Power generation fuel cell cross-scale multi-physical field collaborative modeling method and system

Abstract

The application discloses a power generation fuel cell cross-scale multi-physical field collaborative modeling method and system. The method comprises the following steps: establishing micro, meso and macro models of electrochemical reactions, and developing an algorithm to realize cross-scale parameter transmission; collecting data from various experimental devices and preprocessing; inverting key parameters and calibrating boundary conditions through in-situ observation; adopting a physical field gradient adaptive grid generation and dynamic time step adjustment strategy, combining an MPI parallel architecture to solve multi-physical field control equations, performing simulation analysis under various working conditions, verifying the accuracy of the model on an experimental platform, using an NSGA-III algorithm to optimize key parameters of the fuel cell, predicting the control system dynamic adjustment operation strategy based on the optimized cross-scale model, and evaluating the whole life cycle reliability of the system through machine learning. The method can realize cross-scale collaboration, consider accuracy and efficiency, and adapt to multi-physical field collaborative modeling technology under dynamic working conditions.

Description

Technical Field

[0001] This invention relates to the field of multiphysics simulation technology, and in particular to a method and system for cross-scale multiphysics collaborative modeling of fuel cells for power generation. Background Technology

[0002] Hydrogen energy, as a key component of clean energy, holds immense potential in distributed generation and grid-connected energy storage. Fuel cells are the core equipment for achieving efficient energy conversion. However, the complex coupling effects of multiple physical fields within fuel cells, including electrochemical reactions, mass transport, and thermal stress, exhibit significant differences across different scales, posing a technical challenge to improving fuel cell performance. To increase power density, extend lifespan, and reduce R&D costs, it is essential to accurately understand these multi-scale, multi-physics coupling mechanisms and construct an effective modeling system.

[0003] While some methods have been attempted to model fuel cells using multiphysics, they generally face several major problems. First, most models are limited to a single scale or physical field, failing to effectively bridge different levels from the microscopic to the macroscopic, making it difficult to comprehensively analyze the relationship between system performance and microscopic mechanisms. Second, there is a trade-off between model accuracy and computational efficiency, especially when dealing with cross-scale parameter transfer, which easily leads to error accumulation. The lack of efficient numerical solution algorithms makes real-time simulation particularly difficult.

[0004] Furthermore, existing models are weak in predicting dynamic operating conditions and failure modes. Most models are suitable for simulating steady-state conditions, but perform poorly in capturing dynamic loads and response characteristics across multiple time scales. They also have limited ability to identify critical conditions for common failure modes such as flooding and dry film failure. Additionally, limitations in experimental data due to noise interference and in-situ observation techniques make model parameter calibration difficult, and the lack of a unified data fusion process affects the consistency and reliability of the models.

[0005] Therefore, it is necessary to design a new method that can achieve cross-scale collaboration, balance accuracy and efficiency, and adapt to dynamic operating conditions through multi-physics collaborative modeling technology, in order to solve existing technical problems and provide strong technical support for the design optimization and engineering application of fuel cells. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a cross-scale multi-physics collaborative modeling method and system for fuel cells used in power generation.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a cross-scale multiphysics collaborative modeling method for fuel cells used in power generation, comprising: Electrochemical reaction and transport models at the micro, meso, and macro scales were established using density functional theory, lattice Boltzmann method, and computational fluid dynamics, respectively, and cross-scale parameter transfer was performed to obtain cross-scale models. Multi-source data collected by integrated experimental devices is acquired, the multi-source data is preprocessed, and the key parameters of the cross-scale model are inverted. The boundary conditions of the cross-scale model are calibrated by in-situ observation to obtain the calibrated cross-scale model. The grid is adaptively generated based on the physical field gradient and the time step is dynamically adjusted. The multiphysics control equations corresponding to the calibrated cross-scale model are solved using the MPI parallel computing architecture. Simulation analysis is carried out under various working conditions to output the system performance characteristics. The accuracy of the model was verified on the experimental platform. The key parameters of the fuel cell were optimized using the NSGA-Ⅲ algorithm. The optimized cross-scale model was combined with predictive control to dynamically adjust the operation strategy. The reliability throughout the entire life cycle was evaluated through machine learning.

[0008] The further technical solution is as follows: Electrochemical reaction and transport models at the microscopic, mesoscopic, and macroscopic scales are established using density functional theory, the lattice Boltzmann method, and computational fluid dynamics, respectively, and cross-scale parameter transfer is performed to obtain a cross-scale model, including: Density functional theory was used to calculate the reaction energy barrier on the catalyst surface, and the Butler-Volmer and Nernst-Planck equations were combined to establish an electrochemical reaction and transport model to describe the electrochemical behavior at the microscale, thus forming a microscale model. The lattice Boltzmann method was used to simulate the gas-liquid two-phase flow and mass transport in a porous electrode. Combined with the heat transfer equation, the characteristics of the diffusion layer and its influence on the mass transfer efficiency were analyzed to obtain a mesoscale model. Computational fluid dynamics is used to establish a multiphysics coupled model covering flow field, temperature field, concentration field and electric field to describe the performance and dynamic response of the entire fuel cell stack, forming a macroscopic scale model; The key reaction kinetic parameters of the microscale model are passed to the mesoscale model, and the mass transport parameters of the mesoscale model are mapped to the macroscale model to obtain a cross-scale model.

[0009] The further technical solution is as follows: acquiring multi-source data collected by integrated experimental devices, preprocessing the multi-source data, inverting the key parameters of the cross-scale model, and calibrating the boundary conditions of the cross-scale model through in-situ observations to obtain the calibrated cross-scale model, includes: Multi-source data were collected using an experimental setup, including polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution. The multi-source data is denoised, outliers are removed, and the data format is standardized to obtain preprocessed results. The key parameters in the cross-scale model are retrieved by combining the genetic algorithm and particle swarm optimization algorithm with the preprocessing results. By comparing in-situ observation data with simulation results, the temperature boundary, humidity boundary, and fluid inlet and outlet boundary conditions of the preprocessed results are iteratively adjusted, and after multiple iterations, the deviation between the predicted results and experimental results of the preprocessed results is minimized, so as to obtain the calibrated cross-scale model.

[0010] The further technical solution is as follows: the key parameters include catalyst reaction activity parameters, diffusion layer porosity, and flow channel structure parameters.

[0011] The further technical solution is as follows: The mesh is adaptively generated based on the physical field gradient, and the time step is dynamically adjusted. The multiphysics control equations corresponding to the calibrated cross-scale model are solved using an MPI parallel computing architecture. Simulation analysis is conducted under various operating conditions to output system performance characteristics, including: The grid density is automatically adjusted based on the physical field gradient, with a denser grid in areas where the change range meets the requirements and a sparser grid in areas with a gentler change. Set the time step adjustment range based on residual convergence; The multiphysics control equations corresponding to the calibrated cross-scale model are decomposed into multiple sub-tasks and solved in parallel on a distributed computing architecture using MPI technology. Simulations are performed under different operating conditions to output the multiphysics distribution characteristics, performance indicators, and dynamic response laws of the system, forming simulation results.

[0012] The further technical solution is as follows: The automatic adjustment of mesh density based on the physical field gradient, wherein the mesh is denser in regions where the variation amplitude meets the requirements, and sparser in regions with gentle changes, includes: Based on the gradient distribution characteristics of different physical fields of electrochemical reaction, mass transport, and thermal stress during fuel cell operation, the regions that need to be densified or sparsed are determined to obtain the partitioning results. A high grid density threshold is set for regions where the physical field changes meet the requirements, while a low grid density is set for regions where the physical field changes are gradual, thus forming the grid density requirement. Based on the set grid density requirements, an unstructured adaptive grid is automatically generated from the partitioning results.

[0013] Its further technical solution is as follows: The accuracy of the model is verified on an experimental platform, the key parameters of the fuel cell are optimized using the NSGA-Ⅲ algorithm, the optimized cross-scale model is used for predictive control to dynamically adjust the operating strategy, and the full life-cycle reliability is evaluated through machine learning, including: The simulation results were compared with the actual test data on the experimental platform to obtain the comparison results; Based on the comparison results, the key design parameters of the fuel cell in the calibrated cross-scale model are optimized using multiple objectives to obtain the optimized cross-scale model. The optimized cross-scale model is integrated into the digital twin system, and the optimized cross-scale model predictive control technology is applied to dynamically adjust the operation strategy. By using machine learning algorithms to analyze data from simulated failure modes, a reliability assessment model is established to predict the lifespan and performance degradation trend of fuel cells.

[0014] The further technical solution is as follows: Based on the comparison results, multi-objective optimization is performed on the key design parameters of the fuel cell in the calibrated cross-scale model to obtain the optimized cross-scale model, including: Based on the comparison results, a multi-objective optimization function was constructed using a non-dominated sorting genetic algorithm, with the objectives of maximizing power density, minimizing mass transfer resistance, and minimizing thermal stress. This function was then used to optimize key parameters such as bipolar channel width, diffusion layer porosity, and catalyst loading in the calibrated cross-scale model, thereby obtaining the Pareto optimal solution set and the optimized cross-scale model.

[0015] The further technical solution is as follows: The method of using machine learning algorithms to analyze simulated failure mode data, establish a reliability assessment model, and predict the service life and performance degradation trend of fuel cells includes: By simulating the failure processes of water flooding, dry film, and catalyst agglomeration, corresponding failure characteristic parameters are extracted, and a failure mode recognition model and a reliability assessment model are established based on machine learning algorithms to predict the service life and performance degradation law of fuel cells under different operating conditions.

[0016] This invention also provides a cross-scale multiphysics collaborative modeling system for fuel cells used in power generation, comprising: The cross-scale modeling unit is used to establish electrochemical reaction and transport models at the micro, meso, and macro scales using density functional theory, lattice Boltzmann method, and computational fluid dynamics, respectively, and to perform cross-scale parameter transfer to obtain cross-scale models. The multi-source data fusion unit is used to acquire multi-source data collected by integrated experimental devices, preprocess the multi-source data, invert the key parameters of the cross-scale model, and calibrate the boundary conditions of the cross-scale model through in-situ observations to obtain the calibrated cross-scale model. The high-efficiency numerical solution unit is used to adaptively generate the mesh according to the physical field gradient and dynamically adjust the time step. It uses the MPI parallel computing architecture to solve the multiphysics field control equations corresponding to the calibrated cross-scale model, and carries out simulation analysis under various working conditions to output the system performance characteristics. The model validation and optimization unit is used to verify the accuracy of the model on the experimental platform. It uses the NSGA-Ⅲ algorithm to optimize the key parameters of the fuel cell, combines the optimized cross-scale model predictive control to dynamically adjust the operation strategy, and evaluates the reliability throughout the entire life cycle through machine learning.

[0017] The advantages of this invention compared to existing technologies are as follows: By combining density functional theory, lattice Boltzmann method, and computational fluid dynamics, this invention establishes an electrochemical reaction and transport model spanning microscopic, mesoscopic, and macroscopic scales, and realizes cross-scale parameter transfer. It utilizes various experimental devices to collect and preprocess multi-source data, inverts key parameters, and calibrates boundary conditions through in-situ observations to form a calibrated cross-scale model. It employs physical field gradient adaptive grid generation and dynamic time step adjustment techniques, combined with an MPI parallel computing architecture, to efficiently solve multiphysics control equations, enabling simulation analysis under various operating conditions. Finally, the accuracy of the model is verified on an experimental platform, and the NSGA-III algorithm is used to optimize key parameters of the fuel cell. Combined with the optimized cross-scale model, predictive control dynamically adjusts the operating strategy, while machine learning is applied to evaluate the reliability throughout the entire lifecycle. This solves the problems of existing technologies, such as difficulty in balancing accuracy and efficiency, poor adaptability to dynamic and extreme operating conditions, and insufficient reliability assessment, providing strong technical support for the design optimization and engineering application of fuel cells.

[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the cross-scale multiphysics collaborative modeling method for fuel cells used in power generation provided in an embodiment of the present invention; Figure 2A schematic block diagram of a multi-scale multiphysics collaborative modeling system for fuel cells used in power generation, provided in an embodiment of the present invention. Figure 3 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] Please see Figure 1 , Figure 1This is a flowchart illustrating the multi-scale multiphysics collaborative modeling method for fuel cells used in power generation, provided in an embodiment of the present invention. This method is applied to a server and, by combining density functional theory, the lattice Boltzmann method, and computational fluid dynamics, establishes electrochemical reaction and transport models at micro, meso, and macro scales, respectively, and achieves cross-scale parameter transfer to construct a unified cross-scale model. Multi-source data obtained from integrated experimental devices are used for preprocessing and inversion of key parameters, and in-situ observations are employed to calibrate boundary conditions, ensuring model accuracy. A mesh is adaptively generated based on the physical field gradient, and the time step is dynamically adjusted. The multiphysics control equations are efficiently solved using an MPI parallel architecture, supporting simulation analysis under various operating conditions. Key fuel cell parameters are optimized using the NSGA-III algorithm, and the operating strategy is dynamically adjusted using the optimized cross-scale model predictive control technology. Machine learning is used to evaluate the reliability throughout the entire lifecycle. This method achieves seamless cross-scale integration from micro to macro, balancing accuracy and computational efficiency, and can dynamically adapt to different operating conditions, providing strong technical support for fuel cell design optimization and engineering applications.

[0026] Figure 1 This is a flowchart illustrating the cross-scale multiphysics collaborative modeling method for fuel cells used in power generation provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S110 to S140.

[0027] S110. Electrochemical reaction and transport models at the micro, meso, and macro scales are established using density functional theory, lattice Boltzmann method, and computational fluid dynamics, respectively, and cross-scale parameter transfer is performed to obtain cross-scale models.

[0028] In this embodiment, the cross-scale model refers to the comprehensive description of the entire fuel cell stack performance, from catalyst surface reactions to overall performance, by integrating electrochemical reaction and transport models at different micro, meso, and macro scales and utilizing cross-scale parameter transfer technology. This model can accurately analyze multi-physics coupling mechanisms, providing a complete technical solution for fuel cell performance optimization and reliability assessment.

[0029] In one embodiment, step S110 described above may include steps S111 to S114.

[0030] S111. Density functional theory was used to calculate the reaction energy barrier on the catalyst surface, and the Butler-Volmer and Nernst-Planck equations were combined to establish an electrochemical reaction and transport model to describe the electrochemical behavior at the microscale, thus forming a microscale model.

[0031] In this embodiment, the microscale model primarily focuses on the electrochemical behavior of the catalyst surface. In this step, density functional theory (DFT) is used to calculate the reaction energy barrier at the catalyst surface, and the Butler-Volmer and Nernst-Planck equations are combined to establish an electrochemical reaction and transport model. This allows us to quantify the effects of temperature, pressure, and other factors on catalyst activity, thereby accurately describing the electrochemical processes occurring at the microscale, such as proton conduction pathways and their interactions with electric and concentration fields.

[0032] S112. The lattice Boltzmann method is used to simulate the gas-liquid two-phase flow and mass transport in the porous electrode. Combined with the heat transfer equation, the characteristics of the diffusion layer and its influence on the mass transfer efficiency are analyzed to obtain a mesoscale model.

[0033] In this embodiment, the mesoscale model focuses on analyzing the gas-liquid two-phase flow and mass transport characteristics within the porous electrode. By employing the Lattice Boltzmann Method (LBM), these complex fluid behaviors can be simulated, and the characteristics of the diffusion layer and its impact on mass transfer efficiency can be analyzed in conjunction with heat transfer equations. This mesoscale model helps to understand the mass transport processes occurring within the electrode and how they affect the overall performance of the battery, such as the interaction mechanism between liquid water retention and gas diffusion.

[0034] S113. Utilize computational fluid dynamics to establish a multiphysics coupled model covering the flow field, temperature field, concentration field, and electric field to describe the performance and dynamic response of the entire fuel cell stack, forming a macroscopic scale model.

[0035] In this embodiment, the macroscopic model aims to describe the performance and dynamic response of the entire fuel cell stack, encompassing the multi-physics coupling of flow, temperature, concentration, and electric fields. Using computational fluid dynamics (CFD) technology, corresponding control equations can be established based on the specific structural and operational parameters of the stack, thereby providing the ability to predict the overall output characteristics of the stack, including its hydrothermal management efficiency and dynamic response patterns.

[0036] S114. The key reaction kinetic parameters of the microscale model are transferred to the mesoscale model, and the mass transport parameters of the mesoscale model are mapped to the macroscale model to obtain a cross-scale model.

[0037] To seamlessly integrate the models at different scales, cross-scale parameter transfer is required. This means extracting key reaction kinetic parameters from the microscale model and transferring them to the mesoscale model, while simultaneously mapping mass transport parameters from the mesoscale model to the macroscale model. This cross-scale parameter transfer mechanism ensures continuity and consistency from microscale reactions to macroscale stack performance, ultimately forming a comprehensive cross-scale model that enables accurate analysis and efficient simulation of the full range of fuel cell behavior.

[0038] In this embodiment, a kinetic model of electrochemical reaction on the catalyst surface is established based on density functional theory (DFT). The distribution of active sites and their relationship with proton conduction pathways are analyzed by calculating the reaction energy barrier. The Butler-Volmer equation and the Nernst-Planck equation are used to quantify the specific effects of temperature and pressure on catalyst activity, thereby forming a coupled model of electrochemical reaction with electric field and concentration field at the microscale.

[0039] A coupled model of gas-liquid two-phase flow, mass transport, and charge transfer within a porous electrode is created using the Lattice Boltzmann Method (LBM). Combining the electrical conductivity theory and heat transfer equations of porous media, the effects of diffusion layer porosity and tortuosity on mass transfer efficiency are analyzed, revealing the interaction mechanism between liquid water retention and gas diffusion. A coupled model of mass transport, charge transfer, and temperature field at the mesoscale is then constructed.

[0040] Applying the principles of computational fluid dynamics (CFD), multi-field coupled control equations for the flow field, temperature field, concentration field, and electric field within the fuel cell stack's internal flow channels are established. Based on the stack's structural and operational parameters, a macroscopic overall performance model of the stack is constructed, encompassing its output characteristics, hydrothermal management efficiency, and dynamic response characteristics.

[0041] Develop specific cross-scale parameter mapping algorithms to transfer reaction kinetic parameters at the microscale to the mesoscale model and map mass transport parameters at the mesoscale to the macroscale model, ensuring that the models at the micro, meso, and macroscale levels can be seamlessly connected and together form a complete cross-scale multiphysics coupling modeling system.

[0042] In summary, step S110, by combining DFT, LBM, and CFD methods, established detailed electrochemical reaction and transport models at different scales, and achieved effective connection between models through cross-scale parameter transfer, providing strong technical support for the design optimization of fuel cells.

[0043] S120. Acquire multi-source data collected by integrated experimental devices, preprocess the multi-source data, invert the key parameters of the cross-scale model, and calibrate the boundary conditions of the cross-scale model through in-situ observations to obtain the calibrated cross-scale model.

[0044] In this embodiment, the calibrated cross-scale model refers to a mathematical model that can accurately simulate the actual behavior of a fuel cell after multiple iterations to adjust boundary conditions and key parameters.

[0045] The goal of step S120 is to preprocess multi-source data collected by integrating multiple experimental devices, invert the key parameters of the cross-scale model based on these data, and then calibrate the boundary conditions of the model through in-situ observations to obtain an accurate calibrated cross-scale model.

[0046] In one embodiment, step S130 described above may include steps S131 to S134.

[0047] S131. Collect multi-source data using experimental equipment, wherein the multi-source data includes polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution.

[0048] In this embodiment, various experimental devices are used to collect detailed information about the fuel cell's performance and operating environment. This includes, but is not limited to, polarization curves (used to evaluate the relationship between the cell's current density and voltage), electrochemical impedance spectroscopy (revealing the cell's internal resistance characteristics), temperature distribution (reflecting the cell's thermal condition during operation), and water content distribution (understanding the cell's internal moisture management).

[0049] S132. The multi-source data is denoised, outliers are removed, and the data format is standardized to obtain the preprocessing result.

[0050] In this embodiment, the preprocessing result refers to the high-quality, consistent dataset obtained after denoising, outlier removal, and format standardization of the collected multi-source experimental data.

[0051] Specifically, noise reduction involves using techniques such as wavelet transform to remove noise interference.

[0052] Outlier detection and removal: Statistical methods such as the 3σ criterion are used to identify and remove outliers.

[0053] Data format standardization: Converting data from different sources into a unified format to facilitate subsequent processing.

[0054] Improve the quality of the data to ensure it is suitable for subsequent analysis and model parameter inversion.

[0055] In this embodiment, the preprocessing result refers to the dataset after the above processing, which has higher accuracy and consistency, and provides reliable data support for subsequent steps.

[0056] S133. Using a genetic algorithm and a particle swarm optimization algorithm, the key parameters in the cross-scale model are retrieved by combining the preprocessing results.

[0057] In this embodiment, the key parameters include catalyst reactivity parameters, diffusion layer porosity, and flow channel structure parameters.

[0058] Specifically, the values ​​of unknown parameters in the multi-scale model are determined based on the preprocessed data. Genetic algorithms and particle swarm optimization algorithms are applied in conjunction with the preprocessed data to retrieve key parameters, such as catalyst reactivity parameters, diffusion layer porosity, and flow channel structure parameters. This step aims to make the model calculation results as close as possible to the actual experimental data, thereby improving the model's predictive ability.

[0059] S134. By comparing the in-situ observation data with the simulation results, the temperature boundary, humidity boundary, and fluid inlet and outlet boundary conditions of the preprocessed results are iteratively adjusted, and after multiple iterations, the deviation between the predicted results and the experimental results of the preprocessed results is minimized, so as to obtain the calibrated cross-scale model.

[0060] By comparing experimental data and simulation results, the boundary conditions of the model are adjusted to minimize the deviation between the two.

[0061] Based on in-situ observation data (such as real-time monitored temperature, humidity, and inlet / outlet fluid states), the boundary conditions of the model (such as temperature boundaries, humidity boundaries, and fluid inlet / outlet boundaries) are iteratively adjusted. This process may require multiple iterations to achieve the best match, with the ultimate goal of minimizing the difference between the model's predictions and experimental observations.

[0062] In this embodiment, the calibrated cross-scale model refers to a mathematical model that, after the aforementioned series of processing steps, can more accurately simulate the behavior of fuel cells in the real world. This model not only considers physical phenomena at different scales from the microscopic to the macroscopic, but also achieves high-precision prediction of fuel cell performance by finely adjusting boundary conditions and key parameters, providing strong technical support for fuel cell design optimization and reliability assessment.

[0063] In this embodiment, during the fuel cell performance evaluation and optimization process, firstly, through the data acquisition unit integrating experimental devices such as an electrochemical workstation, an infrared thermal imager, and a neutron imaging device, multi-source experimental data, including polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution, can be acquired. This data covers important performance indicators of the fuel cell under different operating conditions, such as output voltage, current, and power, providing a solid foundation for subsequent analysis.

[0064] Subsequently, a series of processing techniques were applied to the collected raw experimental data, including wavelet transform denoising, outlier detection and removal, and data format standardization, to improve data quality and ensure consistency. This process is crucial for eliminating noise interference and correcting measurement errors, thus contributing to the accuracy of subsequent analysis results.

[0065] Next, a hybrid approach combining genetic algorithms and particle swarm optimization was used to incorporate preprocessed multi-source data to derive key parameters from the cross-scale model. These parameters included catalyst reactivity parameters, porous diffusion layer characteristics, and fuel cell stack flow channel structure parameters. The aim was to minimize the discrepancy between the model's calculated results and actual experimental data, thereby improving the model's prediction accuracy.

[0066] Finally, by comparing and analyzing in-situ observation data with simulation results, key boundary conditions of the model, such as temperature, humidity, and fluid inlet / outlet boundaries, were adjusted. Through continuous iterative calculations, this unit ensures that the model accurately reflects the actual operating environment of the fuel cell, thereby effectively improving its simulation and prediction capabilities. The entire process, from data acquisition to boundary condition calibration, constitutes a closed-loop system, guaranteeing the accuracy and reliability of the cross-scale model.

[0067] S130. Generate a mesh adaptively based on the physical field gradient and dynamically adjust the time step. Use the MPI parallel computing architecture to solve the multiphysics control equations corresponding to the calibrated cross-scale model and carry out simulation analysis under various working conditions to output the system performance characteristics.

[0068] In this embodiment, system performance characteristics refer to the operating performance of the fuel cell under different operating conditions obtained through simulation and experimentation, such as key indicators like power output, efficiency, and stability; it reflects the effectiveness of the system design and its reliability in actual operation.

[0069] In one embodiment, step S130 described above may include steps S131 to S134.

[0070] S131. Automatically adjust the grid density according to the physical field gradient, wherein the grid is denser in areas where the change range meets the requirements, and sparser in areas with gentle changes.

[0071] In one embodiment, step S131 described above may include steps S1311 to S1313.

[0072] S1311. Based on the gradient distribution characteristics of different physical fields of electrochemical reaction, mass transport, and thermal stress during fuel cell operation, determine the regions that need to be densified or sparsed to obtain the partitioning results.

[0073] In this embodiment, the partitioning result refers to the computational grid layout generated after automatically adjusting the grid density according to the physical field gradient, which ensures higher resolution in areas where the physical field changes drastically, thereby improving the accuracy and efficiency of numerical simulation.

[0074] S1312. Set a high grid density threshold for areas where the physical field change amplitude meets the requirements, and a low grid density for areas where the physical field change is gradual, thus forming the grid density requirement. S1313. Based on the set grid density requirements, automatically generate an unstructured adaptive grid from the partitioning results.

[0075] In this embodiment, firstly, in step S131, the system determines the regions to be densified or sparsed based on the gradient distribution characteristics of different physical fields such as electrochemical reactions, mass transport, and thermal stress during fuel cell operation (S1311). This stage requires accurately identifying the regions where the physical field changes most drastically, such as the catalyst layer and flow channel corners; these regions are crucial to the accuracy of the model. Next, in step S1312, the system sets a high grid density threshold for regions where the physical field changes meet the requirements, and sets a lower grid density for regions where the physical field changes are gradual, thus forming a detailed grid density requirement. Finally, in step S1313, based on the above-set grid density requirements, the system automatically generates an unstructured adaptive grid to ensure optimal allocation of computational resources. This adaptive grid technology not only improves computational accuracy but also significantly reduces unnecessary computation, improving overall computational efficiency.

[0076] S132, Set the time step adjustment range based on residual convergence.

[0077] In this embodiment, after mesh generation, step S132 involves dynamic adjustment of the time step. To ensure the stability and efficiency of numerical computation, the system sets a time step adjustment range based on residual convergence. Specifically, when the system detects that the residual of the numerical computation is less than a certain preset threshold, it automatically increases the time step to accelerate the computation process; conversely, if the residual exceeds the threshold, it decreases the time step to ensure the accuracy of the calculation results. This method is particularly suitable for handling transient processes in fuel cells, such as the response characteristics during load changes, and can effectively balance computational speed and accuracy.

[0078] S133. Decompose the multiphysics control equations corresponding to the calibrated cross-scale model into multiple sub-tasks, and solve them in parallel on a distributed computing architecture using MPI technology.

[0079] In this embodiment, step S133 further advances the computation process by decomposing the multiphysics control equations in the parameter-calibrated cross-scale model into multiple independent subtasks, and using Message Passing Interface (MPI) technology to solve them in parallel on a distributed computing architecture. This approach fully utilizes existing high-performance computing resources, significantly reducing computation time. Each computing node is responsible for solving a specific part of the control equations, and then data communication enables collaborative work between the subtasks. This parallel computing strategy not only improves computational efficiency but also makes solving large-scale complex problems possible.

[0080] S134. Simulate different working conditions, output the multi-physics distribution characteristics, performance indicators and dynamic response laws of the system, and form simulation results.

[0081] In this embodiment, the simulation results refer to a detailed description of the multi-physics field distribution characteristics and dynamic response laws inside the fuel cell, including temperature field, concentration field and current density distribution, etc.; it is an important basis for evaluating system performance and optimizing design schemes.

[0082] Finally, in step S134, the system performs extensive simulation analyses under different operating conditions, including but not limited to steady-state conditions, dynamic conditions (such as load step and start-stop cycles), and extreme environmental conditions (such as low-temperature start-up and high-temperature operation). Through these simulation experiments, the system can output the multi-physics distribution characteristics (such as temperature field, concentration field, and current density distribution), performance indicators (such as voltage and power output), and dynamic response laws of the fuel cell, ultimately forming detailed simulation results. These results are crucial for verifying the accuracy and reliability of the model and provide a solid foundation for subsequent parameter optimization and operational strategy adjustments.

[0083] In summary, step S130 and its sub-steps, from mesh generation and time step adjustment to parallel computing and simulation analysis, comprehensively cover the key aspects of efficient numerical solution, ensuring the high accuracy and efficiency of the model and providing strong technical support for fuel cell design optimization and performance evaluation.

[0084] In this embodiment, the mesh density is first automatically adjusted based on the physical field gradient distribution characteristics. Specifically, the mesh is denser in regions where the physical field changes drastically (such as the catalyst layer and flow channel corners), while the mesh is sparser in regions where the physical field is evenly distributed. This method achieves a reasonable allocation of computational resources, improving the computational accuracy in critical areas while reducing the overall computational cost, thus balancing the requirements for computational accuracy and efficiency.

[0085] Secondly, the time step is dynamically adjusted based on the convergence criterion of numerical computation. When handling dynamic conditions, a smaller time step (e.g., 1e-6 to 1e-4 seconds) is used to ensure high-precision calculation results; while under steady-state conditions, a larger time step (e.g., 1e-4 to 1e-2 seconds) is used to improve computational efficiency. In this way, computational efficiency can be dynamically optimized under different operating conditions while ensuring the accuracy of the calculation results.

[0086] Furthermore, based on a distributed computing architecture, the message passing interface (MPI) technology is used to solve the multiphysics control equations for electrochemical reactions, mass transport, heat transfer, and fluid flow in parallel. This not only significantly improves computational speed but also allows for the effective solution of complex problems across scale models. By decomposing the complex multiphysics coupled control equations into multiple subtasks and distributing them to different computing nodes for synchronous execution, and enabling collaborative work between subtasks through data communication, computational efficiency is further improved.

[0087] Finally, by introducing an artificial viscosity term and improving the iterative solution algorithm, the oscillation and divergence problems in the numerical calculation process of the cross-scale model were solved, ensuring the stability and convergence of the numerical solution. Furthermore, based on the constructed cross-scale model and the aforementioned efficient numerical solution algorithm, simulation analyses were conducted under steady-state conditions, dynamic conditions (such as load step and start-stop cycles), and extreme environmental conditions (such as -30℃ cold start and 85℃ high-temperature operation). These analyses can output the multiphysics distribution characteristics, system output performance, and dynamic response laws, providing important basis for subsequent design optimization. Thus, the entire module not only enhances the accuracy and efficiency of the calculations but also greatly promotes the understanding and mastery of the complex phenomena inside fuel cells.

[0088] S140. Verify the accuracy of the model on the experimental platform, optimize the key parameters of the fuel cell using the NSGA-Ⅲ algorithm, combine the optimized cross-scale model with predictive control to dynamically adjust the operation strategy, and evaluate the reliability throughout the entire life cycle through machine learning.

[0089] Step S140 aims to verify the accuracy of the model, optimize the key parameters of the fuel cell through a series of meticulous operations, and dynamically adjust the operating strategy based on the optimized cross-scale model predictive control. Finally, machine learning algorithms are used to evaluate the reliability of the fuel cell throughout its entire life cycle.

[0090] In one embodiment, step S140 described above may include steps S141 to S144.

[0091] S141. Compare the simulation results with the actual test data on the experimental platform to obtain the comparison results.

[0092] In this embodiment, the comparison result refers to the difference analysis between the actual operating data obtained through the experimental testing platform and the data predicted by the simulation model. This aims to verify the accuracy of the model and ensure that the error is within an acceptable range (e.g., ±5%). This result directly reflects the simulation accuracy of the cross-scale multiphysics coupling model for actual working conditions.

[0093] In this phase, a dedicated fuel cell experimental testing platform needs to be built to conduct performance tests under different operating conditions (including steady-state, dynamic, and extreme environmental conditions) and collect experimental data. This data will be compared with the simulation results of the previously established cross-scale multiphysics coupling model. The comparison covers key parameters such as voltage, current, temperature, and water content. The aim is to determine the deviation between the simulation results and the actual test data to verify the accuracy of the model. If the deviation exceeds a predetermined range (e.g., ±5%), the previous steps need to be repeated to calibrate the model and perform parameter inversion.

[0094] S142. Based on the comparison results, perform multi-objective optimization on the key design parameters of the fuel cell in the calibrated cross-scale model to obtain the optimized cross-scale model.

[0095] In this embodiment, the optimized cross-scale model refers to the improved model obtained by adjusting key design parameters using a multi-objective optimization algorithm based on the comparison results. Its purpose is to maximize power density, minimize mass transfer resistance and thermal stress, thereby providing a more accurate and efficient fuel cell performance prediction and optimization scheme. This model not only improves the accuracy and practicality of the original model but also provides a solid foundation for subsequent design optimization and operational strategy adjustments.

[0096] Specifically, based on the comparison results, a non-dominated sorting genetic algorithm is used to construct a multi-objective optimization function with the objectives of maximizing power density, minimizing mass transfer resistance and thermal stress. This function is then used to optimize key parameters such as bipolar channel width, diffusion layer porosity, and catalyst loading in the calibrated cross-scale model, thereby obtaining the Pareto optimal solution set and the optimized cross-scale model.

[0097] Once the model's accuracy was confirmed, the next step was to use the Non-Dominated Sorting Genetic Algorithm (NSGA-III) to perform multi-objective optimization of the key design parameters of the fuel cell based on the aforementioned comparison results. The optimization objectives were to maximize power density, minimize mass transfer resistance, and reduce thermal stress. Specifically, the parameters to be optimized included, but were not limited to, bipolar plate channel width, diffuser porosity, and catalyst loading. By constructing a multi-objective optimization function, a set of Pareto optimal solutions was obtained, thus forming the optimized cross-scale model. This process not only improved the model's accuracy and practicality but also provided a basis for subsequent design optimization.

[0098] S143. Integrate the optimized cross-scale model into the digital twin system, and apply the optimized cross-scale model predictive control technology to dynamically adjust the operation strategy.

[0099] Integrating validated and optimized cross-scale models into a digital twin system enables real-time monitoring of fuel cell operation and dynamic adjustment of operating strategies based on model predictive control algorithms. For example, humidification, coolant flow, or load distribution can be automatically adjusted according to the fuel cell's current operating status to improve system efficiency and adaptability. This approach not only enhances the fuel cell's responsiveness to dynamic loads but also extends its lifespan.

[0100] S144. Utilize machine learning algorithms to analyze data from simulated failure modes, establish a reliability assessment model, and predict the lifespan and performance degradation trend of fuel cells.

[0101] Specifically, by simulating the failure processes of water flooding, dry film, and catalyst agglomeration, corresponding failure characteristic parameters are extracted, and a failure mode recognition model and a reliability assessment model are established based on machine learning algorithms to predict the service life and performance degradation law of fuel cells under different operating conditions.

[0102] The final step involves using machine learning algorithms to analyze simulated failure mode data and build a reliability assessment model. This includes simulating failure processes such as flooding, dry film combustion, and catalyst agglomeration, extracting corresponding failure characteristic parameters, and constructing failure mode identification and reliability assessment models based on this data. This helps predict the lifespan and performance degradation patterns of fuel cells under different operating conditions, providing crucial support for design optimization and operation and maintenance decisions. Furthermore, by continuously accumulating operational data, the predictive accuracy of the model can be further improved, enhancing the overall performance and reliability of the fuel cell.

[0103] In this embodiment, a fuel cell experimental testing platform was built to conduct performance tests under steady-state, dynamic, and extreme conditions. The experimental data was compared with the simulation results to verify the accuracy of the model and ensure that the error between the model calculation results and the actual system operation does not exceed 5%. This step is crucial for confirming the simulation accuracy of the cross-scale multiphysics coupling model for actual operating conditions.

[0104] Based on a multi-objective optimization algorithm, with the optimization objectives of increasing power density, reducing mass transfer resistance, and minimizing thermal stress, key parameters such as bipolar plate flow channel structure, diffuser layer porosity, and catalyst loading are synergistically optimized to output the optimal parameter combination. This process not only enhances the efficiency of the original design but also provides a basis for subsequent design optimization. Simultaneously, the operation strategy optimization unit embeds a cross-scale model into the digital twin system and combines it with model predictive control algorithms to optimize operating strategies such as humidification, coolant flow rate, and load distribution, thereby enhancing the fuel cell's adaptability to dynamic loads and improving overall operating efficiency.

[0105] Finally, by simulating failure processes such as flooding, dry film formation, and catalyst agglomeration, and combining this with machine learning algorithms to identify the critical conditions of key failure modes, a full life-cycle reliability assessment model was constructed to predict the service life and performance degradation patterns of fuel cells. This unit provides crucial support for fuel cell design optimization and operation and maintenance decisions.

[0106] Specifically, firstly, in the model validation phase, a fuel cell experimental test bench was built to conduct experimental tests corresponding to the simulated operating conditions. Experimental data was collected and compared with simulation results to analyze deviations in key parameters such as voltage, current, temperature, and water content. If the deviation exceeds 5%, parameter inversion and boundary condition calibration need to be performed again until the deviation meets the requirements. Next, with the objectives of maximizing power density, minimizing mass transfer resistance, and minimizing thermal stress, a multi-objective optimization function was constructed. The non-dominated sorting genetic algorithm (NSGA-Ⅲ) was used to optimize the key parameters, outputting the Pareto optimal solution set. The optimized cross-scale model was embedded into a digital twin system, combined with model predictive control algorithms, to monitor the fuel cell operating status in real time and dynamically adjust the operating strategy to improve the system's adaptability to dynamic loads and operating efficiency. Finally, in the reliability assessment phase, various failure processes were simulated to extract failure characteristic parameters. Machine learning algorithms were used to construct a failure mode recognition model and a reliability assessment model to predict the service life and performance degradation patterns of the fuel cell under different operating conditions, providing scientific support for fuel cell design optimization and operation and maintenance decisions.

[0107] In this embodiment, to address the pain points of fragmented multiphysics modeling, imbalance between accuracy and efficiency, and poor dynamic operating condition adaptation of fuel cells for power generation, a full-process design involving cross-scale coupled modeling, multi-source data fusion, efficient numerical solution, and closed-loop optimization is adopted to achieve accurate simulation and optimization from microscopic reactions to macroscopic stack performance.

[0108] This embodiment focuses on a 100kW-class proton exchange membrane fuel cell for power generation, and constructs a cross-scale multiphysics collaborative modeling system, the specific structure of which is as follows.

[0109] The cross-scale modeling unit comprises four core units. The microscopic modeling unit, based on density functional theory, uses VASP software to calculate the energy barrier for the hydrogenation reaction on the platinum-based catalyst surface. It then establishes an electrochemical reaction kinetic model using the Butler-Volmer equation, considering the effects of temperature on the activation energy and exchange current density. The model covers key parameters such as catalyst active site density and reaction pathway selectivity. The mesoscopic modeling unit employs the lattice Boltzmann method, building a gas-liquid two-phase flow model within a porous electrode based on the Palabos open-source library. The porosity is set to 0.4–0.7, and the tortuosity to 1.5–2.5. Combining porous media conductivity theory and Fourier's law, a coupled mass transport-charge transfer-temperature field model is constructed. The macroscopic modeling unit, based on ANSYS Fluent software, uses user-defined functions to write multiphysics coupling control equations. The fuel cell stack flow channel adopts a serpentine structure with a width of 2–4 mm and a height of 1–2 mm, constructing a coupled flow field-temperature field-concentration field-electric field model. The cross-scale parameter transfer unit develops a cross-scale parameter mapping model based on artificial neural networks, which maps the microscopic catalyst reaction activity parameters to the mesoscopic porous electrode model and the mesoscopic diffusion layer mass transfer coefficient to the macroscopic stack model, with the mapping error controlled within 3%.

[0110] The multi-source data fusion unit also comprises four functional units. The data acquisition unit integrates an electrochemical workstation, an infrared thermal imager, and an imaging system to acquire multi-dimensional experimental data such as polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution. The data preprocessing unit uses the db4 wavelet basis to denoise the impedance spectroscopy data, removes outliers from the polarization curves using the 3σ criterion, standardizes all data to CSV format, and unifies the sampling frequency to 100Hz. The parameter inversion unit employs a hybrid algorithm combining genetic algorithm and particle swarm optimization, with a population size of 100 and 200 iterations, to invert 20 key parameters, including catalyst exchange current density, diffusion layer porosity, and flow channel resistance coefficient, with an inversion accuracy controlled within 2%. The boundary condition calibration unit calibrates the temperature boundary based on infrared thermal imaging data, with an error controlled within ±0.5℃, and calibrates the water content boundary based on neutron imaging data, with an error controlled within ±5%.

[0111] The high-efficiency numerical solution unit comprises four technical units. The adaptive mesh generation unit uses ANSYS ICEM software to generate unstructured meshes, with a mesh density threshold of 0.1, a catalyst layer mesh size of 1 μm, a flow channel region mesh size of 100 μm, and a total mesh count of 5 million to 10 million. The dynamic time step adjustment unit sets the time step adjustment range to 1e-6 to 1e-2 s, with a residual convergence threshold of 1e-6. When the residual is less than the threshold, the time step is automatically increased; conversely, it is decreased. The multiphysics parallel solver is based on the MPI parallel computing architecture, employing 8 to 16 computing nodes, each configured with a 16-core CPU, to achieve parallel solution of the multiphysics control equations, improving computational efficiency by 8 to 12 times. The numerical stability optimization unit introduces artificial viscosity terms into the control equations and uses the SIMPLEC algorithm to improve the pressure-velocity coupling solution, ensuring the stability of the numerical solution.

[0112] The model validation and optimization unit comprises four application units. The model validation unit constructs a 100kW-level proton exchange membrane fuel cell test bench, enabling wide temperature range adjustment and dynamic loading. Comparison of experimental data and simulation results shows that voltage deviation is controlled within 3%, and power deviation within 4%, meeting the error requirements. The parameter optimization unit uses the NSGA-Ⅲ algorithm to optimize parameters such as bipolar plate channel width, diffusion layer porosity, and catalyst loading. After optimization, power density is increased by 15%, mass transfer resistance is reduced by 20%, and thermal stress is reduced by 18%. The operation strategy optimization unit, based on a model predictive control algorithm, dynamically adjusts humidification and coolant flow rate, reducing voltage fluctuation amplitude under dynamic load by 30%. The reliability assessment unit constructs a failure mode recognition model based on an LSTM neural network, achieving an accuracy rate of over 95%, and the reliability assessment model's predicted service life error is controlled within 8%.

[0113] The above system is used to implement a cross-scale multiphysics collaborative modeling method for fuel cells used in power generation. The specific process is divided into four stages.

[0114] In the cross-scale modeling system construction phase, the microscopic model was first constructed. The energy barrier for the hydroxide reaction on the surface of the platinum-based catalyst was calculated using VASP software to obtain the activation energy at different temperatures. An electrochemical reaction kinetic model was established using the Butler-Volmer equation, and the Nernst-Planck equation was introduced to describe the proton transport process between the catalyst layer and the proton exchange membrane, thus constructing a microscopic-scale electrochemical reaction-electric field-concentration field coupled model. Next, the mesoscopic model was constructed. Based on the palabos open-source library, a gas-liquid two-phase flow transport model within a porous electrode was established using the lattice Boltzmann method. The porosity of the diffusion layer was set to 0.55, and the tortuosity to 1.8. Combining the conductivity theory of porous media and Fourier's law, a mesoscopic-scale mass transport-charge transfer-temperature field coupled model was constructed. Finally, the macroscopic model was constructed. A serpentine flow channel fuel cell stack model was established in ANSYS Fluent, with a channel width of 3 mm and a height of 1.5 mm. Multiphysics coupling control equations were written using user-defined functions to construct a macroscopic-scale flow field-temperature field-concentration field-electric field coupled model. Finally, cross-scale integration was achieved by developing an artificial neural network cross-scale parameter mapping model, which maps the microscopic catalyst exchange current density to the mesoscopic model and the mesoscopic diffusion layer mass transfer coefficient to the macroscopic model, thus completing the seamless integration of microscopic-mesoscopic-macroscopic cross-scale models.

[0115] In the multi-source data fusion and parameter calibration phase, data acquisition was first performed. Polarization curves and electrochemical impedance spectroscopy were acquired using a CHI660E electrochemical workstation, the surface temperature distribution of the fuel cell stack was acquired using a FLIR A655 infrared thermal imager, and water content distribution data was obtained using a neutron imaging system. Data preprocessing followed, using the db4 wavelet basis to reduce noise in the impedance spectrum data, removing outliers from the polarization curves using the 3σ criterion, and standardizing all data to CSV format with a uniform sampling frequency of 100Hz. Next, parameter inversion was performed using a hybrid algorithm combining genetic algorithm and particle swarm optimization (PSO) with a population size of 100 and 200 iterations. Twenty key parameters, including catalyst exchange current density and diffusion layer porosity, were inverted. After inversion, the average deviation between the model's calculated values ​​and experimental data was reduced to within 2%. Finally, boundary condition calibration was completed. The fuel cell stack temperature boundary was calibrated based on infrared thermal imaging data, adjusting the coolant inlet temperature boundary to a set value ±0.5℃. The water content boundary was calibrated based on neutron imaging data to ensure consistency between the water content distribution in the model and experimental observations.

[0116] In the efficient numerical solution and simulation analysis stage, an adaptive mesh was first generated using ANSYS ICEM to create an unstructured mesh. The catalyst layer mesh size was 1 μm, the flow channel region mesh size was 100 μm, and the total mesh size was 8 million, achieving accurate capture of regions with drastic changes in the physical field. Then, a dynamic time step was set, adjustable from 1e-6 to 1e-2 s, with a residual convergence threshold of 1e-6. Under dynamic load abrupt changes, a time step of 1e-6 s was automatically adopted, while under steady-state conditions, a time step of 1e-3 s was used. Next, multi-physics parallel solution was performed using an MPI parallel computing architecture with 12 computing nodes, each with 16 CPU cores, to achieve parallel solution of the multi-physics governing equations, reducing the computation time from 72 hours per node to 8 hours. Finally, simulation analysis was conducted, performing simulations under steady-state, dynamic, and extreme environments, outputting the internal temperature distribution, concentration distribution, current density distribution, and dynamic response curves of the fuel cell stack.

[0117] In the model validation and optimization application phase, model validation was first conducted. Experimental tests were carried out on a 100kW fuel cell test bench under corresponding operating conditions. Comparison of experimental data and simulation results showed that the voltage deviation was 2.8% under steady-state conditions, 3.5% under dynamic conditions, and 4.2% under extreme conditions, all meeting the error requirements. Next, parameter optimization was performed. The NSGA-III algorithm was used to optimize parameters such as bipolar plate channel width, diffusion layer porosity, and catalyst loading, aiming to maximize power density, minimize mass transfer resistance, and minimize thermal stress. The Pareto optimal solution set was obtained, and the power density increased from 0.8 W / cm² to 0.92 W / cm² after selecting the optimal parameter combination. Finally, the operating strategy was optimized. The optimized cross-scale model was embedded into a digital twin system, and the humidification and coolant flow rates were dynamically adjusted using a model predictive control algorithm, reducing the voltage fluctuation amplitude under dynamic load from 15% to 10.5%. Finally, a reliability assessment was completed. By simulating failure processes such as water flooding, dry film, and catalyst agglomeration, failure characteristic parameters were extracted. A failure mode recognition model was built based on an LSTM neural network, with an accuracy rate of 96.3%. The reliability assessment model predicted the fuel cell service life to be 8,000 hours, which deviated from the accelerated aging test results by 2.4%.

[0118] The method in this embodiment achieves accurate analysis of the multi-physics coupling mechanism of fuel cells for power generation by constructing a multi-scale multi-physics collaborative modeling system spanning micro-, meso-, and macro-scales, thus solving problems such as scale fragmentation and the contradiction between accuracy and efficiency in traditional modeling methods. Through multi-source data fusion and efficient numerical solution techniques, the accuracy and computational efficiency of the model are significantly improved. The model verification and optimization module provides full-process technical support for fuel cell parameter design, operation strategy optimization, and reliability assessment, demonstrating significant engineering application value and promising prospects for widespread adoption.

[0119] Therefore, the method in this embodiment aims to solve the problems commonly found in existing fuel cell modeling, such as scale fragmentation, difficulty in balancing accuracy and computational efficiency, poor adaptability to dynamic and extreme conditions, and insufficient reliability assessment. To achieve this goal, the system consists of four core modules: a cross-scale modeling module, a multi-source data fusion module, an efficient numerical solution module, and a model verification and optimization module.

[0120] First, by constructing a multi-scale, multi-physics coupled model spanning from microscopic catalyst layers to mesoscopic porous electrodes and finally to the macroscopic fuel cell stack, the method in this embodiment achieves accurate simulation of physical phenomena at different scales, thus overcoming the scale fragmentation problem in traditional methods. Second, multi-source data fusion technology is employed to integrate various data sources, such as polarization curves, impedance spectra, and in-situ observations, to complete parameter inversion and boundary condition calibration, ensuring the accuracy and applicability of the model. Furthermore, to improve computational efficiency, the method in this embodiment utilizes advanced techniques such as adaptive meshing, dynamic time steps, and parallel computing, significantly improving solution speed and efficiency while ensuring that the model's computational error does not exceed 5%.

[0121] Furthermore, the method in this embodiment is not limited to model building and solving, but also includes structural parameter optimization, operational strategy optimization, and full life-cycle reliability assessment functions. Specifically, the model verification and optimization module can optimize the design parameters and operational strategies of fuel cells, and predict their service life and performance degradation patterns, providing scientific decision support. This enables the method in this embodiment to not only effectively improve the accuracy and computational efficiency of multiphysics modeling of fuel cells, but also adapt to the needs of extreme operating conditions such as -30℃ cold start and 85℃ high temperature.

[0122] In summary, the method of this embodiment provides a comprehensive technical solution, covering the entire process from fuel cell design optimization and performance improvement to life prediction. In particular, it demonstrates excellent capabilities in dealing with dynamic loads and extreme environments, providing strong support for the application and development of fuel cells for power generation.

[0123] The aforementioned multi-physics collaborative modeling method for fuel cells used in power generation establishes an electrochemical reaction and transport model spanning micro, meso, and macro scales by combining density functional theory, lattice Boltzmann method, and computational fluid dynamics, and realizes cross-scale parameter transfer. It utilizes various experimental devices to collect and preprocess multi-source data, inverts key parameters, and calibrates boundary conditions through in-situ observations to form a calibrated multi-scale model. Adaptive grid generation and dynamic time step adjustment techniques based on physics gradients, combined with an MPI parallel computing architecture, are employed to efficiently solve the multi-physics control equations, enabling simulation analysis under various operating conditions. Finally, the model accuracy is verified on an experimental platform, and the NSGA-III algorithm is used to optimize the key parameters of the fuel cell. The optimized multi-scale model is then used to predict and control the dynamic adjustment of the operating strategy, while machine learning is applied to evaluate the reliability throughout the entire lifecycle. This approach addresses the problems of existing technologies, such as difficulty in balancing accuracy and efficiency, poor adaptability to dynamic and extreme operating conditions, and insufficient reliability assessment, providing strong technical support for the design optimization and engineering application of fuel cells.

[0124] Figure 3 This is a schematic block diagram of a multi-scale multiphysics collaborative modeling system 300 for power generation fuel cells provided in an embodiment of the present invention. Figure 3 As shown, corresponding to the above-described cross-scale multiphysics collaborative modeling method for fuel cells used in power generation, this invention also provides a cross-scale multiphysics collaborative modeling system 300 for fuel cells used in power generation. This cross-scale multiphysics collaborative modeling system 300 includes units for executing the above-described cross-scale multiphysics collaborative modeling method for fuel cells used in power generation, and the system can be configured in a server. Specifically, please refer to... Figure 3 The cross-scale multi-physics collaborative modeling system 300 for power generation fuel cells includes a cross-scale modeling unit 301, a multi-source data fusion unit 302, an efficient numerical solution unit 303, and a model verification and optimization unit 304.

[0125] The cross-scale modeling unit 301 is used to establish electrochemical reaction and transport models at the micro, meso, and macro scales respectively using density functional theory, lattice Boltzmann method, and computational fluid dynamics, and to perform cross-scale parameter transfer to obtain the cross-scale model. The multi-source data fusion unit 302 is used to acquire multi-source data collected by integrated experimental devices, preprocess the multi-source data, invert the key parameters of the cross-scale model, and calibrate the boundary conditions of the cross-scale model through in-situ observation to obtain the calibrated cross-scale model. The efficient numerical solution unit 303 is used to adaptively generate a grid based on the physical field gradient and dynamically adjust the time step, use the MPI parallel computing architecture to solve the multiphysics control equations corresponding to the calibrated cross-scale model, and conduct simulation analysis under various operating conditions to output the system performance characteristics. The model verification and optimization unit 304 is used to verify the accuracy of the model on the experimental platform, optimize the key parameters of the fuel cell using the NSGA-Ⅲ algorithm, combine the optimized cross-scale model with predictive control to dynamically adjust the operating strategy, and evaluate the full life cycle reliability through machine learning.

[0126] In one embodiment, the cross-scale modeling unit 301 includes: The microscale modeling module uses density functional theory to calculate the reaction energy barrier on the catalyst surface and combines the Butler-Volmer and Nernst-Planck equations to establish an electrochemical reaction and transport model to describe the electrochemical behavior at the microscale, thus forming a microscale model. The mesoscale modeling module uses the lattice Boltzmann method to simulate the gas-liquid two-phase flow and mass transport within the porous electrode, and combines heat transfer equations to analyze the characteristics of the diffusion layer and its impact on mass transfer efficiency, thus obtaining a mesoscale model. The macroscale modeling module uses computational fluid dynamics to establish a multiphysics coupled model covering the flow field, temperature field, concentration field, and electric field to describe the performance and dynamic response of the entire fuel cell stack, thus forming a macroscale model. The cross-scale parameter transfer module is used to transfer the key reaction kinetic parameters of the microscale model to the mesoscale model and map the mass transport parameters of the mesoscale model to the macroscale model, thus obtaining a cross-scale model.

[0127] In one embodiment, the multi-source data fusion unit 302 includes: The data acquisition module is used to collect multi-source data using experimental equipment, including polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution. The data preprocessing module is used to reduce noise, remove outliers, and standardize the data format to obtain preprocessed results. The parameter inversion module uses a genetic algorithm and a particle swarm optimization algorithm, combined with the preprocessed results, to invert the key parameters in the multi-scale model. The boundary condition calibration module iteratively adjusts the temperature boundary, humidity boundary, and fluid inlet / outlet boundary conditions of the preprocessed results by comparing in-situ observation data with simulation results, and iterates multiple times until the deviation between the predicted and experimental results of the preprocessed results is minimized, thus obtaining a calibrated multi-scale model.

[0128] In one embodiment, the high-efficiency numerical solution unit 303 includes: An adaptive mesh generation module is used to automatically adjust the mesh density based on the physical field gradient, wherein the mesh is denser in regions where the variation range meets the requirements, and sparser in regions with gentle variation. A dynamic time step adjustment module is used to set the time step adjustment range based on residual convergence. A multiphysics parallel solution module is used to decompose the multiphysics control equations corresponding to the calibrated cross-scale model into multiple sub-tasks and solve them in parallel on a distributed computing architecture using MPI technology. A numerical stability optimization module is used to simulate different operating conditions and output the multiphysics distribution characteristics, performance indicators and dynamic response laws of the system to form simulation results.

[0129] In one embodiment, the adaptive mesh generation module is used to determine the regions that need to be densified or sparsed based on the gradient distribution characteristics of different physical fields such as electrochemical reactions, mass transport, and thermal stress during fuel cell operation, so as to obtain the partitioning result; set a high mesh density threshold for regions where the physical field change amplitude meets the requirements, and a low mesh density for regions where the physical field change is gradual, thus forming a mesh density requirement; and automatically generate an unstructured adaptive mesh from the partitioning result according to the set mesh density requirement.

[0130] In one embodiment, the model validation optimization unit 304 includes: The model verification module is used to compare the simulation results with actual test data on an experimental platform to obtain comparison results; the parameter optimization module is used to perform multi-objective optimization of the key design parameters of the fuel cell in the calibrated cross-scale model based on the comparison results to obtain an optimized cross-scale model; the operation strategy optimization module is used to integrate the optimized cross-scale model into the digital twin system and apply the optimized cross-scale model predictive control technology to dynamically adjust the operation strategy; the reliability assessment module is used to analyze the data of simulated failure modes using machine learning algorithms, establish a reliability assessment model, and predict the service life and performance degradation trend of the fuel cell.

[0131] In one embodiment, the parameter optimization module is used to construct a multi-objective optimization function based on the comparison results using a non-dominated sorting genetic algorithm, with the objectives of maximizing power density, minimizing mass transfer resistance and thermal stress, to optimize key parameters such as bipolar channel width, diffusion layer porosity and catalyst loading in the calibrated cross-scale model, thereby obtaining the Pareto optimal solution set and thus obtaining the optimized cross-scale model.

[0132] In one embodiment, the reliability assessment module is used to extract corresponding failure characteristic parameters by simulating the failure processes of water flooding, dry film, and catalyst agglomeration, and to establish a failure mode recognition model and a reliability assessment model based on machine learning algorithms, so as to predict the service life and performance degradation law of fuel cells under different operating conditions.

[0133] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned cross-scale multiphysics collaborative modeling system 300 for power generation fuel cells and each unit can be referred to the corresponding description in the aforementioned method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0134] The aforementioned multi-scale multiphysics collaborative modeling system 300 for fuel cells used in power generation can be implemented as a computer program, which can be used in various applications such as... Figure 3 It runs on the computer device shown.

[0135] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0136] See Figure 3 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0137] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a cross-scale multiphysics collaborative modeling method for fuel cells used in power generation.

[0138] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0139] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a cross-scale multiphysics collaborative modeling method for fuel cells used in power generation.

[0140] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] The processor 502 is used to run the computer program 5032 stored in the memory to implement all the steps of the cross-scale multiphysics collaborative modeling method for fuel cells used for power generation.

[0142] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0143] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0144] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform all steps of the cross-scale multiphysics co-modeling method for fuel cells used in power generation.

[0145] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0146] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0147] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0148] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0149] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0150] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A cross-scale multiphysics collaborative modeling method for fuel cells used in power generation, characterized in that, include: Electrochemical reaction and transport models at the micro, meso, and macro scales were established using density functional theory, lattice Boltzmann method, and computational fluid dynamics, respectively, and cross-scale parameter transfer was performed to obtain cross-scale models. Multi-source data collected by integrated experimental devices is acquired, the multi-source data is preprocessed, and the key parameters of the cross-scale model are inverted. The boundary conditions of the cross-scale model are calibrated by in-situ observation to obtain the calibrated cross-scale model. The grid is adaptively generated based on the physical field gradient and the time step is dynamically adjusted. The multiphysics control equations corresponding to the calibrated cross-scale model are solved using the MPI parallel computing architecture. Simulation analysis is carried out under various working conditions to output the system performance characteristics. The accuracy of the model was verified on the experimental platform. The key parameters of the fuel cell were optimized using the NSGA-Ⅲ algorithm. The optimized cross-scale model was combined with predictive control to dynamically adjust the operation strategy. The reliability throughout the entire life cycle was evaluated through machine learning.

2. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 1, characterized in that, The method employs density functional theory, the lattice Boltzmann method, and computational fluid dynamics to establish electrochemical reaction and transport models at microscopic, mesoscopic, and macroscopic scales, respectively, and performs cross-scale parameter transfer to obtain a cross-scale model, including: Density functional theory was used to calculate the reaction energy barrier on the catalyst surface, and the Butler-Volmer and Nernst-Planck equations were combined to establish an electrochemical reaction and transport model to describe the electrochemical behavior at the microscale, thus forming a microscale model. The lattice Boltzmann method was used to simulate the gas-liquid two-phase flow and mass transport in a porous electrode. Combined with the heat transfer equation, the characteristics of the diffusion layer and its influence on the mass transfer efficiency were analyzed to obtain a mesoscale model. Computational fluid dynamics is used to establish a multiphysics coupled model covering flow field, temperature field, concentration field and electric field to describe the performance and dynamic response of the entire fuel cell stack, forming a macroscopic scale model; The key reaction kinetic parameters of the microscale model are passed to the mesoscale model, and the mass transport parameters of the mesoscale model are mapped to the macroscale model to obtain a cross-scale model.

3. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 1, characterized in that, The process of acquiring multi-source data collected by integrated experimental devices, preprocessing the multi-source data, inverting the key parameters of the cross-scale model, and calibrating the boundary conditions of the cross-scale model through in-situ observations to obtain the calibrated cross-scale model includes: Multi-source data were collected using an experimental setup, including polarization curves, electrochemical impedance spectroscopy, temperature distribution, and water content distribution. The multi-source data is denoised, outliers are removed, and the data format is standardized to obtain preprocessed results. The key parameters in the cross-scale model are retrieved by combining the genetic algorithm and particle swarm optimization algorithm with the preprocessing results. By comparing in-situ observation data with simulation results, the temperature boundary, humidity boundary, and fluid inlet and outlet boundary conditions of the preprocessed results are iteratively adjusted, and after multiple iterations, the deviation between the predicted results and experimental results of the preprocessed results is minimized, so as to obtain the calibrated cross-scale model.

4. The cross-scale multiphysics collaborative modeling method for fuel cells used in power generation according to claim 3, characterized in that, The key parameters include catalyst reactivity parameters, diffusion layer porosity, and flow channel structure parameters.

5. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 1, characterized in that, The process involves adaptively generating a mesh based on the physical field gradient and dynamically adjusting the time step. The multiphysics control equations corresponding to the calibrated cross-scale model are solved using an MPI parallel computing architecture. Simulation analyses are then conducted under various operating conditions to output system performance characteristics, including: The grid density is automatically adjusted based on the physical field gradient, with a denser grid in areas where the change range meets the requirements and a sparser grid in areas with a gentler change. Set the time step adjustment range based on residual convergence; The multiphysics control equations corresponding to the calibrated cross-scale model are decomposed into multiple sub-tasks and solved in parallel on a distributed computing architecture using MPI technology. Simulations are performed under different operating conditions to output the multiphysics distribution characteristics, performance indicators, and dynamic response laws of the system, forming simulation results.

6. The cross-scale multiphysics collaborative modeling method for fuel cells used in power generation according to claim 5, characterized in that, The automatic adjustment of mesh density based on the physical field gradient, wherein the mesh is denser in regions where the variation amplitude meets the requirements, and sparser in regions with gentle variation, includes: Based on the gradient distribution characteristics of different physical fields of electrochemical reaction, mass transport, and thermal stress during fuel cell operation, the regions that need to be densified or sparsed are determined to obtain the partitioning results. A high grid density threshold is set for regions where the physical field changes meet the requirements, while a low grid density is set for regions where the physical field changes are gradual, thus forming the grid density requirement. Based on the set grid density requirements, an unstructured adaptive grid is automatically generated from the partitioning results.

7. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 1, characterized in that, The process involves validating the model's accuracy on an experimental platform, optimizing key fuel cell parameters using the NSGA-III algorithm, dynamically adjusting the operating strategy using the optimized cross-scale model predictive control, and evaluating the full lifecycle reliability through machine learning. This includes: The simulation results were compared with the actual test data on the experimental platform to obtain the comparison results; Based on the comparison results, the key design parameters of the fuel cell in the calibrated cross-scale model are optimized using multiple objectives to obtain the optimized cross-scale model. The optimized cross-scale model is integrated into the digital twin system, and the optimized cross-scale model predictive control technology is applied to dynamically adjust the operation strategy. By using machine learning algorithms to analyze data from simulated failure modes, a reliability assessment model is established to predict the lifespan and performance degradation trend of fuel cells.

8. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 7, characterized in that, The process of performing multi-objective optimization on the key design parameters of the fuel cell in the calibrated cross-scale model based on the comparison results to obtain the optimized cross-scale model includes: Based on the comparison results, a multi-objective optimization function was constructed using a non-dominated sorting genetic algorithm, with the objectives of maximizing power density, minimizing mass transfer resistance, and minimizing thermal stress. This function was then used to optimize key parameters such as bipolar channel width, diffusion layer porosity, and catalyst loading in the calibrated cross-scale model, thereby obtaining the Pareto optimal solution set and the optimized cross-scale model.

9. The cross-scale multiphysics collaborative modeling method for fuel cells for power generation according to claim 7, characterized in that, The method of using machine learning algorithms to analyze simulated failure mode data, establish a reliability assessment model, and predict fuel cell lifespan and performance degradation trends includes: By simulating the failure processes of water flooding, dry film, and catalyst agglomeration, corresponding failure characteristic parameters are extracted, and a failure mode recognition model and a reliability assessment model are established based on machine learning algorithms to predict the service life and performance degradation law of fuel cells under different operating conditions.

10. A multi-scale, multi-physics collaborative modeling system for fuel cells used in power generation, characterized in that, include: The cross-scale modeling unit is used to establish electrochemical reaction and transport models at the micro, meso, and macro scales using density functional theory, lattice Boltzmann method, and computational fluid dynamics, respectively, and to perform cross-scale parameter transfer to obtain cross-scale models. The multi-source data fusion unit is used to acquire multi-source data collected by integrated experimental devices, preprocess the multi-source data, invert the key parameters of the cross-scale model, and calibrate the boundary conditions of the cross-scale model through in-situ observations to obtain the calibrated cross-scale model. The high-efficiency numerical solution unit is used to adaptively generate the mesh according to the physical field gradient and dynamically adjust the time step. It uses the MPI parallel computing architecture to solve the multiphysics field control equations corresponding to the calibrated cross-scale model, and carries out simulation analysis under various working conditions to output the system performance characteristics. The model validation and optimization unit is used to verify the accuracy of the model on the experimental platform. It uses the NSGA-Ⅲ algorithm to optimize the key parameters of the fuel cell, combines the optimized cross-scale model predictive control to dynamically adjust the operation strategy, and evaluates the reliability throughout the entire life cycle through machine learning.

Images