-2 times the sail height), the attenuation coefficient increases to 0.7-0.8 times (as the airflow gradually recovers);

[0201] S524, Wind deflection angle: The wingtip vortex of the fore sail causes the airflow in the wake region to deflect to both sides of the sail. The deflection angle is positively correlated with the angle of attack of the fore sail - when the angle of attack is 10°-15° (optimal range), the wind deflection angle is stable at 3°-5° (the larger the angle of attack, the stronger the vortex, and the larger the deflection angle).

[0202] S53. The model calculates the influence of the fore sail on the aft sail by determining the wake region position and assigning interference coefficients:

[0203] S531. First, based on the installation position of the multiple sails, determine whether the rear sail is in the wake area of ​​the front sail.

[0204] S532. Assign attenuation coefficients according to the position of the rear sail in the wake region;

[0205] Calculate the effective incoming flow parameters for the rear sail: Effective wind speed = actual wind speed × attenuation coefficient (e.g., if the actual wind speed is 10 m / s, the effective wind speed = 10 × 0.65 = 6.5 m / s), Effective wind direction = actual wind direction + deflection angle (e.g., if the actual wind direction is 30°, the effective wind direction = 30° + 4° = 34°).

[0206] S54, the compensation amount of increasing the rear sail angle by 5°-8° and decreasing the tilt angle by 2°-3°, is the result of reverse derivation based on the effective incoming flow parameters. The core objective is to ensure that the actual lift of the rear sail is consistent with that when there is no interference, as detailed below:

[0207] S541, Derivation of increasing the target rotation angle of the sail by 5°-8°;

[0208] The lift formula for an air sail is: L = 0.5ρv 2 SCL;

[0209] Where p is the air density, v is the incoming wind speed, S is the sail area, and CL is the lift coefficient.

[0210] Under undisturbed conditions: Assuming the actual wind speed v1 = 10 m / s, the sail rotation angle is 30° (corresponding to an angle of attack of 12°, CL1 = 1.2), the lift L1 = 0.5 × 1.225 × 10^2 × S × 1.2 = 73.5 S;

[0211] When disturbed: effective wind speed v2=10×0.65=6.5m / s, if the lift L2=L1=73.5S is maintained, then CL2 needs to be lifted: 73.5S=0.5×1.225×6.5^2×S×CL2, solving for CL2≈2.8;

[0212] The relationship between lift coefficient and angle of attack: According to the CL-a (lift coefficient - angle of attack) curve of the sail airfoil, the angle of attack corresponding to CL2 = 2.8 needs to be increased from 12° to 17°-19°. The increase of 5°-7° in the angle of attack needs to be achieved by increasing the sail rotation angle by 5°-8° (the mapping relationship between rotation angle and angle of attack is determined by the sail installation angle. Usually, the change in rotation angle is slightly greater than the change in angle of attack to compensate for mechanical clearance).

[0213] S542. The front sail causes the effective wind direction of the rear sail to deflect by 4° (e.g., actual wind direction 30° → effective wind direction 34°). If the rear sail still maintains its original tilt angle (e.g., 0°), the actual angle of attack will shift by 4° due to the wind direction deflection, exceeding the optimal angle of attack range.

[0214] To counteract a 4° wind deflection, the sail tilt angle needs to be adjusted by 2°-3° in the opposite direction of the wind deflection (e.g., if the wind deflects 4° to the right, the sail should be tilted 2°-3° to the left). By correcting the sail attitude, the sail chord is re-aligned with the effective wind direction, ensuring that the actual angle of attack returns to the optimal range.

[0215] The tilt angle only needs to be reduced by 2°-3° (instead of 4°) because when the sail angle is increased by 5°-8°, the effect of wind deflection is partially offset, and the two work together to achieve precise control of the angle of attack.

[0216] S543. Validation of thrust enhancement effect: Through a comparative test on a 50,000-ton bulk carrier—with a single sail operating, the auxiliary thrust is approximately 80kN. With two sails (bow and hull) and without aerodynamic interference compensation, the total thrust is approximately 145kN (the thrust does not reach 80kN due to interference with the aft sail). After compensation is enabled, the aft sail thrust recovers to 75kN, and the total thrust reaches 155kN, an increase of 6.9% compared to without compensation, consistent with the design target of 5%-10%. If four sails are configured and compensation is coordinated, the total thrust can be increased by 9%-10%.

[0217] Safety basis for roll moment limitation: The safe threshold for roll moment of a 50,000-ton bulk carrier is calculated based on hull stability (approximately 20 kN·m). 5 kN·m is only 25% of the safe threshold, leaving sufficient safety redundancy.

[0218] The model ensures that the roll moment is controllable by calculating the aerodynamic distribution of multiple sails: for example, if the thrust of the port sail is 75kN and the thrust of the starboard sail is 75kN, and the lever arm of both sails is 10m, the roll moment = (75-75)×10 = 0kN·m; if the thrust of a single sail exceeds the tolerance (e.g., 80kN on the port sail and 70kN on the starboard sail), the roll moment = (80-70)×10 = 10kN·m. At this time, the model will automatically adjust the tilt angle of the single sail (reduce the tilt angle of the port sail by 1°, reduce the thrust by 5kN) so that the moment returns to within 5kN·m.

[0219] As described in S51-S54 above, the model is built based on the mechanical structural characteristics of the air sail, ensuring that the target adjustment parameters do not exceed the mechanical physical limits and avoiding structural overload damage. Simultaneously, by controlling the sail adjustment precision through a fixed transmission ratio, it matches the fine-tuning requirements of the optimal angle of attack, preventing the angle of attack from deviating from the efficient range due to parameter deviations. When using a multi-sail configuration, an aerodynamic interference model is introduced to compensate for the wake effect of the forward sail, offsetting the adverse effects of wind speed attenuation and wind direction deflection on the rear sail, thereby increasing the total auxiliary thrust of the multi-sails by 5%-10%. At the same time, the additional roll moment of the hull is controlled to ≤5kN·m to avoid hull stability problems caused by mutual aerodynamic interference between the multi-sails, balancing multi-sail energy efficiency and navigation safety.

[0220] In one embodiment, the method further includes a feedback correction step:

[0221] The actual angle of attack, actual aerodynamic force, and hull attitude parameters of the air sail after adjustment according to the target adjustment parameters are collected in real time and compared with the optimal angle of attack and safety constraint threshold.

[0222] If the deviation between the actual angle of attack and the optimal angle of attack exceeds 5%, or the stress on the sail structure exceeds 80% of the maximum stress threshold, or the roll angle reaches 90% of the stability threshold, the reinforcement learning algorithm is triggered to resolve for the optimal angle of attack and update the target adjustment parameters.

[0223] When the stress of the sail structure exceeds 80% of the maximum stress threshold for three consecutive times, or the roll angle reaches 90% of the stability threshold, an audible and visual warning is triggered; if the parameters continue to exceed the limit for 10 seconds, an emergency command is output to drive the air sail to perform an emergency retraction procedure (adjusting the sail rotation angle to 0° and the sail tilt angle to a horizontal state).

[0224] Example 2, please refer to Figure 2 As shown in this embodiment, a dynamic optimization system for ship air sail navigation efficiency includes:

[0225] Multi-dimensional sensing module: includes a three-dimensional ultrasonic anemometer, wave height sensor, IMU, GPS navigation system, angle sensor, pressure sensor, and strain gauge, used to collect the dynamic parameters and transmit them to the data processing module;

[0226] Data processing module: Electrically connected to the multi-dimensional perception module, used to filter and reduce noise of the collected dynamic parameters, and output clean data to the model building module;

[0227] Model building module: electrically connected to the data processing module, used to build and train the wind-wave-ship-sail multibody dynamics coupling model and quantify the angle of attack offset;

[0228] Algorithm solution module: Electrically connected to the model building module, it has a built-in reinforcement learning algorithm, which is used to input the angle of attack offset and safety constraints to solve for the optimal angle of attack;

[0229] Parameter calculation module: electrically connected to the algorithm solving module and the multi-dimensional sensing module, with built-in geometric change model and aerodynamic disturbance model, used to calculate target adjustment parameters;

[0230] Execution drive module: electrically connected to the parameter calculation module, including hydraulic actuator or electric servo mechanism, used to receive target adjustment parameters and drive the air sail to adjust its attitude;

[0231] Feedback and early warning module: Electrically connected to the multi-dimensional perception module, algorithm solving module, and execution driving module, used to compare actual parameters with thresholds and trigger correction, early warning, or emergency actions;

[0232] The modules communicate with each other via the ship's CAN bus, with a response delay of no more than 200ms.

[0233] Example 3, please refer to Figure 3-4 As shown, the vessel is connected to a base, a motor is fixed inside the base, a rotating main shaft is fixed on the top of the motor, an air-pressurized mainsail is provided outside the rotating main shaft, a supporting inner cylinder is provided inside the air-pressurized mainsail, an upper transmission disk is rotatably provided on the top of the rotating main shaft, a lower transmission disk is rotatably provided on the bottom of the rotating main shaft, a control sub-shaft is rotatably provided inside the lower transmission disk, a tail fin auxiliary sail is provided outside the control sub-shaft, an upper connecting disk is rotatably provided on the top of the control sub-shaft, another control sub-shaft is rotatably connected inside the upper connecting disk, and the control sub-shaft located on the left side is rotatably provided inside the upper transmission disk;

[0234] The air-pressurized mainsail is made of composite material and has the following functions: first, it generates Magnus effect thrust through the wind acting on the surface of the sail, providing auxiliary propulsion for the ship; second, it introduces airflow through the air inlet on its side, and the airflow flows through the air passage inside the sail to the tail of the mainsail, providing the power source for the tail fin auxiliary sail to adjust the wind direction.

[0235] The tail fin auxiliary sail is made of composite material and is used to adjust the wind direction of natural wind and airflow from the main sail, thereby creating a relative thrust that is beneficial to the ship's navigation.

[0236] The upper connecting plate provides stable support for the tail fin auxiliary sail and allows the tail fin auxiliary sail to rotate freely under the drive of the lower transmission plate;

[0237] The base provides support and fixation, and also has an air sail deployment and retraction function;

[0238] The rotating spindle is used to control the motion state of the air-pressurized mainsail to adjust its working angle.

[0239] The control sub-shaft is used to control the motion state of the tail fin auxiliary sail;

[0240] The motor is used on one hand to drive the rotating main shaft to rotate the air-pressurized main sail to adjust the working angle; on the other hand, it is connected to the lower transmission disk through the lower transmission mechanism to drive the lower transmission disk to drive the tail fin auxiliary sail to swing left and right to adjust the wind direction.

[0241] The lower transmission disc is used to adjust the working state of the tail fin auxiliary sail through a transmission mechanism.

[0242] The inner support cylinder is made of steel and serves as a structural support for the entire air sail assembly.

[0243] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamic optimization of energy efficiency of a ship assisted by air sails, characterized in that, The method comprises the following steps: Real-time acquisition of dynamic parameters during ship navigation, wherein the dynamic parameters at least include environmental parameters, ship body attitude parameters, and air sail state parameters; Establishment of a dynamic coupling model by processing the dynamic parameters; Quantification of the dynamic interference of the ship body attitude parameters on the air sail state parameters into an angle of attack offset by using the coupling model; Simulation of the angle of attack offset by using a reinforcement learning algorithm to obtain an optimal angle of attack of the air sail under the current working condition; Calculation of target adjustment parameters of the air sail by a preset geometric change model according to the optimal angle of attack and the environmental parameters, wherein the target adjustment parameters at least include a target turning angle of a sail body and a target inclination angle of a sail surface relative to the ship body; Output of the target adjustment parameters to drive the air sail to adjust the turning angle of the sail body and the inclination angle of the sail surface in real time according to the target adjustment parameters.

2. The method of claim 1, wherein: The real-time acquisition of dynamic parameters during ship navigation specifically comprises: The environmental parameters include real-time wind speed, real-time wind direction, and sea state level; The ship body attitude parameters include roll angle, pitch angle, roll angle velocity, pitch angle velocity, actual ship speed, and actual heading; The air sail state parameters include current angle of attack, current sail body turning angle, real-time aerodynamic force, and sail body structure stress.

3. The method of claim 1, wherein: The dynamic coupling model is a wind-wave-ship-sail multi-body dynamic coupling model, The coupling model further introduces the ship cargo state and the real-time power of the main engine as input variables; The coupling model is trained by historical navigation data.

4. The method of claim 1, wherein: The simulation of the angle of attack offset by using the reinforcement learning algorithm to obtain the optimal angle of attack of the air sail under the current working condition specifically comprises: The reinforcement learning algorithm takes maximizing the air sail aerodynamic efficiency and satisfying the ship navigation safety constraint as a double-objective function, and takes the angle of attack offset as an interference term input into the algorithm; The safety constraint at least includes the air sail stall critical angle of attack, the sail body structure maximum stress threshold, the ship body stability threshold, and the ship main engine load constraint; The weights of the double-objective function are dynamically adjusted according to the sea state level.

5. The method of claim 1, wherein: The calculation of the target adjustment parameters of the air sail by the preset geometric change model specifically comprises: The geometric change model is constructed based on the mechanical structure characteristics of the air sail, and includes the transmission ratio of the sail body turning joint and the geometric constraint of the sail surface and the ship body connecting mechanism.

6. The method of claim 4, wherein: The method further comprises a feedback correction step: Real-time acquisition of the actual angle of attack, actual aerodynamic force, and ship body attitude parameters of the air sail after adjustment according to the target adjustment parameters, and comparison with the optimal angle of attack and the safety constraint threshold.

7. A system for dynamic optimization of energy efficiency of a marine air sail-assisted vessel for implementing a method for dynamic optimization of energy efficiency of a marine air sail-assisted vessel according to any one of claims 1-6, characterized in that, The method comprises the following steps: A multi-dimensional perception module for acquiring the dynamic parameters and transmitting them to a data processing module; A data processing module for filtering and noise reduction processing of the acquired dynamic parameters, and outputting clean data to a model construction module; A model construction module for establishing and training a wind-wave-ship-sail multi-body dynamic coupling model to quantify the angle of attack offset; An algorithm solving module built-in reinforcement learning algorithm for inputting the angle of attack offset and safety constraint to solve the optimal angle of attack; A parameter calculation module built-in geometric change model and aerodynamic interference model for calculating target adjustment parameters; Execution driving module: contains hydraulic actuator or electric servo mechanism, used to receive target adjustment parameters and drive air sail to adjust posture; Feedback warning module: used to compare actual parameters with threshold value, trigger correction, warning or emergency action.

8. A ship comprising a central control system for implementing the method of dynamic optimization of the energy efficiency of a ship air sail according to any one of claims 1-6.

9. The vessel of claim 8, wherein, A base is connected in the ship, a motor is fixedly arranged in the base, a rotating main shaft is fixedly arranged on the top of the motor, an air pressure main sail is arranged outside the rotating main shaft, a supporting inner cylinder is arranged in the air pressure main sail, an upper transmission disc is rotatably arranged on the top of the rotating main shaft, a lower transmission disc is rotatably arranged on the bottom of the rotating main shaft, a control secondary shaft is rotatably arranged in the lower transmission disc, a tail wing auxiliary sail is arranged outside the control secondary shaft, an upper connecting disc is rotatably arranged on the top of the control secondary shaft, another control secondary shaft is rotatably connected in the upper connecting disc, and the control secondary shaft on the left side is rotatably arranged in the upper transmission disc.

Images