System and method for training an ai model for predicting performance of a wind propulsion device

Abstract

A method for training an artificial intelligence model for predicting performance of at least one wind propulsion device (202) on a ship (204) is disclosed. The method comprises configuring a processor (210) to: receive force measurement data indicative of a force generated by the at least one wind propulsion device; receive data of at least one operating parameter indicative of an operating condition associated with at least one of the at least one wind propulsion device and the ship; and use the force measurement data and the data of the at least one operating parameter to train the artificial intelligence model to establish a predictive relationship between performance of the at least one wind propulsion device and different control inputs.

Description

Technical Field

[0001] This disclosure relates to a method for training an artificial intelligence (AI) model for predicting the performance of a wind propulsion device. This disclosure also relates to a system for training an AI model for predicting the performance of a wind propulsion device. Background Technology

[0002] In recent years, the shipping industry has made significant progress in utilizing wind energy to propel vessels such as tankers, cargo ships, passenger ships, and small boats. Wind propulsion devices, such as Magnus rotors or airfoil sails, are increasingly being used to optimize fuel efficiency and supplement traditional propulsion systems, such as ship engines. Typically, wind propulsion devices are mounted vertically on the vessel and generate lift, which is used as the propulsion force to move the vessel in the water. However, it is crucial to accurately predict the performance under different wind directions, speeds, and other conditions to ensure that wind propulsion devices can operate efficiently in diverse marine environments.

[0003] Some existing techniques for predicting the performance of wind turbines are extremely unreliable and error-prone. This is because such techniques rely heavily on physics-based models, such as computational fluid dynamics (CFD) simulations, finite element analysis (FEA), and wind tunnel experiments. These physics-based models simulate the behavior of wind turbines under various conditions (i.e., how they would behave in real-world scenarios) using input parameters such as wind speed, ship characteristics, and environmental conditions. However, these physics-based models typically provide only minimally accurate and realistic results and may require human intervention to assess the impact of input parameters on the predicted performance of wind turbines. Furthermore, some existing techniques combine measurement biases with physics-based models to predict wind turbine performance. However, the overall accuracy and reliability of such predictions are very low.

[0004] Therefore, based on the foregoing discussion, there is a need to overcome the aforementioned shortcomings. Summary of the Invention

[0005] This disclosure seeks to provide a method and system for realistically, accurately, and reliably predicting the performance of wind propulsion devices in an energy-efficient and time-saving manner. The object of this disclosure is achieved by a method and system for training an artificial intelligence (AI) model for predicting the performance of wind propulsion devices, as defined in the appended independent claims, which are to be referenced. Advantageous features are set forth in the appended dependent claims.

[0006] Throughout the description and claims of this specification, the words “comprising,” “including,” “having,” “containing,” and variations thereof such as “containing” and “having” mean “including, but not limited to,” and do not exclude the presence of other components, items, integers, or steps not expressly disclosed. Furthermore, the singular includes the plural unless the context requires otherwise. In particular, where the indefinite article is used, the specification should be understood to consider both the plural and the singular unless the context requires otherwise. Attached Figure Description

[0007] Figure 1 The steps of a method for training an artificial intelligence model for predicting the performance of at least one wind propulsion device on a ship, according to embodiments of the present disclosure, are illustrated; and Figure 2 An exemplary scenario according to an embodiment of the present disclosure is shown, in which a system is employed for training an artificial intelligence model to predict the performance of at least one wind propulsion device on a ship. Detailed Implementation

[0008] The following detailed description illustrates embodiments of the present disclosure and ways in which these embodiments may be implemented. While some modes of implementing the present disclosure have been disclosed, those skilled in the art will recognize that other embodiments for implementing or practicing the present disclosure are also possible.

[0009] In a first aspect, embodiments of this disclosure provide a method for training an artificial intelligence model to predict the performance of at least one wind propulsion device on a ship, wherein the method includes: The processor receives force measurement data, which indicates the force generated by at least one wind-powered propulsion device. The processor receives data of at least one operating parameter, the data of which indicates operating conditions associated with at least one of the at least one wind propulsion device and the vessel; and The processor uses force measurement data and data from at least one operating parameter to train an artificial intelligence model to establish predictive relationships between the performance of at least one wind propulsion device and different control inputs.

[0010] In a second aspect, embodiments of this disclosure provide a system for training an artificial intelligence model to predict the performance of at least one wind propulsion device on a ship, the system comprising: One or more first sensors, the one or more first sensors being used to determine force measurement data, the force measurement data indicating the force generated by at least one wind propulsion device; One or more second sensors, used to determine data for at least one operating parameter, the data of which indicates operating conditions associated with at least one wind propulsion device and at least one vessel; and A processor, operatively coupled to one or more first sensors and one or more second sensors, wherein the processor is configured to: Receive force measurement data; Receive data for at least one operation parameter; and The artificial intelligence model is trained using force measurement data and data from at least one operating parameter to establish predictive relationships between the performance of at least one wind propulsion device and different control inputs.

[0011] This disclosure relates to the aforementioned methods and systems for realistically, accurately, and reliably predicting the performance of at least one wind propulsion device in an energy-efficient and time-saving manner, applicable with minimal modifications from ship to ship. Herein, the methods and systems utilize AI models to provide accurate performance predictions for various control inputs within the control of the wind propulsion device. Furthermore, the methods and systems contribute to informed decision-making for ship maneuvering, support optimal ship power management choices, and contribute to precise route optimization based on predictive relationships. Advantageously, by using AI models, the methods and systems readily improve the overall fuel efficiency of the ship and maximize the energy efficiency of at least one wind propulsion device, thereby intelligently managing available electricity on board (from, for example, fuel sources, batteries, renewable energy). The aforementioned methods and systems enable wind propulsion devices to utilize wind power to propel ships (moving on water along a desired route) with minimal energy consumption, achieving lower overall fuel consumption and cost savings. Therefore, the aforementioned methods and systems contribute to reducing greenhouse gas emissions and promoting more sustainable and eco-friendly navigation practices. Furthermore, when the AI ​​model is trained using a comprehensive dataset (as discussed above), it ensures accurate predictions of the performance of at least one wind propulsion device under various wind and operating conditions. The aforementioned methods and systems are simple, robust, fast, reliable, and easily implemented.

[0012] This method enables the training of an artificial intelligence (AI) model by providing force measurement data and data of at least one operating parameter as input (i.e., training data). During training, the AI ​​model is used to predict the performance of at least one wind propulsion device. It should be understood that the AI ​​model can be designed to predict and evaluate how effectively at least one wind propulsion device can operate on a ship under various conditions, such as atmospheric pressure, wind speed around the ship, ship speed, wind direction, ship direction of movement, etc.

[0013] It is worth noting that the processor controls the overall operation of the system. Optionally, the processor can be communicatively coupled to at least one or more first sensors and one or more second sensors. Optionally, the processor is implemented as a processor of a computing device. Examples of computing devices include, but are not limited to, laptops, desktops, tablets, phablets, personal digital assistants, workstations, and consoles. Alternatively, the processor can be implemented as a cloud server (i.e., a remote server) that provides cloud computing services.

[0014] Throughout this disclosure, the term "ship" refers to a vessel (i.e., a ship) used for navigating on a body of water (e.g., the ocean, the sea, etc.). Ships are typically used for purposes such as transporting cargo and passengers, marine exploration, etc. Furthermore, they are propelled by engines or turbines.

[0015] Throughout this disclosure, the term "wind propulsion device" refers to a device capable of using wind power to propel a vessel across a body of water. Typically, in operation, at least one wind propulsion device captures the kinetic energy of the wind surrounding the vessel, thereby generating aerodynamic or aerostatic forces that propel the vessel forward. This is because when wind flows over the surface of at least one wind propulsion device, a pressure difference is created around it. This pressure difference generates lift or thrust, which is transmitted to the vessel, thus propelling it in the direction of the prevailing wind. It should be understood that wind propulsion devices generally help reduce a vessel's fuel consumption, reduce emissions from the vessel, and minimize the vessel's operating costs, for example, by assisting the vessel's main propulsion system (e.g., an engine or turbine) in reducing fuel consumption, reducing emissions from the vessel, and minimizing operating costs. However, at least one wind propulsion device may also be used as the sole propulsion device for a vessel. Examples of at least one wind propulsion device include, but are not limited to, Magnus rotors (e.g., Flettner rotors) and airfoil sails (e.g., wing sails). For example, airfoil sails can be wing sails, rigid sails, suction sails (also known as turbine sails), Bermuda sails, etc.

[0016] Throughout this disclosure, the term "performance" refers to a relative assessment of at least one wind propulsion device based on the available wind power (wind power available due to airflow) and the wind power actually utilized and converted into mechanical power to propel a ship over water. For example, the performance of at least one wind propulsion device can be high when a major portion of the available wind power is utilized, and low when the available wind power is not effectively utilized. It is understood that the performance of at least one wind propulsion device is predicted in terms of energy saving and time efficiency.

[0017] Throughout this disclosure, the term "control input" refers at least to force measurement data and data for at least one operating parameter. Optionally, the control input can be various inputs related to the operational adjustment of at least one wind propulsion device and its components (i.e., one or more wind-catching components and power transmission mechanisms). Optionally, different control inputs refer to at least one of the following: rotor torque control, rotor braking control, etc. Different control inputs are adjusted to maximize the utilization of wind power. For example, the rotor's rotational speed can be adjusted to utilize more wind power and / or minimize damage under extreme weather conditions (e.g., strong winds, tornadoes, cyclones, etc.). In another example, yaw (referring to the housing of the wind-catching component) can be adjusted to align at least one wind propulsion device with the wind direction to maximize wind power utilization. In yet another example, rotor torque can be controlled to efficiently convert wind power into mechanical power.

[0018] Throughout this disclosure, the term "force measurement data" refers to information relating to the forces exerted by wind on at least one wind-powered propulsion device when the vessel is in operation (i.e., moving in a body of water). In this regard, force measurement data (as discussed below) is obtained, for example, from sensors arranged on at least one wind-powered propulsion device. Force measurement data may include at least one of the magnitude and direction of the force exerted by the wind. The magnitude of the force exerted by the wind may be a value representing the thrust or drag generated by at least one wind-powered propulsion device under various wind conditions. In this document, the term "wind conditions" refers to the state of the wind (i.e., velocity, direction, and / or turbulence) at a given location and time. It should be understood that the force measurement data is received in real time or near real time (by a processor).

[0019] Optionally, one or more first sensors are used to determine the force measurement data, wherein the force measurement data includes at least one of the following: Force measurement data based on pressure, Force measurement data based on strain, Force measurement data based on displacement.

[0020] The term "first sensor" refers to a device capable of detecting and / or measuring the force exerted by wind on at least one wind propulsion device (one or more surfaces). Optionally, at least some of the one or more first sensors are arranged on at least one wind propulsion device. It should be understood that some first sensors may be arranged on at least one wind propulsion device, while the remaining first sensors may be arranged on other different parts of the vessel. Different types of one or more first sensors may be used to measure different types of forces (e.g., pressure-based forces, strain-based forces, displacement-based forces). In this respect, different types of one or more first sensors may be arranged accordingly on different surfaces of at least one wind propulsion device and / or at different locations within at least one wind propulsion device.

[0021] The term "pressure-based force measurement data" refers to information relating to the pressure (i.e., force per unit area) exerted by wind on the surface of at least one wind-powered propulsion device. Optionally, a given first sensor is implemented as a pressure sensor. Pressure sensors and their operation are well known in the art. Typically, when wind flows over the surface of at least one wind-powered propulsion device on which the given first sensor is disposed, the wind generates varying pressure levels at those surfaces, such varying pressure levels being detected / measured by the given first sensor. Furthermore, pressure-based force measurement data can be in the form of numerical values ​​of air pressure, for example, expressed in Pascals (Pa), millimeters of mercury (mmHg), pounds per square inch (psi), or similar units. Examples of one or more first sensors used to measure pressure-based force measurement data include, but are not limited to, piezoresistive pressure sensors, capacitive pressure sensors, microelectromechanical systems (MEMS) pressure sensors, and optical pressure sensors.

[0022] The term "strain-based force measurement data" refers to information relating to the deformation or strain experienced by the surface (i.e., structure or material) of at least one wind propulsion device when a force is applied to it by wind. For example, the strain can be longitudinal strain, transverse strain, volumetric strain, shear strain, etc. The types of strain are well known in the art. It should be understood that one or more first sensors can be used as strain gauges (i.e., strain-based sensors) to measure strain-based force measurement data. Typically, a strain gauge is attached or integrated to the surface of an object where a force is expected to be applied, and when the surface deforms under stress, the strain gauge attached to the surface also deforms, causing a change in the strain gauge's resistance. The strain gauge converts this change in resistance into an electrical signal proportional to the applied strain.

[0023] For example, when at least one wind propulsion device is implemented as a flexible sail-based propulsion device, strain gauges are arranged on the surface of the flexible sail-based propulsion device. In this case, the strain gauges can record 0.1% strain, which means that the surface may have deformed by 0.1% when the flexible sail-based propulsion device interacts with the wind. Examples of such strain or deformation may include, but are not limited to, rotor misalignment in rotor-based propulsion devices, surface damage (e.g., dents, cracks) on suction-based or rigid sail-based propulsion devices, wrinkles in the sail of a flexible sail-based propulsion system, and tangles in the cables of a kite sail-based propulsion system.

[0024] The term "displacement-based force measurement data" refers to information obtained by measuring the movement or positional change of at least one portion of a wind propulsion device under the influence of external wind forces. In this regard, a given first sensor can be implemented as a displacement-based sensor, wherein the displacement of said portion due to external wind forces is detected and quantified. It should be understood that, in cases where the expected force displacement is significant, specific portions of at least one wind propulsion device can be identified, which can serve as a baseline indicating the initial state of at least one wind propulsion device before the force displacement occurs. In other words, the baseline provides a reference point for understanding the subsequent force displacement caused by external wind forces. Therefore, when the displacement-based sensor experiences external wind forces, it measures the subsequent force displacement. It should be understood that the displacement-based sensor can continuously monitor the identified portions of at least one wind propulsion device and generate an electrical signal output corresponding to the magnitude and direction of the corresponding force displacement. The displacement-based sensor can be implemented as a linear variable differential transformer (LVDT), an accelerometer, an optical displacement sensor, etc. Displacement-based force measurement data can also be referred to as displacement / offset-based force measurement data.

[0025] In some implementations, the processor is configured to acquire force measurement data (which refers to information related to the forces exerted by wind on at least one wind propulsion device under different weather conditions) directly from one or more first sensors in real time, and to include such data when training the AI ​​model, thereby improving the efficiency of the AI ​​model in making accurate predictions. In other implementations, the processor is configured to acquire force measurement data from a data warehouse that stores the force measurement data, and the processor is communicatively connected to the data warehouse. The technical benefits of determining the force measurement data are that it helps improve adaptability, data richness, and accuracy when establishing predictive relationships between the performance of at least one wind propulsion device and different control inputs.

[0026] Throughout this disclosure, the term "operating parameter" refers to a variable representing the operating conditions of at least one wind propulsion device and at least one vessel. It should be understood that operating conditions can vary depending on factors such as wind speed, sea state, vessel size, and vessel design. Examples of operating conditions may include, but are not limited to, the minimum wind speed required to propel the vessel, the maximum wind speed tolerable for safe operation of the vessel, the optimal wind speed range for efficient propulsion, the maximum wave height for safe operation, the impact of wind propulsion on vessel speed and course, maneuverability under different wind conditions, the impact on navigation and control systems within the vessel, the temperature affecting equipment and materials within the vessel, corrosion resistance in the marine environment, the effects of salt spray and moisture on components, the compatibility of the wind propulsion system with the vessel structure, aerodynamic or hydrodynamic performance under different conditions, structural integrity under wind loads, emergency procedures in the event of loss of wind power or propulsion failure, and environmental regulations and emission standards. Data on at least one operating parameter can provide information relating to the current state and operational behavior of at least one wind propulsion device and at least one vessel. In this regard, at least one wind propulsion device and at least one vessel are equipped with an operating sensor designed to measure data on at least one operating parameter. Furthermore, such operational sensors can be used to generate data based on measurements captured under different operating conditions.

[0027] Optionally, one or more second sensors are used to determine data for at least one operating parameter, wherein the data for the at least one operating parameter includes at least one of the following: Wind speed and direction data, Wind-catching component data, Weather forecast data, Ship movement data, Ship position data, Ship maneuvering data, Ship engine data.

[0028] The term "second sensor" refers to a device capable of detecting and / or measuring at least one operating parameter associated with at least one wind propulsion unit and at least one vessel under given operating conditions. Optionally, at least some of the one or more second sensors are arranged on at least one wind propulsion unit and / or vessel. It should be understood that some second sensors may be arranged on at least one wind propulsion unit, while the remaining second sensors may be arranged on other different parts of the vessel. Different types of one or more second sensors may be used to measure different types of operating parameters (as discussed above).

[0029] Optionally, one or more second sensors are operatively mounted on the vessel or remotely communicatively coupled to it. In this regard, the one or more second sensors may be physically mounted on at least one wind propulsion unit and / or various parts of the vessel. Furthermore, when one or more second sensors are arranged on at least one of the wind propulsion unit and the vessel itself, the one or more second sensors ensure the acquisition of direct and immediate measurements of at least one operating parameter as real-time data, which may be beneficial for decision-making and control strategies. Alternatively, one or more second sensors may be remotely communicatively coupled to at least one of the wind propulsion unit and the vessel via, for example, known wireless technologies and communication protocols. Furthermore, the one or more second sensors may use telemetry arrangements (e.g., satellite communication networks, cellular networks, etc.) to transmit data to a processor. The technical advantage of remotely placing one or more second sensors is that it reduces the need for physical connections and minimizes potential interference with the operation of the vessel.

[0030] The term "wind speed and direction data" refers to information related to wind speed and direction. In other words, wind speed and direction data includes information related to the speed and direction of the wind flow when a ship is in motion. For example, wind speed can be measured in meters per second (m / s), kilometers per hour (km / h), miles per hour (mph), knots (kt or kn), feet per second (ft / s), etc. Furthermore, wind direction can be measured clockwise starting from north, where 0 degrees or 360 degrees correspond to north, 90 degrees to east, 180 degrees to south, and 270 degrees to west. For example, 10 m / s and 270 degrees could mean a wind speed of 10 meters per second coming from the west. When one or more second sensors are used to measure wind speed and direction data, examples of one or more second sensors may include, but are not limited to, anemometers, wind vanes, and ultrasonic wind measurement devices.

[0031] The term "weather forecast data" refers to predictive information related to meteorological conditions such as wind speed, wind direction, sea state, and temperature. This data helps ship operators and wind propulsion system designers predict and plan optimal navigation conditions, enabling efficient use of wind energy for propulsion while ensuring safety and operational efficiency. Furthermore, weather forecast data may include data related to at least one of temperature forecasts, precipitation forecasts, wind forecasts, humidity forecasts, and atmospheric pressure forecasts. Weather forecast data enables at least one wind propulsion unit and at least one vessel to actively adjust their configuration to prepare for higher wind speeds. Weather forecast data provides information related to expected weather conditions, thereby helping at least one wind propulsion unit predict and adapt to changes in environmental factors. It should be noted that weather forecast data may not be directly measured by onboard sensors (i.e., those secondary sensors deployed within the vessel itself), but may be received from external sources such as weather stations, satellites, or other specialized weather monitoring systems.

[0032] The term "ship movement data" refers to information relating to the motion and behavior of a ship in response to wind and other environmental factors. Ship movement data may include at least one of ship speed, ship heading, ship yaw angle, ship pitch angle, ship course, ship drift, and ship stability. The foregoing terms are well known in the art. When one or more second sensors are used to measure ship movement data, examples of the one or more second sensors may include, but are not limited to, inertial measurement units (IMUs), accelerometers, inclinometers, global positioning systems (GPS), global navigation satellite systems (GNSS), gyroscopes, and magnetic compass sensors.

[0033] The term "ship position data" refers to information relating to a ship's geographical location at a given time. Ship position data provides the ship's geographic coordinates at a specific point in time, which may include the ship's latitude, longitude, and, in some cases, altitude. GPS or other satellite navigation systems are typically used to determine ship position data.

[0034] The term "ship maneuvering data" refers to information related to the movement of a ship during its navigation. Ship maneuvering data provides information related to the dynamic movement of a ship, particularly regarding changes in the ship's course, speed, and other maneuvers. It also provides information about how the ship responds to control inputs, environmental conditions, and navigational requirements. Specifically, ship maneuvering data can provide insights related to the ship's steering mechanism. Advantageously, ship maneuvering data can help assess the effectiveness of at least one wind propulsion device, optimize sail configuration, and ensure safe and efficient maneuvering under various environmental conditions. Optionally, ship maneuvering data includes at least one of rudder angle-related data, ship engine load-related data, ship speed-related data, ship heading-related data, and ship stability-related data. It should be understood that when data including at least one operating parameter of ship maneuvering data is used as input to train an AI model, the relationship between the performance of at least one wind propulsion device and different control inputs can be accurately predicted, thus helping to optimize ship operation in real-world maritime scenarios.

[0035] The term "marine engine data" refers to information relating to the operation and performance of at least one engine located on a ship. Marine engine data may include values ​​for at least one of the following: propulsion power, shaft torque, speed, fuel consumption, and engine health diagnostics. Examples of one or more second sensors used to measure marine engine data in this regard may include, but are not limited to, power transducers, torque sensors, load sensors, tachometers, fuel flow meters, temperature sensors, pressure sensors, exhaust temperature sensors, coolant level sensors, oil level sensors, voltage sensors, hour meters, emission sensors, vibration sensors, speed sensors, engine load sensors, alternator output sensors, and oil quality sensors. It should be understood that marine engine data provides benefits such as performance optimization, planned maintenance, improved fuel efficiency, enhanced safety, compliance assurance, decision support, performance tracking, risk mitigation, and asset management.

[0036] The term "wind catcher" refers to a device capable of capturing wind energy. Examples of wind catchers include, but are not limited to, sails, rotors, or any other structure designed to interact with wind energy to propel a vessel. In this document, the term "wind catcher data" refers to information relating to the performance and behavior of the wind catcher responsible for capturing wind energy. Wind catcher data may include parameters such as wind speed, wind direction, wind pressure, aerodynamic forces acting on the wind catcher, angle of attack, and energy conversion efficiency. Therefore, optionally, wind catcher data may be associated with one or more wind catchers. It should be understood that by analyzing data from each wind catcher, adjustments can be made to optimize the configuration and operation of one or more wind catchers. This ensures maximum propulsion efficiency for the vessel under varying wind conditions. This allows operators to adjust angle of attack, sail settings, or other parameters to fine-tune performance and maintain vessel stability. Furthermore, analyzing wind catcher data enables improved energy conversion and utility, ultimately maximizing the propulsion output of the wind propulsion system.

[0037] Optionally, the data for at least one operating parameter includes ship-to-ship relative wind data. In this document, the term "ship-to-ship relative wind data" refers to information relating to the wind conditions experienced by the vessel and at least one wind propulsion unit on board (i.e., when it is windy and the vessel is in operation, i.e., navigating on water). It is understood that ship-to-ship relative wind data refers to information relating to the wind conditions experienced in real time by at least one wind propulsion unit relative to the movement of the vessel navigating on water. It is noteworthy that the vessel's movement may be in the same direction as the wind flow, in the opposite direction of the wind flow, or at an angle to the wind flow. It is understood that ship-to-ship relative wind data provides information relating to the wind conditions experienced by at least one wind propulsion unit relative to wind speed, wind direction, and the direction of the vessel's movement. It is understood that ship-to-ship relative wind data is used to evaluate the performance of the wind propulsion unit under such wind conditions. In this respect, ship-to-ship relative wind data is measured relative to the position and movement of at least one wind propulsion unit and the vessel, thereby providing information relating to how the wind affects the movement of at least one wind propulsion unit and the vessel. It should be noted that shipboard relative wind data is not an absolute measurement of wind conditions when the ship may be stationary in the water, but rather reflects wind conditions when the ship is in motion. Shipboard relative wind data is obtained by using sensors arranged on the ship and / or at least one wind-powered propulsion device. It should be understood that shipboard relative wind data provides a real-time check of wind speed and direction under various wind conditions.

[0038] Optionally, the data for at least one operating parameter includes ship electrical load balance and generator capacity data. In this regard, the distribution and balance of electrical loads related to the ship's electrical layout may include all equipment that consumes available electricity on board, such as lighting, navigation equipment, communication systems, and other electronic components. Typically, monitoring the electrical load balance provides information about how much electricity the ship is consuming. Furthermore, generator capacity data provides information about the capacity of the ship's generators. Generators are responsible for generating electricity on board, and generator capacity indicates the maximum electrical power that the generator can provide. Advantageously, generator capacity data ensures stable ship movement. It should be understood that when ship electrical load balance and generator capacity data (along with other types of data such as at least one operating parameter discussed earlier) are used to train an AI model, the trained AI model acquires the ability to predict how changes in electrical load and generator capacity may affect the performance of at least one wind propulsion unit. Advantageously, because of this, the operation of at least one wind propulsion unit can be adjusted accordingly to maximize the energy efficiency of at least one wind propulsion unit and ensure the judicious use of available electricity.

[0039] Optionally, the data for at least one operational parameter includes route data and subsequent weather data. The term "route data" refers to information related to the vessel's path during navigation. Optionally, the route data is related to at least one of the vessel's historical route, planned route, and current route. It should be understood that route data includes information related to waypoints, route heading, distances between waypoints, expected arrival times, and hazards along the route. It should also be understood that when route data is fed into an AI model to predict the performance of at least one wind propulsion unit, the route data helps provide recommendations for minimizing fuel consumption, reducing navigation duration, and enhancing the overall operational efficiency of at least one wind propulsion unit. By following an optimized route, vessels can avoid adverse weather conditions, rough seas, and congested waters, thereby achieving cost savings and improved safety. Furthermore, the term "subsequent weather data" refers to historical weather data reconstructed for a past time period. Subsequent weather data includes parameters for said past time period, such as wind speed, wind direction, sea state, and atmospheric conditions. It should be understood that when subsequent weather data (in combination with other types of data as discussed earlier) is fed into an AI model to predict the performance of at least one wind propulsion unit, the subsequent weather data enhances the predictive accuracy of the AI ​​model. This is likely because by training the AI ​​model with comprehensive historical weather data, the AI ​​model can learn patterns and correlations that contribute to more accurate performance predictions. This allows operators to predict and optimize the performance of at least one wind propulsion unit based on predictions from the AI ​​model, leading to more efficient ship operations and improved decision-making.

[0040] It is worth noting that, upon receiving force measurement data and data for at least one operating parameter, the processor is configured to feed the aforementioned data into an AI model to train the data, so that, once the AI ​​model has been trained, a predictive relationship can be established between the performance of at least one wind propulsion device and different control inputs. Optionally, the different control inputs include at least force measurement data and data for at least one operating parameter.

[0041] The term "predictive relationship" can be understood to refer to information about how different control inputs affect the performance of at least one wind propulsion device, based on force measurement data and data from at least one operating parameter. For example, when the control input is set to a first value, a first performance characteristic of at least one wind propulsion device is recorded, and when the control input is set to a second value, a second performance characteristic of at least one wind propulsion device is recorded. The first and second performance characteristics are analyzed, and predictive relationships between performance and different control inputs are established.

[0042] It should be understood that training an AI model involves using force measurement data and data from at least one operating parameter to teach the AI ​​model to recognize patterns and establish relationships between different control inputs and target outputs (i.e., the performance of at least one wind propulsion device). In other words, the force measurement data and data from at least one operating parameter are provided to the AI ​​model by a processor as a training dataset. This training dataset can cover a wide range of scenarios, representing various combinations of wind conditions and ship operating conditions. Furthermore, the training data can be preprocessed by handling missing values, outliers, and any inconsistencies, and can be normalized to numerical features to ensure consistent scaling, which helps the AI ​​model converge faster during training. It should be understood that during training, relevant features (located in the force measurement data and data from at least one operating parameter) that significantly contribute to predicting the performance of at least one wind propulsion device can be identified, and any irrelevant or redundant features that may not contribute meaningfully to the prediction can be removed from the training dataset. Additionally, the training dataset can be divided into a training set and a test set, where the training set is used to teach the AI ​​model, while the test set is retained to evaluate the AI ​​model's performance on new, unseen data. The architecture of the AI ​​model can be selected for prediction, such as a regression model, decision tree, ensemble method, or a more advanced model, such as a neural network, and the training dataset is fed into the chosen AI model. Throughout the training process, the AI ​​model learns to establish predictive relationships. These predictive relationships enable the AI ​​model to predict the performance of at least one wind propulsion device under different conditions and control inputs. Once the AI ​​model has been trained, it is evaluated and validated to ensure its accuracy and generalization ability. The AI ​​model can also be referred to as a machine learning model or a deep learning model.

[0043] It is understandable that when establishing a predictive relationship between the performance of at least one wind propulsion device and different control inputs, force measurement data and data from at least one operating parameter are used to train an AI model, resulting in improved adaptability, data richness, and accuracy of the AI ​​model.

[0044] In one example, given a wind speed of X m / s measured from the ship's bridge, the AI ​​model can predict the sail force generated by various control inputs, taking into account factors such as the rotational speed in the case of a rotary sail. This prediction is based on training the AI ​​model using sail force data under various wind conditions, leveraging the Magnus effect. The advantage in this paper is that the AI ​​model can accurately learn and predict sail forces under real-world airflow disturbances, thus enhancing the realism of the prediction. In another example, when the ship's course is X, the AI ​​model can predict the sail force generated by various control inputs, taking into account changes in course Y degrees. Such changes affect the direction, speed, and disturbance of the relative wind, thus affecting the sail. By integrating knowledge of actual wind conditions, the AI ​​model better understands how course adjustments affect sail force generation. In yet another example, for a ship speed of X knots, the AI ​​model can predict the sail force generated using control inputs when the ship's speed changes from X knots to Y knots. The correlation between ship speed and sail force primarily stems from changes in the characteristics of the sail relative to the wind, such as direction and speed. With insights into real-world wind dynamics, the AI ​​model can accurately measure the impact of changes in ship speed on sail force.

[0045] Optionally, the method further includes training an artificial intelligence model using computational fluid dynamics data associated with at least one of the wind propulsion device and at least one of the vessels. In this document, the term "computational fluid dynamics data" refers to information obtained through simulation of fluid flow around at least one wind propulsion device and the vessel. In this regard, the computational fluid dynamics data can be obtained using a simulation model that represents in real time the at least one wind propulsion device, the vessel, the operating conditions associated with at least one of the wind propulsion device and the vessel, and wind conditions. For example, to obtain computational fluid dynamics data related to a wind propulsion device, simulations using a suitable platform (e.g., Matlab®) can be performed by the following steps: (i) modeling the geometry of the wind propulsion device (e.g., height, width, length, wind contact area) (preferably in 3D modeling); (ii) setting boundary conditions such as wind speed, wind direction, and wind pressure (based on real-time data received from at least one of the first and second sensors); (iii) dividing the air around the wind propulsion device into a fine mesh of small virtual cells to capture detailed flow characteristics (e.g., turbulence of wind flow) using relevant mathematical formulas (e.g., but not limited to the Navier-Stokes equations); and (iv) iteratively solving the relevant mathematical formulas until a reliable solution is obtained (i.e., until the variation between iterations is minimal). Computational fluid dynamics data are then obtained from the solution. It is understood that the computational fluid dynamics data may also provide information related to wind speed, wind direction, high-pressure areas, and low-pressure areas. Simulations of fluid flow around at least one wind propulsion device and a ship are obtained under different control inputs. In this regard, computational fluid dynamics (CFD) data can provide detailed performance of at least one wind propulsion device by modeling / simulating the airflow around at least one component of the device (e.g., a Magnus rotor or an airfoil sail). CFD data can include information related to lift and drag, airflow velocity, and pressure distribution. Furthermore, in addition to force measurement data and data for at least one operating parameter, CFD data is used to train an AI model. During training, the AI ​​model learns the relationship between a combination of feature sets under various conditions and the target output (i.e., the performance of at least one wind propulsion device). Advantageously, such an approach allows the AI ​​model to establish relationships between design parameters, operating conditions, and the resulting performance of at least one wind propulsion device, thereby contributing to highly accurate and informed predictions of real-world scenarios.

[0046] Optionally, the method further includes using predictive relationships to determine at least one wind propulsion device control parameter. In this regard, in addition to predicting at least one wind propulsion performance, predictions from an AI model can be used to optimize at least one wind propulsion device control parameter. The term "control parameter" refers to different control inputs associated with wind-catching components (e.g., sails). For example, the AI ​​model can predict the force generated by the sail under different wind speeds, ship routes, and ship speeds. Based on these predictions, adjustments to sail control inputs (e.g., sail angle, rotation speed (for rotary sails), or deployment strategies) can be dynamically optimized to maximize the efficiency and effectiveness of at least one wind propulsion system. Furthermore, the AI ​​model can predict the impact of sail forces on ship tilt (e.g., given a force X generated by the sail, the ship tilts average Y, and the AI ​​model predicts what the average ship tilt will look like under current conditions using all possible operating inputs) and on power consumption (e.g., when operating the sail under current conditions, consuming X kW of power and generating Y kN of power, the AI ​​model predicts what power the sail will generate if available power is limited or if more power is available). This enables informed decision-making related to the ship's power management and navigation operations. Beneficially, this allows ship operators to make informed decisions in real time, ensuring optimal performance and fuel savings while navigating under varying environmental conditions.

[0047] Optionally, predictions provided by the AI ​​model can be used to optimize the control parameters of at least one wind propulsion device to improve its performance. As a result, additional information related to force measurement data and at least one operating parameter can be generated, and this generated information can enhance data richness for further training of the AI ​​model to improve adaptability and predictive accuracy. Furthermore, the predictive relationship between the performance of at least one wind propulsion device and different control inputs enables the AI ​​model to accurately predict the performance of at least one wind propulsion device under different control inputs.

[0048] This disclosure also relates to the system described above. Various embodiments and variations of the methods disclosed above are applied to this system with necessary modifications.

[0049] Optionally, at least one wind propulsion device includes: One or more wind-catching components, operatively coupled to a processor, for generating thrust using forces produced by wind conditions on the ship; and A power transmission mechanism, which is operatively connected to one or more wind-catching components and a processor.

[0050] In this regard, wind-catching components (such as Magnus rotors or airfoil sails) are designed to capture energy from the wind. When wind flows over these components, it exerts a force on them, generating lift or thrust. The wind-catching components are strategically positioned on the vessel to utilize available wind energy. A processor receives data or information from the wind-catching components. This data may include measurements related to the forces generated by the wind-catching components, as well as other relevant information related to the state and performance of the wind-catching components. Based on the received data, the processor can optimize the configuration or orientation of the wind-catching components to maximize thrust or efficiency. This optimization may involve adjusting the angle or other parameters of the wind-catching components to align with prevailing wind conditions. Furthermore, a power transmission mechanism transfers the mechanical energy generated by the wind-catching components to other components on the vessel. The power transmission mechanism then converts the captured wind energy into a form suitable for propelling the vessel. This may involve converting the rotational motion of the wind-catching components into mechanical energy that can propel the vessel forward.

[0051] Optionally, in this system, the data for at least one operating parameter includes the ship's relative wind data.

[0052] Optionally, in this system, the data for at least one operating parameter includes ship electrical load balance and generator capacity data.

[0053] Optionally, in this system, at least the data for the operating parameters includes route data and subsequent weather reports.

[0054] Optionally, in this system, the force measurement data includes at least one of the following: Force measurement data based on pressure, Force measurement data based on strain, Force measurement data based on displacement.

[0055] Optionally, in this system, the data for at least one operating parameter includes at least one of the following: Wind speed and direction data, Wind-catching component data, Weather forecast data, Ship movement data, Ship position data, Ship maneuvering data, Ship engine data.

[0056] Optionally, in this system, wind-catching component data is associated with one or more wind-catching components.

[0057] Optionally, in this system, one or more second sensors are operatively mounted on the vessel or remotely communicatively connected to the vessel.

[0058] Optionally, in this system, the processor is also configured to train the artificial intelligence model using computational fluid dynamics data associated with at least one of the wind propulsion devices and at least one of the ships.

[0059] Optionally, in this system, the processor is also configured to use predictive relationships to determine at least one wind propulsion device control parameter.

[0060] Detailed description of the attached figures refer to Figure 1 This document illustrates the steps of a method for training an AI model to predict the performance of at least one wind propulsion device on a ship, according to embodiments of the present disclosure. At step 102, force measurement data is received, wherein the force measurement data indicates the forces generated by at least one wind propulsion device. At step 104, data for at least one operating parameter is received, wherein the data for the at least one operating parameter indicates operating conditions associated with at least one wind propulsion device and at least one component of the ship. At step 106, the AI ​​model is trained using the force measurement data and the data for the at least one operating parameter to establish predictive relationships between the performance of at least one wind propulsion device and different control inputs.

[0061] The foregoing steps are merely illustrative, and other alternatives may be provided without departing from the scope of the claims, in which one or more steps may be added, one or more steps may be removed, or one or more steps may be provided in a different order.

[0062] refer to Figure 2 The illustration depicts an exemplary scenario according to an embodiment of the present disclosure, in which a system 200 is employed to train an AI model for predicting the performance of at least one wind propulsion device (depicted as wind propulsion device 202) on a vessel 204. References Figure 2 The system 200 is used in a vessel 204, which includes a wind propulsion device 202. The system 200 includes one or more first sensors (e.g., depicted as first sensors 206a and 206b), one or more second sensors (e.g., depicted as second sensors 208a and 208b), and a processor 210. The processor 210 is operatively coupled to the first sensors 206a, 206b and the second sensors 208a, 208b. The wind propulsion device 202 includes one or more wind-catching components (depicted as wind-catching component 212) and a power transmission mechanism 214. The power transmission mechanism 214 is operatively coupled to the wind-catching component 212 and the processor 210. The processor 210 is configured to perform various operations, as previously described with respect to the first aspect.

[0063] Figure 2These are merely examples and should not unduly limit the scope of the claims herein. Those skilled in the art will recognize many variations, alternatives, and modifications to the embodiments of this disclosure.

Claims

1. A method for training an artificial intelligence model to predict the performance of at least one wind propulsion device (202) on a ship (204), wherein, The method includes: The processor (210) receives force measurement data, which indicates the force generated by the at least one wind propulsion device; The processor receives data for at least one operating parameter, the data of which indicates operating conditions associated with at least one of the at least one wind propulsion device and the vessel; and The processor uses the force measurement data and data from the at least one operating parameter to train the artificial intelligence model to establish a predictive relationship between the performance of the at least one wind propulsion device and different control inputs.

2. The method according to claim 1, wherein, The data for at least one operating parameter includes the ship's relative wind data.

3. The method according to claim 1, wherein, The data for at least one operating parameter includes ship electrical load balance and generator capacity data.

4. The method according to claim 1, wherein, The data indicating the operating conditions associated with the vessel (204) are vessel position data, optimized route data, and subsequent weather data.

5. The method according to any one of the preceding claims, wherein, The force measurement data is determined using one or more first sensors (206a, 206b), wherein the force measurement data includes at least one of the following: Force measurement data based on pressure, Force measurement data based on strain, Force measurement data based on displacement.

6. The method according to any one of the preceding claims, wherein, One or more second sensors (208a, 208b) are used to determine data for the at least one operating parameter, wherein the data for the at least one operating parameter includes at least one of the following: Wind speed and direction data, Wind-catching component data, Weather forecast data, Ship movement data, Ship position data, Ship maneuvering data, Ship engine data.

7. The method according to claim 6, wherein, The wind-catching component data is associated with one or more wind-catching components (212).

8. The method according to claim 6 or 7, wherein, The one or more second sensors (208a, 208b) are operatively mounted on the vessel (204) or remotely communicatively connected to the vessel.

9. The method according to any one of the preceding claims, wherein, The method further includes training the artificial intelligence model using computational fluid dynamics data associated with at least one of the at least one wind propulsion device (202) and the vessel (204).

10. The method according to any one of the preceding claims, wherein, The method further includes using the prediction relationship to determine at least one control parameter for a wind propulsion device.

11. A system (200) for training an artificial intelligence model to predict the performance of at least one wind propulsion device (202) on a ship (204), the system comprising: One or more first sensors (206a, 206b) are used to determine force measurement data, the force measurement data indicating the force generated by the at least one wind propulsion device; One or more second sensors (208a, 208b) are used to determine data of at least one operating parameter, the data of which indicates operating conditions associated with at least one of the at least one wind propulsion device and the vessel; as well as A processor (210), operatively coupled to the one or more first sensors and the one or more second sensors, wherein the processor is configured to: Receive the force measurement data; Receive data for at least one of the operation parameters; as well as The artificial intelligence model is trained using the force measurement data and the data of the at least one operating parameter to establish a predictive relationship between the performance of the at least one wind propulsion device and different control inputs.

12. The system (200) according to claim 11, wherein, The at least one wind propulsion device (202) includes: One or more wind-catching components (212), operatively coupled to the processor (210), for generating thrust using the force produced by the at least one wind propulsion device; and A power transmission mechanism (214) is operatively coupled to the one or more wind-catching components and the processor.

13. The system (200) according to claim 11 or 12, wherein, The force measurement data includes at least one of the following: Force measurement data based on pressure, Force measurement data based on strain, Force measurement data based on displacement.

14. The system (200) according to any one of claims 11 to 13, wherein, The data for the at least one operating parameter includes at least one of the following: Wind speed and direction data, Wind-catching component data, Weather forecast data, Ship movement data, Ship position data, Ship maneuvering data, Ship engine data.

15. The system (200) according to any one of claims 11 to 14, wherein, The processor (210) is also configured to use the prediction relationship to determine at least one wind propulsion device control parameter.

Images