Concentration prediction model generation method and device, concentration prediction method, and vehicle

Abstract

The embodiment of the application is suitable for the hydrogen technology field, and provides a concentration prediction model generation method and device, a concentration prediction method and a vehicle. The method comprises the following steps: obtaining a first concentration time series sequence of hydrogen diffusion under a plurality of first leakage working conditions respectively preset by a hydrogen storage system; generating a hydrogen diffusion model based on the first concentration time series sequence; the hydrogen diffusion model is used for describing the concentration diffusion law of hydrogen in the hydrogen storage system space changing with time; determining a second concentration time series sequence of hydrogen diffusion under each first leakage working condition according to the hydrogen diffusion model; the second concentration time series sequence comprises a plurality of second hydrogen concentrations of hydrogen changing in a second time period at a plurality of preset monitoring positions; the length of the second time period is greater than the length of a first time period; and generating a concentration prediction model based on the second concentration time series sequence corresponding to each first leakage working condition. The above method can improve the accuracy of the concentration prediction model in predicting the hydrogen concentration distribution.

Description

Technical Field

[0001] This application belongs to the field of hydrogen technology, and in particular relates to a method, apparatus, concentration prediction method and vehicle for generating a concentration prediction model. Background Technology

[0002] With the development of clean energy, using hydrogen as a fuel for vehicles is becoming a trend. Typically, vehicles use hydrogen storage systems to store liquid hydrogen. However, due to the low-temperature vaporization, easy diffusion, and explosiveness of liquid hydrogen, in the event of a leak, it will rapidly vaporize and diffuse within the storage system. Therefore, when applying hydrogen energy to vehicles, it is usually necessary to predict the hydrogen concentration distribution after diffusion based on the current hydrogen concentration within the storage system. This allows for early warning when the hydrogen concentration exceeds a warning level, ensuring vehicle and personal safety.

[0003] Currently, methods for predicting hydrogen concentration distribution mainly rely on model simulations based on the fundamental governing equations of fluid mechanics. For example, hydrogen diffusion models are solved based on hydrogen concentration distributions at certain times under multiple pre-defined leakage conditions.

[0004] However, the preset leakage conditions are usually limited and cannot pre-set a large number of actual leakage conditions in the hydrogen storage system. Therefore, the resulting hydrogen diffusion model usually has a high accuracy in predicting the hydrogen concentration distribution at any time under the preset leakage conditions, but a low accuracy in predicting the hydrogen concentration distribution under other leakage conditions. Summary of the Invention

[0005] This application provides a method, apparatus, and vehicle for generating a concentration prediction model, which can solve the problem of low accuracy in predicting hydrogen concentration distribution using hydrogen diffusion models.

[0006] In a first aspect, embodiments of this application provide a method for generating a concentration prediction model, the method comprising:

[0007] Under multiple preset first leakage conditions of the hydrogen storage system, the first concentration time series during hydrogen diffusion is obtained; the first concentration time series includes multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period;

[0008] Based on the first concentration time series, a hydrogen diffusion model is generated; the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time.

[0009] Based on the hydrogen diffusion model, the second concentration time series is determined for each first leakage condition. The second concentration time series includes multiple second hydrogen concentrations at multiple preset monitoring locations during a second time period. The duration of the second time period is longer than that of the first time period.

[0010] A concentration prediction model is generated based on the second concentration time series corresponding to each first leakage condition.

[0011] In one embodiment, a hydrogen diffusion model is generated based on a first concentration time series, including:

[0012] A three-dimensional model of the hydrogen storage system is established, and the three-dimensional model is spatially meshed; the mesh is used to simulate the fluid behavior of hydrogen in the three-dimensional model.

[0013] Based on the preset boundary conditions and the divided three-dimensional model, a hydrodynamic simulation was performed on multiple first hydrogen concentrations at each preset monitoring location to obtain a hydrogen diffusion model.

[0014] In one embodiment, a concentration prediction model is generated based on the second concentration time series corresponding to each first leakage condition, including:

[0015] For any first leakage condition, generate first training data corresponding to the second concentration time series. The first training data includes at least the first input data corresponding to the second concentration time series from the initial time to the first time in the second time period, and the first label data corresponding to the second concentration time series from the first time to the end time.

[0016] The first input data is fed into the initial concentration prediction model for training, and the first predicted concentration time series of hydrogen is predicted from the first time to the end time.

[0017] Based on the first label data and the first predicted concentration time series, the model parameters of the initial concentration prediction model are iteratively updated;

[0018] If the initial concentration prediction model meets the convergence condition during the iterative update process, then the current initial concentration prediction model will be used as the concentration prediction model.

[0019] If the initial concentration prediction model does not meet the convergence condition during the iterative update process, the process returns to the step of inputting the first input data into the initial concentration prediction model for training, predicting the first predicted concentration time series of hydrogen from the first time to the end time, and subsequent steps, until the initial concentration prediction model meets the convergence condition.

[0020] In one embodiment, the duration from the initial time to the first time is greater than the duration from the first time to the end time.

[0021] In one embodiment, the model parameters of the initial concentration prediction model are iteratively updated based on the first label data and the first predicted concentration time series, including:

[0022] Calculate the training loss of the initial concentration prediction model based on the first label data and the first predicted concentration time series.

[0023] The model parameters are updated iteratively based on the training loss.

[0024] In one embodiment, after generating a concentration prediction model based on the second concentration time series corresponding to each first leakage condition, the process includes:

[0025] During the vehicle test, the time series of hydrogen concentrations in the third time period was obtained.

[0026] The second training data is generated based on the third concentration time series; the second training data includes at least the second input data corresponding to the third concentration time series from the initial time to the second time in the third time period, and the second label data corresponding to the third concentration time series from the second time to the end time.

[0027] The model parameters of the concentration prediction model are iteratively updated based on the first and second training data to obtain the updated concentration prediction model.

[0028] Secondly, embodiments of this application provide a concentration prediction model generation apparatus, the apparatus comprising:

[0029] The first acquisition module is used to acquire the first concentration time series of hydrogen diffusion under multiple preset first leakage conditions of the hydrogen storage system; the first concentration time series includes multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period;

[0030] The first generation module is used to generate a hydrogen diffusion model based on the first concentration time series; the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time.

[0031] The first determining module is used to determine the second concentration time series of hydrogen diffusion under each first leakage condition according to the hydrogen diffusion model; the second concentration time series includes multiple second hydrogen concentrations that change at multiple preset monitoring locations within a second time period; the duration of the second time period is longer than the duration of the first time period.

[0032] The second generation module is used to generate a concentration prediction model based on the second concentration time series corresponding to each first leakage condition.

[0033] Thirdly, embodiments of this application provide a concentration prediction method, the method comprising:

[0034] During vehicle operation, the actual hydrogen concentration time series at the current moment is obtained; the actual concentration time series includes the actual hydrogen concentration at the current moment and multiple moments before the current moment.

[0035] The actual concentration time series is input into the concentration prediction model to obtain the third predicted concentration time series of hydrogen in a preset future time period; the concentration prediction model is generated according to the method in the first aspect.

[0036] If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then a preset operation is executed; the preset operation includes at least a hydrogen purging operation.

[0037] In one embodiment, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then a preset operation is performed, including:

[0038] If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then the target accumulation location with the highest predicted hydrogen concentration is determined.

[0039] Based on the target aggregation location, determine the target purging information of the blower system in the vehicle;

[0040] The blower system is controlled to perform hydrogen purging operations based on the target purging information.

[0041] Fourthly, embodiments of this application provide a concentration prediction device, the device comprising:

[0042] The second acquisition module is used to acquire the actual hydrogen concentration time series at the current moment during vehicle operation; the actual concentration time series includes the actual hydrogen concentration at the current moment and multiple moments before the current moment.

[0043] The input module is used to input the actual concentration time series into the concentration prediction model to obtain the third predicted concentration time series of hydrogen in a preset future time period; the concentration prediction model is generated according to the method in the first aspect.

[0044] The execution module is used to perform a preset operation if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold; the preset operation includes at least a hydrogen venting operation.

[0045] Fifthly, embodiments of this application provide a vehicle including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the methods described in the first or third aspects above.

[0046] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described in the first or third aspect above.

[0047] Fifthly, embodiments of this application provide a computer program product that, when run on a vehicle, causes the vehicle to perform the methods described in the first or third aspect.

[0048] The first aspect of this application's embodiments, compared to the prior art, offers the following advantages: First concentration time series data can be obtained for each preset first leakage condition in the hydrogen storage system, allowing for the acquisition of multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period. Then, based on the first concentration time series data, a hydrogen diffusion model is generated to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time. Furthermore, based on the hydrogen diffusion model, a second concentration time series data for hydrogen diffusion under each first leakage condition can be determined. Since the second concentration time series data includes multiple second hydrogen concentrations changing at multiple preset monitoring locations within a second time period, and the duration of the second time period is longer than the duration of the first time period, it can be considered that only a small number of moments (the first time period) of the first hydrogen concentration under multiple preset leakage conditions are needed to obtain the second hydrogen concentration over a large number of moments (the second time period) under multiple leakage conditions. Therefore, when generating a concentration prediction model based on the second concentration time series corresponding to each first leakage condition, the increased data volume of the second concentration time series improves the accuracy of the generated concentration prediction model. Furthermore, since the second concentration time series only needs to be determined based on the first hydrogen concentration at a small number of moments (the first time period), it also reduces the time and manpower costs required for manually generating training data.

[0049] The beneficial effect of the third aspect of this application embodiment compared with the prior art is that, during vehicle operation, by acquiring the actual hydrogen concentration time series corresponding to the actual hydrogen concentration at the current moment and multiple moments before the current moment, and inputting it into the concentration prediction model, a third predicted hydrogen concentration time series within a preset future time period can be obtained. Then, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, a preset operation is performed. For example, a hydrogen release operation. Thus, by predicting in advance whether the hydrogen concentration will be equal to or equal to the preset concentration threshold, a hydrogen release operation can be performed before the predicted hydrogen concentration is greater than or equal to the preset concentration threshold, reducing the hydrogen concentration in the vehicle and improving driving safety. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the implementation of a concentration prediction model generation method according to an embodiment of this application;

[0052] Figure 2 This is a schematic diagram illustrating one implementation of a hydrogen diffusion model generation method in a concentration prediction model generation method provided in an embodiment of this application;

[0053] Figure 3 This is a schematic diagram illustrating one implementation method of generating a concentration prediction model in an embodiment of this application.

[0054] Figure 4 This is a schematic diagram illustrating one implementation method for generating a concentration prediction model in another embodiment of this application.

[0055] Figure 5 This is a schematic diagram of the structure of a concentration prediction model generation device provided in one embodiment of this application;

[0056] Figure 6 This is a flowchart illustrating the implementation of a concentration prediction method according to an embodiment of this application;

[0057] Figure 7 This is a schematic diagram of the structure of a hydrogen storage system and a fan system provided in one embodiment of this application;

[0058] Figure 8 This is a schematic diagram of a hydrogen storage system and a fan system provided in another embodiment of this application;

[0059] Figure 9 This is a flowchart illustrating the implementation of a concentration prediction model generation method and a concentration prediction method according to an embodiment of this application.

[0060] Figure 10 This is a schematic diagram of the structure of a concentration prediction device provided in one embodiment of this application;

[0061] Figure 11 This is a schematic diagram of the structure of a vehicle provided in one embodiment of this application. Detailed Implementation

[0062] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0063] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0064] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0065] With the development of clean energy, using hydrogen as a fuel for vehicle propulsion is becoming a trend. Typically, vehicles use hydrogen storage systems to store liquid hydrogen. However, due to the low-temperature vaporization, easy diffusion, and explosiveness of liquid hydrogen, in the event of a leak, it will rapidly vaporize and diffuse within the storage system.

[0066] Therefore, when applying hydrogen energy to vehicles, it is usually necessary to predict the hydrogen concentration distribution after hydrogen diffusion based on the current hydrogen concentration in the hydrogen storage system, so as to provide early warning when the hydrogen concentration is greater than the warning concentration, and ensure the safety of vehicles and people.

[0067] Currently, methods for predicting hydrogen concentration distribution mainly rely on model simulations based on the fundamental governing equations of fluid mechanics. For example, hydrogen diffusion models are solved based on hydrogen concentration distributions at certain times under multiple pre-defined leakage conditions.

[0068] However, the preset leakage conditions are usually limited and cannot anticipate the actual large-scale leakage conditions of the hydrogen storage system. Therefore, the resulting hydrogen diffusion model typically has a high accuracy rate in predicting the hydrogen concentration distribution at any given time under the preset leakage conditions, but a low accuracy rate in predicting the hydrogen concentration distribution under other leakage conditions.

[0069] Alternatively, a hydrogen diffusion concentration prediction model can be obtained by training the model based on hydrogen concentration data at certain times under multiple leakage conditions. However, the amount of manually determined hydrogen concentration data at certain times under multiple leakage conditions is usually limited. If a large number of hydrogen concentration data points are needed as the initial training data, it will significantly increase manpower and time costs. Furthermore, if only a small number of hydrogen concentration data points under multiple leakage conditions are used as the initial training data, the accuracy of the trained concentration prediction model will be low due to the limited data volume.

[0070] Based on this, in order to generate a concentration prediction model with high accuracy, this application provides a method for generating a concentration prediction model. This method can be applied to vehicles, such as vehicle-mounted terminals, central control terminals, electronic control units (ECUs), and other electronic devices. It can also be applied to tablet computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, and other electronic devices. This application does not impose any restrictions on the specific type of electronic device.

[0071] Please see Figure 1 , Figure 1 The following is a flowchart illustrating the implementation of a concentration prediction model generation method provided in an embodiment of this application. The method includes the following steps:

[0072] S101. Under multiple preset first leakage conditions of the hydrogen storage system, obtain the first concentration time series during hydrogen diffusion; the first concentration time series includes multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period.

[0073] In one embodiment, the hydrogen storage system described above can be a collection of all components through which fuel (hydrogen) passes in a fuel cell vehicle. For example, the hydrogen storage system may include components such as hydrogen storage containers (e.g., hydrogen storage cylinders), pressure regulating devices, valves, pipelines, and accessories, without limitation.

[0074] It is understandable that hydrogen may leak from the hydrogen storage container during vehicle operation and then diffuse within the hydrogen storage system. In this case, a large amount of hydrogen will be distributed within the space occupied by the hydrogen storage system.

[0075] In one embodiment, the first leakage condition described above is used to describe hydrogen leakage information within the hydrogen storage system. Specifically, the first leakage condition includes, but is not limited to, information such as the location, direction, rate, shape and size of the leak, ambient wind speed, and ambient temperature of the hydrogen leak, etc.

[0076] It should be noted that each of the above-mentioned leakage information typically has multiple scenarios, and the resulting first leakage condition, formed by combinations of these scenarios, also usually has multiple possible outcomes. Therefore, manually setting all the first leakage conditions to obtain the first concentration time series would consume a significant amount of manpower and time.

[0077] Therefore, in order to reduce labor and time costs, the first leakage condition mentioned above can be considered as only a part of the total leakage conditions.

[0078] In one embodiment, the hydrogen storage system may have multiple preset monitoring locations, each preset monitoring location including at least one hydrogen concentration sensor to collect the first hydrogen concentration corresponding to the preset monitoring location when hydrogen diffuses.

[0079] Since hydrogen diffusion is a continuous process, each preset monitoring location will collect multiple first hydrogen concentrations within the first time period. The set of multiple first hydrogen concentrations collected from all preset monitoring locations within the first time period constitutes the aforementioned first concentration time series. Based on this, the first concentration time series can be considered to describe the concentration change pattern of hydrogen under the first leakage condition.

[0080] For example, the first concentration time series can be:

[0081] First type of leakage condition: S1

[0082] Time T1:

[0083] Position L1, first hydrogen concentration P11; Position L2, first hydrogen concentration P12; ...; Position Ln, first hydrogen concentration P1n.

[0084] Time T2:

[0085] Position L1, first hydrogen concentration P21; Position L2, first hydrogen concentration P22; ...; Position Ln, first hydrogen concentration P2n. ...

[0087] Time Tm:

[0088] Position L1, first hydrogen concentration Pm1; Position L2, first hydrogen concentration Pm2; ...; Position Ln, first hydrogen concentration Pmn.

[0089] Where n represents the number of preset monitoring locations; the duration between time T1 and time Tm is the first time period. It should be noted that in reality, the total duration of hydrogen diffusion is usually much longer than the first time period. However, since the total duration cannot be specifically defined, in order to obtain the concentration diffusion pattern of hydrogen in the hydrogen storage system over time, and without incurring significant time and manpower costs in generating the first concentration time series, the first time period can be a preset time period of a small duration. For example, half an hour; this is not limited.

[0090] S102. Based on the first concentration time series, generate a hydrogen diffusion model.

[0091] In one embodiment, the aforementioned hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time. That is, it can be considered that after obtaining the hydrogen diffusion model, the first hydrogen concentration at each preset monitoring location at any time under each first leakage condition can be determined based on the hydrogen diffusion model.

[0092] As an example, electronic devices can be based on, for example Figure 2 The S201-S202 steps shown generate a hydrogen diffusion model. Details are as follows:

[0093] S201. Establish a three-dimensional model of the hydrogen storage system and divide the three-dimensional model into spatial meshes; the meshes are used to simulate the fluid behavior of hydrogen in the three-dimensional model.

[0094] In one embodiment, the aforementioned three-dimensional model is a spatial structure model of a hydrogen storage system. The electronic device may have pre-set spatial model data of the hydrogen storage system to build the aforementioned three-dimensional model based on the spatial model data. For example, the aforementioned spatial model data may be CAD design files or 3D scan data, and there is no limitation thereto.

[0095] In one embodiment, the above-mentioned meshing involves selecting an appropriate mesh type, controlling the mesh density and size, and dividing the three-dimensional model into smaller mesh cells. The mesh type, density, and size can be determined based on factors such as the structure of the three-dimensional model and the fluid behavior characteristics of hydrogen, which will not be described in detail here.

[0096] S202. Based on the preset boundary conditions and the divided three-dimensional model, perform hydrodynamic simulation on multiple first hydrogen concentrations at each preset monitoring location to obtain a hydrogen diffusion model.

[0097] In one embodiment, the boundary conditions described above can be set according to actual conditions, and are not limited thereto. For example, boundary conditions include limiting conditions such as fluid velocity, pressure, and temperature, and are not limited thereto.

[0098] It should be noted that when performing fluid dynamics simulations, CFD (Computational Fluid Dynamics) equations can be established first. Then, all the aforementioned initial hydrogen concentrations can be substituted into the CFD equations for solving, yielding the hydrogen diffusion model. The CFD equations are the fundamental equations describing fluid motion.

[0099] S103. Based on the hydrogen diffusion model, determine the second concentration time series for hydrogen diffusion under each first leakage condition; the second concentration time series includes multiple second hydrogen concentrations changing at multiple preset monitoring locations within a second time period; the duration of the second time period is longer than the duration of the first time period.

[0100] As explained above, the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time. Therefore, after inputting the first leakage condition and the first concentration time series within the first time period into the hydrogen diffusion model, the hydrogen diffusion model can automatically output the second hydrogen concentration at each preset monitoring location at each subsequent time. Thus, the aforementioned second concentration time series is obtained.

[0101] For example, for the first leakage condition S1, the first hydrogen concentration at each time point from time T1 to time Tm can be input into the hydrogen diffusion model. Then, the hydrogen diffusion model can output:

[0102] Time Tm+1: Position L1, second hydrogen concentration Pm+1 1; Position L2, second hydrogen concentration Pm+1 2; ..., Time Tx: Position Ln, second hydrogen concentration Px n.

[0103] Where Pxn represents the second hydrogen concentration at the preset monitoring location Ln at time Tx. In this case, the time period from T1 to Tx can be considered the aforementioned second time period, which includes the first time period from T1 to Tm. That is, the first hydrogen concentration from T1 to Tm will be automatically determined as the second hydrogen concentration.

[0104] Based on this, it can be assumed that the first hydrogen concentration at a small number of moments (the first time period) under multiple leakage conditions can be used to obtain the second hydrogen concentration at a large number of moments (the second time period) under multiple leakage conditions. That is, the amount of data increases.

[0105] The second time period can be set according to the actual situation, such as one hour, and there is no limit to this.

[0106] It should be noted that the duration of the second time period mentioned above needs to be longer than the duration of the first time period, so that the concentration prediction model can be generated based on a large number of second concentration time series corresponding to the second time period during the subsequent model training process.

[0107] As an example, the second time period mentioned above can include the first time period. It is understood that the first concentration time series within the first time period represents the actual detected distribution data of hydrogen diffusion, which better reflects the diffusion rules of hydrogen. Therefore, when the second time period includes the first time period, it can be assumed that the concentration prediction model generated based on the second concentration time series of the second time period has a higher prediction accuracy.

[0108] S104. Generate a concentration prediction model based on the second concentration time series corresponding to each first leakage condition.

[0109] In one embodiment, the second concentration time series contains the second hydrogen concentration at a large number of moments (second time periods). Therefore, it can be considered that the concentration prediction model generated based on the second concentration time series generally has a high accuracy in predicting the concentration distribution of hydrogen in future time periods.

[0110] In this embodiment, under each preset first leakage condition of the hydrogen storage system, a first concentration time series can be obtained during hydrogen diffusion to obtain multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period. Then, based on the first concentration time series, a hydrogen diffusion model is generated to describe the concentration diffusion law of hydrogen in the hydrogen storage system space over time. Furthermore, based on the hydrogen diffusion model, a second concentration time series during hydrogen diffusion under each first leakage condition can be determined. Since the second concentration time series includes multiple second hydrogen concentrations changing at multiple preset monitoring locations within a second time period, and the second time period includes the first time period, it can be considered that only a small number of moments (the first time period) of the first hydrogen concentration under multiple leakage conditions are needed to obtain the second hydrogen concentration over a large number of moments (the second time period) under multiple leakage conditions. Therefore, when generating a concentration prediction model based on the second concentration time series corresponding to each first leakage condition, the increased data volume of the second concentration time series improves the accuracy of the generated concentration prediction model. Furthermore, the second concentration time series only needs to be determined based on the first hydrogen concentration at a small number of moments (the first time period), which can also reduce the time and manpower costs of manually generating training data.

[0111] As an example, electronic devices can be based on, for example Figure 3 The steps S301-S305 shown generate the concentration prediction model. Details are as follows:

[0112] S301. For any first leakage condition, generate the first training data corresponding to the second concentration time series.

[0113] In one embodiment, when training a concentration prediction model, the first training data typically needs to be divided into first input data for model processing and first label data for loss calculation with the predicted data output by the model. In this embodiment, the purpose of the concentration prediction model is to predict the hydrogen concentration at various preset monitoring locations over a future time period.

[0114] Therefore, for the aforementioned second concentration time series, since the second hydrogen concentration at each preset monitoring location varies with time, the second concentration time series can be divided into two parts as the aforementioned first training data.

[0115] Specifically, the first training data includes at least the first input data corresponding to the second concentration time series from the initial time to the first time in the second time period, and the first label data corresponding to the second concentration time series from the first time to the end time. Furthermore, the concentration prediction model can predict the first predicted hydrogen concentration time series from the first time to the end time based on the first input data, and iteratively update the model parameters of the concentration prediction model with the actual first label data.

[0116] In order to enable the concentration prediction model to learn the diffusion pattern of hydrogen in this scenario as much as possible during the processing of the first input data, the duration from the initial time to the first time can be set to be greater than the duration from the first time to the end time.

[0117] For example, if the total duration of the second time period is 1 hour, the duration from the initial time to the first time can be set to account for 70% of the total duration, and the duration from the first time to the end time can be set to account for 30% of the total duration, without any limitation.

[0118] S302. Input the first input data into the initial concentration prediction model for training, and predict the first predicted concentration time series of hydrogen from the first time to the end time.

[0119] In one embodiment, the model category of the aforementioned initial concentration prediction model can be a CNN (Convolutional Neural Networks) neural network model, a BP (Back Propagation) neural network model, or the like, and there is no limitation on this.

[0120] The form of the first predicted concentration time series can refer to the form of the first or second concentration time series, and will not be described in detail here. That is, the first predicted concentration time series will include the predicted hydrogen concentrations at each preset monitoring location from the first time to the end time.

[0121] S303. Based on the first label data and the first predicted concentration time series, iteratively update the model parameters of the initial concentration prediction model.

[0122] In one embodiment, the electronic device can calculate the training loss of an initial concentration prediction model based on first label data and a first predicted concentration time series. Then, the model parameters are iteratively updated based on the training loss.

[0123] In one embodiment, the electronic device may use any one of the squared loss function, absolute value loss function, and logarithmic loss function to calculate the training loss. For example, the squared loss function may be used to calculate the training loss.

[0124] Specifically, the squared loss function is the square of the difference between the predicted value and the actual value. Therefore, the second hydrogen concentration in the first-label data at the same time and position in the first predicted concentration time series can be used as the actual value, and the predicted concentration in the first predicted concentration time series example can be used as the predicted value in the above calculation. Then, the average of the sum of the squares of the differences at all times and all positions is used as the final training loss.

[0125] For example, for the second hydrogen concentration Py1 at position L1 at time Ty, the predicted concentration P`y1 at position L1 at time Ty can also be determined from the first predicted concentration time series. Then, the squared difference E between P`y1 and Py1 is calculated. Since there are n preset monitoring positions, the squared differences E corresponding to each of the n preset monitoring positions will be obtained at time Ty. Furthermore, there are usually multiple times between the first time and the end time. For example, there are F times. Therefore, the number of squared differences E obtained in the end will be n*F.

[0126] In one embodiment, the above-mentioned iterative update of model parameters includes, but is not limited to, stochastic gradient descent and learning rate decay, and is not limited thereto.

[0127] S304. If the initial concentration prediction model meets the convergence condition during the iterative update process, then the current initial concentration prediction model shall be used as the concentration prediction model.

[0128] S305. If the initial concentration prediction model does not meet the convergence condition during the iterative update process, return to step S302 and subsequent steps until the initial concentration prediction model meets the convergence condition.

[0129] In one embodiment, the convergence condition includes, but is not limited to, multiple consecutive training losses being less than a preset training loss, or the model being trained a preset number of times; these are not limited. The preset training loss and the preset number of training iterations can be set according to actual circumstances.

[0130] Understandably, if the initial concentration prediction model meets the convergence condition, it can be considered that the current initial concentration prediction model is the successfully trained concentration prediction model. Otherwise, the electronic device can repeat steps S302-S305.

[0131] In another embodiment, after obtaining the concentration prediction model, the portion of the second concentration time series that was not used as the first training data can be used as test data to test the training effect of the concentration prediction model. That is, in step S301, all the second concentration time series can be divided into two parts. The first part of the second concentration time series is used as the first training data to separate the first input data and the first label data for model training in steps S302-S305. Then, the second part of the second concentration time series is used as test data.

[0132] The test data also needs to be divided into test input data and test label data, and the division method can refer to the division method of the first input data and the first label data in step S301 above. Then, the test input data can be input into the concentration prediction model to obtain the test predicted concentration time series. Next, the test training loss between the test predicted concentration time series and the test label data is calculated. If the test training loss is less than a preset loss value, the concentration prediction model is determined to be accurate; otherwise, the concentration prediction model is determined to be incorrect. Then, based on all prediction results, the prediction accuracy of the concentration prediction model is determined. Finally, if the prediction accuracy is greater than a preset accuracy, the concentration prediction model is determined to meet the requirements. Otherwise, if the prediction accuracy is less than or equal to the preset accuracy, the concentration prediction model is determined to not meet the requirements. Furthermore, the electronic device can regenerate the first training data and retrain the concentration prediction model until the concentration prediction model meets the requirements.

[0133] The preset loss value and preset accuracy rate can be set according to the actual situation, and there are no restrictions on them.

[0134] In one embodiment, the example of the test concentration prediction model meeting the requirements is only one way. In another embodiment, it can also be determined based on methods such as cross-validation and model comparison validation, and there is no limitation on this.

[0135] In summary, this is an example of an electronic device training a concentration prediction model based on a data-enhanced second concentration time series. The concentration prediction model trained on a large number of second concentration time series has a high prediction accuracy.

[0136] However, it should be noted that the aforementioned second concentration time series are all pre-determined hydrogen concentration distributions under various first leakage conditions. However, in real-world scenarios, the actual leakage conditions when a hydrogen storage system leaks are usually different from the aforementioned first leakage conditions. Therefore, the accuracy of the predicted concentration time series for future periods based on the concentration prediction model trained under the first leakage conditions may still be low when predicting concentration time series for actual leakage conditions that differ from the first leakage conditions.

[0137] Therefore, in order to improve the accuracy of the concentration prediction model for time series concentration predictions under arbitrary leakage conditions, electronic equipment can also, as shown in the example... Figure 4 Steps S401-S403, as shown, perform a second training of the already trained concentration prediction model. Details are as follows:

[0138] S401. During the vehicle test, obtain the time series of hydrogen concentration in the third time period.

[0139] In one embodiment, the above-described testing process can be either a test conducted on actual roads by a real vehicle or a test conducted in an environmental chamber; there is no limitation on either. During the environmental chamber test, the environmental information of the vehicle being tested is controllable. For example, the temperature and humidity affecting hydrogen diffusion can be generated in the environmental chamber by the personnel themselves, making the test more convenient. Furthermore, testing on actual roads more closely resembles a real environment, resulting in better testing outcomes.

[0140] It should be noted that the purpose of secondary training of the concentration prediction model is to improve the accuracy of predicting concentration time series under actual leakage conditions. However, actual leakage conditions cannot be fully determined. Therefore, it can be considered that the purpose of secondary training is also to improve the accuracy of the concentration prediction model in predicting concentration time series under arbitrary conditions.

[0141] Based on this, during vehicle testing, it is sufficient to obtain the time series of hydrogen concentration in the third time period, without needing to determine the actual leakage condition of the hydrogen storage system at that time.

[0142] S402. Generate the second training data based on the third concentration time series.

[0143] In one embodiment, the generation method of the second training data can be similar to the generation method of the first training data in S301, and will not be described in detail. For example, the second training data may also include at least the second input data corresponding to the third concentration time series from the initial time to the second time in the third time period, and the second label data corresponding to the third concentration time series from the second time to the end time.

[0144] The second time step is similar to the first time step in S301 above, the second input data is similar to the first input data in S301 above, and the second label data is similar to the first label data in S301 above. These will not be described in detail.

[0145] S403. Iteratively update the model parameters of the concentration prediction model based on the first training data and the second training data to obtain the updated concentration prediction model.

[0146] After obtaining the second training data, it can be integrated with the first training data to form new training data, and the concentration prediction model can be iteratively updated. At this point, combining the first and second training data not only allows the concentration prediction model to learn the characteristics of hydrogen concentration diffusion distribution under new, unknown leakage conditions, but also preserves the characteristics of hydrogen concentration diffusion distribution under existing leakage conditions, thus improving the accuracy of the concentration prediction model in predicting concentration time series under any conditions.

[0147] Furthermore, by integrating the first and second training data, the amount of data used for model training will increase significantly. This, in turn, can further improve the prediction accuracy of the iteratively updated concentration prediction model.

[0148] Please see Figure 5 , Figure 5 This is a structural block diagram of a concentration prediction model generation device provided in an embodiment of this application. The modules included in the concentration prediction model generation device in this embodiment are used to perform… Figures 1 to 4 The steps in the corresponding embodiments. Please refer to the details. Figures 1 to 4 as well as Figures 1 to 4 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 5 The concentration prediction model generation device 500 may include: a first acquisition module 510, a first generation module 520, a first determination module 530, and a second generation module 540, wherein:

[0149] The first acquisition module 510 is used to acquire the first concentration time series of hydrogen diffusion under multiple preset first leakage conditions of the hydrogen storage system; the first concentration time series includes multiple first hydrogen concentrations that change at multiple preset monitoring locations within a first time period.

[0150] The first generation module 520 is used to generate a hydrogen diffusion model based on the first concentration time series; the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system over time.

[0151] The first determining module 530 is used to determine the second concentration time series of hydrogen diffusion under each first leakage condition according to the hydrogen diffusion model; the second concentration time series includes multiple second hydrogen concentrations that change at multiple preset monitoring locations within a second time period; the duration of the second time period is longer than the duration of the first time period.

[0152] The second generation module 540 is used to generate a concentration prediction model based on the second concentration time series corresponding to each first leakage condition.

[0153] In one embodiment, the first generation module 520 is further configured to:

[0154] A three-dimensional model of the hydrogen storage system is established, and the three-dimensional model is spatially meshed. The mesh is used to simulate the fluid behavior of hydrogen in the three-dimensional model. Based on the preset boundary conditions and the meshed three-dimensional model, the hydrodynamic simulation is performed on multiple first hydrogen concentrations at each preset monitoring location to obtain the hydrogen diffusion model.

[0155] In one embodiment, the second generation module 540 is further configured to:

[0156] For any given first leakage condition, a first training data set corresponding to the second concentration time series is generated for each first leakage condition. The first training data set includes at least the first input data corresponding to the second concentration time series from the initial time to the first time in the second time period, and the first label data corresponding to the second concentration time series from the first time to the end time. The first input data is input into the initial concentration prediction model for training to predict the first predicted concentration time series of hydrogen from the first time to the end time. The model parameters of the initial concentration prediction model are iteratively updated based on the first label data and the first predicted concentration time series. If the initial concentration prediction model meets the convergence condition during the iterative update process, the current initial concentration prediction model is used as the concentration prediction model. If the initial concentration prediction model does not meet the convergence condition during the iterative update process, the process returns to the step of inputting the first input data into the initial concentration prediction model for training to predict the first predicted concentration time series of hydrogen from the first time to the end time, and subsequent steps, until the initial concentration prediction model meets the convergence condition.

[0157] In one embodiment, the duration from the initial time to the first time is greater than the duration from the first time to the end time.

[0158] In one embodiment, the second generation module 540 is further configured to:

[0159] Based on the first label data and the first predicted concentration time series, calculate the training loss of the initial concentration prediction model; iteratively update the model parameters based on the training loss.

[0160] In one embodiment, the concentration prediction model generation device 500 further includes:

[0161] The third acquisition module is used to acquire the third concentration time series of hydrogen during the third time period during the vehicle test.

[0162] The third generation module is used to generate second training data based on the third concentration time series; the second training data includes at least the second input data corresponding to the third concentration time series from the initial time to the second time in the third time period, and the second label data corresponding to the third concentration time series from the second time to the end time.

[0163] The iterative module is used to iteratively update the model parameters of the concentration prediction model based on the first training data and the second training data, so as to obtain the updated concentration prediction model.

[0164] When it is understood that, Figure 5 In the block diagram of the concentration prediction model generation device shown, each module is used to perform... Figures 1 to 4 The steps in the corresponding embodiments, and for Figures 1 to 4 The steps in the corresponding embodiments have been explained in detail in the above embodiments. Please refer to them for details. Figures 1 to 4 as well as Figures 1 to 4 The relevant descriptions in the corresponding embodiments will not be repeated here.

[0165] In another embodiment, after obtaining the concentration prediction model through the above embodiments, the vehicle can also use the concentration prediction model to predict the hydrogen concentration inside the vehicle in real time during driving, so as to provide early warning or hydrogen emission, ensuring driving safety and personal safety. Specifically, the vehicle can use the concentration prediction model as follows: Figure 6 The steps S601-S603 shown ensure driving safety and personal safety. Details are as follows:

[0166] S601. During vehicle operation, obtain the actual hydrogen concentration time series at the current moment; the actual concentration time series includes the actual hydrogen concentration at the current moment and multiple moments before the current moment.

[0167] In one embodiment, the actual hydrogen concentration at each preset monitoring location in the hydrogen storage system can be acquired when the vehicle is powered on and in motion. Then, as time changes, the actual hydrogen concentration at the current moment and the actual hydrogen concentration at multiple moments prior to the current moment can be used as a time series of the actual concentration at the current moment.

[0168] The actual hydrogen concentration at the preset monitoring location can be collected by the hydrogen concentration sensor described above. Examples of the actual hydrogen concentration collected at each time point can be found in the examples of each time point in the first concentration time series described above, and will not be explained further.

[0169] It should be noted that when the current time is the initial time, the actual hydrogen concentration at each time before the current time can be 0.

[0170] S602. Input the actual concentration time series into the concentration prediction model to obtain the third predicted concentration time series of hydrogen in the preset future time period.

[0171] In one embodiment, the above concentration prediction model is based on Figures 1 to 4 The method shown is used for training. The aforementioned preset future time period can be set according to actual conditions and is not limited thereto. For example, a third predicted concentration time series can be predicted within 1 minute or 10 minutes after the current moment. That is, it describes the concentration distribution of hydrogen gas that will diffuse within the hydrogen storage system within the preset future time period after the current moment.

[0172] S603. If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then a preset operation is executed; the preset operation includes at least a hydrogen discharge operation.

[0173] In one embodiment, the predicted hydrogen concentration is the predicted concentration in the third predicted concentration time series. The preset concentration threshold and preset operation can be set according to actual conditions, and are not limited thereto.

[0174] For example, the preset concentration threshold can be 3%, and the preset operation includes at least hydrogen discharge operation, and may also include early warning operation.

[0175] In one embodiment, the aforementioned warning operation may be one or more of the following: voice broadcast, closing the valve of the hydrogen storage system that supplies hydrogen to the outside, and sending an alarm message; there is no limitation on this. Furthermore, the aforementioned hydrogen purging operation includes, but is not limited to, using a fan system to purge the hydrogen in the hydrogen storage system, or opening the channel between the internal space of the hydrogen storage system and the external environment to discharge the hydrogen in the hydrogen storage system to the external environment.

[0176] The vehicle can stop performing hydrogen emission after a preset duration; alternatively, it can stop performing hydrogen emission when the predicted hydrogen concentrations included in the third predicted concentration time series within a preset future time period are all lower than a preset concentration threshold. There are no restrictions on this.

[0177] In this embodiment, during vehicle operation, a time series of actual hydrogen concentrations is generated by acquiring the actual hydrogen concentrations at the current moment and multiple moments prior to the current moment. This time series is then input into a concentration prediction model to obtain a third predicted hydrogen concentration time series for a preset future time period. Subsequently, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, a preset operation is performed, such as a hydrogen release operation. Thus, by predicting in advance whether the hydrogen concentration will be equal to or equal to the preset concentration threshold, a hydrogen release operation can be performed before the predicted hydrogen concentration exceeds or equals the preset concentration threshold, reducing the hydrogen concentration inside the vehicle and improving driving safety.

[0178] As an example, a fan system can be used inside the vehicle to perform the aforementioned hydrogen purging operation. For instance, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, the target accumulation location with the highest predicted hydrogen concentration is determined. Then, based on the target accumulation location, the target purging information for the fan system in the vehicle is determined, and the fan system is controlled to perform the hydrogen purging operation according to the target purging information.

[0179] The target purging information includes, but is not limited to, the purging angle of the blower system and the first purging wind speed of the blower system. Controlling the purging angle allows adjustment of the blower system's purging position to specifically purge hydrogen from the target accumulation location; and controlling the first purging wind speed allows control of the required energy consumption while maintaining the rate of hydrogen emission.

[0180] Specifically, the fan system may include a rotatable fan, or a fixed fan and a variable direction device. The variable direction device is used to adjust the blowing position of the fan system. The variable direction device can be a rotatable baffle or a rotatable fan blade; there is no limitation on the specific type.

[0181] In this embodiment, since a special rotating structure is required to support the rotation of the fan, the fan system can include a fixed fan (non-rotatable) and a reversing device to simplify the hardware structure. Furthermore, the reversing device is positioned between the fan and the hydrogen storage system to adjust the direction of the airflow generated by the fan.

[0182] Based on this, it can be considered that adjusting the direction-changing angle of the variable direction device is equivalent to controlling the purging angle described above.

[0183] As an example, the vehicle can determine the target turning angle of the variable direction device based on the target aggregation location and a first preset relationship, and define the target turning angle as the target purging information. Then, after adjusting the turning angle of the variable direction device to the target turning angle, the blower is controlled to operate.

[0184] In one embodiment, a first preset relationship is used to characterize the relationship between the reversing angle of the variable reversing device and a preset position range. Based on this, by controlling the reversing angle of the variable reversing device, the airflow direction during blower purging can be controlled to specifically reduce the hydrogen concentration at the target accumulation location.

[0185] The vehicle can pre-store multiple preset mapping relationships between location ranges and turning angles, serving as the first preset relationship. Then, based on the preset location range where the target aggregation location is located, the target turning angle is determined.

[0186] In one embodiment, the vehicle may further include a drive motor for supplying power to the fan to control the operation of the fan, and may also supply power to the directional device (rotatable partition) to control the directional angle of the directional device.

[0187] In another embodiment, in order to reduce the energy consumption of the blower system while rapidly reducing the predicted hydrogen concentration, the vehicle can also reasonably control the purge wind speed during blower operation. For example, the vehicle can determine a first purge wind speed for the blower based on the predicted hydrogen concentration at the target accumulation location and a second preset relationship, and define the first purge wind speed as the target purge information. Then, the purge wind speed during blower operation is controlled to be the first purge wind speed.

[0188] In one embodiment, the second preset relationship is used to characterize the relationship between the blower's purge velocity and a preset concentration range. Based on this, by controlling the blower's purge velocity, the rate of hydrogen discharge from the target accumulation location can be controlled. Furthermore, while ensuring the hydrogen discharge rate, a suitable purge velocity can be selected, reducing the energy consumption of the blower system.

[0189] The vehicle can pre-store multiple preset concentration ranges and purge wind speeds, serving as the second preset relationship. Then, based on the preset concentration range of the predicted hydrogen concentration at the target accumulation location, the first purge wind speed is determined.

[0190] Reference Figure 7 , Figure 7This is a schematic diagram of a hydrogen storage system and a blower system provided in one embodiment of this application. The hydrogen storage system mainly includes a high-pressure hydrogen storage cylinder, a hydrogen concentration sensor, and an ECU (Electronic Control Unit). The valve of the high-pressure hydrogen storage cylinder can be connected to the hydrogen concentration sensor and the ECU, and hydrogen concentration sensors can also be installed at other preset monitoring locations in the hydrogen storage system. The blower system includes a blower and a directional device (e.g., a rotatable partition), and the vehicle may also include a power-driven drive motor. The control commands for the blower system can be generated by the vehicle, for example, by the vehicle's Hardware Security Module (HSM) or the ECU. In this embodiment, the ECU is used as an example for explanation.

[0191] When a hydrogen leak occurs, the hydrogen concentration sensor transmits the detected actual hydrogen concentration to the ECU. The ECU combines the actual hydrogen concentration at the current moment with the actual hydrogen concentrations at previous moments to form a time series of actual concentration data. This actual concentration time series is then input into a concentration prediction model to obtain a third predicted concentration time series. Finally, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, a preset operation is executed.

[0192] For example, the ECU can first determine the target accumulation location with the highest predicted hydrogen concentration. Then, when performing a preset hydrogen purging operation at the target accumulation location, the ECU can determine the target turning angle of the variable direction device based on the target accumulation location. Simultaneously, it determines the first purging speed of the blower based on the predicted hydrogen concentration at the target accumulation location. Then, after controlling the turning angle of the variable direction device to the target turning angle, it controls the blower to operate at the first purging speed to quickly purge the hydrogen from the hydrogen storage system.

[0193] Based on the above explanation, the vehicle can perform the hydrogen purging operation by adjusting the steering angle while maintaining a constant purge airflow. Alternatively, it can perform the hydrogen purging operation by adjusting the purge airflow while maintaining a constant purge direction. Alternatively, it can perform the hydrogen purging operation by simultaneously adjusting both the steering angle and the purge airflow; there is no limitation on which approach is used.

[0194] Understandably, performing the hydrogen release operation reduces the actual hydrogen concentration at the current moment. Consequently, when the concentration prediction model subsequently makes predictions based on the reduced actual hydrogen concentration at the current moment, the predicted hydrogen concentration included in the third predicted concentration time series within the preset future time period will also decrease.

[0195] In another embodiment, the direction in which the fan blows can be the same as the windward direction in which the vehicle is moving. See details for further information. Figure 8 Example, Figure 8This is a schematic diagram of a hydrogen storage system and a fan system provided in one embodiment of this application. It is understood that when the vehicle is in motion, an airflow will be generated to purge the interior of the hydrogen storage system. Therefore, it can be assumed that during vehicle operation, the hydrogen inside the storage system will be naturally discharged under the action of the airflow.

[0196] Based on this, in order to further reduce the energy consumption required while lowering the predicted hydrogen concentration within the hydrogen storage system, the vehicle can first obtain its driving status. Then, when the vehicle is in a driving state, the first purging air speed is reduced to a second purging air speed. And when the vehicle is in a parked state, the first purging air speed used when the blower is operating is maintained.

[0197] The second purge velocity can be set according to actual conditions and is not limited thereto. For example, the second purge velocity can be determined based on vehicle speed. For instance, multiple vehicle speed ranges and corresponding wind speed reduction values ​​for each speed range can be preset. Then, the current vehicle speed is determined within its speed range to identify the corresponding target wind speed reduction value. Finally, the difference between the first purge velocity and the target wind speed reduction value is used as the second purge velocity. In other words, the second purge velocity can be reasonably adjusted according to the vehicle speed. This ensures that the second purge velocity can quickly expel hydrogen from the hydrogen storage system while reducing the energy consumption required by the blower system.

[0198] Among them, the larger the vehicle speed range, the greater the decrease in wind speed, and the lower the energy consumption required for the second purging wind speed when the fan is working.

[0199] Reference Figure 9 , Figure 9 This is a flowchart illustrating a concentration prediction model generation method and a concentration prediction method implementation method, provided in another embodiment of this application. The vehicle can obtain the first concentration time series during hydrogen diffusion under multiple preset first leakage conditions of the hydrogen storage system. Furthermore, a three-dimensional model of the hydrogen storage system is established, and the three-dimensional model is spatially meshed. Then, based on preset boundary conditions and the meshed three-dimensional model, hydrodynamic simulations can be performed on multiple first hydrogen concentrations at each preset monitoring location to obtain a hydrogen diffusion model that describes the concentration diffusion law of hydrogen in the hydrogen storage system over time. That is, the hydrogen diffusion model is obtained by solving the CFD equations.

[0200] Subsequently, based on the hydrogen diffusion model, the vehicle can determine the second concentration time series during the second time period under each first leakage condition. Then, a large number of second concentration time series are used as the first training data to initially train a concentration prediction model that meets the convergence conditions.

[0201] Furthermore, to improve the accuracy of the concentration prediction model for the third predicted concentration time series under any leakage condition, the vehicle can acquire the third concentration time series of hydrogen during the third time period during vehicle testing. Then, second training data is generated based on the third concentration time series, and the model parameters of the concentration prediction model are iteratively updated in combination with the first training data to obtain the updated concentration prediction model.

[0202] Finally, when the iteratively updated concentration prediction model is applied to vehicles, it can predict a third predicted hydrogen concentration time series within a preset future time period based on the actual hydrogen concentration time series at the current moment. Furthermore, if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, a preset operation is executed. For example, a hydrogen release operation is performed to proactively avoid the risk of hydrogen accumulation and improve driving safety.

[0203] Please see Figure 10 , Figure 10 This is a structural block diagram of a concentration prediction device provided in an embodiment of this application. The modules included in the concentration prediction device in this embodiment are used to perform… Figure 6 and Figure 9 The steps in the corresponding embodiments. Please refer to the details. Figure 6 and Figure 9 as well as Figure 6 and Figure 9 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 10 The concentration prediction device 1000 may include: a second acquisition module 1010, an input module 1020, and an execution module 1030, wherein:

[0204] The second acquisition module 1010 is used to acquire the actual concentration time series of hydrogen at the current moment during vehicle operation; the actual concentration time series includes the actual hydrogen concentration at the current moment and multiple moments before the current moment.

[0205] The input module 1020 is used to input the actual concentration time series into the concentration prediction model to obtain the third predicted concentration time series of hydrogen in a preset future time period; the concentration prediction model is generated according to the method of the first aspect.

[0206] The execution module 1030 is used to perform a preset operation if any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold; the preset operation includes at least a hydrogen venting operation.

[0207] In one embodiment, the execution module 1030 is further configured to:

[0208] If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then the target accumulation location with the highest predicted hydrogen concentration is determined; based on the target accumulation location, the target purging information of the blower system in the vehicle is determined; and the blower system is controlled to perform hydrogen discharge operation according to the target purging information.

[0209] Figure 11 This is a structural block diagram of a vehicle provided in one embodiment of this application. For example... Figure 11 As shown, the vehicle 1100 in this embodiment includes: a processor 1110, a memory 1120, and a computer program 1130 stored in the memory 1120 and executable on the processor 1110, such as a program for a concentration prediction model generation method or a concentration prediction method. When the processor 1110 executes the computer program 1130, it implements the steps in the various embodiments of the concentration prediction model generation methods described above, for example... Figure 1 S101 to S104 shown or Figure 6 S601 to 603 are shown. Alternatively, the processor 1110 implements the above when executing computer program 1130. Figure 5 or Figure 10 The functions of each module in the corresponding embodiments, for example, Figure 5 or Figure 10 For details on the functions of each module shown, please refer to [link / reference]. Figure 5 or Figure 10 The relevant descriptions in the corresponding embodiments.

[0210] For example, computer program 1130 can be divided into one or more modules, one or more modules are stored in memory 1120 and executed by processor 1110 to implement the concentration prediction model generation method or concentration prediction method provided in the embodiments of this application. One or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of computer program 1130 in vehicle 1100. For example, computer program 1130 can implement the concentration prediction model generation method or the concentration prediction method provided in the embodiments of this application.

[0211] Vehicle 1100 may include, but is not limited to, processor 1110 and memory 1120. Those skilled in the art will understand that... Figure 11 This is merely an example of vehicle 1100 and does not constitute a limitation on vehicle 1100. It may include more or fewer components than shown, or combine certain components, or different components. For example, a vehicle may also include input / output devices, network access devices, buses, etc.

[0212] The processor 1110 may be a central processing unit, or it may be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0213] The memory 1120 can be an internal storage unit of the vehicle 1100, such as a hard drive or memory of the vehicle 1100. The memory 1120 can also be an external storage device of the vehicle 1100, such as a plug-in hard drive, smart memory card, flash memory card, etc., equipped on the vehicle 1100. Furthermore, the memory 1120 can include both internal storage units and external storage devices of the vehicle 1100.

[0214] This application provides a computer-readable storage medium, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the concentration prediction model generation method or concentration prediction method as described in the above embodiments.

[0215] This application provides a computer program product that, when run on a vehicle, causes the vehicle to execute the concentration prediction model generation method or concentration prediction method described in the above embodiments.

[0216] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating a concentration prediction model, characterized in that, The method includes: Under multiple preset first leakage conditions of the hydrogen storage system, the first concentration time series during hydrogen diffusion is obtained; the first concentration time series includes multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period; Based on the first concentration time series, a hydrogen diffusion model is generated; the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system space over time. Based on the hydrogen diffusion model, a second concentration time series is determined for each of the first leakage conditions; the second concentration time series includes multiple second hydrogen concentrations changing at multiple preset monitoring locations within a second time period; the duration of the second time period is longer than the duration of the first time period. A concentration prediction model is generated based on the second concentration time series corresponding to each of the first leakage conditions.

2. The method according to claim 1, characterized in that, The step of generating a hydrogen diffusion model based on the first concentration time series includes: A three-dimensional model of the hydrogen storage system is established, and the three-dimensional model is divided into spatial meshes; the meshes are used to simulate the fluid behavior of hydrogen in the space of the three-dimensional model. Based on the preset boundary conditions and the divided three-dimensional model, a hydrodynamic simulation is performed on multiple first hydrogen concentrations at each preset monitoring location to obtain the hydrogen diffusion model.

3. The method according to claim 1, characterized in that, The step of generating a concentration prediction model based on the second concentration time series corresponding to each of the first leakage conditions includes: For any of the first leakage conditions, generate first training data for each of the first leakage conditions corresponding to the second concentration time series; the first training data includes at least the first input data corresponding to the second concentration time series from the initial time to the first time in the second time period, and the first label data corresponding to the second concentration time series from the first time to the end time. The first input data is input into the initial concentration prediction model for training, and the first predicted concentration time series of hydrogen gas from the first time to the end time is predicted. Based on the first label data and the first predicted concentration time series, the model parameters of the initial concentration prediction model are iteratively updated; If the initial concentration prediction model meets the convergence condition during the iterative update process, then the current initial concentration prediction model is used as the concentration prediction model. If the initial concentration prediction model does not meet the convergence condition during the iterative update process, then return to the step of inputting the first input data into the initial concentration prediction model for training and predicting the first predicted concentration time series of hydrogen from the first time to the end time, and subsequent steps, until the initial concentration prediction model meets the convergence condition.

4. The method according to claim 3, characterized in that, The duration from the initial time to the first time is greater than the duration from the first time to the end time.

5. The method according to claim 3, characterized in that, The step of iteratively updating the model parameters of the initial concentration prediction model based on the first label data and the first predicted concentration time series includes: Based on the first label data and the first predicted concentration time series, calculate the training loss of the initial concentration prediction model; The model parameters are iteratively updated based on the training loss.

6. The method according to claim 3, characterized in that, After generating the concentration prediction model based on the second concentration time series corresponding to each of the first leakage conditions, the process includes: During the vehicle test, the third concentration time series of the hydrogen gas was obtained in the third time period; The second training data is generated based on the third concentration time series; the second training data includes at least the second input data corresponding to the third concentration time series from the initial time to the second time in the third time period, and the second label data corresponding to the third concentration time series from the second time to the end time. The model parameters of the concentration prediction model are iteratively updated based on the first training data and the second training data to obtain the updated concentration prediction model.

7. A concentration prediction method, characterized in that, The method includes: During vehicle operation, the actual hydrogen concentration time series at the current moment is obtained; the actual concentration time series includes the actual hydrogen concentration at the current moment and multiple moments prior to the current moment. The actual concentration time series is input into the concentration prediction model to obtain the third predicted concentration time series of hydrogen in a preset future time period; the concentration prediction model is generated according to the method described in any one of claims 1-6. If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then a preset operation is performed; the preset operation includes at least a hydrogen purging operation.

8. The method according to claim 7, characterized in that, If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then a preset operation is performed, including: If any predicted hydrogen concentration in the third predicted concentration time series is greater than or equal to a preset concentration threshold, then the target accumulation location with the highest predicted hydrogen concentration is determined. Based on the target aggregation location, the target purging information of the vehicle's blower system is determined; The blower system is controlled to perform the hydrogen discharge operation based on the target purging information.

9. A concentration prediction model generation device, characterized in that, The device includes: The first acquisition module is used to acquire the first concentration time series of hydrogen during diffusion under multiple preset first leakage conditions of the hydrogen storage system; the first concentration time series includes multiple first hydrogen concentrations changing at multiple preset monitoring locations within a first time period; The first generation module is used to generate a hydrogen diffusion model based on the first concentration time series; the hydrogen diffusion model is used to describe the concentration diffusion law of hydrogen in the hydrogen storage system space over time. The first determining module is used to determine, according to the hydrogen diffusion model, a second concentration time series during hydrogen diffusion under each of the first leakage conditions; the second concentration time series includes multiple second hydrogen concentrations changing at multiple preset monitoring locations within a second time period; the duration of the second time period is longer than the duration of the first time period; The second generation module is used to generate a concentration prediction model based on the second concentration time series corresponding to each of the first leakage conditions.

10. A vehicle comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-6 or 7-8.

Images