A method and device for determining fuel cell optimized operating conditions

Abstract

This article relates to the field of fuel cells and provides a method and device for determining the optimized operating conditions of a fuel cell. The method includes: matching a target fuel cell response model according to the fuel type and size of the fuel cell to be analyzed; determining multiple groups of input parameter values ​​for the fuel cell to be analyzed; inputting each group of input parameter values ​​into the target fuel cell response model to obtain the response parameter value of the fuel cell to be analyzed under each input parameter value; determining the optimized operating conditions of the fuel cell to be analyzed based on the response parameter value of the fuel cell to be analyzed under each group of input parameter values; the fuel cell response model is determined by training parameters in a PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model, and a dimension recovery module. This article is applicable to fuel cells of various fuel types and sizes, realizes the universality of the model architecture, and can also improve the prediction speed and accuracy of the fuel cell response parameters.

Description

Technical Field

[0001] This article relates to the field of fuel cells, and in particular to a method and device for determining the optimal operating conditions of a fuel cell. Background Art

[0002] A fuel cell uses hydrogen, methane, methanol, and other fuels to directly convert chemical energy into electrical energy through an electrochemical reaction. Unlike traditional thermal power generation, fuel cell power generation offers advantages such as smokelessness, low noise, high efficiency, and low carbon emissions. It has found applications in the automotive, aerospace, and satellite industries.

[0003] Solid oxide fuel cells (SOFCs) are high-temperature fuel cells that use solid oxide as an electrolyte and offer advantages such as high efficiency, high energy density, and low pollution. SOFCs operate at relatively high temperatures, typically between 800°C and 1000°C, which enables high power generation efficiency but also presents challenges for their application.

[0004] Controlling fuel cell operating conditions is a crucial tool for achieving efficient and stable operation. Fuel cell operating conditions include fuel flow, voltage, temperature, and other operating conditions, and fuel cell performance varies under different operating conditions. Therefore, controlling fuel cell operating conditions can optimize performance and increase both efficiency and lifespan.

[0005] The following two methods are used in the prior art to predict the optimal operating conditions of fuel cells:

[0006] One approach is to predict the optimal fuel cell operating conditions based on a physical model of the fuel cell. This approach requires building a complex physical model to describe the multiple physical processes within the fuel cell. These physical processes and their interactions are often very complex, making it difficult to build an accurate and fast-running physical model. Furthermore, building a physical model requires a large amount of experimental data, and separate experiments must be conducted for different fuel cell types and specifications, which is time-consuming and costly.

[0007] The other method is to train a model based on machine learning or deep learning algorithms and use the model to predict operating conditions. This method of training models has the problem of low efficiency and also requires a large amount of experimental data. In addition, it is necessary to establish models of different architectures for different types and specifications of fuel cells. Therefore, it is time-consuming and costly and is not universal. Summary of the Invention

[0008] This paper aims to solve the problems of the non-universality of the model architecture, low model establishment efficiency and poor prediction accuracy in the existing methods based on fuel cell physical models and machine learning or deep learning algorithm training models, and using models to predict operating conditions.

[0009] In order to solve the above technical problems, this paper provides a method for determining the optimal operating condition of a fuel cell, comprising:

[0010] Matching a target fuel cell response model from a fuel cell response model library based on the fuel type and size of the fuel cell to be analyzed, wherein the fuel cell response model library stores fuel cell response models of various fuel types and preset sizes established based on multi-physics fuel cell reference models of various fuel types;

[0011] Determining multiple sets of input parameter values ​​of the fuel cell to be analyzed, wherein the input parameters include: actual battery performance parameters and operating conditions;

[0012] Inputting each set of input parameter values ​​into the target fuel cell response model to obtain the response parameter value of the fuel cell to be analyzed under each input parameter value;

[0013] determining an optimized operating condition of the fuel cell to be analyzed according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values;

[0014] Among them, the fuel cell response model is determined by training the parameters in the PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model and a dimension recovery module; the local battery model is used to determine the local current density according to the input parameters; the ideal battery model is used to determine the ideal current according to the input parameters; the dimensionless processing module is used to perform dimensionless processing on the input parameters, the local current density and the ideal current to obtain a dimensionless vector; the neural network model is used to calculate the dimensionless battery response parameters based on the dimensionless vector; the dimension recovery module is used to perform dimension recovery processing on the dimensionless battery response parameters.

[0015] As a further embodiment of this invention, both the local battery model and the ideal battery model are zero-dimensional models.

[0016] As a further embodiment of the present invention, the response model training process of each preset size fuel cell of the same fuel type includes:

[0017] Establish a multi-physics fuel cell reference model for this fuel type;

[0018] Using a multi-physics fuel cell reference model of the fuel type, a plurality of samples are determined for each preset size, each sample including: an input parameter value and a response parameter value;

[0019] Construct a PINN model based on input parameters, response parameters and fuel type;

[0020] According to the principle of consistency between the response parameter value predicted by the PINN model and the response parameter value in the sample, a loss function is constructed;

[0021] Use samples of various preset sizes and loss functions to train the parameters of the PINN model;

[0022] According to the parameters obtained by training the samples of each predicted size and the PINN model, the fuel cell response model of each preset size of the fuel type is obtained.

[0023] As a further embodiment of the present invention, when the fuel is hydrogen, the multi-physics fuel cell reference model is a two-dimensional multi-physics fuel cell reference model using hydrogen as fuel, and the response parameters include the output current of the fuel cell and the minimum mole fraction of hydrogen in the fuel gas;

[0024] When the fuel is a carbon-based fuel, the multi-physics fuel cell reference model is a three-dimensional multi-physics fuel cell reference model using carbon-based fuel as fuel, and the response parameters include the output current of the fuel cell, the reference values ​​of the highest local chemical equilibrium oxygen partial pressure of the fuel electrode and the highest local carbon deposition rate.

[0025] As a further embodiment of the present invention, determining multiple sets of input parameter values ​​of the fuel cell to be analyzed includes:

[0026] Acquiring calibration data of the fuel cell to be analyzed, wherein the calibration data includes volt-ampere characteristic curves at different fuel flow rates and temperatures;

[0027] Calibrate the actual parameter values ​​of the fuel cell performance of the fuel cell to be analyzed using the target fuel cell response model according to the calibration data;

[0028] Sampling to obtain multiple groups of operating condition values ​​according to the operating condition range of the fuel cell to be analyzed;

[0029] A set of input parameter values ​​is composed of the actual parameter values ​​of the battery performance after calibration and each set of operating condition values.

[0030] In a further embodiment of the present invention, the response parameter includes at least an output current; and according to the calibration data, calibrating the battery performance reference parameter value of the fuel cell to be analyzed using the target fuel cell response model includes:

[0031] Determining multiple sets of calibration operating condition values ​​and standard values ​​of output current based on the calibration data;

[0032] Initialize battery performance reference parameter values;

[0033] Calculating actual battery performance parameter values ​​according to the battery performance reference parameter values;

[0034] Inputting the calibration operating condition values ​​and actual battery performance parameter values ​​into the target fuel cell response model to obtain a predicted value of the output current;

[0035] Calculating the mean square error between the predicted value of the output current and the standard value of the relevant output current;

[0036] Adjusting the battery performance reference parameter value, and repeating the steps of calculating the battery performance actual parameter value and subsequent steps;

[0037] The actual parameter value of the battery performance with the smallest mean square error is taken as the actual parameter value of the battery performance after calibration.

[0038] As a further embodiment of this invention, the method further comprises:

[0039] Update the battery performance reference parameter values ​​using the following method:

[0040] Adding a change amount to the battery performance reference parameter value to obtain an updated battery performance reference parameter value, wherein the change amount is determined according to the operating time of the fuel cell to be analyzed under the optimized operating condition;

[0041] Based on the updated battery performance reference parameter values, multiple sets of input parameter values ​​are re-determined, and the optimized operating conditions are re-determined based on the re-determined multiple sets of input parameter values.

[0042] As a further embodiment of this invention, the method further comprises:

[0043] Update the battery performance reference parameter values ​​using the following method:

[0044] Obtaining updated battery performance reference parameter values ​​using the quantitative relationship between the actual battery performance parameters and the operating conditions;

[0045] Based on the updated battery performance reference parameter values, multiple sets of input parameter values ​​are re-determined, and the optimized operating conditions are re-determined based on the re-determined multiple sets of input parameter values.

[0046] As a further embodiment of the present invention, before determining the multiple sets of input parameter values ​​of the fuel cell to be analyzed, the method further includes:

[0047] Searching for batch information of the fuel cell to be analyzed from a historical database, wherein the historical database stores batch information of fuel cells whose optimized operating conditions have been determined;

[0048] If the search is successful, the optimized operating conditions of the successfully found fuel cell are assigned to the fuel cell to be analyzed.

[0049] As a further embodiment of the present invention, determining the optimized operating condition of the fuel cell to be analyzed according to the response parameters of the fuel cell to be analyzed under each set of input parameter values ​​includes:

[0050] Calculating scores for each response parameter under each set of input parameter values ​​according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values;

[0051] Calculate the score of each group of input parameter values ​​based on the score of each response parameter under each group of input parameter values;

[0052] A set of input parameter values ​​and their corresponding output response parameter values ​​with the highest scores are screened out as the optimized operating conditions of the fuel cell to be analyzed.

[0053] Another aspect of the present invention provides a fuel cell optimized operating condition determination device, comprising:

[0054] a model matching unit for matching a target fuel cell response model from a fuel cell response model library based on the fuel type and size of the fuel cell to be analyzed, wherein the fuel cell response model library stores fuel cell response models of various fuel types and preset sizes established based on multi-physics fuel cell reference models of various fuel types;

[0055] An input parameter value determination unit, configured to determine a plurality of sets of input parameter values ​​of the fuel cell to be analyzed, wherein the input parameters include: actual battery performance parameters and operating conditions;

[0056] A prediction unit, configured to input each set of input parameter values ​​into the target fuel cell response model to obtain a response parameter value of the fuel cell to be analyzed under each input parameter value;

[0057] an optimization unit, configured to determine an optimized operating condition of the fuel cell to be analyzed based on the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values;

[0058] Among them, the fuel cell response model is determined by training the parameters in the PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model and a dimension recovery module; the local battery model is used to determine the local current density according to the input parameters; the ideal battery model is used to determine the ideal current according to the input parameters; the dimensionless processing module is used to perform dimensionless processing on the input parameters, the local current density and the ideal current to obtain a dimensionless vector; the neural network model is used to calculate the dimensionless battery response parameters based on the dimensionless vector; the dimension recovery module is used to perform dimension recovery processing on the dimensionless battery response parameters.

[0059] On the other hand, this document also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in any one of the aforementioned embodiments is implemented.

[0060] On the other hand, the present invention further provides a computer storage medium having a computer program stored thereon, which implements the method described in any of the above embodiments when the computer program is executed by a processor of a computer device.

[0061] The method and device for determining the optimal operating conditions of a fuel cell, provided herein, proposes a PINN model comprising a local battery model, an ideal battery model, a dimensionless processing module, a neural network model, and a dimension recovery module. The model uses the fuel cell's operating conditions and actual battery performance parameters as input parameters, and the fuel cell's output response parameters as output. By training the parameters in the PINN model, a fuel cell response model is obtained, which is applicable to fuel cells of various fuel types and sizes, achieving the universality of the model architecture. Furthermore, by adding the local battery model and the ideal battery model to the fuel cell response model, the fuel cell response parameters can be accurately predicted in a relatively short period of time, thereby determining the optimal operating conditions. This improves the prediction speed and accuracy of the fuel cell response parameters, offering the advantages of fast and accurate prediction. Furthermore, based on the battery response parameters at each input parameter value, the optimal operating conditions can be accurately determined. These optimized operating conditions can be used for real-time control of the fuel cell to meet specific operating conditions.

[0062] In order to make the above and other purposes, features and advantages of this article more obvious and easy to understand, the following specifically cites preferred embodiments and provides detailed descriptions in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of this article or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of this article. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] Figure 1 A schematic diagram of the PINN model of the embodiment of this article is shown;

[0065] Figure 2 A flowchart showing the training process of the response model of fuel cells of various preset sizes of the same fuel type according to the embodiment of this invention is shown;

[0066] Figure 3A A flow chart showing a method for determining an optimized operating condition of a fuel cell according to an embodiment of the present invention is shown;

[0067] Figure 3B Another flow chart of the method for determining the optimal operating condition of a fuel cell according to an embodiment of the present invention is shown;

[0068] Figure 4A flow chart showing a process for determining input parameters of a fuel cell to be analyzed according to an embodiment of the present invention is shown;

[0069] Figure 5 A flow chart showing a process for determining actual parameters of the battery performance of the fuel cell to be analyzed in an embodiment of this invention is shown;

[0070] Figure 6 Another flow chart of the method for determining the optimal operating condition of a fuel cell according to an embodiment of this invention is shown;

[0071] Figure 7 A flow chart showing a process for determining the optimal operating condition of a fuel cell to be analyzed in an embodiment of this invention is shown;

[0072] Figure 8 A structural diagram of a fuel cell optimization operating condition determination device according to an embodiment of the present invention is shown;

[0073] Figure 9A A comparison of the experimentally determined volt-ampere characteristic curve of a solid oxide fuel cell at 720°C and the volt-ampere characteristic curve predicted based on the fuel cell response model in this paper is shown;

[0074] Figure 9B A comparison of the experimentally determined volt-ampere characteristic curve of a solid oxide fuel cell at 750°C and the volt-ampere characteristic curve predicted based on the fuel cell response model in this paper is shown;

[0075] Figure 9C The contour map of predicted values ​​of fuel utilization (uF), power generation (P), and current (I) of the solid oxide fuel cell at different hydrogen fuel flow rates and different operating voltages is shown;

[0076] Figure 9D The power generation efficiency (η) and the minimum hydrogen mole fraction (x) of the solid oxide fuel cell at different hydrogen fuel flow rates and different operating voltages are shown. H2 )’s predicted value contour map;

[0077] Figure 10A A comparison of measured and predicted currents and corresponding voltages and efficiencies is shown;

[0078] Figure 10B A comparison of measured and predicted fuel utilization and corresponding voltage and efficiency is shown;

[0079] Figure 11 The diagram shows the structure of the computer device according to the embodiment of this article.

[0080] Description of the accompanying symbols:

[0081] 101. Local battery model;

[0082] 102. Ideal battery model;

[0083] 103. Dimensionless processing module;

[0084] 104. Neural network model;

[0085] 105. Dimension recovery module;

[0086] 801, model matching unit;

[0087] 802. Input parameter value determination unit;

[0088] 803, optimization unit;

[0089] 1102. Computer equipment;

[0090] 1104, processor;

[0091] 1106. Memory;

[0092] 1108, driving mechanism;

[0093] 1110, input / output module;

[0094] 1112. Input device;

[0095] 1114. Output device;

[0096] 1116. Presentation equipment;

[0097] 1118. Graphical User Interface;

[0098] 1120, network interface;

[0099] 1122, communication link;

[0100] 1124. Communication bus. DETAILED DESCRIPTION

[0101] The following will be combined with the accompanying drawings to clearly and completely describe the technical solutions in the embodiments of this document. Obviously, the embodiments described are only part of the embodiments of this document, not all of the embodiments. Based on the embodiments of this document, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this document.

[0102] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of this document and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or equipment.

[0103] This specification provides method operation steps as described in the embodiments or flowcharts, but more or fewer operation steps may be included based on routine or non-creative work. The order of steps listed in the embodiments is only one way of executing the steps among many orderings and does not represent the only execution order. When a system or device product is actually executed, the method can be executed in the order shown in the embodiments or the drawings or in parallel.

[0104] A physics-informed neural network (PINN) is a method that incorporates physical laws into neural networks. This approach can leverage physics knowledge to enhance the neural network's predictive capabilities when data is insufficient. PINN has been successfully applied in fields such as fluid dynamics, materials science, and robotics, improving model accuracy and reliability. However, prior art has not incorporated physical laws into fuel cell response or operating condition prediction models.

[0105] In addition, in the existing technology, the model is usually based on the fuel cell physical model or based on machine learning and deep learning algorithm training model, and the model is used to predict the operating conditions. This method of building a model has problems such as the model architecture is not universal, the model building efficiency is low, and the prediction result accuracy is poor.

[0106] In order to solve the above technical problems, in one embodiment of this invention, a PINN model for establishing a fuel cell response model is provided, such as Figure 1 As shown, the PINN model includes: a local battery model 101, an ideal battery model 102, a dimensionless processing module 103, a neural network model 104 and a dimension recovery module 105.

[0107] The local battery model 101 is used to determine the local current density based on input parameters.

[0108] The ideal battery model 102 is used to determine the ideal current based on the input parameters. For hydrogen fuel, the ideal current includes: the ideal average current density and the ideal outlet hydrogen mole fraction.

[0109] The dimensionless processing module 103 is used to perform dimensionless processing on the input parameters and the output parameters of the local battery model 101 and the ideal battery model 102 to obtain dimensionless vectors.

[0110] The neural network model 104 is used to calculate the dimensionless response parameters of the fuel cell according to the dimensionless vector.

[0111] The dimension recovery module 105 is used to perform dimension recovery processing on the dimensionless response parameters of the fuel cell.

[0112] The PINN model proposed in this embodiment is applicable to fuel cells of various fuel types and sizes, and can achieve the versatility of the fuel cell response model architecture. At the same time, using the PINN model to establish a fuel cell response model can also improve the efficiency and accuracy of fuel cell response parameter prediction. Specifically, the local battery model predicts the battery response when the reactants (fuel and air) are sufficient and the performance is only subject to the battery performance. The ideal battery model predicts the battery response when the battery performance is ideal and the performance is only subject to the supply of reactants (fuel, air). In fact, the battery is constrained by both the reactant supply and the battery performance. Under the same conditions, the output current does not exceed the output of the local battery model or the ideal battery model. It can be seen that by adding the local battery model and the ideal battery model to the fuel cell response model, the output range of the fuel cell response model can be limited, providing a numerical reference, preventing the fuel cell response model from violating the basic principles of physics (chemistry), thereby improving the prediction speed and accuracy of the fuel cell response parameters.

[0113] In detail, the establishment process of the local battery model 101 and the ideal battery model 102 can refer to the existing technology. In specific implementation, in order to improve the running speed of the PINN model, the local battery model 101 and the ideal battery model 102 are both zero-dimensional models.

[0114] Taking hydrogen fuel as an example, the output of the zero-dimensional local battery model is the local current density, and the output of the zero-dimensional ideal battery model is the ideal average current density and the ideal outlet hydrogen mole fraction.

[0115] Specifically, the implementation of the zero-dimensional local battery model is as follows:

[0116] Calculate the voltage setting value U set , use the bisection method to iteratively solve the equation U(j local )=U set , obtain the local current density j local ,in,

[0117] U set =OCV-η tot ;

[0118]

[0119] U(j local )=E int -ηact,an -η act,ca -η ohm ;

[0120]

[0121]

[0122]

[0123] η act,an Use the bisection method to iteratively solve equation j an (η act,an )=j local get:

[0124]

[0125] η act,ca Use the bisection method to iteratively solve equation j ca (η act,ca )=j local get:

[0126]

[0127] η ohm By j local and r ohm Direct multiplication yields: η ohm =j local ·r ohm .

[0128] Among them, η tot is the total overpotential, E0 is the standard Nernst electromotive force, R gas T is the ideal gas constant R gas The product of absolute temperature T, F is the Faraday constant, x H2,in is the hydrogen mole fraction at the fuel inlet, x O2,in is the oxygen mole fraction at the air inlet, j lim,an is the anode diffusion limiting current density, j local is the local current density, j ex,an is the anode exchange current density, β an is the anode charge transfer coefficient, n an is the charge transferred by the anode elementary reaction, η act,an is the anode activation overpotential, r ohm is the surface resistivity in ohms.

[0129] The implementation of the zero-dimensional ideal battery model is:

[0130] Calculate the theoretical current when hydrogen in the fuel gas is oxidized to the Nernst equilibrium state while the oxidizing gas composition remains unchanged Theoretical current when oxygen in the oxidizing gas is reduced to the Nernst equilibrium state when the composition of the fuel gas remains unchanged And obtain the ideal battery fuel outlet hydrogen mole fraction during the calculation process

[0131] Calculating ideal current Then calculate the ideal average current density j ideal =I ideal / A, A is the macroscopic active area of ​​the battery.

[0132] The Nernst equilibrium state refers to the Nernst voltage calculated based on the fuel gas composition and the oxidizing gas composition and the voltage setting value U set The specific calculation method for equal status is as follows:

[0133]

[0134]

[0135] in, and The calculation method is as follows:

[0136]

[0137]

[0138] Among them, Q fuel,in is the fuel inlet flow rate, Q oxid,in is the air inlet flow rate, is the hydrogen mole fraction at the fuel inlet, is the oxygen mole fraction at the air inlet, is the ideal cell fuel outlet hydrogen mole fraction, is the oxygen mole fraction at the air outlet of the ideal cell.

[0139] The dimensionless processing module 103 and the dimension recovery module 105 can refer to existing dimensionless algorithms, and the specific implementation process is not limited in this article. Dimensionless processing can reduce the difficulty of neural network training and enhance the ability of the neural network to adapt to different actual battery performance parameters and operating conditions.

[0140] The neural network model 104 may include independent neural network units with the same number as the output response parameters. Each neural network unit may adopt an existing multi-layer perception neural network model, which is not specifically limited in this article. Figure 1 The neural network units described in the figure are only examples and do not limit the actual number of neurons. By setting the neural network units with the same number of output response parameters, the difficulty of training the neural network units can be reduced.

[0141] In specific applications, a fuel cell response model library must be pre-established based on the PINN model. This library contains fuel cell response models for each fuel type and preset size, established based on the multi-physics fuel cell reference model for each fuel type. These fuel cell response models for each fuel type and preset size are used to predict the response parameters of a fuel cell of the same fuel type and preset size to be analyzed. Of course, in specific implementations, fuel cell response models for the required fuel type and size can also be pre-established based on the required fuel type and size.

[0142] In detail, the fuel types described herein include but are not limited to hydrogen and carbon-based (carbon monoxide, hydrogen, methane and other gases) fuels.

[0143] Input parameters are controllable parameters that affect the battery response parameters, including actual battery performance parameters and operating conditions. Operating conditions include fuel flow, voltage, and temperature. Fuel cells of different fuel types have different actual battery performance parameters. These parameters are calculated from reference battery performance parameters. The specific calculation process can be calculated using various methods based on existing technologies. Taking hydrogen fuel as an example, actual battery performance parameters include fuel gas hydrogen mole fraction, oxidizing gas oxygen mole fraction, specific ohmic resistance, anode exchange current density, cathode exchange current density, anode limiting current density, cathode limiting current density, etc.

[0144] Cell response parameters are crucial to the actual performance (output power) and safety of fuel cells. When the fuel is hydrogen, these parameters include the fuel cell's output current and the minimum hydrogen mole fraction in the fuel gas. When the fuel is carbon-based, these parameters include the fuel cell's output current, the maximum local equilibrium oxygen partial pressure at the fuel electrode, and the maximum local carbon deposition rate.

[0145] The output response parameters of fuel cells of different fuel types, as well as the local battery model and the ideal battery model are different. Therefore, the PINN model of the corresponding fuel type can be constructed by adjusting the output items, the local battery model and the ideal battery model. Then, the fuel cell response model of the preset size of each fuel type can be trained based on the PINN of each fuel type.

[0146] Specifically, such as Figure 2 As shown, the training process of the response model of each preset size fuel cell of the same fuel type includes:

[0147] Step 201: Establish a multi-physics fuel cell reference model for the fuel type.

[0148] Specifically, the multi-physics fuel cell reference model for each fuel type is determined based on known physical laws. The specific development process can be referenced in existing technologies. When hydrogen is used as fuel, the multi-physics fuel cell reference model is a two-dimensional multi-physics fuel cell reference model using hydrogen as fuel. When carbon-based fuels are used, the multi-physics fuel cell reference model is a three-dimensional multi-physics fuel cell reference model using carbon-based fuels as fuel.

[0149] In step 202 , a multi-physics fuel cell reference model of the fuel type is used to determine a plurality of samples for each preset size, each sample including an input parameter value and a response parameter value.

[0150] When implementing this step, for each preset size, multiple groups of input parameter values ​​are first sampled within the input parameter range, and the preset size and each group of input parameter values ​​are input into the multi-physics fuel cell reference model to obtain the response parameter value.

[0151] Step 203: construct a PINN model based on the input parameters, response parameters and fuel type.

[0152] When this step is implemented, the local battery model and the ideal battery model are determined according to the fuel type; the neural network model is determined according to the input parameters and the response parameters; and the PINN model of the fuel type is constructed based on the determined local battery model, the ideal battery model and the neural network model, as well as the predetermined dimensionless processing module and the dimension recovery module.

[0153] Step 204 : constructing a loss function based on the consistency principle between the response parameter value predicted by the PINN model and the response parameter value in the sample.

[0154] Step 205 : Using samples of various preset sizes and loss functions, the parameters in the PINN model, ie, the parameters in the neural network model in the PINN model, are trained.

[0155] During this step, the parameters of the PINN model are adjusted to minimize the error between the response parameter values ​​predicted by the PINN model and the response parameter values ​​in the sample. After training is complete, the parameters of the PINN model are no longer changed.

[0156] Step 206 : Obtain a fuel cell response model of each preset size of the fuel type based on the parameters obtained by training the samples of each predicted size and the PINN model.

[0157] In this embodiment, the response parameter values ​​output by the multi-physics fuel cell reference model are used as standard values, and the fuel cell response model is trained by continuously adjusting the deviation between the PINN prediction results and the standard values. The fuel cell response model has the advantage of fast calculation speed compared to the multi-physics fuel cell reference model. Specifically, on a computer device, the multi-physics fuel cell reference model established using the commercial software ANSYS Fluent takes about 2-30 minutes per calculation example, depending on the spatial resolution (the number of grids in the finite element method); when using a two-dimensional simplified model, each calculation example takes 2-5 seconds; when using the fuel cell response model proposed in this case, each calculation example takes 0.5-0.8 milliseconds. In usage scenarios such as optimizing operating conditions, the number of calculation examples that need to be calculated is 10-1000. Therefore, the fuel cell response model proposed in this case has an obvious advantage in fast calculation speed.

[0158] The following uses hydrogen fuel as an example to illustrate the parameter training process in the PINN model:

[0159] (1) Divide the input parameter batches in the sample. Specifically, all input parameter combinations are randomly divided into n batches of no more than 32 combinations. batch batches.

[0160] (2) Calculate the PINN model deviation. Specifically, for each batch of input parameters, the response parameter value predicted by the PINN model is used, and the deviation between the predicted response parameter value and the response parameter value in each batch of samples is calculated.

[0161] Taking hydrogen fuel cells as an example, the deviation value is calculated as follows: Calculate the dimensionless total current y I The mean square error between the predicted value and the reference value and the minimum mole fraction of hydrogen in the dimensionless fuel gas The mean square error between the predicted value and the reference value

[0162] (3) Adjust the PINN model parameters.

[0163] According to the intermediate variables of the PINN model calculation process, the output error of the current neural network is calculated using the error back propagation principle. The gradient G of the current neural network parameters I , calculate the output error of the hydrogen partial pressure neural network Gradient of the neural network parameters for the current hydrogen partial pressure Use Adam optimizer according to the gradient G I , the learning rate Lr adjusts the parameters of the primary current neural network, and uses the Adam optimizer according to the gradient The learning rate Lr adjusts the parameters of the hydrogen partial pressure neural network once. The learning rate Lr is adjusted according to the number of training rounds epoch by the cosine annealing method with preheating:

[0164]

[0165] (4) Determine whether to traverse all batches. Specifically, when this step is performed for the first time, the batch count i batch is 1, and each time this step is performed, the batch count i batch Increase by 1; if i batch ≥n batch , then all batches have been traversed and the output is “yes”; otherwise, all batches have not been traversed and the output is “no”.

[0166] (5) Determine whether the PINN model training is completed. Specifically, when this step is performed for the first time, the number of training rounds epoch is 1, and each time this step is performed, the number of training rounds epoch increases by 1; if the number of training rounds exceeds 200, the PINN model training is completed and the batch count is reset to zero. batch :=0, output “yes”, otherwise PINN model training is not completed and output “no”.

[0167] Based on the establishment of the fuel cell response model library, the fuel cell optimization operating conditions can be determined, specifically, Figure 3A As shown, the method for determining the optimal operating condition of a fuel cell includes:

[0168] Step 301 : Match a target fuel cell response model from a fuel cell response model library according to the fuel type and size of the fuel cell to be analyzed.

[0169] When this step is implemented, the fuel type and size are searched from the fuel response model library, and the fuel cell response model corresponding to the found fuel type and size is used as the target fuel cell response model.

[0170] Step 302: Determine multiple sets of input parameter values ​​for the fuel cell to be analyzed. The input parameters include actual battery performance parameters and operating conditions.

[0171] Once the actual battery performance parameters are determined, they are fixed. To ensure that the predicted output response parameter values ​​are consistent with the actual output response parameter values, the actual battery performance parameters must be calibrated. Multiple sets of operating condition values ​​can be sampled based on the operating condition range. A set of input parameter values ​​is derived from the calibrated actual battery performance parameter values ​​and a set of operating condition values.

[0172] Taking a hydrogen fuel cell as an example, the actual battery performance parameter range and operating condition range are shown in Table 1. In one embodiment, after the actual battery performance parameters are calibrated, Latin Hypercube Sampling is used to uniformly randomly sample within the operating condition range to obtain 50,000 random input parameter combinations.

[0173] Table 1

[0174]

[0175] Step 303 : Input each set of input parameter values ​​into the target fuel cell response model to obtain the response parameter value of the fuel cell to be analyzed under each input parameter value.

[0176] Step 304 : determining the optimized operating condition of the fuel cell to be analyzed based on the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values.

[0177] The fuel cell response model established in this embodiment, based on a physical neural network, can quickly and accurately predict the output response parameters of the fuel cell under analysis. This model offers a speed-up ratio approximately 10,000 times greater than existing approaches for determining fuel cell response. Based on the prediction results, optimized operating conditions can be quickly and accurately determined. These optimized operating conditions can be used for real-time fuel cell control to meet specific operating requirements.

[0178] like Figure 3B As shown, the method for determining the optimal operating condition of a fuel cell includes:

[0179] Step 301 ′: matching a target fuel cell response model from a fuel cell response model library according to the fuel type and size of the fuel cell to be analyzed.

[0180] Step 302': Determine a set of input parameter values. Specifically, this step includes calibrating actual battery performance parameters and specifying / adjusting operating conditions.

[0181] Step 303 ′: input the set of input parameter values ​​into the target fuel cell response model to obtain response parameter values.

[0182] Step 304 ′: evaluate the response parameter value to obtain a score.

[0183] Step 305 ′: determine whether the score meets the preset conditions. If so, determine that the current input parameter value and its corresponding response parameter value are the optimized operating conditions. Otherwise, continue to specify / adjust the operating conditions.

[0184] Figure 3B Compared with the embodiment shown Figure 3A The illustrated embodiment can further improve the efficiency of determining the optimal operating condition.

[0185] In one embodiment of this invention, Figure 4 As shown, the above step 302 determines the multiple sets of input parameter values ​​of the fuel cell to be analyzed, including:

[0186] Step 401: Obtain calibration data of the fuel cell to be analyzed, wherein the calibration data includes volt-ampere characteristic curves at different fuel flow rates and temperatures. The calibration data can be obtained through experiments or determined based on a multi-physics fuel cell reference model.

[0187] Step 402 : Calibrate the actual parameter values ​​of the fuel cell performance of the fuel cell to be analyzed using the target fuel cell response model according to the calibration data.

[0188] When implementing this step, firstly, the battery performance reference parameter values ​​of the fuel cell to be analyzed are calibrated using the target fuel cell response model according to the calibration data, and then the actual battery performance parameter values ​​after calibration are calculated based on the battery performance reference parameter values.

[0189] Step 403: sampling and obtaining a plurality of operating condition values ​​according to the operating condition range of the fuel cell to be analyzed.

[0190] During specific implementation, the operating condition value can also be determined in the following manner:

[0191] When this step is executed for the first time, the desired operating conditions are set to the expected values ​​and the operating conditions to be optimized are set to the default values. Each subsequent time this step is executed, the constrained sequential least squares quadratic programming (SLSQP) method is used to maximize the operating condition score and adjust the operating conditions to be optimized. The constraint is that the operating conditions must not exceed the established allowable range.

[0192] Step 404 : forming a set of input parameter values ​​from the actual parameter values ​​of the calibrated battery performance and a set of operating condition values.

[0193] This embodiment calibrates the battery performance reference parameter values ​​using the target fuel cell response model, and obtains the calibrated battery performance actual parameter values ​​based on the calibrated battery performance reference parameter values, thereby ensuring the accuracy of battery performance and further ensuring that the predicted output response parameter values ​​are consistent with the actual ones.

[0194] In one embodiment of this invention, Figure 5 As shown, the above step 402 calibrates the actual parameter values ​​of the battery performance of the fuel cell to be analyzed using the target fuel cell response model according to the calibration data, including:

[0195] Step 501: Determine multiple sets of operating condition values ​​and standard values ​​of output current for calibration based on calibration data.

[0196] Step 502: Initialize battery performance reference parameter values.

[0197] When this step is implemented, the battery performance reference parameter value may be set as an initial value, or may be randomly set within a reasonable range.

[0198] In detail, the battery performance reference parameters for different types of fuels are different. Taking a hydrogen fuel cell as an example, the battery performance reference parameters are shown in Table 2.

[0199] Table 2

[0200]

[0201] Step 503: Calculate the actual battery performance parameter value based on the battery performance reference parameter value.

[0202] Specifically, taking hydrogen fuel as an example, the actual parameters of battery performance are realized as follows:

[0203]

[0204]

[0205]

[0206] Among them, E act,an is the activation energy of the anode electrochemical reaction, R gas is the ideal gas constant, is the hydrogen mole fraction at the anode-electrolyte interface, is the water vapor mole fraction at the anode-electrolyte interface, E act,ca is the activation energy of the cathode electrochemical reaction, E act,elyt is the activation energy of ionic conduction.

[0207] Step 504 : Input the calibration operating condition values ​​and the actual battery performance parameter values ​​into the target fuel cell response model to obtain a predicted value of the output current.

[0208] When this step is performed, each calibration operating condition value and the actual battery performance parameter value are input as input parameter values ​​into the target fuel cell response model to obtain a predicted value of the output current.

[0209] Step 505 : Calculate the mean square error between the predicted value of the output current and the standard value of the relevant output current.

[0210] Specifically, the mean square error between the output current prediction value and the standard value is calculated using the following formula:

[0211]

[0212] I sim =IU set / R leak ;

[0213] Among them, I sim is the output current prediction value, I exp is the standard value of output current, R leak is the equivalent leakage resistance, I is the internal output current, that is, the measured current I sim and the leakage current.

[0214] When this step is implemented, if the mean square error calculated in this step is less than a preset threshold, the current actual parameter value of the battery performance is used as the actual parameter value after calibration.

[0215] Step 506: Adjust the battery performance reference parameter value, and repeat step 503 and subsequent steps.

[0216] When this step is implemented, the sequential least squares quadratic programming (SLSQP) method is used to iteratively adjust the values ​​allowed to change in the battery performance reference parameters.

[0217] Step 507 : The actual battery performance parameter value with the minimum mean square error is used as the calibrated actual battery performance parameter value.

[0218] In one embodiment of the present invention, considering that fuel cells of the same batch have similar optimized operating conditions, each time after the optimized operating conditions of the fuel cell to be analyzed are determined, the batch information and the optimized operating conditions of the fuel cell are stored in a historical database. Before determining the multiple sets of input parameter values ​​of the fuel cell to be analyzed in step 302, the following steps are further included:

[0219] Batch information of the fuel cell to be analyzed is searched from a historical database, wherein the historical database stores batch information of fuel cells for which optimized operating conditions have been determined, and the batch information includes but is not limited to fuel cell manufacturer, fuel cell type, size, production time, place of origin, etc.

[0220] If the search is successful, the optimized operating conditions of the fuel cell that was successfully found are assigned to the fuel cell to be analyzed, and step 302 and subsequent steps are not performed. If the search fails, step 302 is continued.

[0221] This embodiment can improve the efficiency of determining the optimal operating conditions of the fuel cell to be analyzed.

[0222] After the fuel cell is operated under the optimized working condition for a period of time, its battery performance reference parameter value, that is, the actual battery performance parameter value will change. In order to ensure that the fuel cell always operates according to the optimal working condition, in one embodiment of this invention, Figure 6 As shown, after the fuel cell operates at the optimized operating conditions for a period of time, the following steps are also included:

[0223] Step 601 : Adding a change amount to a battery performance reference parameter value to obtain an updated battery performance reference parameter value, wherein the change amount is determined based on the operating time of the fuel cell to be analyzed under the optimized operating condition.

[0224] When this step is implemented, the quantitative relationship between the battery performance reference parameter and the operating conditions can be used to obtain an updated battery performance reference parameter value.

[0225] Step 602 : re-determine the optimized operating condition based on the updated battery performance reference parameter value and the re-determined multiple sets of input parameter values.

[0226] Specifically, the actual parameter values ​​of the battery performance are first recalibrated, and then multiple sets of input parameter values ​​are determined based on the calibrated actual parameter values ​​of the battery performance. Then, the above steps 203 and 204 are executed to obtain a new optimized operating condition.

[0227] In one embodiment of this invention, Figure 7 As shown, the above step 204 determines the optimized operating condition of the fuel cell to be analyzed according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values, including:

[0228] Step 701 : Calculate the score of each response parameter under each set of input parameter values ​​according to the response parameter value of the fuel cell to be analyzed under each set of input parameter values.

[0229] When implementing this step, unilateral constrained optimization can be adopted, and the optimization constraint is that the thermodynamic equilibrium oxygen partial pressure of the anode is lower than the thermodynamic equilibrium oxygen partial pressure of the Ni-NiO system.

[0230] Taking hydrogen fuel as an example, the response parameters are the output current of the fuel cell and the minimum mole fraction of hydrogen in the fuel gas;

[0231] A specific implementation process of this step includes:

[0232] The output current rating is calculated using the following formula:

[0233] score P =-(P / P set -1) 2 , P=U set I sim ;

[0234] The score for the minimum hydrogen mole fraction in the fuel gas is calculated using the following formula:

[0235]

[0236] Where P is the output power, P set is the set value, U set is the set voltage, I simis the output current, score P For rating, is the minimum mole fraction of hydrogen in the fuel gas, is the safety critical value, score safety For rating.

[0237] Step 702 : Calculate the score of each set of input parameter values ​​according to the score of each response parameter under each set of input parameter values.

[0238] When this step is implemented, the score of each group of input parameter values ​​is obtained by calculating the sum of the scores of each response parameter under each group of input parameter values.

[0239] Taking hydrogen fuel as an example, the scoring formula for each group of input parameter values ​​is: score = score P +score safety .

[0240] Step 703 : Filter out a set of input parameter values ​​and their corresponding output response parameter values ​​with the highest scores as the optimized operating conditions of the fuel cell to be analyzed.

[0241] When this step is implemented, input parameter values ​​and their corresponding output response parameter values ​​with scores greater than a predetermined value (eg, -0.01) may be screened out, and the screened out input parameter values ​​and their corresponding output response parameter values ​​may be used as the optimal operating conditions.

[0242] This embodiment can screen out the optimal operating conditions that meet the power requirements and maximize the power generation efficiency, and can ensure the safety and economy of the fuel cell operation.

[0243] In one embodiment of this invention, taking hydrogen fuel as an example, the input parameters input to the dimensionless processing module 103 are shown in Table 1, which also includes the local current density j output by the local battery model 101. local , the ideal average current density j output by the ideal battery model 102 ideal and ideal outlet hydrogen mole fraction Among them, the ideal average current density j ideal By the ideal current I ideal Calculated, The calculation formula is:

[0244] The internal processing logic of the dimensionless processing module 103 is implemented using the following formula:

[0245]

[0246]

[0247]

[0248]

[0249]

[0250]

[0251]

[0252]

[0253]

[0254]

[0255]

[0256]

[0257]

[0258]

[0259]

[0260] x 16 =-c;

[0261]

[0262]

[0263]

[0264]

[0265]

[0266]

[0267] in, w is the width of the fuel cell, and L is the length of the fuel cell.

[0268] The dimensionless processing module 103 outputs a dimensionless vector x=[x1, x2, ..., x 22 ] to the neural network model 104.

[0269] The internal processing logic of the dimension recovery module 105 is implemented using the following formula:

[0270] I=(1+y I )·j ref·w·L;

[0271]

[0272] Among them, y I is the dimensionless current, is the dimensionless mole fraction of hydrogen.

[0273] Based on the same inventive concept, this article also provides a fuel cell optimized operating condition determination device, as described in the following embodiments. Since the principle of solving the problem by the fuel cell optimized operating condition determination device is similar to that of the fuel cell optimized operating condition determination method, the implementation of the fuel cell optimized operating condition determination device can refer to the fuel cell optimized operating condition determination method, and the repeated parts will not be repeated. Specifically, Figure 8 As shown, the fuel cell optimized operating condition determination device includes:

[0274] A model matching unit 801 is configured to match a target fuel cell response model from a fuel cell response model library based on the fuel type and size of the fuel cell to be analyzed, wherein the fuel cell response model library stores fuel cell response models of various fuel types and preset sizes established based on multi-physics fuel cell reference models of various fuel types;

[0275] An input parameter value determination unit 802 is configured to determine multiple sets of input parameter values ​​for the fuel cell to be analyzed, wherein the input parameters include actual battery performance parameters and operating conditions; a prediction unit is configured to input each set of input parameter values ​​into the target fuel cell response model to obtain a response parameter value of the fuel cell to be analyzed under each input parameter value;

[0276] An optimization unit 803 is configured to determine an optimized operating condition of the fuel cell to be analyzed based on the response parameters of the fuel cell to be analyzed under each set of input parameter values;

[0277] Among them, the fuel cell response model is determined by training the parameters in the PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model and a dimension recovery module; the local battery model is used to determine the local current density according to the input parameters; the ideal battery model is used to determine the ideal current according to the input parameters; the dimensionless processing module is used to perform dimensionless processing on the input parameters, the local current density and the ideal current to obtain a dimensionless vector; the neural network model is used to calculate the dimensionless battery response parameters based on the dimensionless vector; the dimension recovery module is used to perform dimension recovery processing on the dimensionless battery response parameters.

[0278] In order to more clearly illustrate the prediction accuracy of the fuel cell response model established in this paper, two specific examples are given below.

[0279] Example 1: A 10×10cm piece of steel at 720°C and 750°C 2 The output response parameters of the solid oxide fuel cell at 0.3, 0.5, and 1.0 SLM hydrogen fuel flow rates were predicted, and the predicted results were plotted. Figures 9A-9D , Figure 9A The figure shows the comparison between the experimentally determined volt-ampere characteristic curve of the solid oxide fuel cell at 720°C and the volt-ampere characteristic curve predicted based on the fuel cell response model in this paper. Figure 9B The figure shows the comparison between the experimentally determined volt-ampere characteristic curve of the solid oxide fuel cell at 750°C and the volt-ampere characteristic curve predicted based on the fuel cell response model in this paper. Figure 9C The predicted values ​​of fuel utilization (uF), power generation (P), and current (I) of the solid oxide fuel cell at different hydrogen fuel flow rates and different operating voltages are shown in the contour map. Figure 9D The power generation efficiency (η) and the minimum hydrogen mole fraction (x) of the solid oxide fuel cell at different hydrogen fuel flow rates and different operating voltages are shown. H2 ), where Figure 9C and Figure 9D The thick solid curve in the middle is the peak power voltage at each hydrogen fuel flow rate. Figure 9C and Figure 9D It can be seen that the fuel cell response model is accurate in prediction and the output current prediction value is close to the measured value.

[0280] Example 2: Predicted and measured operating conditions for the solid oxide fuel cell (SOFC) at 720°C, 750°C, and 780°C. Specifically, Table 3 shows the optimization results for six steady-state operating conditions at 720°C, 750°C, and 780°C that meet efficiency, power, and safety requirements.

[0281] Table 3

[0282]

[0283] Under these 6 working conditions, Figure 10A A comparison of the measured and predicted currents and the corresponding voltages and efficiencies is shown. Figure 10B A comparison of measured and predicted fuel utilization and corresponding voltage and efficiency is shown. Figure 10A and Figure 10B The mean absolute percentage error (MAPE) of the data is 2.21%.

[0284] As can be seen from Examples 1 and 2, the above-mentioned embodiments constructed according to the method described in this article can accurately predict the output response performance of industrial-sized solid oxide fuel cells under given temperatures and fuel flow rates, have accurate working condition interpolation and extrapolation functions, and achieve fast and universal solid oxide fuel cell optimized working condition prediction and optimization based on accurate prediction.

[0285] In one embodiment of the present invention, a computer device is also provided, such as Figure 11 As shown, computer device 1102 may include one or more processors 1104, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. Computer device 1102 may also include any memory 1106 for storing any type of information, such as code, settings, data, and the like. For example, and without limitation, memory 1106 may include any one or more combinations of the following: any type of RAM, any type of ROM, a flash memory device, a hard disk, an optical disk, and the like. More generally, any memory may use any technology to store information. Specifically, memory 1106 may store a program that implements any of the aforementioned methods. Furthermore, any memory may provide volatile or non-volatile retention of information. Furthermore, any memory may represent a fixed or removable component of computer device 1102. In one embodiment, when processor 1104 executes associated instructions stored in any memory or combination of memories, computer device 1102 may perform any operation of the associated instructions. The computer device 1102 also includes one or more drive mechanisms 1108 for interacting with any storage, such as a hard disk drive mechanism, an optical disk drive mechanism, and the like.

[0286] The computer device 1102 may also include an input / output module 1110 (I / O) for receiving various inputs (via input devices 1112) and for providing various outputs (via output devices 1114). A specific output mechanism may include a presentation device 1116 and an associated graphical user interface 1118 (GUI). In other embodiments, the input / output module 1110 (I / O), input devices 1112, and output devices 1114 may not be included, and the computer device 1102 may simply be a computer device in a network. The computer device 1102 may also include one or more network interfaces 1120 for exchanging data with other devices via one or more communication links 1122. One or more communication buses 1124 couple the components described above together.

[0287] The communication link 1122 may be implemented in any manner, for example, via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 1122 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0288] The embodiments of this document further provide a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described in any of the aforementioned embodiments are executed.

[0289] The embodiments of this document also provide a computer-readable instruction, wherein when a processor executes the instruction, the program therein causes the processor to execute the steps of the method described in any of the aforementioned embodiments.

[0290] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0291] It should also be understood that in the embodiments herein, the term "and / or" merely describes an association between associated objects, indicating that three possible relationships exist. For example, "A and / or B" could represent: A alone, A and B simultaneously, or B alone. Furthermore, the character " / " in this document generally indicates an "or" relationship between the associated objects.

[0292] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the composition and steps of each example according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this document.

[0293] Those skilled in the art will clearly understand that, for the sake of convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0294] In the several embodiments provided herein, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices, or units, or can be an electrical, mechanical, or other form of connection.

[0295] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments herein.

[0296] In addition, the functional units in the various embodiments herein may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0297] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0298] This article uses specific embodiments to illustrate the principles and implementation methods of this article. The description of the above embodiments is only used to help understand the methods and core ideas of this article. At the same time, for those skilled in the art, based on the ideas of this article, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation to this article.

Claims

1. A method for determining an optimal operating condition of a fuel cell, characterized in that: include: Matching a target fuel cell response model from a fuel cell response model library based on the fuel type and size of the fuel cell to be analyzed, wherein the fuel cell response model library stores fuel cell response models of various fuel types and preset sizes established based on multi-physics fuel cell reference models of various fuel types; Determining multiple sets of input parameter values ​​of the fuel cell to be analyzed, wherein the input parameters include: actual battery performance parameters and operating conditions; Inputting each set of input parameter values ​​into the target fuel cell response model to obtain the response parameter value of the fuel cell to be analyzed under each input parameter value; determining an optimized operating condition of the fuel cell to be analyzed according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values; Among them, the fuel cell response model is determined by training the parameters in the PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model and a dimension recovery module; the local battery model is used to determine the local current density according to the input parameters; the ideal battery model is used to determine the ideal current according to the input parameters; the dimensionless processing module is used to perform dimensionless processing on the input parameters, the local current density and the ideal current to obtain a dimensionless vector; the neural network model is used to calculate the dimensionless response parameters of the fuel cell according to the dimensionless vector; the dimension recovery module is used to perform dimension recovery processing on the response parameters of the dimensionless fuel cell.

2. The method according to claim 1, wherein The local battery model and the ideal battery model are both zero-dimensional models.

3. The method according to claim 1, wherein The training process for the response model of fuel cells of various preset sizes for the same fuel type includes: Establish a multi-physics fuel cell reference model for this fuel type; Using a multi-physics fuel cell reference model of the fuel type, a plurality of samples are determined for each preset size, each sample including: an input parameter value and a response parameter value; Construct a PINN model based on input parameters, response parameters and fuel type; According to the principle of consistency between the response parameter value predicted by the PINN model and the response parameter value in the sample, a loss function is constructed; Use samples of various preset sizes and loss functions to train the parameters of the PINN model; The fuel cell response model of each preset size of the fuel type is obtained based on the parameters obtained by training the samples of each predicted size and the PINN model.

4. The method according to claim 3, wherein When the fuel is hydrogen, the multi-physics fuel cell reference model is a two-dimensional multi-physics fuel cell reference model using hydrogen as fuel, and the response parameters include the output current of the fuel cell and the minimum mole fraction of hydrogen in the fuel gas; When the fuel is a carbon-based fuel, the multi-physics fuel cell reference model is a three-dimensional multi-physics fuel cell reference model using carbon-based fuel as fuel, and the response parameters include the output current of the fuel cell, the highest local chemical equilibrium oxygen partial pressure of the fuel electrode, and the highest local carbon deposition rate.

5. The method according to claim 1, wherein Determine multiple sets of input parameter values ​​for the fuel cell to be analyzed, including: Acquiring calibration data of the fuel cell to be analyzed, wherein the calibration data includes volt-ampere characteristic curves at different fuel flow rates and temperatures; Calibrate the actual parameter values ​​of the fuel cell performance of the fuel cell to be analyzed using the target fuel cell response model according to the calibration data; Sampling to obtain multiple groups of operating condition values ​​according to the operating condition range of the fuel cell to be analyzed; A set of input parameter values ​​is composed of the actual parameter values ​​of the battery performance after calibration and each set of operating condition values.

6. The method according to claim 5, wherein The response parameter includes at least: output current; based on the calibration data, using the target fuel cell response model to calibrate the actual parameter value of the battery performance of the fuel cell to be analyzed, including: Determining multiple sets of calibration operating condition values ​​and standard values ​​of output current based on the calibration data; Initialize battery performance reference parameter values; Calculating actual battery performance parameter values ​​according to the battery performance reference parameter values; Inputting the calibration operating condition values ​​and actual battery performance parameter values ​​into the target fuel cell response model to obtain a predicted value of the output current; Calculating the mean square error between the predicted value of the output current and the standard value of the relevant output current; Adjusting the battery performance reference parameter value, and repeating the steps of calculating the battery performance actual parameter value and subsequent steps; The actual parameter value of the battery performance with the smallest mean square error is taken as the actual parameter value of the battery performance after calibration.

7. The method according to claim 6, wherein The method further comprises: Update the battery performance reference parameter values ​​using the following method: Adding a change amount to the battery performance reference parameter value to obtain an updated battery performance reference parameter value, wherein the change amount is determined according to the operating time of the fuel cell to be analyzed under the optimized operating condition; Based on the updated battery performance reference parameter values, multiple sets of input parameter values ​​are re-determined, and the optimized operating conditions are re-determined based on the re-determined multiple sets of input parameter values.

8. The method according to claim 6, wherein The method further comprises: Update the battery performance reference parameter values ​​using the following method: Obtaining updated battery performance reference parameter values ​​using the quantitative relationship between the actual battery performance parameters and the operating conditions; Based on the updated battery performance reference parameter values, multiple sets of input parameter values ​​are re-determined, and the optimized operating conditions are re-determined based on the re-determined multiple sets of input parameter values.

9. The method according to claim 1, wherein Before determining multiple sets of input parameter values ​​of the fuel cell to be analyzed, it also includes: Searching for batch information of the fuel cell to be analyzed from a historical database, wherein the historical database stores batch information of fuel cells whose optimized operating conditions have been determined; If the search is successful, the optimized operating conditions of the successfully found fuel cell are assigned to the fuel cell to be analyzed.

10. The method according to claim 1, wherein Determining the optimized operating condition of the fuel cell to be analyzed according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values ​​includes: Calculating scores for each response parameter under each set of input parameter values ​​according to the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values; Calculate the score of each group of input parameter values ​​based on the score of each response parameter under each group of input parameter values; A set of input parameter values ​​and their corresponding output response parameter values ​​with the highest scores are screened out as the optimized operating conditions of the fuel cell to be analyzed.

11. A fuel cell optimal operating condition determination device, characterized in that: include: a model matching unit for matching a target fuel cell response model from a fuel cell response model library based on the fuel type and size of the fuel cell to be analyzed, wherein the fuel cell response model library stores fuel cell response models of various fuel types and preset sizes established based on multi-physics fuel cell reference models of various fuel types; An input parameter value determination unit, configured to determine a plurality of sets of input parameter values ​​of the fuel cell to be analyzed, wherein the input parameters include: actual battery performance parameters and operating conditions; A prediction unit, configured to input each set of input parameter values ​​into the target fuel cell response model to obtain a response parameter value of the fuel cell to be analyzed under each input parameter value; an optimization unit, configured to determine an optimized operating condition of the fuel cell to be analyzed based on the response parameter values ​​of the fuel cell to be analyzed under each set of input parameter values; Among them, the fuel cell response model is determined by training the parameters in the PINN model, and the PINN model includes: a local battery model, an ideal battery model, a dimensionless processing module, a neural network model and a dimension recovery module; the local battery model is used to determine the local current density according to the input parameters; the ideal battery model is used to determine the ideal current according to the input parameters; the dimensionless processing module is used to perform dimensionless processing on the input parameters, the local current density and the ideal current to obtain a dimensionless vector; the neural network model is used to calculate the dimensionless response parameters of the fuel cell according to the dimensionless vector; the dimension recovery module is used to perform dimension recovery processing on the response parameters of the dimensionless fuel cell.

12. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 10 is implemented.

13. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor of a computer device, the method according to any one of claims 1 to 10 is implemented.

Images