A hydrogen leakage monitoring and early warning method and related device

Abstract

The application discloses a hydrogen leakage monitoring and early warning method and related devices, first, the fluid flow mathematical physics model is corrected by using the hydrogen leakage monitoring sensor measurement data of the reduced model test bench and the simulation results of the fluid dynamics simulation test, the accuracy of the fluid flow mathematical physics model and the rationality of the monitoring point setting are guaranteed, then the loss function of the initial physical information neural network model is constructed by using the corrected target fluid flow mathematical physics model, the initial physical information neural network model is trained based on the simulation data of the hydrogen leakage event of the target fluid flow mathematical physics model under the target hydrogen leakage monitoring field scene and the loss function, the target physical information neural network model is obtained, finally, the target hydrogen leakage monitoring field scene is monitored and early warned by using the target physical information neural network model. The technical problems of poor accuracy and insufficient real-time of the existing hydrogen leakage monitoring and early warning method are solved.

Description

Technical Field

[0001] This invention relates to the field of chemical safety monitoring technology, and in particular to a method and device for monitoring and early warning of hydrogen leaks. Background Technology

[0002] Hydrogen energy is a clean energy source and an important component of the energy structure. However, hydrogen leaks can easily cause accidents such as explosions and fires. Therefore, the monitoring, early warning, and prevention of hydrogen leaks are of paramount importance.

[0003] Hydrogen gas possesses characteristics such as a high diffusion coefficient, light molecular weight, wide explosion limits, and low minimum ignition energy. However, conventional sensors can only detect hydrogen concentration at a single point in space, failing to provide a clear visualization of the temporal and spatial evolution of hydrogen concentration in three-dimensional space, let alone predict the subsequent development of hydrogen leaks under different scenarios, thus offering limited guidance for emergency response after a leak. Furthermore, for equipment such as hydrogen fuel cell vehicles and mobile emergency power supplies, the operating and refueling processes involve numerous scenarios and complex boundary conditions. Simply using numerical simulations to build models for detecting hydrogen leaks results in poor accuracy and insufficient real-time performance. Therefore, improving the accuracy and real-time performance of hydrogen leak monitoring and early warning is a crucial technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] This invention provides a hydrogen leak monitoring and early warning method and related device to solve the technical problems of poor accuracy and insufficient real-time performance of existing hydrogen leak monitoring and early warning methods.

[0005] In view of this, the first aspect of the present invention provides a method for monitoring and early warning of hydrogen leakage, comprising:

[0006] Based on the target hydrogen leak monitoring site scenario, a scaled-down model test bench was established, and hydrogen leak monitoring sensors were arranged at preset positions on the scaled-down model test bench.

[0007] The scaled-down model test bench was modeled and fluid dynamics simulation experiments were conducted in computational fluid dynamics simulation software to obtain simulation results;

[0008] Based on the measurement data from the hydrogen leak monitoring sensor and the simulation results, a mathematical and physical model for the target fluid flow is determined.

[0009] The loss function for constructing the initial physical information neural network model based on the target fluid flow mathematical and physical model;

[0010] Based on the simulation data of hydrogen leakage events in the target hydrogen leakage monitoring site scenario using the target fluid flow mathematical physics model and the loss function, the initial physical information neural network model is trained to obtain the target physical information neural network model.

[0011] Based on the target physical information neural network model, hydrogen leakage monitoring and early warning are performed on the target hydrogen leakage monitoring site.

[0012] Optionally, the scaled-down model test bench is modeled and subjected to fluid dynamics simulation experiments in computational fluid dynamics simulation software to obtain simulation results, including:

[0013] S1. Construct the geometric model of the scaled-down model test bench in computational fluid dynamics simulation software;

[0014] S2. Mesh the geometric model;

[0015] S3. Establish the initial fluid flow mathematical and physical model of the geometric model after mesh generation;

[0016] S4. Set the boundary conditions and initial conditions for the initial fluid flow mathematical and physical model;

[0017] S5. Perform mesh independence verification on the geometric model. If the verification passes, perform flow field simulation on the geometric model to obtain simulation results. If the verification fails, return to step S2.

[0018] Optionally, based on the measurement data from the hydrogen leak monitoring sensor and the simulation results, a mathematical and physical model of the target fluid flow is determined, including:

[0019] T1. Compare the measurement data of the hydrogen leak monitoring sensor with the simulation results of the monitoring points corresponding to the initial fluid flow mathematical and physical model;

[0020] T2. Determine whether the relative deviation between the simulation result data and the measurement data is less than the error threshold. If yes, proceed to step T3; otherwise, proceed to step T4.

[0021] T3. Determine the current position of the hydrogen leak monitoring sensor as the target position, and the initial fluid flow mathematical and physical model as the target fluid flow mathematical and physical model;

[0022] T4. Based on the measurement data, correct the initial fluid flow mathematical and physical model, adjust the placement of the hydrogen leak monitoring sensor, and return to step S4.

[0023] Optionally, the mathematical and physical model of the target fluid flow is:

[0024]

[0025] in, For the density of the fluid, Let t be the velocity vector of the fluid, t be time, and p be the pressure of the fluid. The dynamic viscosity coefficient of the fluid. For volume forces, Let be the specific heat capacity of the fluid, T be the temperature of the fluid, k be the thermal conductivity of the fluid, and Q be the heat source term per unit volume. The term represents the heat source for chemical reactions per unit volume. Let be the volume fraction or mass fraction of the i-th component, and D be the diffusion coefficient of the component. Let i be the source term of the i-th component.

[0026] Optionally, the loss function of the initial physical information neural network model is:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. All are weighting coefficients. The number of monitoring points, Let j be the predicted value of the physical quantity at the j-th monitoring point. Let j be the true value of the physical quantity at the j-th monitoring point. Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

[0034] Optionally, based on the target physical information neural network model, hydrogen leakage monitoring and early warning are performed on the target hydrogen leakage monitoring site scenario, and the process further includes:

[0035] Based on the measurement data of the hydrogen leakage monitoring sensor in the target hydrogen leakage monitoring site scenario, the mathematical and physical model of the target fluid flow is corrected;

[0036] The target physical information neural network model is corrected based on the corrected target fluid flow mathematical and physical model;

[0037] The corrected target physical information neural network model is used to perform hydrogen leakage monitoring and early warning for the target hydrogen leakage monitoring site.

[0038] Optionally, based on the target physical information neural network model, hydrogen leakage monitoring and early warning are performed on the target hydrogen leakage monitoring site scenario, and the process further includes:

[0039] Based on the prediction results of the target physical information neural network model, it is determined whether there are false alarms in the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site. If so, the false alarm hydrogen leak monitoring sensors are corrected or replaced.

[0040] A second aspect of the present invention provides a hydrogen leak monitoring and early warning device, comprising:

[0041] The scaled-down model building module is used to establish a scaled-down model test bench based on the target hydrogen leak monitoring site scenario, and to arrange hydrogen leak monitoring sensors at preset positions on the scaled-down model test bench.

[0042] The fluid dynamics simulation module is used to model the scaled-down test bench and conduct fluid dynamics simulation experiments in computational fluid dynamics simulation software to obtain simulation results;

[0043] The model calibration module is used to determine the mathematical and physical model of the target fluid flow based on the measurement data of the hydrogen leakage monitoring sensor and the simulation results.

[0044] The loss function construction module is used to construct the loss function of the initial physical information neural network model based on the target fluid flow mathematical and physical model;

[0045] The prediction network training module is used to train the initial physical information neural network model based on the simulation data of the hydrogen leakage event in the target hydrogen leakage monitoring site scenario and the loss function, using the mathematical and physical model of the target fluid flow and the loss function, to obtain the target physical information neural network model.

[0046] The monitoring module is used to perform hydrogen leakage monitoring and early warning based on the target physical information neural network model of the target hydrogen leakage monitoring site.

[0047] Optionally, the fluid dynamics simulation module is specifically used to perform the following steps:

[0048] S1. Construct the geometric model of the scaled-down model test bench in computational fluid dynamics simulation software;

[0049] S2. Mesh the geometric model;

[0050] S3. Establish the initial fluid flow mathematical and physical model of the geometric model after mesh generation;

[0051] S4. Set the boundary conditions and initial conditions for the initial fluid flow mathematical and physical model;

[0052] S5. Perform mesh independence verification on the geometric model. If the verification passes, perform flow field simulation on the geometric model to obtain simulation results. If the verification fails, return to step S2.

[0053] Optionally, the model correction module is specifically used to perform the following steps:

[0054] T1. Compare the measurement data of the hydrogen leak monitoring sensor with the simulation results of the monitoring points corresponding to the initial fluid flow mathematical and physical model;

[0055] T2. Determine whether the relative deviation between the simulation result data and the measurement data is less than the error threshold. If yes, proceed to step T3; otherwise, proceed to step T4.

[0056] T3. Determine the current position of the hydrogen leak monitoring sensor as the target position, and the initial fluid flow mathematical and physical model as the target fluid flow mathematical and physical model;

[0057] T4. Based on the measurement data, correct the initial fluid flow mathematical and physical model, adjust the placement of the hydrogen leak monitoring sensor, and return to step S4.

[0058] Optionally, the mathematical and physical model of the target fluid flow is:

[0059]

[0060] in, For the density of the fluid, Let t be the velocity vector of the fluid, t be time, and p be the pressure of the fluid. The dynamic viscosity coefficient of the fluid. For volume forces, Let be the specific heat capacity of the fluid, T be the temperature of the fluid, k be the thermal conductivity of the fluid, and Q be the heat source term per unit volume. The term represents the heat source for chemical reactions per unit volume. Let be the volume fraction or mass fraction of the i-th component, and D be the diffusion coefficient of the component. Let i be the source term of the i-th component.

[0061] Optionally, the loss function of the initial physical information neural network model is:

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. All are weighting coefficients. The number of monitoring points, Let j be the predicted value of the physical quantity at the j-th monitoring point. Let j be the true value of the physical quantity at the j-th monitoring point. Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

[0069] Optionally, the monitoring module is further configured to:

[0070] Based on the measurement data of the hydrogen leakage monitoring sensor in the target hydrogen leakage monitoring site scenario, the mathematical and physical model of the target fluid flow is corrected;

[0071] The target physical information neural network model is corrected based on the corrected target fluid flow mathematical and physical model;

[0072] The corrected target physical information neural network model is used to perform hydrogen leakage monitoring and early warning for the target hydrogen leakage monitoring site.

[0073] Optionally, it also includes a false alarm handling module, which is used for:

[0074] Based on the prediction results of the target physical information neural network model, it is determined whether there are false alarms in the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site. If so, the false alarm hydrogen leak monitoring sensors are corrected or replaced.

[0075] A third aspect of the present invention provides a hydrogen leak monitoring and early warning device, the device comprising a processor and a memory:

[0076] The memory is used to store program code and transmit the program code to the processor;

[0077] The processor is used to execute any one of the hydrogen leakage monitoring and early warning methods described in the first aspect according to the instructions in the program code.

[0078] A fourth aspect of the present invention provides a computer-readable storage medium for storing program code for executing the hydrogen leakage monitoring and early warning method according to any one of the first aspects.

[0079] As can be seen from the above technical solutions, the hydrogen leakage monitoring and early warning method provided by the present invention has the following advantages:

[0080] The hydrogen leak monitoring and early warning method provided by this invention first uses measurement data from hydrogen leak monitoring sensors on a scaled-down model test bench and simulation results from fluid dynamics simulation experiments to calibrate the mathematical and physical model of fluid flow, ensuring the accuracy of the fluid flow mathematical and physical model and the rationality of the monitoring point settings. Then, using the calibrated target fluid flow mathematical and physical model, a loss function is constructed for an initial physical information neural network model. Based on simulation data of hydrogen leak events in a target hydrogen leak monitoring scenario and the loss function, the initial physical information neural network model is trained to obtain the target physical information neural network model. Finally, the target physical information neural network model is used to perform hydrogen leak monitoring and early warning for the target hydrogen leak monitoring scenario. This invention combines fluid dynamics simulation with a physical information neural network model, enabling rapid and accurate hydrogen leak monitoring and early warning for hydrogen leak monitoring scenarios, solving the technical problems of poor accuracy and insufficient real-time performance in existing hydrogen leak monitoring and early warning methods.

[0081] Meanwhile, the hydrogen leak monitoring and early warning method provided by this invention adjusts the arrangement position of the hydrogen leak monitoring sensor when the relative deviation between the simulation results of the fluid dynamics simulation test and the measurement data of the hydrogen leak monitoring sensor is greater than the error threshold. The method then corrects the mathematical and physical model of fluid flow based on the measurement data of the hydrogen leak monitoring sensor after the adjustment, until the relative deviation between the simulation results of the fluid dynamics simulation test and the measurement data of the hydrogen leak monitoring sensor is within the error threshold. This achieves bidirectional correction of the position of the hydrogen leak monitoring sensor and the mathematical and physical model of fluid flow.

[0082] The hydrogen leak monitoring and early warning method provided by this invention can use a target physical information neural network model to judge false alarms of hydrogen leak monitoring sensors in hydrogen leak monitoring scenarios, thereby reducing the false alarm rate of hydrogen leaks in hydrogen leak monitoring scenarios.

[0083] The hydrogen leak monitoring and early warning method provided by this invention can correct the mathematical and physical model of the target fluid flow and the target physical information neural network model based on the measurement data of the hydrogen leak monitoring sensor in the target hydrogen leak monitoring scene, thereby further improving the accuracy of hydrogen leak monitoring and early warning of the target physical information neural network model. Attached Figure Description

[0084] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, 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 the present invention. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0085] Figure 1This is a flowchart illustrating a hydrogen leak monitoring and early warning method provided in an embodiment of the present invention;

[0086] Figure 2 This is a schematic diagram of the fluid dynamics simulation test process provided in the embodiments of the present invention;

[0087] Figure 3 This is a schematic diagram of the fluid dynamics simulation test process and the dual calibration process of the hydrogen leakage monitoring sensor position and the mathematical and physical model of fluid flow on the scaled-down model test bench provided in the embodiments of the present invention.

[0088] Figure 4 This is a schematic diagram of the physical information neural network model provided in an embodiment of the present invention;

[0089] Figure 5 This is a schematic diagram of the structure of a hydrogen leak monitoring and early warning device provided in an embodiment of the present invention;

[0090] Figure 6 This is a schematic diagram of the structure of a hydrogen leak monitoring and early warning device provided in an embodiment of the present invention. Detailed Implementation

[0091] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0092] For easier understanding, please refer to Figure 1 and Figure 2 This invention provides an embodiment of a hydrogen leak monitoring and early warning method, comprising:

[0093] Step 101: Based on the target hydrogen leak monitoring site scenario, establish a scaled-down model test bench and place hydrogen leak monitoring sensors at preset positions on the scaled-down model test bench.

[0094] It should be noted that, firstly, a scaled-down model test bench, identical to the target hydrogen leak monitoring scenario, is constructed according to a preset ratio. Then, hydrogen leak monitoring sensors are placed at predetermined locations within the scaled-down model test bench. Based on the high diffusivity and buoyancy of hydrogen in air, the sensors are positioned above the leak point and downstream of the gas flow. The hydrogen leak monitoring sensors can detect parameters such as hydrogen concentration, gas velocity in the flow field, and flow field temperature.

[0095] Step 102: Model the scaled-down test bench in computational fluid dynamics simulation software and conduct fluid dynamics simulation experiments to obtain simulation results.

[0096] It should be noted that computational fluid dynamics (CFD) simulation software is used. In this invention, CFD simulation software is used to model a scaled-down test bench and conduct fluid dynamics simulation experiments to obtain simulation results. The basic principle of CFD simulation is to numerically solve the differential equations controlling the fluid flow, obtain the discrete distribution of the fluid flow field in a continuous region, and thus approximately simulate the fluid flow situation.

[0097] Specifically, such as Figure 2 As shown, the CFD simulation process includes the following steps:

[0098] S1. Construct the geometric model of the scaled-down model test bench in computational fluid dynamics simulation software.

[0099] S2. Mesh the geometric model.

[0100] S3. Establish the initial mathematical and physical model of fluid flow after mesh generation. The initial mathematical and physical model of fluid flow used in this invention is the differential equation governing the fluid flow. This differential equation adopts the Navier-Stokes equations, specifically expressed as:

[0101]

[0102] in, For the density of the fluid, Let t be the velocity vector of the fluid, t be time, and p be the pressure of the fluid. The dynamic viscosity coefficient of the fluid. For volume forces, Let be the specific heat capacity of the fluid, T be the temperature of the fluid, k be the thermal conductivity of the fluid, and Q be the heat source term per unit volume. The term representing the heat source per unit volume of a chemical reaction indicates the heat released or absorbed during the chemical reaction. Let be the volume fraction or mass fraction of the i-th component, and D be the diffusion coefficient of the component. For the i-th component, the source term includes the rate of formation or consumption of the chemical reaction.

[0103] S4. Set the boundary conditions and initial conditions for the initial fluid flow mathematical physics model.

[0104] S5. Verify the mesh independence of the geometric model. If the verification passes, perform flow field simulation on the geometric model to obtain simulation results. If the verification fails, return to step S2. The mesh independence verification passes when the simulation results do not change with the mesh. If the simulation results change with the mesh, the mesh independence verification fails, and you need to return to step S2 to re-mesh the model.

[0105] The simulation results obtained by modeling a scaled-down test bench using CFD simulation software and conducting fluid dynamics simulation experiments include the velocity, fluid pressure, fluid temperature, and fluid component concentration in the x, y, and z directions of the fluid flow within the flow field.

[0106] Step 103: Based on the measurement data and simulation results of the hydrogen leakage monitoring sensor, determine the mathematical and physical model of the target fluid flow.

[0107] It should be noted that the measurement data and simulation results of the hydrogen leak monitoring sensors arranged in the scaled-down model test bench are compared. Based on the comparison results, the rationality of the current arrangement of the hydrogen leak monitoring sensors and the rationality of the fluid flow mathematical and physical model used in the current CFD simulation are determined, thereby identifying the target location of the hydrogen leak monitoring sensors and the target fluid flow mathematical and physical model. Specifically, such as... Figure 3 As shown, the process of determining the target location of the hydrogen leak monitoring sensor and the mathematical and physical model of the target fluid flow includes the following steps:

[0108] T1. Compare the measurement data of the hydrogen leak monitoring sensor with the simulation results of the monitoring points corresponding to the initial fluid flow mathematical and physical model.

[0109] T2. Determine whether the relative deviation between the simulation result data and the measurement data is less than the error threshold. If yes, proceed to step T3; otherwise, proceed to step T4. The error threshold is set to 5%.

[0110] T3. Determine the current position of the hydrogen leak monitoring sensor as the target position, and use the initial fluid flow mathematical and physical model as the target fluid flow mathematical and physical model.

[0111] T4. Based on the measurement data, correct the initial fluid flow mathematical and physical model, adjust the placement of the hydrogen leak monitoring sensor, and return to step S4. Specifically, when the relative deviation between the simulation results and the measurement data is not within the error threshold, adjust the parameters of the initial fluid flow mathematical and physical model according to the measurement data of the hydrogen leak monitoring sensor. Simultaneously, adjust the placement of the hydrogen leak monitoring sensor in the model test rig according to the location of higher hydrogen concentration in the CFD simulation results, and return to step S4 in step 102 until the relative deviation between the simulation results and the measurement data is less than the error threshold. The target fluid flow mathematical and physical model obtained after adjusting the parameters of the initial fluid flow mathematical and physical model has the same representation as the initial fluid flow mathematical and physical model, only the parameter values ​​have changed. When the relative deviation between the simulation results of the fluid dynamics simulation experiment and the measurement data of the hydrogen leak monitoring sensor is greater than the error threshold, the placement of the hydrogen leak monitoring sensor is adjusted, and the mathematical and physical model of fluid flow is corrected based on the measurement data of the hydrogen leak monitoring sensor after the adjustment, until the relative deviation between the simulation results of the fluid dynamics simulation experiment and the measurement data of the hydrogen leak monitoring sensor is within the error threshold, thus realizing the bidirectional correction of the position of the hydrogen leak monitoring sensor and the mathematical and physical model of fluid flow.

[0112] Step 104: Construct the loss function of the initial physical information neural network model based on the mathematical and physical model of the target fluid flow.

[0113] It should be noted that the Physics-Informed Neural Networks (PINN) model is a machine learning model that combines deep learning and knowledge of physics. The structure of the PINN model used in this invention is as follows: Figure 4 As shown, the neural network is a fully connected layer network with four input neurons, corresponding to the fluid flow time information (t) and coordinate information (x, y, z), respectively. The output consists of six neurons, corresponding to the fluid flow velocity in the x, y, and z directions (u, v, w), fluid pressure (p), fluid temperature (T), and fluid component concentration (X), respectively. In this invention, a loss function for the physical information neural network model is constructed using the mathematical physics model of the target fluid flow. Physical information constraints are introduced into the loss function of the physical information neural network model, allowing the neural network to be trained both at the data level and at the physical law level. The network parameters are updated through backpropagation of the loss function. Therefore, the loss function of the physical information neural network model in this invention includes a data term loss function and a physical law term loss function, which can be expressed as:

[0114]

[0115] in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. The weighting coefficients of the loss function for the data items are between 0.1 and 50.

[0116] Data Item Loss Function Defined as the mean square error between the predicted value and the actual value (i.e., the value of the CFD simulation result), it can be expressed as:

[0117]

[0118] in, The number of monitoring points, For the predicted value of the physical quantity (velocity, pressure, temperature, or concentration) at the j-th monitoring point, This represents the true value of the physical quantity at the j-th monitoring point. A monitoring point is a point where the values ​​of the physical quantities in the flow field are known.

[0119] Physical law term loss function Defined as:

[0120]

[0121] in, Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . These are all weighting coefficients, with values ​​ranging from 0.1 to 50.

[0122] , and The definitions are as follows:

[0123]

[0124]

[0125]

[0126] in, The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

[0127] Step 105: Based on the simulation data and loss function of the hydrogen leakage event in the target hydrogen leakage monitoring site scenario using the mathematical and physical model of the target fluid flow, train the initial physical information neural network model to obtain the target physical information neural network model.

[0128] It should be noted that a CFD simulation of a hydrogen leak event under the target hydrogen leak monitoring scenario is performed using a mathematical and physical model of the target fluid flow, yielding corresponding simulation results. The simulation results and the loss function constructed in step 104 are then used to train the initial physical information neural network model. Specifically, the training process is as follows:

[0129] (1) Set an appropriate number of hidden layers and the number of neurons in each hidden layer. Depending on the complexity of the equipment scenarios involved, such as hydrogen fuel cell vehicles and hydrogen fuel mobile emergency power supplies, the number of hidden layers in this invention is set to 5-50 layers, and the number of neurons in each hidden layer is set to 50-500.

[0130] (2) Configure the tanh function as the activation function of the initial physical information neural network model.

[0131] (3) Configure a first-order optimization algorithm optimizer (SGD, Adam, RMSprop Adam, etc.) as the optimizer for the initial physical information neural network model.

[0132] (4) The initial learning rate of the initial physical information neural network model is set to 0.0001~0.01, and the number of training iterations is set to 100,000~2,000,000.

[0133] (5) Input the simulation results obtained by CFD simulation of hydrogen leakage events in the target hydrogen leakage monitoring site into the initial physical information neural network model, and use the back propagation of the loss function constructed in step 104 to update the network parameters of the initial physical information neural network model.

[0134] (6) Adjust the hyperparameters of the initial physical information neural network model, including the learning rate, batch size, and number of iterations. The learning rate is a parameter that controls the step size of the model weight update. If the learning rate is too large, it may cause the model to be unstable or even diverge during training. If the learning rate is too small, it may cause the training process to be slow or get stuck in a local optimum. Therefore, the learning rate needs to be adjusted to an appropriate size during model training. The batch size determines the number of samples used each time the model parameters are updated. A larger batch size can provide more accurate gradient estimation, but the computational cost is higher and more memory may be required. A smaller batch size may train faster, but may cause training instability. Therefore, the choice of batch size needs to be balanced between computational efficiency and gradient estimation accuracy. The number of iterations (Epochs) is the number of times the entire dataset is used to train the model. Too many iterations may lead to overfitting, while too few may lead to underfitting. Therefore, the number of iterations needs to be adjusted to an appropriate size during model training.

[0135] (7) Adjust the number of hidden layers and neurons of the initial physical information neural network model according to different application requirements (different computational accuracy, computational speed, and computational resources) until a satisfactory prediction result is obtained, and obtain the target physical information neural network model.

[0136] Step 106: Based on the target physical information neural network model, perform hydrogen leakage monitoring and early warning for the target hydrogen leakage monitoring site.

[0137] It should be noted that after obtaining the target physical information neural network model, the target physical information neural network model is applied to the target hydrogen leakage monitoring site scenario. The target physical information neural network model is used to predict the hydrogen concentration information of the target hydrogen leakage monitoring site scenario to achieve hydrogen leakage monitoring and early warning of the target hydrogen leakage monitoring site scenario.

[0138] The hydrogen leak monitoring and early warning method provided by this invention first uses measurement data from hydrogen leak monitoring sensors on a scaled-down model test bench and simulation results from fluid dynamics simulation experiments to calibrate the mathematical and physical model of fluid flow, ensuring the accuracy of the fluid flow mathematical and physical model and the rationality of the monitoring point settings. Then, using the calibrated target fluid flow mathematical and physical model, a loss function is constructed for an initial physical information neural network model. Based on simulation data of hydrogen leak events in a target hydrogen leak monitoring scenario and the loss function, the initial physical information neural network model is trained to obtain the target physical information neural network model. Finally, the target physical information neural network model is used to perform hydrogen leak monitoring and early warning for the target hydrogen leak monitoring scenario. This invention combines fluid dynamics simulation with a physical information neural network model, enabling rapid and accurate hydrogen leak monitoring and early warning for hydrogen leak monitoring scenarios, solving the technical problems of poor accuracy and insufficient real-time performance in existing hydrogen leak monitoring and early warning methods.

[0139] In one embodiment, after step 106, the following may also be included:

[0140] Step 107: Based on the measurement data of the hydrogen leakage monitoring sensor in the target hydrogen leakage monitoring field scenario, correct the mathematical and physical model of the target fluid flow.

[0141] Step 108: Correct the target physical information neural network model based on the corrected target fluid flow mathematical and physical model.

[0142] Step 109: Use the corrected target physical information neural network model to perform hydrogen leakage monitoring and early warning for the target hydrogen leakage monitoring site.

[0143] It should be noted that the initial target fluid flow mathematical and physical model was obtained based on a scaled-down model test bench. The data from this scaled-down model test bench may contain errors compared to the actual data from the target hydrogen leak monitoring site. Therefore, to further improve the accuracy of hydrogen leak monitoring and early warning, after applying the initially obtained target physical information neural network model to the target hydrogen leak monitoring site, the target fluid flow mathematical and physical model can be corrected and updated based on the measurement data from the hydrogen leak monitoring sensors in that site. Then, the corrected target fluid flow mathematical and physical model is used to correct and update the target physical information neural network model. Finally, the corrected target physical information neural network model is used to perform hydrogen leak monitoring and early warning for the target hydrogen leak monitoring site, thus ensuring the fit between the prediction model and the actual application scenario, and further improving the accuracy and reliability of the hydrogen leak monitoring and early warning provided by the target physical information neural network model.

[0144] In one embodiment, after step 106, the following may also be included:

[0145] Step 110: Based on the prediction results of the target physical information neural network model, determine whether there are false alarms in the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site. If so, correct or replace the false alarm hydrogen leak monitoring sensors.

[0146] It should be noted that by comparing the hydrogen leak prediction results of the target physical information neural network model with the measurement data of the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site, if the prediction result of the target physical information neural network model indicates that no hydrogen leak has occurred, but the hydrogen leak monitoring sensor has issued a hydrogen leak alarm, it can be determined that the hydrogen leak monitoring sensor at the corresponding location has issued a false alarm. The location of the false alarm hydrogen leak monitoring sensor can be used to investigate and correct or replace the false alarm hydrogen leak monitoring sensor. Using the target physical information neural network model to identify false alarms of hydrogen leak monitoring sensors at the hydrogen leak monitoring site can reduce the false alarm rate of hydrogen leaks at the hydrogen leak monitoring site.

[0147] Step 110 can be performed after step 109 or before step 107.

[0148] For easier understanding, please refer to Figure 5 This invention provides an embodiment of a hydrogen leak monitoring and early warning device, comprising:

[0149] The scaled-down model building module is used to build a scaled-down model test bench based on the target hydrogen leak monitoring site scenario, and to place hydrogen leak monitoring sensors at preset positions on the scaled-down model test bench.

[0150] The fluid dynamics simulation module is used to model a scaled-down test bench and conduct fluid dynamics simulation experiments in computational fluid dynamics simulation software to obtain simulation results;

[0151] The model calibration module is used to determine the mathematical and physical model of the target fluid flow based on the measurement data and simulation results of the hydrogen leak monitoring sensor.

[0152] The loss function construction module is used to construct the loss function of the initial physical information neural network model based on the mathematical and physical model of the target fluid flow.

[0153] The prediction network training module is used to train the initial physical information neural network model based on the simulation data and loss function of the hydrogen leakage event in the target hydrogen leakage monitoring site scenario, based on the mathematical and physical model of the target fluid flow, to obtain the target physical information neural network model.

[0154] The monitoring module is used to monitor and warn of hydrogen leaks at the target hydrogen leak monitoring site based on a neural network model of the target's physical information.

[0155] The fluid dynamics simulation module is specifically used to perform the following steps:

[0156] S1. Construct the geometric model of the scaled-down model test bench in computational fluid dynamics simulation software;

[0157] S2. Mesh the geometric model;

[0158] S3. Establish the initial mathematical and physical model of fluid flow after the geometric model is meshed;

[0159] S4. Set the boundary conditions and initial conditions for the initial fluid flow mathematical physics model;

[0160] S5. Verify the mesh independence of the geometric model. If the verification passes, perform flow field simulation on the geometric model to obtain simulation results. If the verification fails, return to step S2.

[0161] The model calibration module is specifically used to perform the following steps:

[0162] T1. Compare the measurement data of the hydrogen leak monitoring sensor with the simulation results of the monitoring points corresponding to the initial fluid flow mathematical and physical model;

[0163] T2. Determine whether the relative deviation between the simulation result data and the measurement data is less than the error threshold. If yes, proceed to step T3; otherwise, proceed to step T4.

[0164] T3. Determine the current position of the hydrogen leak monitoring sensor as the target position, and use the initial fluid flow mathematical and physical model as the target fluid flow mathematical and physical model;

[0165] T4. Based on the measurement data, revise the initial mathematical and physical model of fluid flow, adjust the placement of the hydrogen leak monitoring sensor, and return to step S4.

[0166] The mathematical and physical model of the target fluid flow is as follows:

[0167]

[0168] in, For the density of the fluid, Let t be the velocity vector of the fluid, t be time, and p be the pressure of the fluid. The dynamic viscosity coefficient of the fluid. For volume forces, Let be the specific heat capacity of the fluid, T be the temperature of the fluid, k be the thermal conductivity of the fluid, and Q be the heat source term per unit volume. The term represents the heat source for chemical reactions per unit volume. Let be the volume fraction or mass fraction of the i-th component, and D be the diffusion coefficient of the component. Let i be the source term of the i-th component.

[0169] The loss function of the initial physical information neural network model is:

[0170]

[0171]

[0172]

[0173]

[0174]

[0175]

[0176] in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. All are weighting coefficients. The number of monitoring points, Let j be the predicted value of the physical quantity at the j-th monitoring point. Let j be the true value of the physical quantity at the j-th monitoring point. Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

[0177] The monitoring module is also used for:

[0178] Based on the measurement data of hydrogen leakage monitoring sensors in the target hydrogen leakage monitoring field scenario, the mathematical and physical model of the target fluid flow is corrected;

[0179] The target physical information neural network model is corrected based on the corrected target fluid flow mathematical and physical model.

[0180] A corrected target physical information neural network model is used to monitor and warn of hydrogen leaks at the target hydrogen leak monitoring site.

[0181] It also includes a false alarm handling module, which is used for:

[0182] Based on the prediction results of the target physical information neural network model, it is determined whether there are false alarms in the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site. If so, the false alarm hydrogen leak monitoring sensors are corrected or replaced.

[0183] For easier understanding, please refer to Figure 6 This invention provides an embodiment of a hydrogen leak monitoring and early warning device, the device including a processor and a memory:

[0184] The memory is used to store program code and transfer the program code to the processor;

[0185] The processor is used to execute any of the hydrogen leak monitoring and early warning methods provided in this invention according to the instructions in the program code.

[0186] The present invention also provides an embodiment of a computer-readable storage medium for storing program code for executing any of the hydrogen leakage monitoring and early warning methods provided in the present invention.

[0187] The hydrogen leak monitoring and early warning device, equipment, and computer-readable storage medium provided in this invention are used to execute the hydrogen leak monitoring and early warning method provided in this invention. Their principles and the technical effects achieved are the same as those of the hydrogen leak monitoring and early warning method provided in this invention, and will not be repeated here.

[0188] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 the present invention.

Claims

1. A method for monitoring and early warning of hydrogen leakage, characterized in that, include: Based on the target hydrogen leak monitoring site scenario, a scaled-down model test bench was established, and hydrogen leak monitoring sensors were arranged at preset positions on the scaled-down model test bench. The scaled-down model test bench was modeled and fluid dynamics simulation experiments were conducted in computational fluid dynamics simulation software to obtain simulation results; Based on the measurement data from the hydrogen leak monitoring sensor and the simulation results, a mathematical and physical model for the target fluid flow is determined. The loss function for constructing the initial physical information neural network model based on the mathematical and physical model of the target fluid flow; Based on the simulation data of hydrogen leakage events in the target hydrogen leakage monitoring site scenario using the target fluid flow mathematical physics model and the loss function, the initial physical information neural network model is trained to obtain the target physical information neural network model. Based on the target physical information neural network model, hydrogen leakage monitoring and early warning are performed on the target hydrogen leakage monitoring site scenario; The loss function of the initial physical information neural network model is: in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. All are weighting coefficients. The number of monitoring points, Let j be the predicted value of the physical quantity at the j-th monitoring point. Let j be the true value of the physical quantity at the j-th monitoring point. Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

2. The hydrogen leakage monitoring and early warning method according to claim 1, characterized in that, The scaled-down model test bench was modeled and fluid dynamics simulation experiments were conducted in computational fluid dynamics simulation software to obtain simulation results, including: S1. Construct the geometric model of the scaled-down model test bench in computational fluid dynamics simulation software; S2. Mesh the geometric model; S3. Establish the initial fluid flow mathematical and physical model of the geometric model after mesh generation; S4. Set the boundary conditions and initial conditions for the initial fluid flow mathematical and physical model; S5. Perform mesh independence verification on the geometric model. If the verification passes, perform flow field simulation on the geometric model to obtain simulation results. If the verification fails, return to step S2.

3. The hydrogen leakage monitoring and early warning method according to claim 2, characterized in that, Based on the measurement data from the hydrogen leak monitoring sensor and the simulation results, a mathematical and physical model of the target fluid flow is determined, including: T1. Compare the measurement data of the hydrogen leak monitoring sensor with the simulation results of the monitoring points corresponding to the initial fluid flow mathematical and physical model; T2. Determine whether the relative deviation between the simulation result data and the measurement data is less than the error threshold. If yes, proceed to step T3; otherwise, proceed to step T4. T3. Determine the current position of the hydrogen leak monitoring sensor as the target position, and the initial fluid flow mathematical and physical model as the target fluid flow mathematical and physical model; T4. Based on the measurement data, correct the initial fluid flow mathematical and physical model, adjust the placement of the hydrogen leak monitoring sensor, and return to step S4.

4. The hydrogen leakage monitoring and early warning method according to claim 1, characterized in that, The mathematical and physical model of the target fluid flow is as follows: in, For the density of the fluid, Let t be the velocity vector of the fluid, t be time, and p be the pressure of the fluid. The dynamic viscosity coefficient of the fluid. For volume forces, Let be the specific heat capacity of the fluid, T be the temperature of the fluid, k be the thermal conductivity of the fluid, and Q be the heat source term per unit volume. The term represents the heat source for chemical reactions per unit volume. Let be the volume fraction or mass fraction of the i-th component, and D be the diffusion coefficient of the component. Let i be the source term of the i-th component.

5. The hydrogen leakage monitoring and early warning method according to claim 1, characterized in that, Based on the target physical information neural network model, hydrogen leak monitoring and early warning are performed on the target hydrogen leak monitoring site scenario, followed by: Based on the measurement data of the hydrogen leakage monitoring sensor in the target hydrogen leakage monitoring site scenario, the mathematical and physical model of the target fluid flow is corrected; The target physical information neural network model is corrected based on the corrected target fluid flow mathematical and physical model; The corrected target physical information neural network model is used to perform hydrogen leakage monitoring and early warning for the target hydrogen leakage monitoring site.

6. The hydrogen leakage monitoring and early warning method according to claim 1, characterized in that, Based on the target physical information neural network model, hydrogen leak monitoring and early warning are performed on the target hydrogen leak monitoring site scenario, followed by: Based on the prediction results of the target physical information neural network model, it is determined whether there are false alarms in the hydrogen leak monitoring sensors at the target hydrogen leak monitoring site. If so, the false alarm hydrogen leak monitoring sensors are corrected or replaced.

7. A hydrogen leak monitoring and early warning device, characterized in that, include: The scaled-down model building module is used to establish a scaled-down model test bench based on the target hydrogen leak monitoring site scenario, and to arrange hydrogen leak monitoring sensors at preset positions on the scaled-down model test bench. The fluid dynamics simulation module is used to model the scaled-down test bench and conduct fluid dynamics simulation experiments in computational fluid dynamics simulation software to obtain simulation results; The model calibration module is used to determine the mathematical and physical model of the target fluid flow based on the measurement data of the hydrogen leakage monitoring sensor and the simulation results. The loss function construction module is used to construct the loss function of the initial physical information neural network model based on the target fluid flow mathematical and physical model; The prediction network training module is used to train the initial physical information neural network model based on the simulation data of the hydrogen leakage event in the target hydrogen leakage monitoring site scenario and the loss function, using the mathematical and physical model of the target fluid flow and the loss function, to obtain the target physical information neural network model. The monitoring module is used to perform hydrogen leakage monitoring and early warning based on the target physical information neural network model of the target hydrogen leakage monitoring site. The loss function of the initial physical information neural network model is: in, The loss function for the initial physical information neural network model. For the data item loss function, The loss function is the physical law term. All are weighting coefficients. The number of monitoring points, Let j be the predicted value of the physical quantity at the j-th monitoring point. Let j be the true value of the physical quantity at the j-th monitoring point. Let be the loss function of the fluid partial differential equation. Let the boundary condition loss function of the fluid system be . Let the initial condition loss function of the fluid system be . The number of data points used to calculate the loss function of the fluid partial differential equation. The number of data points used to calculate the boundary condition loss function of the fluid system. The number of data points used to calculate the initial condition loss function of the fluid system. This is an implicit scheme of the Navier-Stokes equations, where u is the velocity vector of the fluid. The component in the x-direction, v is the velocity vector of the fluid. The component in the y-direction, w is the velocity vector of the fluid. The component in the z-direction, where T is the fluid temperature and X is the fluid component concentration.

8. A hydrogen leak monitoring and early warning device, characterized in that, The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the hydrogen leakage monitoring and early warning method according to any one of claims 1-6 according to the instructions in the program code.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code for executing the hydrogen leakage monitoring and early warning method according to any one of claims 1-6.

Images