Apparatus and method for providing specialized artificial neural network model for hydrogen fuel supply

Abstract

The apparatus for providing a specialized artificial neural network model corresponding to a process of fueling hydrogen for hydrogen-fueled mobility devices, according to the present invention, includes a specialized artificial neural network for each charging station, each dispenser, or each mobility device, for supplying hydrogen fuel to mobility devices, and provides a specialized artificial neural network model that is specialized so as to acquire control or prediction information related to a process of supplying hydrogen fuel to a mobility device, on the basis of identification information regarding a charging station, a dispenser, or a mobility device.

Description

Device and method for providing a specialized artificial neural network model for hydrogen fuel supply

[0001] The present invention relates to a control technology for hydrogen fueling of hydrogen fuel mobility, and more particularly, to a hydrogen fueling process that increases the efficiency of hydrogen fueling and the convenience, speed, and real-time nature of fueling, and a platform for the process, and to an apparatus and method for providing a specialized artificial neural network model for hydrogen fueling.

[0002] The material described in this section merely provides background information for the present embodiment and does not constitute prior art.

[0003] Hydrogen vehicles, or hydrogen electric vehicles, are zero-emission vehicles that run on electricity generated by combining high-pressure hydrogen stored onboard with air. They are also called fuel cell electric vehicles (FCEVs). Most FCEVs use hydrogen as their energy source, generating electricity through a fuel cell system. FCEVs not only emit only pure water (H2O) during the electricity generation process, but also have the ability to remove ultrafine dust from the atmosphere during operation, making them a promising future eco-friendly form of mobility. Given the inexhaustible availability of hydrogen as a fuel and the environmental friendliness of the energy production process, they are gaining widespread attention as a technology with potential for widespread industrial application.

[0004] Hydrogen-fueled mobility refers to mobility that uses hydrogen as an energy source or fuel to generate electricity and use it to drive an electric motor. In addition to the hydrogen-powered vehicles described above, hydrogen-fueled mobility also includes aerial mobility, industrial trucks, trains, ships, and aircraft, as well as devices that use hydrogen as fuel to generate electricity and use it to drive vehicles.

[0005] Most hydrogen fuel cell vehicles use high-pressure hydrogen safely stored in hydrogen fuel tanks and oxygen supplied through an air supply system to deliver it to the fuel cell stack, where the electrochemical reaction between the hydrogen and oxygen produces electrical energy. This electrical energy is converted into kinetic energy via the drive motor, propelling the vehicle. While in motion, hydrogen fuel cell vehicles have the advantage of emitting only pure water through the exhaust.

[0006] Meanwhile, the concept of a hydrogen fuel cell vehicle (HV)—not a hydrogen electric vehicle—also refers to a vehicle that uses hydrogen as fuel. A HV uses the heat generated by directly combusting hydrogen in an internal combustion engine (ICE) to drive an electric motor. The method of supplying hydrogen for HVs is not significantly different from that for hydrogen electric vehicles.

[0007] The control technique for supplying hydrogen to a vehicle that uses hydrogen as fuel ultimately aims to control the temperature and pressure of the compressed hydrogen storage system (CHSS) on the fuel cell side to operate under the limit temperature and limit pressure conditions for the safety of hydrogen fuel supply.

[0008] The hydrogen fueling process, control techniques, and protocols for conventional hydrogen fuel cell vehicles were developed when wired and wireless communication technologies and control computing techniques were not yet mature. Therefore, they do not adequately reflect recent advancements in information and communication technology (ICT). Therefore, conventional hydrogen fueling technologies for hydrogen fuel cell vehicles are inefficient, slow, and unsuitable for large-scale hydrogen fueling.

[0009] The purpose of the present invention to solve the above problems is to propose a hydrogen fueling process that can improve the safety, compatibility, efficiency, and reliability of the hydrogen fueling process of hydrogen fueled mobility.

[0010] Another object of the present invention is to improve the speed and real-time performance of the hydrogen fueling process and to improve the accuracy of the control and prediction process of the hydrogen fueling process by utilizing additional means such as an artificial neural network model and / or model predictive control.

[0011] Another object of the present invention is to propose a process for applying an optimized communication protocol and fueling protocol that can utilize conventional or advanced communication media to enable hydrogen fuel mobility and dispensers to effectively achieve hydrogen fueling goals.

[0012] Another object of the present invention is to propose an optimized fuel supply process according to the surrounding circumstances and conditions when supplying hydrogen fuel.

[0013] In order to achieve the above object, a device for providing a specialized artificial neural network model for a process of fueling hydrogen fueled mobility according to one embodiment of the present invention may include a memory storing at least one command; a processor executing at least one command; and a charging station, dispenser, or artificial neural network model specialized for each mobility, the charging station electronically communicating with the processor and fueling hydrogen fueled mobility.

[0014] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention may, through at least one command, obtain identification information regarding a charging station, dispenser, or mobility. Based on the identification information, the processor may provide a specialized artificial neural network model so that the specialized artificial neural network model obtains control or prediction information related to the process of supplying hydrogen to the mobility.

[0015] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can train a basic model of the specialized artificial neural network model to learn a function of generating control or prediction information related to a process of refueling hydrogen to a mobility using measurement data of a charging station, dispenser, or mobility obtained in the process of refueling hydrogen to the mobility.

[0016] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can store a model in which a basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility, and can provide a specialized artificial neural network model based on identification information.

[0017] The processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain a basic model from a cloud or local server. The processor can then feed back the updated and stored specialized artificial neural network model to the cloud or local server.

[0018] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain a new basic model implemented using federated learning from the cloud or local server after feeding back the specialized artificial neural network model to the cloud or local server.

[0019] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can provide a specialized artificial neural network model based on identification information so that control or prediction information can be used to control or predict a process in which a dispenser supplies hydrogen to a mobility.

[0020] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can provide a specialized artificial neural network model based on a fuel supply protocol corresponding to a process in which a dispenser supplies hydrogen to a mobility or a communication protocol between the mobility and the dispenser.

[0021] A method for providing a specialized artificial neural network model according to one embodiment of the present invention may include the steps of: obtaining an artificial neural network model specialized for each charging station, dispenser, or mobility that fuels hydrogen fueled mobility; obtaining identification information about the charging station, dispenser, or mobility; and providing an artificial neural network model specialized for the specialized artificial neural network model to obtain control or prediction information related to a process of fueling hydrogen to the mobility based on the identification information.

[0022] A device for controlling a process of supplying hydrogen fuel to hydrogen fueled mobility according to one embodiment of the present invention may include a memory storing at least one command; a processor executing at least one command; and a charging station, dispenser, or artificial neural network model specialized for each mobility.

[0023] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can obtain control or prediction information related to a process of supplying hydrogen fuel to a mobility from a specialized artificial neural network model determined based on identification information about a charging station, a dispenser, or a mobility, by at least one command.

[0024] A device for controlling a hydrogen fuel supply process according to one embodiment of the present invention may include a plurality of artificial neural network model candidates specialized for a plurality of types of mobility.

[0025] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may select one of a plurality of artificial neural network model candidates as a specialized artificial neural network model based on first identification information about mobility.

[0026] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can transmit first identification information about mobility and second identification information about a charging station or dispenser to a cloud or local server, and can receive a specialized artificial neural network model from the cloud or local server based on the first identification information and the second identification information.

[0027] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may receive a specialized artificial neural network model from the mobility based on first identification information about the mobility.

[0028] A processor of a device controlling a hydrogen fueling process according to one embodiment of the present invention can acquire a basic model of a specialized artificial neural network model. The processor can train the basic model to learn the function of generating control or predictive information related to the process of hydrogen fueling a vehicle using measurement data from a charging station, dispenser, or vehicle obtained during the process of hydrogen fueling a vehicle.

[0029] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may store a model in which a basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility.

[0030] The processor of a device controlling a hydrogen fueling process according to one embodiment of the present invention can obtain a basic model from a cloud or local server. The processor can then feed back an updated and stored specialized artificial neural network model to the cloud or local server.

[0031] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can obtain control or prediction information related to a process of supplying hydrogen fuel to a mobility using a specialized artificial neural network model determined based on a fuel supply protocol corresponding to a process of supplying hydrogen fuel to a mobility by a dispenser or a communication protocol between the mobility and the dispenser.

[0032] A hydrogen fuel supply control device according to one embodiment of the present invention may be disposed in a dispenser that supplies hydrogen to a mobility vehicle, or may be implemented as a controller that is not disposed in the dispenser but is capable of electronically communicating with the dispenser and influencing fueling of the dispenser.

[0033] According to one embodiment of the present invention, a method for fueling hydrogen fueled mobility may include the steps of: obtaining an artificial neural network model specialized for a charging station, dispenser, or mobility that fuels the mobility with hydrogen; and obtaining control or prediction information related to a process of fueling the mobility with hydrogen from the specialized artificial neural network model determined based on identification information about the charging station, dispenser, or mobility.

[0034] The step of obtaining a specialized artificial neural network model may include a step of selecting one of a plurality of artificial neural network model candidates specialized for each of a plurality of mobility types as the specialized artificial neural network model based on first identification information about the mobility.

[0035] The step of obtaining a specialized artificial neural network model may include a step of transmitting first identification information about the mobility and second identification information about the charging station or dispenser to a cloud or local server; and a step of receiving a specialized artificial neural network model from the cloud or local server based on the first identification information and the second identification information.

[0036] The step of obtaining a specialized artificial neural network model may include a step of receiving a specialized artificial neural network model from the mobility based on first identification information about the mobility.

[0037] A method for supplying hydrogen to hydrogen fuel mobility according to one embodiment of the present invention may further include: a step of setting a current specialized artificial neural network model used in a step of obtaining control or prediction information as a basic model; a step of training the basic model to learn a function of generating control or prediction information related to a process of supplying hydrogen to the mobility by using measurement data of a charging station, a dispenser, or the mobility obtained in the process of supplying hydrogen to the mobility; and a step of storing a model in which the basic model is updated through training as a new specialized artificial neural network model.

[0038] A method for supplying hydrogen to hydrogen fuel mobility according to one embodiment of the present invention may further include a step of obtaining a basic model from a cloud or local server; and a step of feeding back a new specialized artificial neural network model to the cloud or local server.

[0039] The step of obtaining a specialized artificial neural network model may include a step of obtaining a specialized artificial neural network model determined based on a fueling protocol corresponding to a process in which the dispenser supplies hydrogen to the mobility or a communication protocol between the mobility and the dispenser.

[0040] According to one embodiment of the present invention, the safety, compatibility, efficiency and reliability of hydrogen fueling can be improved in the hydrogen fueling process of hydrogen fueled mobility.

[0041] According to one embodiment of the present invention, the speed and real-time performance of the hydrogen fueling process can be improved, and the accuracy of the control and prediction process of the hydrogen fueling process can be improved by utilizing additional means such as an artificial neural network model and / or model predictive control.

[0042] According to one embodiment of the present invention, an optimized fuel supply process can be implemented according to the surrounding circumstances and conditions when supplying hydrogen fuel.

[0043] FIG. 1 is a conceptual diagram illustrating a hydrogen fuel supply and control process for hydrogen fueled mobility to which one embodiment of the present invention is applied.

[0044] FIG. 2 is a conceptual diagram illustrating an example of state changes occurring during a hydrogen fuel supply process for a hydrogen vehicle / hydrogen mobility to which one embodiment of the present invention is applied.

[0045] FIG. 3 is a conceptual diagram illustrating a platform or test platform to which a hydrogen fuel supply process according to one embodiment of the present invention is applied.

[0046] FIG. 4 is a conceptual diagram illustrating a model prediction control (MPC) process according to one embodiment of the present invention.

[0047] FIG. 5 is a conceptual diagram illustrating a platform or test platform for simulation used in a hydrogen fuel supply process according to one embodiment of the present invention.

[0048] FIG. 6 is a conceptual diagram illustrating a concept of modeling, predicting, or controlling a hydrogen fuel supply process using an artificial neural network (ANN) in a hydrogen fuel supply process according to one embodiment of the present invention.

[0049] FIG. 7 is a conceptual diagram illustrating a concept of controlling a hydrogen fuel supply process to reach a target output using model predictive control in a hydrogen fuel supply process according to one embodiment of the present invention.

[0050] FIG. 8 is a flowchart illustrating the training and inference process of an artificial neural network when an artificial neural network is used for model prediction control in a hydrogen fuel supply process according to one embodiment of the present invention.

[0051] FIG. 9 is a conceptual diagram illustrating a process of implementing and providing a specialized artificial neural network model for a hydrogen fuel supply process according to one embodiment of the present invention.

[0052] FIG. 10 is a flowchart illustrating a hydrogen fuel supply method and / or a control method for a hydrogen fuel supply process according to one embodiment of the present invention.

[0053] Figure 11 is a flowchart illustrating a process in which a specialized artificial neural network model is trained and updated based on the measurement data of Figure 9.

[0054] FIG. 12 is a conceptual diagram illustrating an example of a generalized hydrogen fuel supply control device, hydrogen fuel supply control system, hydrogen fuel supply simulation device, hydrogen fuel supply simulation system, hydrogen fuel supply test platform, hydrogen fuel supply test system, device or computing system that provides a specialized artificial neural network model for a hydrogen fuel supply process, capable of performing at least a part of the processes of FIGS. 1 to 11.

[0055] In addition to the above purpose, other objects and features of the present invention will become apparent through the description of embodiments with reference to the attached drawings.

[0056] The present invention is susceptible to various modifications and embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the present invention. Throughout the description of each drawing, similar reference numerals have been used to designate similar components.

[0057] Terms such as "first," "second," "A," and "B" may be used to describe various components, but these components should not be limited by these terms. These terms are used solely to distinguish one component from another. For example, without departing from the scope of the present invention, the first component could be referred to as the "second component," and similarly, the second component could also be referred to as the "first component." The term "and / or" includes any combination of multiple related items listed or any one of multiple related items listed.

[0058] In the embodiments of the present application, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.” Furthermore, in the embodiments of the present application, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.”

[0059] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0060] The terminology used in this application is only used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0061] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0062] Some terms used in this specification are defined as follows:

[0063] Hydrogen fueled vehicles generally include both hydrogen electric vehicles or hydrogen fuel cell electric vehicles (FCEVs) that use fuel cells, as well as ICE (Internal Combustion Engine)-based vehicles that use hydrogen as fuel.

[0064] Additionally, hydrogen fueled mobility refers to mobility that uses hydrogen as a fuel, and mobility can include automobiles in the conventional sense, automobiles, and hybrid vehicles that can utilize both human power and fuel.

[0065] In the following embodiments, a hydrogen fueling protocol and / or a communication protocol for hydrogen fueling may be applied in hydrogen fuel mobility.

[0066] The hydrogen fluid fuel may include gaseous hydrogen fuel or liquid hydrogen fuel.

[0067] Compressed Hydrogen Storage System (CHSS) refers to a device that compresses and stores hydrogen as part of the vehicle / mobility side.

[0068] A pressure relief device (PRD) is placed in the CHSS and is capable of isolating the stored hydrogen from the rest of the fuel system and environment, or, conversely, releasing the hydrogen to the outside.

[0069] Hydrogen charging basically refers to the process of supplying high-pressure hydrogen from a dispenser at a hydrogen charging station and compressing and storing it in a vehicle's tank. In the context of supplying hydrogen fuel to a hydrogen electric vehicle, "hydrogen charging" can be used interchangeably with the term "fueling." That is, in this specification, "fueling" can refer to fuel supply, hydrogen charging, or charging, and "charging" can refer to the charging of hydrogen fuel. For example, a fueling protocol can be referred to as a "charging protocol," a fueling session can be referred to as a "charging session," and a fueling method can be referred to as a "hydrogen charging method" or "fueling method."

[0070] Pressure Ramp Rate (PRR) is expressed in MPa / min and refers to the pressure increase rate of CHSS.

[0071] Average Pressure Ramp Rate (APRR) refers to the average pressure ramp rate from the beginning to the end of hydrogen fueling.

[0072] Pre-cooling refers to the process of cooling hydrogen in advance at a hydrogen charging station before charging.

[0073] A dispenser is a component that delivers pre-cooled hydrogen to the CHSS. The dispenser is placed at a hydrogen charging station and can perform hydrogen charging operations between the station's hydrogen storage tank and the vehicle's CHSS.

[0074] The nozzle is a device that is connected to the dispenser and is coupled to the receptacle of the hydrogen electric vehicle and allows the delivery of hydrogen fuel.

[0075] A fueling session may be used to mean communication sessions that occur across use cases for hydrogen fueling.

[0076] Interoperability can refer to the state in which components of a system can work together to achieve the intended function of the entire system. Information interoperability can refer to the ability of two or more networks, systems, devices, applications, or components to share information securely and effectively and easily with little or no user inconvenience.

[0077] Correlation / Association may include the process of establishing a relationship between two peer communication entities.

[0078] Command and control communication may refer to communication between an electric vehicle hydrogen fueling device and a hydrogen electric vehicle that exchanges information necessary for starting, controlling, and terminating the hydrogen fueling process.

[0079] Meanwhile, in the detailed description below, for the convenience of explanation, embodiments related to hydrogen electric vehicles or fuel cell electric vehicles (FCEVs) may be illustrated, but it will be apparent to those skilled in the art that the spirit of the present invention can be applied to various types of hydrogen fueled mobility. Hydrogen fueled mobility refers to mobility that uses hydrogen as an energy source or generates electric energy using hydrogen as fuel and uses this to drive an electric motor. In addition to hydrogen electric vehicles, hydrogen fueled mobility may include aerial mobility, as well as industrial trucks, trains, ships, and aircraft, which produce electric energy using hydrogen as fuel and use this to drive devices.

[0080] In addition, the two-way communication process for hydrogen fuel supply of the present invention can be partially applied not only to hydrogen fuel mobility but also to buildings or facilities that use hydrogen as an energy source.

[0081] Additionally, in the following description, hydrogen fuel may include at least one of gaseous hydrogen and liquid hydrogen, and may basically mean compressed hydrogen, but is not limited thereto.

[0082] In addition, vehicles that utilize the hydrogen fuel supply communication two-way process are described with a focus on hydrogen electric vehicles (FCEVs) for convenience of explanation, but are not limited to this configuration and may also include hybrid EV (electric vehicle) vehicles that use hydrogen as fuel, and ICE (internal combustion engine) vehicles.

[0083] In the following specification, some or all of the communication method, communication protocol negotiation method, hydrogen charging (fueling) protocol negotiation method, and hydrogen charging (fueling) parameter negotiation method performed in hydrogen fuel mobility may be performed by an electronic control unit (ECU), a communication device, or a communication control device in hydrogen fuel mobility.

[0084] In the following specification, some or all of the processes of the communication method, the communication protocol negotiation method, the hydrogen charging (fueling) protocol negotiation method, the hydrogen charging (fueling) parameter negotiation method, the hydrogen charging (fueling) method, and the hydrogen charging (fueling) control method performed in the dispenser may be performed by a controller, an electronic control unit, a communication unit, or a communication control unit of the dispenser. In addition, some processes of the above methods may be performed by a controller, an electronic control unit, a communication unit, or a communication control unit of a charging station associated with the dispenser.

[0085] Meanwhile, even if a technology was known before the filing date of the present application, it may be included as a part of the composition of the present application invention if necessary, and this will be described in this specification to the extent that it does not obscure the spirit of the present invention. However, in describing the composition of the present application, a detailed description of matters that were known before the filing date of the present application and that can be clearly understood by those skilled in the art may obscure the spirit of the present invention, and therefore, an excessively detailed description of the known technology will be omitted. For example, a technology that uses a thermodynamic model for hydrogen charging control, a technology that applies a model prediction control (MPC) technique for generalized dynamic control, a technology that configures and controls an artificial neural network for training and inference of the artificial neural network, etc., can utilize technology known before the filing date of the present application, and at least some of these known technologies can be applied as element technologies necessary for practicing the present invention.

[0086] However, the purpose of the present invention is not to claim rights to these known technologies, and the contents of the known technologies may be included as part of the present invention within a scope that does not deviate from the purpose of the present invention.

[0087] Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the attached drawings.

[0088] FIG. 1 is a conceptual diagram illustrating a hydrogen fuel supply and control process for hydrogen fueled mobility to which one embodiment of the present invention is applied.

[0089] Referring to Fig. 1, pre-cooled hydrogen gas from a hydrogen charging station (Station, 200) is supplied to a hydrogen vehicle (Hydrogen fueled mobility / vehicle, 300) via a dispenser (100). At this time, the hydrogen fuel supply process can be described by parameters including the average pressure increase rate (APRR).

[0090] Hydrogen storage systems typically installed in vehicles can be broadly categorized into high-pressure hydrogen tanks, pressure control devices, high-pressure piping, and external frames. High-pressure hydrogen tanks have been developed and commercialized with capacities ranging from tens to hundreds of liters. For vehicles, small and lightweight storage tanks are connected in parallel to achieve high capacity.

[0091] A high-pressure hydrogen tank is widely known as a compressed hydrogen storage system (CHSS) (310). In this specification, for the convenience of explanation, the expression “tank” means CHSS (310), and for the convenience of explanation, the expressions “tank” and CHSS (310) may be used interchangeably.

[0092] In a typical hydrogen storage system, hydrogen storage is controlled by utilizing a boss unit that allows hydrogen gas to enter and exit the storage tank (310). Since hydrogen injection and use cannot be performed simultaneously, a valve, a pressure reducing device, and various sensors for measurement are attached to one boss unit to control hydrogen storage.

[0093] The interface between the hydrogen charging station (200) and the hydrogen fuel mobility (300) is handled by the dispenser (100), where the vehicle storage tank information and the fuel supply information of the charging station (200) are integrated to control the target pressure and injection speed, etc. An example of the control logic currently used is the control logic that follows the SAE J2601 (2020-05) standard.

[0094] In the prior art, there are two methods for transmitting information from a hydrogen fuel mobility (300) to a dispenser (100): a communication method and a non-communication method. Even when communication is used, in the prior art, the temperature and pressure values ​​of the vehicle storage tank (310) are simply transmitted in one direction to the dispenser (100), and the dispenser (100) does not actively utilize the information, but only uses it as a safety standard, such as an emergency stop at a temperature and pressure limit.

[0095] All charging logic for safe and rapid charging is managed in the dispenser (100), and the vehicle storage tank (310) has only the minimum safety management device that automatically releases hydrogen in conditions such as overheating through a pressure relief device (PRD) (320) without an active safety management method.

[0096] In order to cope with the phenomenon of the temperature of hydrogen gas rising during the hydrogen fuel supply described later in Fig. 2, the charging station (200) includes a high-pressure hydrogen storage unit (220) and a pre-cooler (210). The pre-cooler (210) supplies hydrogen gas to the hydrogen electric vehicle (300) via the dispenser (100) in a state where the temperature of the hydrogen gas is lowered through pre-cooling.

[0097] In one embodiment of the present invention, a basic configuration similar to that of the prior art is used, but a hydrogen fuel supply controller (110) inside a dispenser (100) actively controls a hydrogen fueling process by utilizing state information such as temperature and pressure received from a hydrogen fuel mobility (300) and a charging station (200), and state of charge information such as a state of charge (SOC, State of Charge) of a CHSS (310).

[0098] According to one embodiment of the present invention, the charging speed is controlled in real time by utilizing real-time temperature data of the vehicle storage tank (310), and is designed to operate at the highest charging speed that satisfies conditions below the safety limit, thereby reducing the charging time within the available range.

[0099] The charging protocol of the prior art has a problem in that the boundary conditions for safety are excessively set, so that the temperature of most storage tanks (310) is measured to be around 40 to 50°C at the time of charging completion, and thus the charging is supplied after excessive pre-cooling.

[0100] According to one embodiment of the present invention, the cooling load of the charging station (200) can be optimized by actively controlling the amount of pre-cooling required and the amount provided, thereby increasing the operating efficiency of the hydrogen charging station (200).

[0101] Conventional protocols centered on lightweight hydrogen electric vehicles have the problem that all variables must be reset and reflected in the standard when charging new mobility vehicles.

[0102] According to one embodiment of the present invention, a control technique that updates the logic itself through a certain learning and training process when applying a new device to an ANN-based learnable charging logic can be widely applied to various mobility areas.

[0103] The only measure to prevent overheating of the storage tank of a conventional hydrogen electric vehicle is to release gas through a pressure relief device / PRD (Pressure Relief Device) (320) when overheating exceeds a certain temperature.

[0104] According to one embodiment of the present invention, by installing a cooling system (330) to be described later in the storage tank (310) itself, the charging speed can be increased, and at the same time, overheating of the storage tank (310) can be actively dealt with, thereby improving the safety of hydrogen fuel mobility (300).

[0105] According to one embodiment of the present invention, it is possible to safely charge / supply hydrogen fuel while improving the efficiency of the hydrogen charging / supply process and improving the speed and real-time nature of the hydrogen charging / supply process.

[0106] According to one embodiment of the present invention, a real-time hydrogen charging control technique based on model predictive control (MPC) can be provided.

[0107] According to one embodiment of the present invention, a control technique based on an artificial neural network (ANN) model can be provided with improved accuracy in predicting hydrogen charging results. By incorporating real-time measured values ​​into an ANN model that utilizes actual fuel supply data along with theoretical simulation results, the accuracy of the prediction results can be improved.

[0108] According to one embodiment of the present invention, the efficiency of hydrogen charging control can be improved by integrating and managing actual measured data and status information predicted from a model using an intelligent meta system (IMS).

[0109] As described above, in the conventional hydrogen fuel supply process, the dispenser (100) is in charge of controlling between the hydrogen fuel mobility (300) and the hydrogen charging station (200), and the dispenser (100) is equipped with a protocol for injecting hydrogen into the hydrogen fuel mobility (300) according to a set rule, thereby overseeing the control.

[0110] An example of a protocol mounted on the dispenser (100) is a protocol based on the international standard SAE-J2601 (2020-05), which is also applied to embodiments of the present invention within the scope consistent with the purpose of the present invention.

[0111] For the minimum requirements for safety, simulations are conducted through thermodynamic modeling for various situations, and the parameters derived from this are utilized in a table-based single injection method and an MC-formula-based partial real-time correction method.

[0112] Minimum safety requirements / requirements include upper limits for temperature and pressure conditions of CHSS (310) and guidelines for state of charge (SOC).

[0113] Simulations can be performed through thermodynamic modeling using boundary conditions including Best to Worst Case.

[0114] This configuration can also be applied to the configuration of an embodiment of the present invention within the scope consistent with the purpose of the present invention.

[0115] Even if the configuration of Fig. 1 is followed, the following problems are found in the prior art that does not actively control the state value in the dispenser (100). The following problems are also revealed in the prior art that relies on simulation using a simple thermodynamic model.

[0116] The injection rate is predetermined based on the assumption of the worst-case boundary conditions (excessive boundary conditions), which leads to unnecessary pre-cooling and a decrease in the overall filling rate. In conventional technology, the injection rate is simply determined by the average pressure ramp rate (APRR), which can hinder proactive response to changing conditions. Unnecessary pre-cooling can also lead to excessive energy and operating costs.

[0117] Since it relies on simulation-based results, there are limitations on the capacity and shape of the applicable storage tank (310), and in the case of a new system, separate resources are required for new development and application, which limits the scope of application.

[0118] Thermodynamic models take a lot of time to derive mathematical calculation results, so they are limited in their application when there are no pre-calculated variables, as they indirectly utilize variables derived through the model. In addition, there is a lack of flexibility in the detailed adjustments to the method itself.

[0119] The table-based method of the prior art has a very low efficiency because it does not utilize the temperature of the pre-cooled hydrogen provided at the charging station (200) or the temperature of the storage tank (310) measured at the hydrogen fuel mobility (300), and has a problem in that it is difficult to flexibly respond to changes in the surrounding environment.

[0120] The MC-Formula-based method of the prior art corrects the precooling temperature in real time, but has the problem of being difficult to expand due to the complex calculation and application method and limitations in the scope of application.

[0121] The protocol was developed with the primary goal of completing safe charging, so there is no alternative to actively control unexpected situations such as excessive pre-cooling or overheating of the storage tank (310), which causes problems such as increased operating costs due to overcooling and delayed charging due to overheating.

[0122] The present invention is characterized by being derived to solve the problems of the prior art, and is characterized by reducing dependence on simulation and attempting to actively control state variables by reflecting real-time measurement data.

[0123] FIG. 2 is a conceptual diagram illustrating an example of state changes that occur during a hydrogen fuel supply process for hydrogen fuel mobility (300) to which one embodiment of the present invention is applied.

[0124] Referring to Fig. 2, when hydrogen is injected into a hydrogen storage tank (310), the internal temperature rises due to compression heat, and as a result, the temperature of the hydrogen gas inside the storage tank (310) rises.

[0125] Temperature control in the hydrogen fuel supply process is achieved by supplying pre-cooled hydrogen gas and controlling the internal temperature of the storage tank (310) to be 85°C or lower at the time of final charging completion.

[0126] The storage tank (310) is configured to have low heat transfer efficiency of the carbon fiber covering the dome and body of the tank (310) in order to block heat exchange between the external atmosphere and the hydrogen gas stored inside during driving.

[0127] When the temperature of the hydrogen gas inside the storage tank (310) rises during the fuel supply process, the temperature rise revealed on the surface of the hydrogen storage tank (310) is insignificant compared to the temperature rise inside until the charging is completed due to the low heat transfer characteristics of the storage tank (310).

[0128] These characteristics prevent heat exchange with the outside air, which can alleviate the rapid temperature rise inside the hydrogen storage tank (310) that occurs during charging, from occurring, so separate temperature management measures are required.

[0129] However, the prior art does not include a separate cooling means other than supplying pre-cooled hydrogen gas from a charging station (200).

[0130] The hydrogen buffer time is managed by controlling the precooling and hydrogen injection speed at the hydrogen charging station (200) to keep the temperature below 85°C, which is the upper limit of the temperature management of the vehicle hydrogen storage tank, but there is no separate temperature management measure for the storage tank (310) of the hydrogen fuel mobility (300).

[0131] Due to this, especially in the summer when the outside temperature is high, it is difficult to control the temperature of the vehicle storage tank (310) at the charging station (200), which causes problems such as charging delays.

[0132] In the embodiment of the present invention, the characteristic curve of FIG. 2 is installed as a basic model, but unlike the conventional technology, real-time data according to variables (outside temperature, air pressure, weather conditions, etc.) appearing in the actual surrounding environment are taken into consideration to optimize the control of the operating load of the charging station (200) in the Phase I precooling stage and the charging speed (pressure increase rate, PRR) occurring in the fuel supply process of Phase II to Phase IV, and optimal control conditions suitable for the actual environment can be derived.

[0133] FIG. 3 is a conceptual diagram illustrating a platform or test platform to which a hydrogen fuel supply process according to one embodiment of the present invention is applied.

[0134] Referring to FIG. 3, an embodiment is shown in which an artificial neural network model (120) is placed in addition to a hydrogen fuel supply controller (110) within a dispenser (100).

[0135] When the artificial neural network model (120) performs a model prediction control (MPC) process to predict the next state value, hydrogen fuel supply control by the ANN-MPC method can be implemented. Assuming that a fuel supply protocol that allows ANN-MPC-based hydrogen fuel supply control is used in FIG. 3, the ANN-MPC-based hydrogen fuel supply protocol is typically implemented in the dispenser (100) between the charging station (200) and the mobility (300). In an alternative embodiment of the present invention, the ANN-MPC-based hydrogen fuel supply protocol can be implemented in the hydrogen fuel supply controller (110) or the artificial neural network model (120). Additionally, in an alternative embodiment of the present invention, the ANN-MPC based hydrogen fueling protocol, hydrogen fueling controller (110), or artificial neural network model (120) may be implemented as part of a controller that is capable of electronically communicating with the dispenser (100) and influencing the hydrogen fueling process within the dispenser (100), even if not disposed within the dispenser (100).

[0136] Referring again to FIG. 3, the artificial neural network model (120) can receive data from a charging station (200) and / or mobility (300) using real-time communication, and use the data to transmit information for controlling hydrogen fuel supply to the hydrogen fuel supply controller (110) for optimal hydrogen fuel supply.

[0137] At this time, real-time communication data is composed of static and dynamic data. Static data is composed of data such as the capacity, configuration, or type of equipment of a storage tank that does not change over time, and dynamic data may refer to data such as temperature or pressure that may change during the fuel supply process.

[0138] Static data may include, for example, the volume of the tank of the mobility (300) or the charging station (200), the pressure rating, the maximum pressure allowed, the type of the tank, the number of tanks, the size of the tank, the serial number of the tank, the manufacturer of the tank, the usage data of the tank, the allowable temperature range of the hydrogen fluid in the tank, etc.

[0139] Dynamic data may include, for example, fueling commands (e.g., start, stop, pause, abandon, increase flow, decrease flow, tank change commands, etc.), control parameters for fueling (e.g., average pressure ramp rate (APRR), pressure ramp rate (PR), etc.), real-time measurements of pressure and / or temperature within the tank, including fluctuations in temperature and / or pressure within the tank (which may indicate a leak and / or impending failure), state of charge (SOC) of the tank at the charging station or mobility side, ambient temperature outside the tank, and / or real-time measurements of the flow rate of hydrogen fluid flowing into the tank.

[0140] Referring again to FIG. 3, a hydrogen fuel supply control method based on an ANN-MPC-based fuel supply protocol can set basic conditions required for hydrogen fuel supply using static data before starting hydrogen fuel supply, and perform real-time control based on dynamic data during the hydrogen fuel supply process.

[0141] Referring to Fig. 3, an exception and problem situation handling module (400) can be added. The ANN-MPC model basically assumes real-time two-way communication, and can exhibit optimal performance under conditions where real-time two-way communication is possible.

[0142] However, various situations that deviate from real-time communication conditions may occur during the verification process before starting hydrogen fueling or during actual hydrogen fueling, as follows.

[0143] (1) No communication situation: When the mobility (300) or charging station (200) is not equipped with a communication function, the communication function is broken, or the communication function is broken during fuel supply and communication is impossible.

[0144] (2) Communication error: Communication with the mobility (300) or charging station (200) is possible, but parts, sensors, or equipment are not functioning properly, causing some of the data to be missing or to have errors.

[0145] (3) Non-certification of equipment, parts, or vehicles: When communication with mobility (300) or charging station (200) is possible and data is collected normally, but the reliability of data transmitted from mobility (300) or equipment is low, requiring verification or limited use.

[0146] ④ Uncertified communication protocol: Communication with mobility (300) or charging station (200) is possible, but the communication method is not specified in the standard or is not certified in the country, so safety, reliability, and stability are not secured.

[0147] Although the real-time communication-based control of ANN-MPC may have optimal performance in terms of functionality, in order to apply it to an actual field, a method that can effectively, step by step, and differentially respond to communication problems that may occur in various forms is required, and a means to solve such problems is proposed in the embodiment of FIG. 9 or FIG. 10 described below.

[0148] FIG. 4 is a conceptual diagram illustrating a model prediction control (MPC) process according to one embodiment of the present invention.

[0149] In the control method using MPC, the accuracy of control increases as the accuracy of the prediction model increases. The control method illustrated in Fig. 4 is ANN-MPC, which uses an artificial neural network model (120) as a prediction model and a feedback control loop including a hydrogen fuel supply controller (110) as a control method.

[0150] Referring to FIG. 4, the temperature, pressure, and outside temperature of the mobility (300) and the charging station (200) are used as inputs, and the temperature and pressure of the storage tank (310) of the mobility (300) are set as outputs.

[0151] In FIG. 4, an embodiment is shown in which the temperature and pressure of the hydrogen mobility (300) are predicted as output, but in an alternative embodiment of the present invention, an embodiment in which the temperature and pressure of the storage tank of the charging station (200) are predicted as output can be implemented.

[0152] When a new input is given, the output can be predicted by an artificial neural network model (120) that learns the correlation between input and output using simulation data by a theoretical model and data measured in an actual field.

[0153] When the SOC value (target value) specified in the controller (110) is given, the pressure ramp rate (PRR), which is the charging speed to achieve the target value, can be predicted. When PRR_pd is given as a control parameter, the artificial neural network model (120) can predict SOC_pd. SOC_pd is input to the controller (110) to replace the initial target SOC_sp. At this time, the SOC_m of the mobility (300) side where PRR_c is actually applied and hydrogen fuel supply is performed is transmitted to the artificial neural network model (120) so that the artificial neural network model (120) can be trained to improve the prediction accuracy of SOC_pd as an objective function or loss function.

[0154] Referring to FIGS. 3 and 4 together, real-time temperature, pressure, supply rate (which may mean PRR) on the side of the charging station (200), and abnormal signals may be transmitted to the artificial neural network model (120). In an alternative embodiment of the present invention, real-time temperature, pressure, supply rate (which may mean PRR) on the side of the charging station (200), and abnormal signals may also be transmitted to the hydrogen fuel supply controller (110).

[0155] Real-time temperature, pressure, and abnormal signals of the mobility (300) can be transmitted to the artificial neural network model (120). In an alternative embodiment of the present invention, real-time temperature, pressure, and abnormal signals of the mobility (300) can also be transmitted to the hydrogen fuel supply controller (110).

[0156] The hydrogen fuel supply controller (110) generates an actual control command, and hydrogen fuel supply of the dispenser (100) is performed based on the control command, and part of the control command, for example, a temperature down command, can also be transmitted to the mobility (300) side.

[0157] At this time, the command or information related to the control of hydrogen fuel supply may include a pressure ramp rate (PRR) for hydrogen fuel supply in the tank (310) of the mobility (300), a control command for the pressure ramp rate, and / or status information on the dispenser (100) side as a result of executing the control command. In addition, the information related to the control of hydrogen fuel supply may include information on variables that may affect changes in the state of hydrogen in the vehicle tank (310) of the mobility (300), such as a real-time pressure ramp rate (PRR) derived from a feedback control process or a mass flow rate of compressed hydrogen (kg / s) [m_dot], and / or information on the result of executing control by the variables. These hydrogen fuel supply control-related information may also affect the weights or parameters of the hidden layer of the artificial neural network model (120).

[0158] A hydrogen fuel supply control request based on a hydrogen fuel supply protocol may include a control command for a pressure ramp rate (PRR) for charging hydrogen in a vehicle tank (310).

[0159] Information related to hydrogen fuel supply control may include at least one of information on a hydrogen fuel supply control request and status information according to the execution result of the hydrogen fuel supply control request, and a control request for a change in the status of hydrogen may include a control request for at least one of the temperature and pressure of hydrogen in the vehicle tank (310).

[0160] Changes in the state of hydrogen in the vehicle tank (310) may include at least one of temperature, pressure, and state of charge (SOC).

[0161] As a theoretical simulation, a thermodynamic model such as H2FillS (Hydrogen Filling Simulation) can be used, for example. The thermodynamic model can include a model and / or software designed to track and report transient changes in at least one of hydrogen temperature, pressure, and mass flow and / or transient changes in the state of hydrogen within a vehicle tank when fueling a hydrogen fueled vehicle / mobility. Of course, the spirit of the present invention is not limited to embodiments of a specific thermodynamic model.

[0162] The hydrogen fueling protocol may include, for example, the hydrogen charging protocol defined in SAE J2601.

[0163] A thermodynamic model can generate output data based on modeling and simulation when input data, such as those of an artificial neural network, is input. The parameters of the thermodynamic model can be adjusted according to the hydrogen charging protocol, and different hydrogen charging protocols with the same input data can produce different output data.

[0164] In one embodiment of the present invention, on-site data collected by a test platform is provided as input and output data of an artificial neural network, instead of input / output data of a thermodynamic model, so that the artificial neural network can be trained. That is, some of the on-site data collected by the test platform can be provided as input data for the artificial neural network, and other parts can be provided as ground truth data corresponding to the output data of the artificial neural network.

[0165] The internal parameters of an artificial neural network can be trained without initialization, or they can be initialized to predetermined values ​​and then trained based on field data. For example, the input and output data (ground truth data) of an artificial neural network can be given and initially trained based on a thermodynamic model, thereby initializing the internal parameters of the artificial neural network.

[0166] Learning in artificial neural networks does not necessarily have to be deep learning; it can also be shallow learning.

[0167] A test platform according to one embodiment of the present invention may rely on dynamic field data to optimize a hydrogen fueling process.

[0168] One embodiment of the present invention can predict the next state using an artificial neural network-based model predictive control (MPC) technique. In this case, the artificial neural network model (120) can be trained using theoretical results, on-site data, or both.

[0169] The artificial neural network model (120) may be a model trained to receive information related to hydrogen fuel supply control and information on the state of hydrogen in a vehicle tank (310) related to the result of executing a hydrogen fuel supply control request, and to predict changes in the state of hydrogen in the vehicle tank (310).

[0170] The artificial neural network model (120) may be a model trained to receive information related to hydrogen fuel supply control, the current state of hydrogen in the vehicle tank (310), the state of hydrogen supplied from the dispenser (100) to the mobility (300), and the ambient temperature, and to predict changes in the state of hydrogen in the vehicle tank (310) in the future.

[0171] The artificial neural network model (120) may be a model trained with the function of predicting changes in the state of hydrogen in the vehicle tank (310) according to each target state of the state of hydrogen in the vehicle tank (310) and each hydrogen fuel supply protocol, which are related to information on hydrogen fuel supply control.

[0172] The artificial neural network model (120) receives information related to hydrogen fuel supply control and information on the state of hydrogen in the vehicle tank (310), and trains a function to predict changes in the state of hydrogen in the vehicle tank (310) by using field data on changes in the state of hydrogen in the vehicle tank as ground truth data, thereby updating parameters in the model (120).

[0173] The artificial neural network model (120) may be a model trained to predict changes in the state of hydrogen in a vehicle tank (310) using a model prediction control technique.

[0174] In cases where the pre-cooling function of the charging station (200) or the cooling system (310) of the mobility (300) is not available, the control of the hydrogen fuel supply can be performed by reflecting these factors.

[0175] Big data is collected for each type and individual ID of the dispenser (100), type and individual ID of the charging station (200), type and individual ID of the mobility (300), control target state (temperature, pressure, SOC), initial state (temperature, pressure), and type of hydrogen fuel supply protocol, and the test platform is trained using dynamic changes in field data corresponding to each case, thereby deriving standardized items that can optimize and accurately describe the hydrogen fuel supply process.

[0176] One embodiment of the present invention can improve the reliability of a hydrogen fuel supply process by collecting and processing on-site hydrogen fuel supply data.

[0177] Field data on changes in the state of hydrogen within the vehicle tank (310) can be first acquired from the mobility (300) side. That is, field data on changes in the state of hydrogen within the vehicle tank (310) can be acquired from the mobility (300) side regardless of whether a control request is transmitted to the hydrogen vehicle. Alternatively, in another embodiment of the present invention, the control request may include a field data request, and the mobility (300) side may acquire field data in response to the control request / field data request.

[0178] Acquisition of field data can be performed not only on the mobility (300) side but also on the charging station side, and similarly, field data on the charging station side can be collected regardless of the control request, and the control request can include a field data request, and field data on the status of hydrogen stored in the charging station and / or hydrogen supplied from the charging station to the dispenser can be acquired in response to the control request / field data request on the charging station side.

[0179] On-site data refers to data obtained from a charging station (200) and / or mobility (300) during the actual hydrogen fuel supply / supply process. In this case, on-site data may include data obtained in cases where some or all of the environment is a test environment, in addition to cases where the environment is an actual charging station or an actual vehicle.

[0180] Field data may include data obtained from a test environment or execution device that may be equipped with a test environment or response (feedback) capable device corresponding to the hydrogen vehicle side, a test environment or response (feedback) capable device corresponding to the hydrogen storage cylinder and dispenser side of the charging station, and / or a module A corresponding to the dispensing control system side.

[0181] At this time, the field data may include data obtained from a fuel supply field or test environment in which all devices are a test environment composed of the above three test environments or devices, or which includes any one of them.

[0182] For example, field data may include data obtained from a test environment or fuel supply site comprised of simulation models linked to module B (150) and module C (160) illustrated in FIG. 5.

[0183] At this time, a device capable of responding (feedback) may mean a device that includes a database built based on actual field data and can respond according to a case requested by module A.

[0184] Field data may include static data and dynamic data, and may be classified into static data (type of vehicle tank (310), volume of vehicle tank (310), number of vehicle tank (310) modules / banks, etc.) and dynamic data (vehicle tank (310) hydrogen temperature, pressure, etc.).

[0185] A charging station (200) may be equipped with multiple hydrogen storage cylinders and operated as a bank system. Real-time on-site information requested and received from the charging station (200) by the dispenser (100) may include temperature and pressure information for each bank.

[0186] Multiple banks can be switched and connected to the dispenser (100) according to a request from the dispenser (100) and / or a selection from the charging station (200). At this time, temperature and pressure information for each bank included in the real-time field information that the dispenser (100) requests and receives feedback from the charging station (200) can influence the selection and / or switching of the banks.

[0187] At this time, the status of at least one bank in the bank system can be adjusted based on the temperature and pressure information for each bank included in the real-time field information that the dispenser (100) requests and receives feedback from the charging station (200).

[0188] For example, as a preparatory step, at least one or more banks within the charging station (200) may be adjusted to have a chargeable temperature and pressure based on current status information about the banks. Such adjustment may be performed at the request of the dispenser (100) or by the control logic of the charging station (200).

[0189] FIG. 5 is a conceptual diagram illustrating a platform or test platform for simulation used in a hydrogen fuel supply process according to one embodiment of the present invention.

[0190] Referring to FIG. 5, a mobility part (300a) for simulating actual mobility (300), a charging station part (200a) for simulating an actual charging station (200), and a dispenser part (100a) for simulating an actual dispenser (100) are illustrated.

[0191] In Fig. 5, the mobility part (300a), the charging station part (200a), and the dispenser part (100a) may be implemented by merging, combining, or competing between a simulation model and a data-based model based on real-time on-site dynamic data.

[0192] In order to analyze the process state change between the dispenser (100) and the charging station (200), a learning model based on field data between the dispenser (100) and the mobility (300) can serve as a reference for the interaction process between the dispenser part (100a) and the mobility part (300a).

[0193] Conversely, in order to analyze the process state change between the dispenser (100) and the mobility (300), a learning model based on field data between the dispenser (100) and the charging station (200) can serve as a reference for the interaction process between the dispenser part (100a) and the charging station part (200a).

[0194] In an alternative embodiment of the present invention, the integrated management system controlling the simulation of FIG. 5 may be implemented in the form of a cloud and / or remote server.

[0195] In addition, although an embodiment centered on an artificial neural network model (120) is illustrated in the present application specification, other embodiments of the present invention need not be limited by the artificial neural network model (120) or the model predictive control technique. In another embodiment of the present invention, after hydrogen fuel supply control-related information and / or feedback control based on a hydrogen fuel supply protocol is transmitted, real-time field data is fed back as a response, and the hydrogen fuel supply control-related information and / or feedback control may be updated based on the real-time field data, or the next hydrogen fuel supply control-related information and / or feedback control may be generated. According to another embodiment of the present invention, a hydrogen fuel supply protocol that does not rely on model predictive control or an artificial neural network model (120), and a test method and a test platform for testing a hydrogen fuel supply system may also be proposed.

[0196] When the charging station part (200a) functions as a simulation model, the mobility part (300a) can function as a modulation and optimization means for the temperature of the CHSS (310).

[0197] When the mobility part (300a) functions as a simulation model, the charging station part (200a) can function as a means for stabilizing and controlling the precooling temperature.

[0198] The integrated management system controlling the simulation of Fig. 5 can obtain the difference between the simulation results for changes in the state of hydrogen in the vehicle tank corresponding to hydrogen charging control-related information and field data for changes in the state of hydrogen in the vehicle tank.

[0199] An integrated management system controlling the simulation of FIG. 5 can update a model (120) for a hydrogen fueling process for a hydrogen vehicle corresponding to information related to hydrogen charging control based on differences between simulation results and field data.

[0200] An integrated management system controlling the simulation of FIG. 5 can obtain simulation results for changes in the state of hydrogen in a vehicle tank in response to a control request by using a thermodynamic model that tracks transient changes in at least one of temperature, pressure, and mass flow of hydrogen in a vehicle tank.

[0201] The integrated management system controlling the simulation of Fig. 5 can obtain simulation results for changes in the state of hydrogen in a vehicle tank in response to a hydrogen charging control request by inputting a future predicted hydrogen charging control request into a model for a hydrogen fueling process using a model prediction control (MPC) technique.

[0202] The integrated management system controlling the simulation of FIG. 5 can, after updating the model (120), replace the hydrogen fuel mobility (300) side with the model (120) to obtain at least one of simulation data and field data of a change in the state of hydrogen supplied to the dispenser (100) from the station (200) corresponding to the second control request between the dispenser (100) supplying hydrogen to the hydrogen fuel mobility (300) and the station (200) supplying hydrogen to the dispenser (100).

[0203] The integrated management system controlling the simulation of FIG. 5 can obtain at least one of simulation data and field data of changes in the state of hydrogen in a vehicle tank of a hydrogen fuel mobility (300) that receives hydrogen from a dispenser (100) and responds to a third control request between the hydrogen fuel mobility (300) and the dispenser (100) by replacing the station (200) side with the model (120) after updating the model (120).

[0204] Conventional hydrogen fueling processes or hydrogen charging control techniques have difficulties in controlling the final SOC, final nozzle, and CHSS (310) temperature / pressure as desired. Furthermore, theoretical simulation-based hydrogen charging techniques often fail to match actual field data due to pressure variability, unstable flow rates, and high environmental variability.

[0205] The discrepancy between simulation results and actual field data is easily affected by the characteristics of the device and the diversity of the environment, which are difficult to sufficiently consider in theoretical simulations. Even when using the same hydrogen charging protocol, the final field data may differ depending on the initial value or final target value. Conversely, even when assuming the same final target value or initial value, the final field data may differ depending on different hydrogen charging protocols.

[0206] To solve these problems of the prior art, one embodiment of the present invention may adopt utilization of real-time field data, two-way communication between each device, predictive control techniques, integrated control of the entire system including stations and vehicles, and utilization and standardization of hydrogen fuel supply data based on user needs.

[0207] Additionally, one embodiment of the present invention may enable enhancement of an existing hydrogen fueling protocol, standardization of on-site hydrogen fueling data format, and compilation and diagnosis of dynamic field data.

[0208] In relation to enhancing existing hydrogen charging protocols, one embodiment of the present invention may include the following:

[0209] Existing hydrogen charging protocols can be selected for testing.

[0210] One embodiment of the present invention executes artificial neural network-based model predictive control (ANN-MPC) and executes an existing hydrogen charging protocol under the same charging conditions, thereby comparing the results of charging control using ANN-MPC with the results of charging control using the existing hydrogen charging protocol, thereby strengthening or improving the existing hydrogen charging protocol.

[0211] The supervisory system (130) can embed a protocol (131) to be tested and can control charging based on the embedded protocol and compare the control value with the 'predicted output' input from the ANN-MPC.

[0212] The supervisory system (130) or controller can transmit hydrogen charging control related information to the hydrogen fuel mobility (300) via the communication interface (140, 160) by executing an existing hydrogen charging protocol, and can obtain field data on changes in the state of hydrogen in the vehicle tank via the communication interface (140, 160).

[0213] The supervisory system (130) or controller can obtain a hydrogen charging control sequence by a model (120) for a hydrogen fueling process (if using the ANN-MPC technique, future time series control sequence inputs can be predicted), and based on the hydrogen charging control sequence by the model (120) for a hydrogen fueling process, an existing hydrogen charging protocol can be enhanced.

[0214] The supervisory system (130) or controller can obtain result data by executing a hydrogen charging control sequence by a model for a hydrogen fueling process, and based on a comparison result between the result data by executing the hydrogen charging control sequence and field data by an existing hydrogen charging protocol, can strengthen a part of an existing hydrogen charging protocol using the hydrogen charging control sequence.

[0215] At this time, the charging control result using the artificial neural network model (120) can be obtained by operating the model (120) or can be obtained from a pre-built database. The basic specifications used in the charging protocol (131) to be tested are not changed, and the details, i.e., the charging table or logic, can be strengthened / improved by referring to the charging control values ​​of the artificial neural network model (120). At this time, the charging control result using the artificial neural network model (120) and the field data obtained as the result of executing the charging protocol (131) to be tested are compared, and if the performance of the charging control result using the artificial neural network model (120) is superior, the details of the charging protocol (131) to be tested can be partially improved. A comparison between the charging control result and the execution result of the charging protocol (131) to be tested can also be performed for all or part of the hydrogen fuel supply process.

[0216] In one embodiment of the present invention, a Lumped Thermodynamic Model for an artificial neural network for dispenser-vehicle interaction can be utilized by considering the following features.

[0217] - 0-Dimensional Unsteady State Mass & Energy Balance

[0218] - 1-Dimensional Heat Transfer for Vehicle Tank Wall

[0219] - CoolProp for Evaluation of Hydrogen Properties

[0220] One embodiment of the present invention can perform comparative analysis between theoretical simulation results and real on-site data.

[0221] One embodiment of the present invention can perform predictive analysis under specific conditions in addition to the conditions assumed in the hydrogen charging protocol.

[0222] FIG. 6 is a conceptual diagram illustrating a concept of modeling, predicting, or controlling a hydrogen fuel supply process using an artificial neural network (ANN) in a hydrogen fuel supply process according to one embodiment of the present invention.

[0223] Referring to Figure 6, measurement values ​​of the current state are input into the input layer.

[0224] At this time, the ambient temperature Tamb (ambient temperature), pre-cooling temperature Tpre (pre-cooled gas temperature), and pre-cooling pressure Tpre (pre-cooled gas pressure) can be measured at the nozzle of the dispenser (100) or the charging station (200).

[0225] Hydrogen gas temperature T CHSS , hydrogen gas pressure P CHSS is a value measured on the CHSS (310) side of hydrogen fuel mobility (300), and the actual measured value can be input to the input layer.

[0226] During the training process of an artificial neural network, the actual current measurement value is passed to the input layer, and the next measurement value is passed to the output layer, where it can be utilized as ground truth data in the learning process of the artificial neural network. At this time, the learning process of the artificial neural network can be a process of learning a function that can predict the next measurement value of the output layer based on a combination of input measurement values. The correlation between the data input to the input layer and the data provided to the output layer is learned, and through this, predictions using real dynamic fueling data along with theoretical results are possible.

[0227] In the inference or output process using an artificial neural network, the actually measured on-site measurement value is passed to the input layer, and a prediction value for the next measurement value can be obtained as an output by the operation of the artificial neural network.

[0228] The learning process of the artificial neural network used in the embodiment of the present invention may be either shallow learning or deep learning, and the artificial neural network may be a type of neural network that suits the purpose of the present invention among known neural networks.

[0229] The values ​​input through the input layer are passed to the output layer after going through the weight-based operation of the hidden layer.

[0230] The state values ​​(predicted values ​​for the next state) output by the output layer can be used to derive state-of-charge variables, for example, state-of-charge (SOC), using at least part of a thermodynamic model.

[0231] In the embodiment of the present invention, hybrid control combining a theoretical simulation model and an artificial neural network is also possible, so that a desired result can be achieved even through learning using a small amount of data, and performance suitable for the purpose of the present invention can be derived even through a lightweight artificial neural network.

[0232] The real-time pressure increase rate (PRR) or mass flow rate of compressed hydrogen (kg / s) [m_dot] derived from the feedback control process can affect the weights or parameters of the hidden layer of the artificial neural network.

[0233] The artificial neural network-based hydrogen charging technique of the present invention can improve the accuracy of model-based charging result predictions. By utilizing actual fuel supply data alongside theoretical simulation results, real-time measurements can be reflected, further improving the accuracy of prediction results.

[0234] While the control protocol of the prior art calculates and predicts results through simulations tailored to individual situations, the embodiment of the present invention differs in that it utilizes a process of improving accuracy through repeated training for various situations.

[0235] Due to these differences, in the embodiment of the present invention, as various theoretical values ​​and empirical results are added, the accuracy is gradually improved through updates, and even if a new fuel supply process using a new storage tank (310) configuration or a change in flow rate is introduced, the function can be updated in the model by adding actual data and training, so that it can be widely applied to various mobility fields.

[0236] As a hydrogen charging control technique for a hydrogen charging test for hydrogen fuel mobility (300) according to one embodiment of the present invention, model prediction control (MPC) may be used.

[0237] FIG. 7 is a conceptual diagram illustrating a concept of controlling a hydrogen fuel supply process to reach a target output using model predictive control in a hydrogen fuel supply process according to one embodiment of the present invention.

[0238] Referring to Figure 7, Model Predictive Control (MPC) is a control method that uses a model that simulates a process to predict outputs based on future inputs, then optimizes these outputs to derive control inputs. To ensure compliance with various boundary conditions during process control, the process response must remain within a defined range. Compared to other control methods, MPC can effectively ensure that the process response remains within a defined range.

[0239] MPC plans inputs for a prediction horizon to achieve a process response that approaches the target output. It then obtains the corresponding response through a prediction model, using only the control input as the control input. Once the plant receives output, it repeats the same process based on that output.

[0240] In an embodiment of the present invention, the accuracy of the hydrogen charging model is secured to a considerable level, and the future charging result is predicted from the hydrogen charging model and the current measurement value, and the temperature T of the hydrogen gas of CHSS (310) is determined based on the predicted value and the charging value. CHSS or pressure P of hydrogen gas CHSS The pressure ramp rate / rise rate (PRR) can be controlled in real time to reach the optimal filling target without violating constraints such as specific variables.

[0241] Referring to FIG. 7, in an embodiment of the present invention, an MPC-based control is used to calculate a future output value based on a current measurement value and a model's predicted value, and an operation parameter / variable can be adjusted so that the predicted future response moves to a setpoint (or target) in an optimal manner.

[0242] For example, at the current time i, n model predictions can be derived. These n model-based predictions form a prediction horizon.

[0243] Each model-based prediction, or prediction horizon, corresponds to a control horizon. That is, the n control commands / control actions required to achieve n model predictions can form a control horizon.

[0244] In practice, the first of the n model prediction and control actions derived at the current time i, the i+1 control action, can be transmitted to the system. As time passes, a new set of n model prediction and control actions are derived at the current time i+1, forming a new Prediction Horizon and Control Horizon, respectively.

[0245] The technique of controlling the system by expanding / moving the Horizon in this way is called MPC, and in the embodiment of the present invention, MPC-based control can be executed using measured values ​​and predicted values ​​for state information (state values) including the temperature and pressure of hydrogen gas of CHSS (310).

[0246] FIG. 8 is a flowchart illustrating the training and inference process of an artificial neural network when an artificial neural network is used for model prediction control in a hydrogen fuel supply process according to one embodiment of the present invention.

[0247] Referring to FIG. 8, a training process of an artificial neural network considering an artificial neural network-model predictive control (ANN-MPC) technique according to one embodiment of the present invention is illustrated.

[0248] In Fig. 8, we assume an artificial neural network that has learned the function of obtaining an MPC-based Prediction Horizon and a Control Horizon, and in particular, an artificial neural network that has learned the function of obtaining n future predictions and corresponding control commands so that the process of a future response reaching a set point is optimized by MPC.

[0249] Referring to FIGS. 6 to 8 together, in an embodiment of the present invention, a control system can be configured based on an ANN model (120), and a test platform system for real-time control based on model predictive control can be configured by ensuring the accuracy of the ANN model (120).

[0250] The test platform system for real-time control is a method of controlling the charging speed / pressure increase rate / pressure increase rate by predicting future charging results and comparing them with actual measured values. By separately setting restrictions, control time intervals, and sensitivity within the system logic, control can be achieved within optimal values.

[0251] First, optimal control is performed based on real-time data of the charging station (200) and hydrogen fuel mobility (300), but when a specific event situation occurs in the operation of the system, the precooling temperature of the precooler (210) and the cooling system of the hydrogen fuel mobility (300) can be directly controlled, thereby increasing the overall efficiency of the hydrogen fuel supply process.

[0252] Referring to Figure 8, a control process is initiated by receiving a specific SOCsp from a customer (t=0, S710). For example, the current SOC may be 50% and the SOCsp may be 85%.

[0253] SOC(t) is T CHSS (t), and P CHSS It is given as a function of (t), and this process can be performed based on a general dynamic model.

[0254] If the current SOC(t) is greater than or equal to SOCsp (S720), hydrogen charging can be stopped. If the current SOC(t) is less than SOCsp (S720), i=t is set, and Moving Horizon Prediction using an artificial neural network is performed (S730).

[0255] Step S730 can be performed by generating an MPC prediction using an artificial neural network (120), etc. In step S740, it can be determined whether the obtained n predictions are optimized predictions that meet the intended purpose.

[0256] If the obtained n predictions are optimized predictions, a control command PRR(t) is obtained based on the n predictions and control commands, and PRR(t) can be applied to the dispenser (100)-vehicle tank (310) (S750).

[0257] After that, increase the time t and obtain a new measurement value T CHSS (t), and P CHSS (t) is obtained and passed as input to step S720.

[0258] If the n predictions obtained in step S730 are not optimized predictions, step S730 may be performed again to obtain n new predictions and control commands.

[0259] In step S730 of FIG. 8, state prediction values ​​(T, P) that satisfy temperature and pressure limits for all arbitrary i,k can be generated.

[0260] Based on the current time i(=t), n state prediction values ​​and corresponding control commands can be derived.

[0261] Step S740 of FIG. 8 can be understood as a process of searching for a set of n predictions that minimizes a cost function indicating whether the final control objective, SOCsp, has been reached.

[0262] As an output of hydrogen fuel mobility (300), state measurement values ​​including temperature and pressure of CHSS (310) can be provided as feedback input to an artificial neural network model (120).

[0263] The state measurement values ​​including the temperature and pressure of the pre-cooled hydrogen gas as an output of the hydrogen charging station (200) can be provided as feedback input to the artificial neural network model (120).

[0264] The artificial neural network model (120) transmits the predicted output to the hydrogen fuel supply controller (110), and the Supervisory System (130) can input future inputs obtained through simulation or model-based prediction to the artificial neural network model (120) via module A (140).

[0265] The ANN-MPC-based control process is a control technique that utilizes simulation and actual measurement data together, and is a control technique that performs simulation at least partially using an artificial neural network model (120) and uses the prediction results in the control process.

[0266] An embodiment of the present invention aims to configure a real-time data-based integrated control protocol for hydrogen charging, and the system is implemented by utilizing various element technologies.

[0267] The protocol installed in the dispenser (100) can control the charging speed / pressure increase rate / pressure increase rate (PRR or [m_dot]) as output by using the data of the pre-cooled hydrogen gas provided from the charging station (200) and the data of the storage tank (310) provided from the hydrogen fuel mobility (300) as real-time input values ​​through the installed model.

[0268] When an event such as an external environmental change occurs, the pre-cooling temperature of the charging station (200) and the cooling system of the hydrogen fuel mobility (300) can be directly controlled to control the charging speed / pressure increase rate / pressure increase rate (PRR or [m_dot]) and process efficiency in general.

[0269] To complement the control protocol, the pre-cooling system / pre-cooler (210) of the hydrogen charging station (200) may be independently installed with its own cooling stabilization system.

[0270] In terms of temperature stabilization, the cooling stabilization system of the pre-cooler (210) can be independently controlled, and the target value of the control can be changed integrally in the protocol of the dispenser (100).

[0271] To improve the economic efficiency of the charging station (200) and to complement the functions of the integrated control protocol, the pre-cooler (210) may be provided with additional functions related to temperature stabilization.

[0272] The precooling temperature varies depending on the initial temperature and flow rate of hydrogen gas supplied to the precooler (210), and to compensate for this, a novel precooler structure for stabilizing the temperature is proposed as an embodiment of the present invention.

[0273] A pre-cooler (210) according to one embodiment of the present invention may include control logic for its own temperature control and linkage with a protocol.

[0274] A forced cooling system may be installed in the storage tank (310) of the hydrogen fuel mobility (300), and the charging speed may be improved by cooling some of the compression heat generated during hydrogen charging, and the operation / control of the forced cooling system of the storage tank (310) may also be involved in the protocol.

[0275] In one embodiment of the present invention, a temperature management function may be provided to the storage tank (310) of the hydrogen fuel mobility (300) to improve the hydrogen charging speed and complement the functions of the integrated control protocol.

[0276] In one embodiment of the present invention, the vehicle storage tank (310) may configure a system for self-cooling to increase the overall charging speed and improve the safety of hydrogen fuel mobility (300), and may include control logic for self-driving of the system and linkage with a protocol.

[0277] In one embodiment of the present invention, through such integrated control, not only can the current charging efficiency be improved, but preparation for the next charging can also be smoothly supported.

[0278] In the case of T40 where the precooling temperature of the pre-cooler (210) is set to -40°C, if the precooling temperature has achieved the target value but the outside temperature is higher than the set value and the temperature rise on the storage tank (310) side is greater than expected, a control signal or information on the current status can be transmitted to the hydrogen fuel mobility (300) / storage tank (310) side so that the self-cooling system of the storage tank (310) can be driven.

[0279] Conversely, if the precooling temperature of the pre-cooler (210) is set to -40°C but is judged to be excessive cooling when considering the external environment and actual data, the target value of the precooling temperature can be adjusted (for example, -35°C).

[0280] In cases where additional temperature control of the pre-cooling target temperature and the storage tank (310) side is required, control information or control commands can be transmitted from the dispenser (100) to both the hydrogen fuel mobility (300) and the charging station (200).

[0281] According to an embodiment of the present invention, the self-cooling system of the hydrogen fuel mobility (300) and the charging station (200) may be independently controlled, or may be controlled by transmitting a signal from the dispenser (100).

[0282] An integrated control method for hydrogen charging according to one embodiment of the present invention may further include a step of evaluating whether a measurement value of a current state satisfies a constraint condition.

[0283] The constraint may be that the temperature and pressure of the compressed hydrogen storage system on the hydrogen vehicle side do not exceed the limit temperature and limit pressure, respectively.

[0284] According to one embodiment of the present invention, it is possible to safely charge / supply hydrogen fuel while improving the efficiency of the hydrogen charging / supply process and improving the speed and real-time nature of the hydrogen charging / supply process.

[0285] According to one embodiment of the present invention, a test method and test platform capable of precisely modeling a hydrogen charging / supply process based on real-time on-site dynamic data can be implemented.

[0286] According to one embodiment of the present invention, a test method and a test platform can be implemented that provide a model capable of precisely controlling a hydrogen charging / supplying process by considering the difference between modeling and simulation results by a theoretical model and real-time field data, or by considering both modeling and simulation results and real-time field data.

[0287] According to one embodiment of the present invention, a test technique for real-time hydrogen charging control based on model predictive control (MPC) can be implemented.

[0288] According to one embodiment of the present invention, a test technique for control with improved accuracy in predicting hydrogen charging results based on an artificial neural network (ANN) model can be implemented.

[0289] By inputting a charge amount (e.g., SOC) directly set by the user, hydrogen can be charged / supplied until the target amount is reached through control logic based on real-time communication and calculation using a sequence such as that in FIG. 8.

[0290] Various variables such as SOC, target pressure, time, and temperature can be used to set the charge amount, and the charge speed can be controlled to increase efficiency.

[0291] In a hydrogen fuel supply method according to one embodiment of the present invention, a control level can be determined based on a communication level.

[0292] A method for fueling hydrogen in a hydrogen-fueled mobility according to one embodiment of the present invention may include: determining a communication level of an available communication protocol; determining a control level related to application of artificial neural network-based control or model predictive control to a hydrogen fueling process based on the communication level; and performing a hydrogen fueling process based on the determined control level.

[0293] In a hydrogen fuel supply method according to one embodiment of the present invention, a step (S1030) of performing a process of supplying hydrogen fuel based on a determined control level may include a step of predicting current state information of mobility based on an artificial neural network model into which previous state information is input when the available communication protocol is No Communication (when communication between mobility and dispenser is impossible) and the fuel supply protocol requires current state information of mobility for hydrogen fuel supply control; a step of providing the predicted current state information of mobility to the fuel supply protocol as information for hydrogen fuel supply control; and a step of performing a process of supplying hydrogen fuel based on a control parameter that the fuel supply protocol generates based on the current state information.

[0294] At this time, the charging station (200) can transmit ambient temperature and pre-cooled conditions to the dispenser (200). The mobility (300) is required to transmit CHSS conditions to the dispenser (200), but since a communication function is not provided, the dispenser (200) cannot refer to the dynamic data of the mobility (300).

[0295] At this time, the predicted temperature and pressure values ​​of the CHSS predicted by moving horizon prediction can be treated as the correct temperature and pressure values ​​of the CHSS and provided as inputs to an artificial neural network model. In this way, the hydrogen fueling process using the ANN-MPC method can be implemented even in situations of communication failure.

[0296] That is, at this time, the predicted values ​​of the temperature and pressure of the CHSS on the mobility (300) side can be estimated as the correct values ​​of the temperature and pressure of the CHSS, thereby operating the artificial neural network model. However, as described below, an embodiment can also be implemented in which the reliability or weight when estimating the predicted values ​​as the correct values ​​can be selectively adjusted rather than being assigned 100%.

[0297] At this time, SOCsp is the initially set target value, and control parameters for optimal fuel supply that comply with constraints may be required. Constraints include, for example, T CHSS The temperature shall be less than 85 [℃], P CHSS It can be set to be less than 87.5 [MPa], SOC to be up to 100 [%], etc.

[0298] In a hydrogen fuel supply method according to one embodiment of the present invention, the step of determining a communication level of an available communication protocol may determine the communication level based on whether communication is possible between mobility and a dispenser or a charging station that supplies hydrogen to the mobility in the available communication protocol, whether static data of the mobility, the dispenser, or the charging station can be collected, whether dynamic data of the mobility, the dispenser, or the charging station can be collected, reliability of the collected static data, reliability of the collected dynamic data, whether the collected static data can be used for a safety function, or whether the collected dynamic data can be used for a safety function.

[0299] In a hydrogen fuel supply method according to one embodiment of the present invention, the step of determining a control level may include determining a control level including a method of acquiring data on the mobility side when communication is impossible in an available communication protocol, a method of processing or predicting dynamic data on the mobility side required in a control process when communication is impossible in an available communication protocol, a method of collecting static data on the mobility side, a dispenser side that supplies hydrogen to the mobility, or a charging station side, a method of collecting dynamic data on the mobility side, the dispenser side, or the charging station side, a method of using static data on the mobility side, the dispenser side, or the charging station side for controlling hydrogen fuel supply, or a method of using dynamic data on the mobility side, the dispenser side, or the charging station side for controlling hydrogen fuel supply.

[0300] Level Description: Non-Communication Level 0: When the vehicle or charging station does not have communication equipment, or when communication is impossible due to a malfunction. Communication Level 1: Communication is possible, but is not used for safety functions and can be used for data collection. If risk mitigation measures due to data defects are provided, it can be used to improve the control function of the fuel supply process. Level 2: Communication is possible, and static data is reliable, so it can be used directly for safety functions. Dynamic data is the same as Level 1. Communication (ANN model basic method). Level 3: Communication is possible, and both static and dynamic data are reliable, so it can be used directly for safety and control functions related to fuel supply.

[0301] Another embodiment of a hydrogen fuel supply method according to the communication level of the present invention is disclosed in Table 1. Table 1 is an example of defining communication reliability levels differentially from 0 to 3 based on various situations that may occur, such as communication failure, failure, and lack of data safety / reliability. The communication levels in Table 1 can refer to the communication levels included in ISO 19885-1, which is currently under development.

[0302] Level of classification Simulation Basic information Charging conditions Relearning conditions Non-communication Level 0 Based on the total capacity of the vehicle storage tank and the basic information of the station, worst-case conditions are assumed. After checking the initial vehicle residual pressure and total capacity, charging is performed at a constant speed with the lowest flow rate as the worst-case conditions. Non-communication Analysis and selective reflection only of actual problem situations that occur during the charging process. Communication Level 1 Generate basic simulation conditions based on static data of vehicles and charging stations, but apply conservatively to individual static conditions, and apply conservative standards to dynamic conditions. Use the received dynamic data as a reference (for checking in case of an abnormal signal), and conservatively predict the temperature and pressure of the vehicle based on the temperature and pressure of the breakaway. Relearn based on conservatively operated dynamic data (breakaway standard), accumulate the received actual dynamic data, and derive correction factors through periodic comparative learning. Level 2 Apply the static data of vehicles and charging stations directly individually, but apply conservative standards to dynamic data. Use the received dynamic data as a reference (for checking in case of an abnormal signal), and conservatively predict the temperature and pressure of the vehicle based on the temperature and pressure of the breakaway. Use the conservatively operated dynamic data (breakaway standard) Baseline) Baseline relearning, accumulation of actual dynamic data received, and correction factor derivation through periodic comparative learning Level 3 Basic simulation using individual static and dynamic data of vehicles and charging stations Real-time control operation optimized for the situation based on real-time communication data All recorded fuel supply data base relearning

[0303] Table 2 shows the control levels that designate static and dynamic variables according to the reliability level of communication, the allowable range of use for unreliable data, and alternative variables. While some functions are limited at low communication levels, the operation of the fueling protocol is possible at each communication level. The ANN-MPC-based hydrogen fueling protocol proposed in one embodiment of the present invention corresponds to Level 3 defined in Tables 1 and 2, and it can be assumed that both static and dynamic data transmitted and received between the mobility (300) and the charging station (200) are reliable and can be directly utilized.

[0304] In a hydrogen fuel supply method according to one embodiment of the present invention, the step of determining a control level may be performed when a communication level of an available communication protocol allows two-way communication between mobility and a dispenser or charging station that supplies hydrogen to the mobility and can collect static data and dynamic data of the mobility, the dispenser, or the charging station, and the control level may be determined so as to use the static data and dynamic data of the mobility, the dispenser, or the charging station as control input data required by the fuel supply protocol and generate control parameters required by the fuel supply protocol.

[0305] That is, if the reliability of the collected data is limited depending on the communication level, each collected data in the hydrogen fuel supply control process can be used for limited purposes based on its reliability.

[0306] At this time, the communication level and corresponding control level may vary over time and with environmental changes. For example, if the communication level is 3 at the start of hydrogen fueling, and the control level is set accordingly, and hydrogen fueling is performed, the communication level may be lowered during the process due to deterioration of the communication environment or failure of some equipment.

[0307] At this point, the control level can be reset to match the lowered communication level. Once the communication level is restored, the control level can also be restored to match the restored communication level.

[0308] Initially, when the communication level is 3, CHSS dynamic data on the mobility (300) side is acquired through two-way communication, and an ANN-MPC-based hydrogen fuel supply protocol is provided. When the communication level is changed to 0, CHSS dynamic data on the mobility (300) side can be replaced using predicted values ​​thereafter. At this time, the recently collected data in the communication level 3 state can be reflected as the initial values ​​and the subsequent process can be performed.

[0309] Meanwhile, detailed embodiments of the present invention may be implemented as a control method for hydrogen fuel supply by combining the detailed embodiments with each other as long as they do not conflict with each other.

[0310] For example, in the case of non-communication between the mobility and the dispenser, the first embodiment can predict the current state information of the mobility based on an artificial neural network model into which previous state information is input.

[0311] In the case of no communication between the mobility and dispenser, the second embodiment may utilize the measurement data in a limited manner as shown in Tables 1 and 2, or may select a conventional fueling protocol that does not utilize the measurement data.

[0312] These two different embodiments may be implemented alternatively or complementarily depending on the circumstances.

[0313] In the case of no communication, only the first no communication embodiment may be adopted and the second no communication embodiment of Tables 1 and 2 may not be adopted. Conversely, in the case of no communication, only the second no communication embodiment of Tables 1 and 2 may be adopted and the first no communication embodiment may not be adopted.

[0314] Meanwhile, the first non-communication embodiment and the second non-communication embodiment of Tables 1 and 2 may be implemented complementarily. That is, as in the first embodiment, the state prediction value on the mobility (300) side may be utilized as if it were an actual state value, but as shown in the second embodiment of Tables 1 and 2, if the reliability is low, it may be utilized only in a limited manner, and thus the state prediction value may be assigned a low reliability or a low weight and may be utilized only for a limited purpose in the control process.

[0315] Meanwhile, a simulation model can also be implemented to implement control levels based on the communication level. In this case, the initial model reflects basic mobility and charging station information and sets simulation conditions. Later, the communication level information is reflected, and the basic model is retrained based on specialized information such as mobility and charging stations, thereby optimizing the artificial neural network model.

[0316] Confirmation / reconfirmation of communication level and setting / resetting of control level can be performed as follows.

[0317] ① When a vehicle enters a charging station, the communication method between the vehicle and the dispenser is verified. Depending on the applicable standard, the communication method can take various forms, including non-communication, IrDA, wired, and Bluetooth. However, the communication method does not determine the level of communication. Even if the communication method is verified, communication may be unavailable for various reasons, in which case it is considered non-communication.

[0318] ② Check the communication level according to the confirmed communication method, and select a model that matches the communication level among the specialized ANN models installed in the dispenser.

[0319] ③ After the protocol is determined, static data is transmitted for communication methods of Levels 1, 2, and 3, and then a preparatory process for charging begins. This process includes communication checks and various preparatory steps for hydrogen transport between the mobility vehicle and the charging station. For non-communication methods of Level 0, the static data transmission and reception process is omitted.

[0320] ④ The fuel supply process proceeds. In the case of non-communication methods, charging proceeds according to pre-determined rules. In the case of communication methods, a charging control function utilizing dynamic data is activated. At this time, in Lv. 1 and 2, the reliability of the received dynamic data is not guaranteed, so the ANN model's own thermodynamic model predicts the mobility temperature and pressure results. In the case of Lv. 3, the received dynamic data is directly utilized.

[0321] ⑤ Charging ends when the target value is reached. The non-communication method does not accumulate data for retraining the ANN model, except for information about the problem situation. Data from the communication method is sent to the ANN model for retraining, and the data itself is accumulated and utilized as big data.

[0322] The table-based method used in the existing SAE J2601 fuel supply protocol uses an APRR-based method that charges at a single rate, while the MC-formula method uses a PRR-based method that controls the charging rate by reflecting some real-time data at set time intervals.

[0323] The method proposed in one embodiment of the present invention can propose an optimization control method that complements the existing protocol as follows.

[0324] The primary protocol installed in the dispenser (100) can receive information necessary for operation from the mobility (300) and the charging station (200) and set an initial target value to supply hydrogen fuel.

[0325] An artificial neural network model (120) is optionally additionally installed in a dispenser (100) equipped with a primary protocol to improve speed, and all information transmitted to the primary protocol can be simultaneously transmitted to the artificial neural network model (120). This information may vary depending on the communication protocol or format of the mobility (300) and the charging station (200), and may include all static and dynamic data that can be transmitted.

[0326] Based on the information transmitted to the artificial neural network model (120), the temperature and pressure values ​​of the storage tank on the mobility (300) side after a certain time can be predicted, and the information calculated based on this can be transmitted to the primary protocol.

[0327] For example, when the primary protocol is a protocol that performs the entire hydrogen fueling process to achieve a target using a single control parameter, in one embodiment of the present invention, control assistance information may be provided to reset the control parameter by predicting the next state based on the current state for each time interval. In an alternative embodiment of the present invention, control sequences up to an intermediate target for reaching the final target rather than a time interval may be distinguished, and for each distinguished control sequence, control assistance information may be provided to reset the control parameter by predicting the state up to the next control sequence. Such embodiments of the present invention may be interpreted as secondary control or control assistance for the primary protocol.

[0328] If the primary protocol is a single-rate charging format such as the SAE J2601 table method that does not utilize dynamic data, the application table can be changed to a table that can reflect predicted and calculated results, or it can be directly reset with a modified target pressure and modified charging rate.

[0329] When the primary protocol utilizes some or all of the dynamic data for real-time control, the charging speed of the primary protocol can be reset by providing a predicted result value after an arbitrary time and suggesting a charging speed to achieve it.

[0330] In a hydrogen fuel supply method according to one embodiment of the present invention, the step of providing control assistance information may further include a step of inputting fuel supply data into a control assistance model so that the control assistance model relearns the fuel supply data.

[0331] The result data of fuel supply that has gone through the fuel supply optimization process of the artificial neural network model (120) is re-learned in the artificial neural network model (120) to improve the prediction accuracy of the artificial neural network model (120), and the re-learning process can be repeatedly performed using the data obtained in the fuel supply process for future mobility (300).

[0332] In a hydrogen fuel supply method according to another embodiment of the present invention, the step of providing control assistance information may include a step of obtaining a reset second fuel supply control parameter for a second hydrogen fuel supply process based on mobility, a dispenser that supplies hydrogen to the mobility, a charging station where the dispenser is placed, or an environmental factor affecting the hydrogen fuel supply process; and a step of providing the second fuel supply control parameter as control assistance information to the first fuel supply protocol.

[0333] At this time, if an artificial neural network model (120) is optionally added to a dispenser equipped with a primary protocol, and direct control of the control logic of the primary protocol is not possible due to structural or security factors, the speed optimization process can be performed indirectly by providing analyzed result data.

[0334] The dispenser is equipped with a primary protocol to utilize information on mobility and charging stations to perform charging, and the fuel supply data generated as a result of fuel supply is transferred to an artificial neural network model (120) to be accumulated, learned, and analyzed.

[0335] The artificial neural network model (120) can improve its reliability by learning from external data, and can simulate temperature and pressure changes in mobility fuel supply situations by utilizing a built-in thermodynamic model for its own predictive model. Therefore, the generated fuel supply data can be used to provide users with the following information.

[0336] Learning and analyzing fuel supply data can provide correction factors for manual control of the primary protocol in operation. Information related to optimizing fuel supply speed by charging station, mobility type, and season can be provided to users. Based on this information, users can manually control the protocol's fuel supply speed.

[0337] Since this embodiment of the present invention provides optimized control assistance information based on the specific conditions of each charging station (200), dispenser (100), or mobility (300) as the primary control information of the primary protocol, the control assistance information at this time can be understood as control reinforcement information.

[0338] In a hydrogen fuel supply method according to another embodiment of the present invention, the step of providing control assistance information may include a step of obtaining control information on the temperature of a hydrogen storage tank of a charging station where a dispenser for supplying hydrogen to mobility is arranged, the pressure of the hydrogen storage tank of the charging station, or the precooling temperature for the hydrogen storage tank of the charging station; and a step of providing the obtained control information to the charging station.

[0339] In this case, in the case of the general fuel supply protocol, strict and conservative standards are applied to the temperature and pressure increase to ensure safety even in the worst case scenario, and accordingly, in many cases, the temperature and pressure of the mobility storage tank are below the predicted value even after the fuel supply is terminated. In this regard, it is possible to change the fuel supply speed, but if the current fuel supply speed does not pose any particular problem to the consumer, or if the fuel supply protocol installed in the dispenser is not controllable, the operational efficiency can be improved by adjusting the precooling temperature of the charging station (200) or the storage combination and size of the storage tank of the charging station. In the case of the precooling temperature, the efficiency of the charging station can be improved by lowering the load of the cooler below the allowable temperature range of the mobility storage tank by predicting it using an artificial neural network model (120). Storage tanks are mainly divided into low-pressure, medium-pressure, and high-pressure tanks and operated. When a large amount of fuel supply data is analyzed with an artificial neural network model (120), an optimal point for the type of storage tank utilized and the pressure application configuration can be suggested, thereby increasing the efficiency of the charging station operation. The step of providing control assistance information at this time can be understood as an embodiment of optimizing the process for achieving the final control goal through indirect control or indirect control assistance, such as transmitting a pre-cooling command to the charging station (20) side without being able to directly participate in the control process.

[0340] FIG. 9 is a conceptual diagram illustrating a process of implementing and providing a specialized artificial neural network model for a hydrogen fuel supply process according to one embodiment of the present invention.

[0341] A device for providing a specialized artificial neural network model for a process of fueling hydrogen fueled mobility according to one embodiment of the present invention may include a memory storing at least one command; a processor executing at least one command; and a charging station, dispenser, or artificial neural network model specialized for each mobility, the charging station electronically communicating with the processor and fueling hydrogen fueled mobility.

[0342] A device providing a specialized artificial neural network model according to one embodiment of the present invention may have a structure similar to the computing system (3000) illustrated in FIG. 12, which will be described later.

[0343] A device providing a specialized artificial neural network model according to one embodiment of the present invention may be implemented as a local server or a cloud server (1000), or may be implemented in the form of an on-device on the mobility (1300) or charging station / dispenser (1100) side.

[0344] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain identification information about a charging station / dispenser (1100) or a mobility (1300) by at least one command (S2110, S2130, S2150). Based on the identification information, the processor can provide a specialized artificial neural network model (S2230, S2250, S2260) so that the specialized artificial neural network model obtains control or prediction information related to a process (S2400) of supplying hydrogen to the mobility (1300).

[0345] At this time, the identification information is not limited to identifiers in the general sense. Identification information may include the fuel supply profile, the hydrogen storage capacity of the charging station / dispenser / mobility (including the type or condition of the hydrogen storage container), the fuel supply capacity of the charging station / dispenser, functions related to fuel supply, the size of the dispenser / mobility (including the class and model of the mobility), the brand, etc.

[0346] The specialized artificial neural network model may ultimately be provided to the charging station / dispenser (1100) (S2230, S2260). However, the specialized artificial neural network model does not necessarily have to be placed within the charging station / dispenser (1100). In an alternative embodiment of the present invention, the specialized artificial neural network model may be capable of electronic communication with the charging station / dispenser (1100) so as to influence the control of the fuel supply process of the charging station / dispenser (1100).

[0347] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can train a basic model of the specialized artificial neural network model (S2500). The basic model of the specialized artificial neural network model can be trained to learn a function of generating control or prediction information related to the process of supplying hydrogen to the mobility (1300) by using measurement data of a charging station, dispenser, or mobility (1300) obtained during the process of supplying hydrogen to the mobility (1300) (S2440) (S2500).

[0348] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can store a model in which a basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility (1300).

[0349] The processor of the device providing the specialized artificial neural network model according to one embodiment of the present invention can provide the updated and stored specialized artificial neural network model in response to a subsequent request from a new mobility or charging station / dispenser (1100) (based on new identification information).

[0350] That is, through the learning process of a specialized artificial neural network model, an updated model can be implemented by starting from a basic model and learning actual measured data for each specialized condition (S2500).

[0351] The processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain a basic model from a cloud server or a local server (1000). The processor can feed back the updated and stored specialized artificial neural network model to the cloud server or local server (1000).

[0352] That is, when the device providing the specialized artificial neural network model is mobility (1300) or a charging station / dispenser (1100), the device can obtain a basic model from a cloud server or local server (1000) and feed back an updated model through learning using actual measured data to the cloud server or local server (see FIG. 11 described later).

[0353] The processor of the device providing the specialized artificial neural network model according to one embodiment of the present invention can obtain a new basic model implemented using federated learning from the cloud server or local server (1000) after feeding back the specialized artificial neural network model to the cloud server or local server (1000).

[0354] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can provide a specialized artificial neural network model based on identification information so that control or prediction information can be used to control or predict a process in which a dispenser (1100) supplies hydrogen to a mobility (1300) (S2230, S2260).

[0355] In one embodiment of the present invention, when the device providing the specialized artificial neural network model is a cloud server or a local server (1000), the specialized artificial neural network model can be provided to the charging station / dispenser (1100) so that it can be used to control or predict the process of supplying hydrogen fuel to the charging station or dispenser (S2230). In this case, the specialized artificial neural network model can be provided to the charging station / dispenser (1100) via mobility (1300) (S2250) (S2260).

[0356] In an alternative embodiment, if the device providing the specialized artificial neural network model is a mobility (1300), the mobility (1300) can receive the specialized artificial neural network model from a cloud server or local server (1000) (S2250). In addition, the mobility (1300) can provide the specialized artificial neural network model to the charging station or dispenser (1100) so that the specialized artificial neural network model can be used to control or predict the process of supplying hydrogen fuel to the charging station or dispenser (1000) (S2260).

[0357] As an alternative embodiment, when a device providing a specialized artificial neural network model is placed on the side of a charging station or dispenser (1100), the specialized artificial neural network model can be provided or applied to a controller or control part on the side of the dispenser so that it can be used to control or predict the process of supplying hydrogen fuel (S2300).

[0358] A processor of a device providing a specialized artificial neural network model according to one embodiment of the present invention can provide a specialized artificial neural network model based on a fuel supply protocol corresponding to a process (S2400) in which a dispenser supplies hydrogen to a mobility (1300) or a communication protocol between the mobility (1300) and the dispenser.

[0359] That is, the availability of provision and / or utilization of specialized artificial neural network models may be related to the fuel supply protocol and / or communication protocol. Interoperability or compatibility between dispensers and mobility (1300) may be considered, and among the functions supported by the compatible fuel supply protocol and / or communication protocol between dispensers and mobility (1300), whether bidirectional communication between dispensers and mobility (1300) is possible, and / or whether Advanced Control technology using artificial neural network models can be applied, etc. may be considered.

[0360] A method for providing a specialized artificial neural network model according to one embodiment of the present invention may include steps (S2230, S2250, S2260) of obtaining an artificial neural network model specialized for each charging station / dispenser (1100) or mobility (1300) that fuels hydrogen fueled mobility (1300); steps (S2110, S2130, S2150) of obtaining identification information for the charging station / dispenser (1100) or mobility (1300); and steps (S2230, S2250, S2260) of providing a specialized artificial neural network model so that the specialized artificial neural network model obtains control or prediction information related to a process (S2400) of fueling hydrogen to the mobility (1300) based on the identification information.

[0361] The basic prediction model of the artificial neural network model is implemented based on data through theoretical simulation, and the reliability of the specialized artificial neural network model can be improved and refined by additionally learning actual measured data (S2440).

[0362] In the simulation stage, information such as the combination and capacity of the actual mobility storage tank can be used, and in the training and update stage (S2500) to improve reliability, data actually measured (S2440) in the process of supplying fuel to the mobility (1300) can be used.

[0363] For mobility vehicles with known storage tank capacity or specifications, an artificial neural network model can be trained using actual measured data and fueled using an optimized protocol. However, if the artificial neural network model from the previous vehicle is directly applied to a new mobility vehicle, the fundamental fuel supply profile may differ, leading to errors in the fueling process and significant limitations in the fueling speed.

[0364] Accordingly, a local server or cloud server (1000) can receive fuel supply profiles for each mobility type from a mobility manufacturer server (1400) (S2010), and use these to train an artificial neural network model. The initial model can be updated to reflect actually measured data (S2020).

[0365] In the future, hydrogen fuel cell vehicles with various storage tank capacities and combinations of storage tanks will be manufactured by various manufacturers. As each vehicle will have different characteristics, such as temperature changes during the fueling process, the development of specialized artificial neural network models for each type and version of the vehicle will be required.

[0366] According to one embodiment of the present invention, the fuel supply speed can be maximized by controlling the hydrogen fuel supply process using an artificial neural network model specialized according to the type and version of mobility, etc.

[0367] In the case of hydrogen charging stations, they are currently operated at 60g / s for fuel supply, but depending on the standard development and component development process, they may support maximum capacities such as 120g / s, 200g / s, and 300g / s, or even select maximum flow limits depending on the equipment status or external environment within the same charging station, or operate in various combinations with multiple dispensers. Therefore, a method that can reflect this in an artificial neural network model is required. According to one embodiment of the present invention, in response to this need, an optimized / specialized artificial neural network model can be provided based on various equipment characteristics and information of a hydrogen charging station.

[0368] A hydrogen fuel supply method or a hydrogen fuel supply control method according to one embodiment of the present invention can implement, store, manage, and utilize different specialized artificial neural network models according to the maximum flow rate, pre-cooling temperature, etc. (at the time of supply) of a hydrogen charging station, and various hydrogen storage tank specifications according to the type and version of mobility.

[0369] In a hydrogen fuel supply method or a hydrogen fuel supply control method according to one embodiment of the present invention, the basic configuration and logic of a specialized artificial neural network model implemented for each charging station specification and mobility type can be standardized to be identical.

[0370] That is, a common basic model can be provided before optimization / specialization learning. At this time, a common model obtained as a result of fuel supply simulation faithful to the basic theory before optimization / specialization learning is developed, and the common model is subdivided and specialized according to the detailed conditions of individual mobility / charging station / station, so that a specialized artificial neural network model can be implemented. For convenience of explanation, the terms common model and specialized model are described, but not all models are always either a common model or a specialized model. A model currently defined as a specialized model may be treated as a common model when the type, version, fuel supply function, or characteristics of mobility are subdivided in the future, and a subdivided specialized model can be derived from this common model.

[0371] An individual specialized model trained using simulation results and / or actual measured data optimized for each type of mobility and charging station conditions can be provided as a specialized artificial neural network model to a mobility (1300) or charging station / dispenser (1100) (S2230, S2250, S2260).

[0372] According to one embodiment of the present invention, when a vehicle enters a charging station, an individual specialized artificial neural network model is selected according to the type and version of the vehicle, and after fuel supply, the individual specialized artificial neural network model retrains actual measurement data obtained during the fuel supply process to improve reliability.

[0373] According to one embodiment of the present invention, as individual information that influences the determination of a specialized artificial neural network model related to hydrogen fuel supply of mobility, primary classification information according to the type of mobility, information on details of hydrogen storage tanks related to hydrogen fuel supply, other facilities, information on fuel supply protocols, and information on preferred fuel supply protocols, etc., may be classified within the "fuel supply profile" category and stored on the mobility side since components within the mobility may be different even if the mobility is of the same type. Such individual information may be used as "identification information" to determine a specialized artificial neural network model.

[0374] Referring again to FIG. 9, in one embodiment of the present invention, a local server or cloud server (1000) can, through cooperation with a mobility manufacturer server (1400), identify individual conditions such as maximum flow rate and pre-cooling temperature conditions when supplying hydrogen at a charging station, including basic fuel supply profiles such as characteristics according to various hydrogen storage tank configurations, capacities, and loading structures for each type of mobility.

[0375] ② Although not illustrated in Figure 9, the local server or cloud server (1000) may include a simulation module and a learning module. The information in ① may be transmitted to the simulation module (S2010). The simulation module sets simulation evaluation items appropriate for the type of mobility and the characteristics of the charging station, and can automatically generate various theoretical simulation results based on the set criteria.

[0376] ③ The generated theoretical data is transmitted to an artificial neural network learning module within a local server or cloud server (1000). A specialized artificial neural network basic model specific to the mobility type can be implemented through training using a portion of the transmitted data. The remaining portion of the transmitted data can be utilized to evaluate the suitability of the training.

[0377] The amount of data used when implementing the specialized artificial neural network basic model can be adjusted by the user depending on the type of mobility applied or the characteristics of the charging station.

[0378] The base model can be retrained and updated based on simulation data or by reflecting actual measurement results obtained in a test environment. Information from the updated model can then be shared with the mobility manufacturer server (1400) (S2020).

[0379] ④ The generated basic model is installed in the dispenser (1100) and applied to the fuel supply process (S2400) (S2300), and the basic model can be updated (S2500) by actual measurement data obtained (S2440) as a result of the fuel supply process (S2400).

[0380] At this time, the data actually measured (S2440) can be shared by at least one of the mobility manufacturer server (1400), local server or cloud server (1000), mobility (1300), or charging station / dispenser (1100).

[0381] In one embodiment of the present invention, data is shared with a local server or a cloud server (1000), and the data is transmitted to an artificial neural network learning module within the local server or the cloud server (1000), and an artificial neural network model specialized for each detailed condition can be retrained, and an existing model can be further refined to create a new specialized model according to the situation.

[0382] In an alternative embodiment of the present invention, a portion of the retraining and updating process (S2500) of a specialized artificial neural network model may be performed on the mobility (1300) or charging station / dispenser (1100) side, and weights / parameters or changes in weights / parameters for the updated model may be fed back to a local server or cloud server (1000). This process may partially borrow from known technologies, and for example, technologies such as federated learning may be utilized.

[0383] In this way, through the process of retraining and updating an artificial neural network model using actual measurement data, an artificial neural network model can be specialized for each individual condition, and the reliability and stability of the specialized artificial neural network model can be improved.

[0384] As described above, the local server or cloud server (1000) may include a simulation module that generates theoretical data and an artificial neural network learning module that manages learning of an artificial neural network model.

[0385] Both the simulation module and the artificial neural network learning module are implemented in the form of a software package, and can be operated through a hardware configuration such as that shown in Fig. 12, which will be described later, during actual operation.

[0386] When implementing the initial model of an artificial neural network, users can set the amount of data and the required level of reliability. They can also select the amount of data and the required level of reliability during the retraining process. Users can also be mobility manufacturers, charging station equipment manufacturers, or government agencies, academic institutions, or private associations that manage related matters.

[0387] Referring to FIG. 9, in an alternative embodiment of the present invention, the roles of implementing, storing, managing, and applying artificial neural network models specialized for each actor illustrated in FIG. 9 can be defined.

[0388] ① Dispenser (1100): The artificial neural network model can be installed in the dispenser (1100) or placed in a close relationship so as to influence the fuel supply control part of the dispenser (1100).

[0389] - The dispenser manufacturer can receive a specialized artificial neural network basic model from the mobility manufacturer, or can implement an artificial neural network model on its own based on the fuel supply profile information of each type of mobility and load it into the dispenser (1100).

[0390] - Even if the specialized artificial neural network basic model is stored in a mobility or external organization, it can be received, stored, and used in the dispenser (1100).

[0391] - All models and data related to the operation of the charging station, as well as information related to the artificial neural network model, can be stored and utilized in bulk in the dispenser (1100).

[0392] ② Individual mobility (1300)

[0393] - The specialized artificial neural network basic model implemented by the mobility manufacturer is stored in the storage within the mobility (1300), and when the mobility (1300) enters a charging station, the basic model can be transmitted to the dispenser (1100).

[0394] ③ Charging station operator

[0395] - The operator who manages the entire charging station, including the dispenser, can create a specialized artificial neural network basic model or receive and manage a model from a mobility manufacturer or dispenser manufacturer.

[0396] - It can store and accumulate all data related to charging stations as well as the specialized artificial neural network basic model.

[0397] ④ Third-party organizations (government agencies, private organizations)

[0398] Third-party stakeholders, excluding mobility manufacturers, dispenser manufacturers, and charging station operators, can build and store specialized artificial neural network models. These entities can be government agencies, associations, or for-profit or non-profit private companies.

[0399] With the consent and authorization of stakeholders, information related to the specialized artificial neural network basic model can be stored in bulk on any server (e.g., a cloud server). Mobility manufacturers, dispensers, and charging station operators can periodically update all generated information to any server.

[0400] Referring to FIG. 9, in an alternative embodiment of the present invention, each actor can participate in the process of updating and utilizing a stored artificial neural network model as follows.

[0401] - The specialized artificial neural network basic model is used to fuel individual mobility at a charging station / dispenser (1100), and the model can be updated to improve the reliability of the artificial neural network model through retraining using the generated data.

[0402] - The subject of retraining and model update may be a charging station / dispenser (1100), a local server, or a cloud server (1000), and mobility (1300) may partially perform the role.

[0403] If retraining and model updates are proactively performed at the charging station / dispenser (1100), the dispenser manufacturer responsible for operating the dispenser (1100) or a delegated charging station operator may participate in the retraining and model updates. Furthermore, the mobility manufacturer or a third party may assume the same role. Individual operating scenarios may be exemplified as follows:

[0404] ① Dispenser: The main entity that utilizes and maintains the basic storage model.

[0405] - Perform actual hydrogen fuel supply to mobility based on a specialized artificial neural network basic model provided by or implemented directly by mobility, and self-relearn and update the model based on the data obtained at this time.

[0406] - By synthesizing the operation data of multiple charging stations with the same dispenser installed, the model can be optimized to optimize the fuel supply speed, and at the same time, the accumulation of multiple data can be utilized to analyze problems related to dispenser operation and optimize operation.

[0407] ② Individual mobility

[0408] - If a basic artificial neural network model is installed in the mobility, it is transmitted to the dispenser to perform fuel supply, and the obtained data can be stored in the mobility manufacturer's central server (1400) through the mobility management remote system, and the model can be updated by performing retraining at the same location.

[0409] - The updated model is then transmitted back to the mobility, and the artificial neural network model can be updated by converging all data from the same type of mobility operating in the same area.

[0410] - If mobility manufacturers directly manage their fuel supply models and optimize them to improve speed, they may benefit from access to detailed information about their mobility.

[0411] - Mobility or mobility manufacturers can directly manage data, enabling them to identify failure rates and problems with mobility components.

[0412] ③ Charging station operator

[0413] - The entire process can be managed by the operator who manages the entire charging station, including the dispenser, and relearning and updating are possible using data from the entire charging station, including mobility and dispensers.

[0414] - Accumulated data can be used to analyze problems and optimize operations not only for individual charging stations but also for all charging stations operated by the charging station operator.

[0415] ④ Third-party organizations (government, private)

[0416] - The third party may be government or private, but since the rights and responsibilities of multiple stakeholders coexist, it may be most efficient to operate as an individual country's standards and safety-related agency, organization, or association that can integrate them.

[0417] - Fuel supply models and generated data related to mobility, dispensers, and charging stations can be collectively collated and utilized at the regional and national level, and upgrades can be carried out for individual charging stations in bulk.

[0418] - The generated and collected data can be used to optimize artificial neural network models, but can also be used to derive failure rates for individual charging stations and mobility components, resolve problems, and derive optimal operating conditions for charging stations.

[0419] Referring to FIG. 9, in an alternative embodiment of the present invention, an artificial neural network utilization scenario involving actors illustrated in FIG. 9 can be exemplified as follows.

[0420] ① When an artificial neural network model is installed and operated in mobility

[0421] - Mobility manufacturers can implement a basic model by inputting fuel supply profiles for each type of mobility into artificial neural network learning modules within their own servers.

[0422] The basic model of the artificial neural network can be transmitted to hydrogen fuel mobility, which can then store it, and the stored basic model can be transmitted to the dispenser via a hydrogen charging station.

[0423] The dispenser receives an artificial neural network model from the mobility device and utilizes it to perform fueling. The resulting data can be fed back from the hydrogen fueled mobility device to the mobility manufacturer. The mobility manufacturer, upon receiving the fueling data, can utilize its own artificial neural network model learning module to retrain the basic artificial neural network model and update the model. This updated model can then be transmitted to the same type of hydrogen fueled mobility device.

[0424] When a vehicle of the same type re-enters the hydrogen charging station, the updated model held by the vehicle can be transmitted to the dispenser (S2260). The dispenser can then use the received model to supply fuel. As this process is repeated, retraining using actual fuel supply data is performed, gradually improving the reliability of the artificial neural network model.

[0425] - Mobility manufacturers can continuously update their artificial neural network models to optimize fuel delivery speeds by leveraging fuel supply data for the same type of mobility operating in each region or country.

[0426] ② When an artificial neural network model is installed and operated in the dispenser

[0427] - Dispenser manufacturers can operate their own artificial neural network model development tools within the dispenser, or integrate them through the manufacturer's central server for data management and maintenance.

[0428] Dispenser manufacturers can obtain individual fuel supply profiles for each type of mobility expected to use hydrogen charging stations from mobility manufacturers. Based on this information, dispenser manufacturers can leverage their own proprietary neural network model learning modules to create a basic neural network model.

[0429] The basic model is transmitted to the dispenser, and after the actual hydrogen fuel supply process is performed in the dispenser, the resulting fuel supply data can be obtained (S2440).

[0430] - Actual measured data can be fed back from the dispenser to the dispenser manufacturer, and the dispenser or dispenser manufacturer server can independently or cooperatively perform retraining of the artificial neural network model, and the artificial neural network model can be updated.

[0431] The updated model can be used again in the actual hydrogen fueling process, and the data obtained again can be used to retrain and update the model, thereby improving the reliability and stability of the artificial neural network model.

[0432] - Dispenser manufacturers can optimize artificial neural network models for individual charging stations, comprehensively manage data and models by region or country, and optimize fuel supply speed by performing model updates tailored to regional, national, and environmental characteristics.

[0433] FIG. 10 is a flowchart illustrating a hydrogen fuel supply method and / or a control method for a hydrogen fuel supply process according to one embodiment of the present invention.

[0434] A method according to one embodiment of the present invention can be performed by a device that controls a process of supplying hydrogen as fuel to hydrogen fueled mobility according to one embodiment of the present invention.

[0435] A method according to one embodiment of the present invention comprises a device for controlling a process of supplying hydrogen to a hydrogen-fueled mobility according to one embodiment of the present invention, the device including: a memory for storing at least one command; a processor for executing at least one command; and a charging station, a dispenser, or an artificial neural network model specialized for each mobility for supplying hydrogen to the mobility.

[0436] A device for controlling a process of supplying hydrogen fuel to a hydrogen-fueled mobility according to one embodiment of the present invention may have a structure similar to the computing system (3000) illustrated in FIG. 12, which will be described later.

[0437] A hydrogen fuel supply control device according to one embodiment of the present invention may be disposed in a dispenser that supplies hydrogen to a mobility vehicle, or may be implemented as a controller that is not disposed in the dispenser but is capable of electronically communicating with the dispenser and influencing fueling of the dispenser.

[0438] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can obtain control or prediction information related to a process of supplying hydrogen fuel to a mobility from a specialized artificial neural network model determined based on identification information about a charging station, a dispenser, or a mobility, by at least one command (S2300).

[0439] At this time, the specialized artificial neural network model stored in the dispenser to control the process of supplying hydrogen fuel may be an artificial neural network model specialized for each mobility, or may be an artificial neural network model specialized for each charging station or dispenser.

[0440] A device for controlling a hydrogen fuel supply process according to one embodiment of the present invention may include a plurality of artificial neural network model candidates specialized for a plurality of types of mobility.

[0441] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may select one of a plurality of artificial neural network model candidates as a specialized artificial neural network model based on first identification information about mobility.

[0442] That is, in the first embodiment where the dispenser acquires / stores / implements / manages a specialized artificial neural network model, the dispenser can select an optimal model based on the identification information of the mobility from a state where multiple specialized model candidates have been stored in advance.

[0443] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can transmit first identification information about mobility and second identification information about a charging station or dispenser to a cloud or local server, and can receive a specialized artificial neural network model from the cloud or local server based on the first identification information and the second identification information.

[0444] That is, in the second embodiment in which the dispenser acquires / stores / implements / manages a specialized artificial neural network model, the dispenser acquires first identification information of the mobility, and can receive an optimal specialized artificial neural network model from a cloud server or a local server based on a combination of mobility-dispenser identification information.

[0445] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may receive a specialized artificial neural network model from the mobility based on first identification information about the mobility.

[0446] That is, in the third embodiment where the dispenser acquires / stores / implements / manages a specialized artificial neural network model, the dispenser can receive a specialized artificial neural network model from the mobility based on the first identification information of the mobility. At this time, the specialized artificial neural network model can be selected based on not only the first identification information of the mobility but also the second identification information of the charging station / dispenser. The specialized artificial neural network model can be provided to the dispenser from a cloud server or a local server via the mobility. The second identification information of the charging station / dispenser can be directly transmitted from the charging station / dispenser to the cloud server or the local server, or can be transmitted to the cloud server or the local server via the mobility.

[0447] A processor of a device controlling a hydrogen fueling process according to one embodiment of the present invention can acquire a basic model of a specialized artificial neural network model. The processor can train the basic model to learn the function of generating control or predictive information related to the process of hydrogen fueling a vehicle using measurement data from a charging station, dispenser, or vehicle obtained during the process of hydrogen fueling a vehicle.

[0448] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may store a model in which a basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility.

[0449] That is, the dispenser can receive a basic model from a cloud server, a local server, or mobility, and store / manage an updated model by training it based on actual data from the fuel supply process.

[0450] The processor of a device controlling a hydrogen fueling process according to one embodiment of the present invention can obtain a basic model from a cloud or local server. The processor can then feed back an updated and stored specialized artificial neural network model to the cloud or local server.

[0451] A processor of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can obtain control or prediction information related to a process of supplying hydrogen fuel to a mobility using a specialized artificial neural network model determined based on a fuel supply protocol corresponding to a process of supplying hydrogen fuel to a mobility by a dispenser or a communication protocol between the mobility and the dispenser.

[0452] That is, the availability and / or utilization of specialized artificial neural network models may be related to fuel supply protocols and / or communication protocols. Interoperability or compatibility between dispensers and mobility may be considered, and among the functions supported by compatible fuel supply protocols and / or communication protocols between dispensers and mobility, the availability of bidirectional communication between dispensers and mobility, and / or the applicability of advanced control technologies utilizing artificial neural network models may be considered.

[0453] Referring again to FIGS. 9 and 10, a method for fueling hydrogen fueled mobility according to an embodiment of the present invention may include a step (S2200) of obtaining an artificial neural network model specialized for a charging station, dispenser, or mobility that fuels hydrogen to the mobility; and a step (S2300) of obtaining control or prediction information related to a process of fueling hydrogen to the mobility from the specialized artificial neural network model determined based on identification information about the charging station, dispenser, or mobility.

[0454] According to one embodiment of the present invention, a method for supplying hydrogen fuel to a mobility that uses hydrogen as a fuel may further include a step (S2400) of supplying hydrogen fuel to the mobility using control or prediction information.

[0455] The step of obtaining a specialized artificial neural network model (S2200) may include a step of selecting one of a plurality of artificial neural network model candidates specialized for each of a plurality of mobility types as the specialized artificial neural network model based on first identification information about the mobility.

[0456] The step of obtaining a specialized artificial neural network model (S2200) may include a step of transmitting first identification information about the mobility and second identification information about the charging station or dispenser to a cloud or local server (S2130); and a step of receiving a specialized artificial neural network model from the cloud or local server based on the first identification information and the second identification information (S2230). In this case, depending on the embodiment, the specialized artificial neural network model may be transmitted to the dispenser via the mobility (S2250, S2260).

[0457] The step of obtaining a specialized artificial neural network model (S2200) may include a step of receiving a specialized artificial neural network model from the mobility based on first identification information about the mobility (S2260).

[0458] Figure 11 is a flowchart illustrating a process in which a specialized artificial neural network model is trained and updated based on the measurement data of Figure 9.

[0459] Referring to FIG. 11, a method for supplying hydrogen to hydrogen fuel mobility according to an embodiment of the present invention may further include a step (S2520) of setting a current specialized artificial neural network model used in a step (S2300) of obtaining control or prediction information as a basic model; a step (S2540) of training the basic model to learn a function of generating control or prediction information related to a process of supplying hydrogen to mobility using measurement data of a charging station, dispenser, or mobility obtained (S2440) in a process of supplying hydrogen to mobility; and a step (S2560) of storing or managing a model in which the basic model is updated through training as a new specialized artificial neural network model.

[0460] A method for supplying hydrogen to hydrogen fuel mobility according to one embodiment of the present invention may further include a step of obtaining a basic model from a cloud or local server (S2230, S2250, S2260); and a step of feeding back a new specialized artificial neural network model to the cloud or local server (S2580).

[0461] The step of obtaining a specialized artificial neural network model (S2200) may include a step of obtaining a specialized artificial neural network model determined based on a fuel supply protocol corresponding to a process in which the dispenser supplies hydrogen to the mobility or a communication protocol between the mobility and the dispenser.

[0462] FIG. 12 is a conceptual diagram illustrating an example of a generalized hydrogen fuel supply control device, hydrogen fuel supply control system, hydrogen fuel supply simulation device, hydrogen fuel supply simulation system, hydrogen fuel supply test platform, hydrogen fuel supply test system, device or computing system that provides a specialized artificial neural network model for a hydrogen fuel supply process, capable of performing at least a part of the processes of FIGS. 1 to 11.

[0463] Referring to FIG. 12, a computing system (3000) according to one embodiment of the present invention may be configured to include a processor (3100), a memory (3200), a communication interface (3300), a storage device (3400), an input interface (3500), an output interface (3600), and a bus (3700).

[0464] Additionally, the computing system (3000) may further include an artificial neural network model (3800) capable of electronically communicating with at least one of a processor (3100), a memory (3200), or a storage device (3400) via a bus (3700).

[0465] In a computing system (3000) according to one embodiment of the present invention, the artificial neural network model (3800) may be a neural network model specialized for each type, specification, or performance of a charging station / dispenser / mobility.

[0466] A computing system (3000) according to one embodiment of the present invention may include at least one processor (3100) and a memory (3200) that stores instructions that instruct the at least one processor (3100) to perform at least one step. At least some steps of a method according to one embodiment of the present invention may be performed by the at least one processor (3100) loading and executing instructions from the memory (3200).

[0467] The processor (3100) may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed.

[0468] Each of the memory (3200) and the storage device (3400) may be configured with at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (3200) may be configured with at least one of a read-only memory (ROM) and a random access memory (RAM).

[0469] Additionally, the computing system (3000) may include a communication interface (3300) that performs communication via a wireless network.

[0470] Additionally, the computing system (3000) may further include a storage device (3400), an input interface (3500), an output interface (3600), etc.

[0471] Additionally, each component included in the computing system (3000) can communicate with each other by being connected by a bus (3700).

[0472] Examples of the computing system (3000) of the present invention may include a communicable desktop computer, a laptop computer, a notebook, a smart phone, a tablet PC, a mobile phone, a smart watch, smart glasses, an e-book reader, a portable multimedia player (PMP), a portable game console, a navigation device, a digital camera, a digital multimedia broadcasting (DMB) player, a digital audio recorder, a digital audio player, a digital video recorder, a digital video player, a PDA (Personal Digital Assistant), etc.

[0473] Referring to FIG. 12 and FIG. 9 together, a device for providing a specialized artificial neural network model for a process of fueling hydrogen fueled mobility according to an embodiment of the present invention may include a memory (3200) for storing at least one command; a processor (3100) for executing at least one command; and a charging station, dispenser, or artificial neural network model (3800) specialized for each mobility, the processor electronically communicating with the processor for fueling hydrogen fueled mobility.

[0474] A processor (3100) of a device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain identification information about a charging station / dispenser (1100) or a mobility (1300) by at least one command (S2110, S2130, S2150). Based on the identification information, the processor (3100) can provide a specialized artificial neural network model (S2230, S2250, S2260) so that the specialized artificial neural network model obtains control or prediction information related to a process (S2400) of supplying hydrogen to the mobility (1300).

[0475] A processor (3100) of a device providing a specialized artificial neural network model according to one embodiment of the present invention can train a basic model of the specialized artificial neural network model (S2500). The basic model of the specialized artificial neural network model can be trained to learn a function of generating control or prediction information related to the process of supplying hydrogen to the mobility (1300) by using measurement data of a charging station, dispenser, or mobility (1300) obtained during the process of supplying hydrogen to the mobility (1300) (S2440) (S2500).

[0476] The processor (3100) of the device providing a specialized artificial neural network model according to one embodiment of the present invention can store a model in which a basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility (1300).

[0477] The processor (3100) of the device providing the specialized artificial neural network model according to one embodiment of the present invention can provide the updated and stored specialized artificial neural network model in response to a subsequent request from a new mobility or charging station / dispenser (1100) (based on new identification information).

[0478] That is, through the learning process of a specialized artificial neural network model, an updated model can be implemented by starting from a basic model and learning actual measured data for each specialized condition (S2500).

[0479] The processor (3100) of the device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain a basic model from a cloud server or a local server (1000). The processor (3100) can feed back the updated and stored specialized artificial neural network model to the cloud server or the local server (1000).

[0480] The processor (3100) of the device providing a specialized artificial neural network model according to one embodiment of the present invention can obtain a new basic model implemented using federated learning from the cloud server or local server (1000) after feeding back the specialized artificial neural network model to the cloud server or local server (1000).

[0481] The processor (3100) of the device providing a specialized artificial neural network model according to one embodiment of the present invention can provide a specialized artificial neural network model based on identification information so that control or prediction information can be used to control or predict the process of supplying hydrogen to the mobility (1300) by the dispenser (1100) (S2230, S2260).

[0482] Referring to FIG. 12 and FIG. 10 together, a device for supplying hydrogen fuel or a device for controlling a hydrogen fuel supply process mounted on a mobility (300) that uses hydrogen as fuel according to one embodiment of the present invention may include a memory (3200) that stores at least one command; and a processor (3100) that executes at least one command.

[0483] At this time, the device may further include a charging station, a dispenser, or an artificial neural network model (3800) specialized for each mobility to refuel the mobility with hydrogen.

[0484] At this time, the processor (3100) can obtain control or prediction information related to the process of supplying hydrogen to the mobility from a specialized artificial neural network model determined based on identification information about the charging station, dispenser, or mobility by at least one command (S2300).

[0485] A processor (3100) of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention may select one of a plurality of artificial neural network model candidates as a specialized artificial neural network model based on first identification information about mobility.

[0486] The processor (3100) of the device for controlling the process of supplying hydrogen fuel according to one embodiment of the present invention can transmit first identification information about mobility and second identification information about a charging station or dispenser to a cloud or local server, and can receive a specialized artificial neural network model from the cloud or local server based on the first identification information and the second identification information.

[0487] A processor (3100) of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can receive a specialized artificial neural network model from mobility based on first identification information about the mobility.

[0488] A processor (3100) of a device controlling a process for supplying hydrogen fuel according to one embodiment of the present invention can acquire a basic model of a specialized artificial neural network model. The processor (3100) can train the basic model to learn a function for generating control or prediction information related to the process of supplying hydrogen to a mobility using measurement data of a charging station, dispenser, or mobility obtained during the process of supplying hydrogen to a mobility.

[0489] The processor (3100) of the device controlling the process of supplying hydrogen fuel according to one embodiment of the present invention can store a model in which the basic model is updated through training as an artificial neural network model specialized for each charging station, dispenser, or mobility.

[0490] The processor (3100) of a device controlling a hydrogen fueling process according to one embodiment of the present invention can obtain a basic model from a cloud or local server. The processor (3100) can then feed back an updated and stored specialized artificial neural network model to the cloud or local server.

[0491] A processor (3100) of a device for controlling a process of supplying hydrogen fuel according to one embodiment of the present invention can obtain control or prediction information related to a process of supplying hydrogen fuel to a mobility using a specialized artificial neural network model determined based on a fuel supply protocol corresponding to a process of supplying hydrogen fuel to a mobility by a dispenser or a communication protocol between the mobility and the dispenser.

[0492] A hydrogen fuel supply control device according to one embodiment of the present invention may be disposed in a dispenser that supplies hydrogen to a mobility vehicle, or may be implemented as a controller that is not disposed in the dispenser but is capable of electronically communicating with the dispenser and influencing fueling of the dispenser.

[0493] The operations of the method according to an embodiment of the present invention can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes any type of recording device that stores information readable by a computer system. Furthermore, a computer-readable recording medium can be distributed across network-connected computer systems, allowing the computer-readable program or code to be stored and executed in a distributed manner.

[0494] Additionally, the computer-readable recording medium may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. The program instructions may include not only machine language codes produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.

[0495] While some aspects of the present invention have been described in the context of a device, they may also represent a description of a corresponding method, wherein a block or device corresponds to a method step or a feature of a method step. Similarly, aspects described in the context of a method may also be described as a corresponding block or item or a feature of a corresponding device. Some or all of the method steps may be performed by (or using) a hardware device, such as, for example, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, at least one or more of the most important method steps may be performed by such a device.

[0496] In embodiments, a programmable logic device (e.g., a field-programmable gate array) may be used to perform some or all of the functions of the methods described herein. In embodiments, the field-programmable gate array may operate in conjunction with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by some hardware device.

[0497] Although the present invention has been described above with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various modifications and changes may be made to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below.

Claims

1. Memory that stores at least one command; a processor for executing at least one of the above instructions; and A station, dispenser, or artificial neural network model specialized for hydrogen fueled mobility, in electronic communication with said processor and supplying hydrogen to said hydrogen fueled mobility; Including, The above processor, Obtaining identification information for the charging station, the dispenser, or the mobility; Based on the above identification information, the specialized artificial neural network model is provided so that the specialized artificial neural network model obtains control or prediction information related to the process of supplying hydrogen to the mobility. A device that provides a specialized artificial neural network model.

2. In paragraph 1, The above processor, The basic model of the specialized artificial neural network model is trained to learn the function of generating control or prediction information related to the process of fueling hydrogen to the mobility by using measurement data of the charging station, the dispenser, or the mobility obtained in the process of fueling hydrogen to the mobility. The above basic model is updated through training and stored as a specialized artificial neural network model specialized for each charging station, dispenser, or mobility. Based on the above identification information, providing the stored specialized artificial neural network model, A device that provides a specialized artificial neural network model.

3. In paragraph 2, The above processor, Obtain the above basic model from the cloud or local server, Feeding back the above updated and stored specialized artificial neural network model to the cloud or local server. A device that provides a specialized artificial neural network model.

4. In paragraph 3, The above processor, After feeding back the specialized artificial neural network model to the cloud or the local server, a new basic model implemented using federated learning is obtained from the cloud or the local server. A device that provides a specialized artificial neural network model.

5. In paragraph 1, The above processor, Based on the above identification information, the specialized artificial neural network model is provided so that the control or prediction information can be used to control or predict the process of the dispenser supplying hydrogen to the mobility. A device that provides a specialized artificial neural network model.

6. In paragraph 1, The above processor, Providing the specialized artificial neural network model based on the fuel supply protocol corresponding to the process of the above dispenser supplying hydrogen to the above mobility or the communication protocol between the above mobility and the above dispenser. A device that provides a specialized artificial neural network model.

7. A device that controls the process of supplying hydrogen to hydrogen fueled mobility. Memory that stores at least one instruction; a processor for executing at least one of the above instructions; and A charging station, dispenser, or artificial neural network model specialized for each mobility that supplies hydrogen to the above mobility; Including, The processor obtains control or prediction information related to the process of fueling the mobility with hydrogen from the specialized artificial neural network model determined based on identification information about the charging station, the dispenser, or the mobility. Control device for the hydrogen fuel supply process.

8. In paragraph 7, Contains multiple artificial neural network model candidates specialized for multiple mobility types, The above processor, Selecting one of the plurality of artificial neural network model candidates as the specialized artificial neural network model based on the first identification information for the above mobility. Control device for the hydrogen fuel supply process.

9. In paragraph 7, The above processor, Transmitting first identification information for the above mobility and second identification information for the above charging station or the above dispenser to a cloud or local server, Based on the first identification information and the second identification information, the specialized artificial neural network model is provided from a cloud or local server. Control device for the hydrogen fuel supply process.

10. In paragraph 7, The above processor, From the above mobility, the specialized artificial neural network model is provided based on the first identification information for the above mobility. Control device for the hydrogen fuel supply process.

11. In paragraph 7, The above processor, Obtain the basic model of the above specialized artificial neural network model, The basic model is trained to learn the function of generating control or prediction information related to the process of supplying hydrogen to the mobility by using measurement data of the charging station, the dispenser, or the mobility obtained in the process of supplying hydrogen to the mobility, The above basic model is updated through training and stored as a specialized artificial neural network model specialized for each charging station, dispenser, or mobility. Control device for the hydrogen fuel supply process.

12. In paragraph 11, The above processor, Obtain the above basic model from the cloud or local server, Feeding back the above updated and stored specialized artificial neural network model to the cloud or local server. Control device for the hydrogen fuel supply process.

13. In paragraph 7, The above processor, Obtaining control or prediction information related to the process of supplying hydrogen to the mobility by using the specialized artificial neural network model determined based on the fuel supply protocol corresponding to the process of supplying hydrogen to the mobility by the dispenser or the communication protocol between the mobility and the dispenser. Control device for the hydrogen fuel supply process.

14. A method for fueling hydrogen for hydrogen fueled mobility, A step of obtaining a charging station, dispenser, or artificial neural network model specialized for each mobility that supplies hydrogen to the above mobility; and A step of obtaining control or prediction information related to the process of fueling hydrogen to the mobility from the specialized artificial neural network model determined based on identification information about the charging station, the dispenser, or the mobility; Including, A method for fueling hydrogen for hydrogen fuel mobility.

15. In paragraph 14, The steps for obtaining the above specialized artificial neural network model are: A step of selecting one of a plurality of artificial neural network model candidates specialized for each of a plurality of mobility types as the specialized artificial neural network model based on the first identification information for the above mobility; A method for fueling hydrogen into hydrogen fuel mobility, comprising:

16. In paragraph 14, The steps for obtaining the above specialized artificial neural network model are: A step of transmitting first identification information for the above mobility and second identification information for the above charging station or the above dispenser to a cloud or local server; and A step of receiving the specialized artificial neural network model from a cloud or local server based on the first identification information and the second identification information; Including, A method for fueling hydrogen for hydrogen fuel mobility.

17. In paragraph 14, The steps for obtaining the above specialized artificial neural network model are: A step of receiving the specialized artificial neural network model from the above mobility based on first identification information about the mobility; Including, A method for fueling hydrogen for hydrogen fuel mobility.

18. In paragraph 14, A step of setting the current specialized artificial neural network model used in the step of obtaining the above control or prediction information as a basic model; A step of training the basic model to learn a function of generating control or prediction information related to the process of supplying hydrogen to the mobility by using measurement data of the charging station, the dispenser, or the mobility obtained in the process of supplying hydrogen to the mobility; and A step of saving the above basic model as a new specialized artificial neural network model updated through training; Including more, A method for fueling hydrogen for hydrogen fuel mobility.

19. In paragraph 18, A step of obtaining the above basic model from a cloud or local server; and A step of feeding back the new specialized artificial neural network model to the cloud or local server; Including more, A method for fueling hydrogen for hydrogen fuel mobility.

20. In paragraph 14, The steps for obtaining the above specialized artificial neural network model are: A step of obtaining the specialized artificial neural network model determined based on a fuel supply protocol corresponding to a process of supplying hydrogen to the mobility by the dispenser or a communication protocol between the mobility and the dispenser; Including, A method for fueling hydrogen for hydrogen fuel mobility.

Images