CN115017803B - Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack - Google Patents
Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack Download PDFInfo
- Publication number
- CN115017803B CN115017803B CN202210595253.2A CN202210595253A CN115017803B CN 115017803 B CN115017803 B CN 115017803B CN 202210595253 A CN202210595253 A CN 202210595253A CN 115017803 B CN115017803 B CN 115017803B
- Authority
- CN
- China
- Prior art keywords
- fuel cell
- cell stack
- porous structure
- thermal mass
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/30—Hydrogen technology
- Y02E60/50—Fuel cells
Abstract
The invention provides a design method for local and global directional regulation and control of thermal mass resistance in a porous structure of a fuel cell stack, which is based on the idea of combining deep learning and topology optimization, combines a deep convolutional neural network model with a multi-scale model and a topology optimization method, avoids the problem of high calculation cost caused by repeated large-scale finite volume solution in the topology optimization process, can quickly realize local random area and global directional regulation and control of thermal mass resistance in the porous structure of the fuel cell stack under the constraint of multi-dimensional large variables, realizes small volume and light weight of the fuel cell stack, and greatly improves the comprehensive performance.
Description
Technical Field
The invention belongs to the field of aerospace new energy, and particularly relates to a design method for locally and globally regulating and controlling internal thermal mass resistance of a porous structure of a fuel cell stack.
Background
The single cell of the fuel cell consists of a membrane electrode, a bipolar plate and a sealing piece, wherein the membrane electrode comprises a gas diffusion layer, a catalytic layer and a proton exchange membrane. The gas diffusion layer is an important component of the membrane electrode, plays a role of supporting the catalyst layer and is also a channel for reaction gas and product water. The main catalysts of the cathode and the anode of the proton exchange membrane fuel cell are mainly platinum and platinum carbon particles, and the proton exchange membrane has the characteristics of higher proton conductivity, low gas or fuel permeability, small electroosmosis coefficient of water, better chemical and electrochemical stability, good mechanical strength, lower cost and the like. The primary functions of the bipolar plates are to distribute fuel and oxidant in the cells, separate the cells in the stack, conduct current, transport generated water and moisture, cool the stack, etc. The anode gas diffusion layer is mainly used for transmitting hydrogen and water; the anode catalytic layer is mainly formed by oxidizing hydrogen under the action of platinum catalysis to generate protons; a proton exchange membrane, which is used for conducting protons generated by the anode to the cathode; the cathode catalytic layer is mainly formed by combining oxygen with protons under the catalytic action of platinum to generate reduction reaction to generate product water; the cathode gas diffusion layer is mainly used for transmitting oxygen and product water.
The fuel cell gas diffusion layer is a hydrophobic porous dielectric material positioned between the catalytic layer and the flow field plates that acts as a carrier for water-gas transport, heat transfer, and electron transfer, providing structural support during stack assembly and operation. The gas diffusion layer is typically composed of a macroporous substrate layer and a microporous layer. Wherein, the substrate layer is generally composed of carbon fibers which are in anisotropic random stacking in a plane and are in direct contact with the flow field plate; the microporous layer is formed by mixing carbon-based powder and a water repellent and is in direct contact with the catalytic layer.
The gas diffusion layer adopts a porous foam structure, so that the water-gas transportation characteristic, the temperature uniformity and the heat dissipation performance can be obviously improved while the structural strength is ensured, and the porosity, the pore diameter, the thickness and the specific microstructure of the porous structure affect the performance indexes of the gas diffusion layer. Most of the existing simulation researches at present simulate the porous foam by adopting a volume average method, and the optimization indexes are mainly macro-structure parameters such as the porosity, the pore diameter and the like of the structure, and the macro-meso-micro multi-scale optimization of the porous structure cannot be performed, so that the performance of the porous structure in the fuel cell stack still has room for improvement at present.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a design method for local and global directional regulation and control of the internal thermal mass resistance of a porous structure of a fuel cell stack, which can quickly realize local and arbitrary area and global directional regulation and control of the internal thermal mass resistance of the porous structure of the fuel cell stack under the constraint of a multi-dimensional large variable.
The technical scheme of the invention is as follows:
the method for designing the local and global directional regulation and control of the internal thermal mass resistance of the porous structure of the fuel cell stack comprises the following steps:
step 1: establishing a multi-scale model of the fuel cell stack, wherein the structure in the model comprises a polymer electrolyte membrane, a catalyst layer, a porous layer, a gas diffusion layer and a cooling flow passage, and the porous layer and the gas diffusion layer adopt porous medium framework structures;
step 2: acquiring multi-scale porous structure information in a real fuel cell stack, and performing numerical simulation based on the multi-scale model in the step 1 to obtain a numerical simulation result of corresponding thermal mass transport characteristics;
step 3: repeating the step 2 for fuel cell stacks of different models to obtain sample data; training a deep convolutional neural network model by using sample data, wherein the model input is multi-scale porous structure information, and the model outputs corresponding thermal mass transport characteristic data;
step 4: setting physical property parameters of materials of all parts in the fuel cell stack, and initializing a physical field, wherein the physical field comprises speed, temperature, pressure and substance concentration;
step 5: carrying out complete optimization analysis on the selected porous structure of the fuel cell stack by using the deep convolutional neural network model obtained by training in the step 3 and adopting a multidirectional orthogonal punishment material density method to obtain and output a structural material density distribution map and a density gradient distribution map of each iteration step in the optimization process, and adopting a trained convolutional neural network to replace a conventional solver of a flowing heat and mass transfer problem in the optimization process to rapidly predict the physical field and calculate a cost function;
step 6: based on the result of the step 5, analyzing sensitivity in topology optimization by adopting an accompanying method, inputting predicted physical field information into an accompanying solver, predicting accompanying variables in the accompanying solver, carrying out sensitivity analysis, solving the sensitivity, and calculating the derivative of an optimization target on the pseudo density of the material;
step 7: and (3) inputting the derivative obtained in the step (6) into a moving asymptote method for optimization, optimizing the structure topology under the constraint condition of the set porosity and structural rigidity by taking the maximum mass transfer and heat transfer performance and the minimum pressure drop as optimization targets, and filtering the density field obtained by optimization by adopting a Helmholtz filter, wherein the density fields before and after filtration are all values of 0-1, so as to obtain the final optimized configuration.
Further, in step 2, the process of obtaining the multi-scale porous structure information inside the real fuel cell stack is as follows:
the method comprises the steps of integrally scanning a fuel cell stack bipolar plate, a gas diffusion layer and a catalytic layer multi-scale porous structure through a micron X-ray three-dimensional CT imaging system to form a series of two-dimensional image layer structures, reconstructing the series of two-dimensional image layer structures through combination of Mimics Research and ImageJ software to form Stl format files comprising porous framework structure information, and then carrying out local trimming on the porous framework structure through Cinema 4D software to finally obtain the real fuel cell stack bipolar plate, the gas diffusion layer and the catalytic layer multi-scale porous structure information.
Further, in step 2, a numerical simulation is performed to obtain a numerical simulation result of the corresponding thermal mass transport characteristic, where the process is as follows:
and introducing the finally obtained true multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack into grid division software for grid division, establishing a multi-scale model aiming at the true physical and chemical reaction thermal mass transportation process of the multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack, applying boundary conditions and constraint conditions, carrying out detailed numerical simulation, and obtaining a numerical simulation result of corresponding thermal mass transportation characteristics.
Further, in the step 3, the deep convolutional neural network model is formed by alternately connecting four convolutional layers and four pooling layers, and finally connecting a full connecting layer; wherein the core size of the convolution layer is 5 multiplied by 5, the convolution step length is 1, the pooling adopts a maximum pooling method, the core size of the pooling layer is 2 multiplied by 2, the pooling step length is 2, the filling parameter is 2, the activation function adopts ReLU, the full connection layer is a one-dimensional variable containing 21600 elements, and the mean square error function is set as a loss function.
Further, in step 3, preprocessing the obtained multi-scale porous structure information and thermal mass transport characteristic data of the fuel cell stack to obtain sample data; the multi-scale porous structure information of the fuel cell stack comprises design domain size, boundary conditions, porous structure porosity and porous structure physical model structure data; the pretreatment process comprises the following steps: under the principle of ensuring that the contained characteristic information is unchanged, the scanned series of porous medium cross-sectional views is converted into a 100 x 100 three-dimensional tensor, and further changes the structure of the training data, the 100 x 100 three-dimensional tensor is converted into a format of 1000000 x 1 and is input into a convolutional neural network, each point in the tensor specifically characterizes the pseudo-density at each coordinate.
Further, in step 3, the deep convolutional neural network model learns and replaces the heat and mass transfer performance in the fuel cell stack with a function of the microstructure of the porous medium, and optimizes with the local and global maximum mass transfer and heat transfer performance and minimum pressure drop of the porous medium of the microporous layer and the gas diffusion layer in the fuel cell stack module as optimization targets.
Further, a self-oriented online learning optimization algorithm is adopted to accelerate non-gradient topological optimization, the obtained density distribution map and density gradient distribution map of the porous structure material in each step of iteration are used as references to guide the dynamic generation of new training data of the deep learning network, the old deep learning network is further trained, a trained deep neural network model is obtained, and the neural network provides better prediction in a region which is closer to an optimal solution until convergence
Further, after the optimized configuration is obtained, model processing is carried out, a three-dimensional model is represented by using grids, regularization processing is carried out, a 3D printing technology is further adopted on the model, when slicing is printed, the model is cut into a group of parallel plane shapes, and after slicing layering is completed, path planning is carried out on the obtained section profile information; based on the 3D printing structure, the catalytic layer structure is subjected to directional synthesis of holes by adopting a micropore sintering technology.
Advantageous effects
The design method for local and global directional regulation and control of the internal thermal mass resistance of the porous structure of the fuel cell stack, which is provided by the invention, is based on the idea of combining deep learning and topology optimization, combines a deep convolutional neural network model with a multi-scale model and a topology optimization method, avoids the problem of high calculation cost caused by repeated large-scale finite volume solution in the topology optimization process, can rapidly realize local random area and global directional regulation and control of the internal thermal mass resistance of the porous structure of the fuel cell stack under the constraint of multi-dimensional large variables, realizes small volume and light weight of the fuel cell stack, and greatly improves the comprehensive performance.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the invention will become apparent and may be better understood from the following description of embodiments taken in conjunction with the accompanying drawings in which:
FIG. 1 Fuel cell stack Structure and porous component (a) Fuel cell stack porous structural monolith component (b) porous bipolar plate (c) proton exchange Membrane (d) gas diffusion layer GDL (carbon paper/carbon cloth)
FIG. 2 design flow combining deep learning and topology optimization
FIG. 3 deep learning model
FIG. 4 topology optimization flow for porous structures
FIG. 5 is a directional design effect diagram (a) optimization of polar plate runner porous structure (b) gas diffusion layer (c) catalytic layer
Detailed Description
The following detailed description of embodiments of the invention is exemplary and intended to be illustrative of the invention and not to be construed as limiting the invention.
The design method for local and global directional regulation and control of the thermal mass resistance in the porous structure of the fuel cell stack, which is provided by the embodiment, is based on the idea of combining deep learning and topology optimization, and combines a deep convolutional neural network model with a multi-scale model and a topology optimization method, so that the problem of high calculation cost caused by repeated large-scale finite volume solution in the topology optimization process is avoided, and local and random area and global directional regulation and control of the thermal mass resistance in the porous structure of the fuel cell stack under the constraint of multi-dimensional large variables can be rapidly realized.
In the optimization process, a trained convolutional neural network model is adopted as a substitute original physical field solver, and the rapid and accurate prediction of the convolutional neural network is used for substituting for large-scale solving of a conventional flow heat transfer problem, so that the sensitivity of an objective function value, an objective function and corresponding physical constraint is rapidly evaluated; and taking the obtained density distribution map and density gradient distribution map of the porous structure material iterated in each step as references, carrying out performance evaluation by adopting a limited volume method again aiming at a structure with better performance of an objective function, accurately solving a physical field of the structure, inputting the physical field into a convolutional neural network, guiding dynamic generation of new training data of a deep learning network, further training an old deep learning network, obtaining a trained deep neural network model, and enabling the neural network to provide better prediction in a region which is closer to an optimal solution until convergence.
The embodiment specifically comprises the following steps:
step 1: a multi-scale model of the fuel cell stack is established, wherein the model comprises a polymer electrolyte membrane, a catalyst layer, a porous layer, a gas diffusion layer and a cooling runner, and the porous layer and the gas diffusion layer adopt porous medium framework structures.
Step 2: and (3) acquiring multi-scale porous structure information in the real fuel cell stack, and carrying out numerical simulation based on the multi-scale model in the step (1) to obtain a numerical simulation result of the corresponding thermal mass transport characteristic.
The method comprises the steps of integrally scanning a fuel cell stack bipolar plate, a gas diffusion layer and a catalytic layer multi-scale porous structure through a micron X-ray three-dimensional CT imaging system to form a series of two-dimensional image layer structures, reconstructing the series of two-dimensional image layer structures through combination of Mimics Research and ImageJ software to form Stl format files comprising porous framework structure information, and then carrying out local trimming on the porous framework structure through Cinema 4D software to finally obtain the real fuel cell stack bipolar plate, the gas diffusion layer and the catalytic layer multi-scale porous structure information.
And introducing the finally obtained true multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack into grid division software for grid division, establishing a multi-scale model aiming at the true physical and chemical reaction thermal mass transportation process of the multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack, applying boundary conditions and constraint conditions, carrying out detailed numerical simulation, and obtaining a numerical simulation result of corresponding thermal mass transportation characteristics.
The numerical simulation analysis aims at a typical fuel cell stack porous structure, a fuel cell stack porous slice structure is known through scanning cross sections of electron microscope with different dimensions, a fuel cell stack multi-scale model is built, the model comprises a polymer electrolyte membrane, a catalyst layer, a porous layer, a gas diffusion layer and a cooling flow channel, wherein the porous layer and the gas diffusion layer adopt porous medium framework structures; and acquiring multi-scale porous structure information in the real fuel cell stack, and carrying out numerical simulation based on the multi-scale model to obtain a numerical simulation result of the corresponding thermal mass transport characteristic.
Step 3: repeating the step 2 for fuel cell stacks of different models to obtain sample data; training a deep convolutional neural network model by using sample data, wherein the model input is multi-scale porous structure information, and the model outputs corresponding thermal mass transport characteristic data.
And (3) repeating the step (2) for fuel cell stacks of different types to obtain a large number of numerical simulation results of corresponding thermal mass transport characteristics under different structures, and preprocessing the data to obtain training data for training the deep convolutional neural network.
The fuel cell stack multi-scale porous structure information includes design domain size, boundary conditions, porous structure porosity, and porous structure physical model structure data. The pretreatment process of the obtained multi-scale porous structure information and thermal mass transport characteristic data of the fuel cell stack is as follows: under the principle of ensuring that the contained characteristic information is unchanged, the scanned series of porous medium cross-sectional views is converted into a 100 x 100 three-dimensional tensor, and further changes the structure of the training data, the 100 x 100 three-dimensional tensor is converted into a format of 1000000 x 1 and is input into a convolutional neural network, each point in the tensor specifically characterizes the pseudo-density at each coordinate.
The constructed deep convolutional neural network model is formed by alternately connecting four convolutional layers and four pooling layers, and finally connecting a full connecting layer; wherein the core size of the convolution layer is 5 multiplied by 5, the convolution step length is 1, the pooling adopts a maximum pooling method, the core size of the pooling layer is 2 multiplied by 2, the pooling step length is 2, the filling parameter is 2, the activation function adopts ReLU, the full connection layer is a one-dimensional variable containing 21600 elements, and the mean square error function is set as a loss function. In the optimization process, the optimization targets of the maximum mass transfer and heat transfer performance and the minimum pressure drop of the local and full fields in the module are learned and replaced by the functions of the design variables by using a deep learning model, so that the problem of high calculation cost caused by repeated large-scale finite volume method solution in the optimization process is avoided.
Step 4: physical property parameters of materials of each component in the fuel cell stack are set, and physical field initialization such as speed, temperature, pressure, substance concentration and the like is performed.
Step 5: and 3, learning the local and full-field heat and mass transfer characteristics in the porous structure of the fuel cell stack by using the deep convolutional neural network model obtained by training in the step 3, replacing the characteristic with a function of a design variable, namely a porous medium microstructure, adopting an orthotropic punishment material density method, taking the local and global maximum mass transfer and heat transfer performance and the minimum pressure drop of the porous medium of the microporous layer and the gas diffusion layer in the fuel cell stack module as optimization targets, carrying out complete optimization analysis on the selected porous structure of the fuel cell stack, obtaining and outputting a structural material density distribution map and a density gradient distribution map of each iteration step in the optimization process, adopting a trained convolutional neural network to replace a conventional solver of a flow heat and mass transfer problem in the optimization process, replacing a large-scale solution of the conventional flow heat transfer problem by quick accurate prediction of the convolutional neural network, and carrying out quick prediction on a new structural physical field and calculating a cost function.
And accelerating non-gradient topological optimization by adopting a self-oriented online learning optimization algorithm, taking the obtained density distribution map and density gradient distribution map of the porous structure material iterated in each step as a reference, guiding the dynamic generation of new training data of the deep learning network, further training the old deep learning network, and obtaining a trained deep neural network model, so that the neural network provides better prediction in a region which is closer to an optimal solution until convergence.
The self-oriented online learning optimization algorithm finds possible optimal design according to the prediction of the deep learning model, dynamically generates new training points on the basis of optimization, performs performance evaluation through a limited volume method, and feeds back and inputs the new training points to the convolutional neural network as additional training data for training. This self-optimizing online learning cycle is repeated until convergence. The iterative learning scheme utilizes the searching capability of a heuristic method and the efficient prediction of a deep learning model, and compared with a gradient-based method, the algorithm does not depend on gradient information of a topological optimization problem objective function.
Step 6: based on the result of the step 5, sensitivity is analyzed in topology optimization by adopting a concomitant method, predicted physical field information such as speed, temperature, pressure, material concentration and the like is input into a concomitant solver, a concomitant variable is predicted in the concomitant solver, sensitivity analysis is performed, sensitivity is solved, and the derivative of an optimization target on the material pseudo density is calculated.
Step 7: and (3) inputting the derivative obtained in the step (6) into a moving asymptote method for optimization, optimizing the structure topology under the constraint condition of certain porosity and structural rigidity by taking the maximum mass transfer and heat transfer performance and the minimum pressure drop as optimization targets, and filtering the density field obtained by optimization by adopting a Helmholtz filter, wherein the density fields before and after filtration are all values of 0-1, so as to obtain the final optimized configuration.
And then, realizing the trans-scale forming of the porous structure of the fuel cell stack by a method combining a 3D printing technology and a micropore sintering technology.
The 3D printing manufacturing physical forming process comprises the following steps:
1) Model generation, namely importing the optimized structural model into three-dimensional Computer Aided Design (CAD) or other 3D software: the three-dimensional model printed by 3D is represented by using grids, such as STL or OBJ format files, regularization treatment is carried out before printing, so that the problems of degraded triangles, self-intersections, gaps and the like of unstructured triangular surfaces are prevented, the self-intersection and non-manifold models enable slicing algorithms in a layered printing process to be unstable, and a geometric regularization method represented by layered depth normal images is adopted to carry out robust and efficient restoration on the three-dimensional model;
2) And (5) slice calculation: based on the prepared input model, the model is preprocessed into appropriate data for directing the operation of the 3D printing machine. The pattern is here cut into a set of parallel planar shapes;
3) Path planning: after slicing and layering are completed on the input model, path planning is carried out on the obtained section profile information, and a Zig-Zag scanning filling algorithm is adopted, so that a unidirectional scanning mode is improved;
4) 3D printing: the designed porous structure inherits the advantages of smoothness, connectivity and the like of the three-period minimum curved surface, almost no support is needed when a small-size model is printed, and the support material needed in the process of printing a large-size model can be smoothly led out through communicated holes;
5) Based on the 3D printing structure, the catalytic layer structure adopts a micropore sintering technology to realize directional synthesis of holes.
Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives, and variations may be made in the above embodiments by those skilled in the art without departing from the spirit and principles of the invention.
Claims (8)
1. A design method for local and global directional regulation and control of thermal mass resistance in a porous structure of a fuel cell stack is characterized by comprising the following steps: the method comprises the following steps:
step 1: establishing a multi-scale model of the fuel cell stack, wherein the structure in the model comprises a polymer electrolyte membrane, a catalyst layer, a porous layer, a gas diffusion layer and a cooling flow passage, and the porous layer and the gas diffusion layer adopt porous medium framework structures;
step 2: acquiring multi-scale porous structure information in a real fuel cell stack, and performing numerical simulation based on the multi-scale model in the step 1 to obtain a numerical simulation result of corresponding thermal mass transport characteristics;
step 3: repeating the step 2 for fuel cell stacks of different models to obtain sample data; training a deep convolutional neural network model by using sample data, wherein the model input is multi-scale porous structure information, and the model outputs corresponding thermal mass transport characteristic data;
step 4: setting physical property parameters of materials of all parts in the fuel cell stack, and initializing a physical field, wherein the physical field comprises speed, temperature, pressure and substance concentration;
step 5: carrying out complete optimization analysis on the selected porous structure of the fuel cell stack by using the deep convolutional neural network model obtained by training in the step 3 and adopting a multidirectional orthogonal punishment material density method to obtain and output a structural material density distribution map and a density gradient distribution map of each iteration step in the optimization process, and adopting a trained convolutional neural network to replace a conventional solver of a flowing heat and mass transfer problem in the optimization process to rapidly predict the physical field and calculate a cost function;
step 6: based on the result of the step 5, analyzing sensitivity in topology optimization by adopting an accompanying method, inputting predicted physical field information into an accompanying solver, predicting accompanying variables in the accompanying solver, carrying out sensitivity analysis, solving the sensitivity, and calculating the derivative of an optimization target on the pseudo density of the material;
step 7: and (3) inputting the derivative obtained in the step (6) into a moving asymptote method for optimization, optimizing the structure topology under the constraint condition of the set porosity and structural rigidity by taking the maximum mass transfer and heat transfer performance and the minimum pressure drop as optimization targets, and filtering the density field obtained by optimization by adopting a Helmholtz filter, wherein the density fields before and after filtration are all values of 0-1, so as to obtain the final optimized configuration.
2. The method for designing local and global directional regulation of thermal mass resistance in a porous structure of a fuel cell stack according to claim 1, wherein the method comprises the following steps: in the step 2, the process of obtaining the multi-scale porous structure information inside the real fuel cell stack is as follows:
the method comprises the steps of integrally scanning a fuel cell stack bipolar plate, a gas diffusion layer and a catalytic layer multi-scale porous structure through a micron X-ray three-dimensional CT imaging system to form a series of two-dimensional image layer structures, reconstructing the series of two-dimensional image layer structures through combination of Mimics Research and ImageJ software to form Stl format files comprising porous framework structure information, and then carrying out local trimming on the porous framework structure through Cinema 4D software to finally obtain the real fuel cell stack bipolar plate, the gas diffusion layer and the catalytic layer multi-scale porous structure information.
3. The method for designing local and global directional regulation of thermal mass resistance in a porous structure of a fuel cell stack according to claim 2, wherein the method comprises the following steps: in the step 2, the numerical simulation is carried out, and the process of obtaining the numerical simulation result of the corresponding thermal mass transport characteristic is as follows:
and introducing the finally obtained true multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack into grid division software for grid division, establishing a multi-scale model aiming at the true physical and chemical reaction thermal mass transportation process of the multi-scale porous structures of the bipolar plate, the gas diffusion layer and the catalytic layer of the fuel cell stack, applying boundary conditions and constraint conditions, carrying out detailed numerical simulation, and obtaining a numerical simulation result of corresponding thermal mass transportation characteristics.
4. The method for designing local and global directional regulation of thermal mass resistance in a porous structure of a fuel cell stack according to claim 1, wherein the method comprises the following steps: in the step 3, the deep convolutional neural network model is formed by alternately connecting four convolutional layers and four pooling layers, and finally connecting a full connecting layer; wherein the core size of the convolution layer is 5 multiplied by 5, the convolution step length is 1, the pooling adopts a maximum pooling method, the core size of the pooling layer is 2 multiplied by 2, the pooling step length is 2, the filling parameter is 2, the activation function adopts ReLU, the full connection layer is a one-dimensional variable containing 21600 elements, and the mean square error function is set as a loss function.
5. The method for designing local and global directional control of thermal mass resistance in a porous structure of a fuel cell stack according to claim 1 or 4, wherein the method comprises the following steps: in the step 3, preprocessing the obtained multi-scale porous structure information and thermal mass transport characteristic data of the fuel cell stack to obtain sample data; the multi-scale porous structure information of the fuel cell stack comprises design domain size, boundary conditions, porous structure porosity and porous structure physical model structure data; the pretreatment process comprises the following steps: under the principle of ensuring that the contained characteristic information is unchanged, the scanned series of porous medium cross-sectional views is converted into a 100 x 100 three-dimensional tensor, and further changes the structure of the training data, the 100 x 100 three-dimensional tensor is converted into a format of 1000000 x 1 and is input into a convolutional neural network, each point in the tensor specifically characterizes the pseudo-density at each coordinate.
6. The method for designing local and global directional regulation of thermal mass resistance in a porous structure of a fuel cell stack according to claim 1, wherein the method comprises the following steps: in step 3, the deep convolutional neural network model learns and replaces the heat and mass transfer performance in the fuel cell stack with a function of the microstructure of the porous medium, and optimizes by taking the local and global maximum mass and heat transfer performance and minimum pressure drop of the porous medium of the microporous layer and the gas diffusion layer in the fuel cell stack module and the electric stack as optimization targets.
7. The method for designing local and global directional control of thermal mass resistance in a porous structure of a fuel cell stack according to claim 6, wherein the method comprises the steps of: the self-oriented online learning optimization algorithm is adopted to accelerate non-gradient topological optimization, the obtained density distribution map and density gradient distribution map of the porous structure material iterated in each step are used as references to guide dynamic generation of new training data of the deep learning network, the old deep learning network is further trained, a trained deep neural network model is obtained, and the neural network provides better prediction in a region which is closer to an optimal solution until convergence.
8. The method for designing local and global directional regulation of thermal mass resistance in a porous structure of a fuel cell stack according to claim 1, wherein the method comprises the following steps: after the optimized configuration is obtained, model processing is carried out, a three-dimensional model is represented by using grids, regularization processing is carried out, a 3D printing technology is further adopted on the model, the model is cut into a group of parallel plane shapes when slicing is printed, and path planning is carried out on the obtained section profile information after slicing layering is completed; based on the 3D printing structure, the catalytic layer structure is subjected to directional synthesis of holes by adopting a micropore sintering technology.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210595253.2A CN115017803B (en) | 2022-05-28 | 2022-05-28 | Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210595253.2A CN115017803B (en) | 2022-05-28 | 2022-05-28 | Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115017803A CN115017803A (en) | 2022-09-06 |
| CN115017803B true CN115017803B (en) | 2023-10-13 |
Family
ID=83071129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210595253.2A Active CN115017803B (en) | 2022-05-28 | 2022-05-28 | Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115017803B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280305A (en) * | 2018-01-30 | 2018-07-13 | 西安交通大学 | Radiating element cooling duct rapid topology optimum design method based on deep learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2803066A1 (en) * | 2010-07-29 | 2012-02-02 | Exxonmobil Upstream Research Company | Methods and systems for machine-learning based simulation of flow |
-
2022
- 2022-05-28 CN CN202210595253.2A patent/CN115017803B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280305A (en) * | 2018-01-30 | 2018-07-13 | 西安交通大学 | Radiating element cooling duct rapid topology optimum design method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| Tihao Yang等.Numerical Study of flow and heat transfer in a three-dimensional metal foam considering different direction micropores in skeleton structure.International Coummnication in Heat and Mass Transfer.2022,106052. * |
| 王泽林 等.基于深度学习驱动的Ｌ型定向热疏导机理.航空学报.2021,第42卷(第6期),124242. * |
| 王浩然 ; 吴私 ; 杨林林 ; 兰海兵 ; 陈超 ; 孙伟 ; 宋学官 ; .高温质子交换膜燃料电池电堆端板拓扑优化.大连理工大学学报.2020,(02),全文. * |
| 阎杰 ; 杨永竹 ; 谢军 ; 陈月尧 ; 马宏 ; .离散体结构拓扑优化综述.科学技术与工程.2020,(24),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115017803A (en) | 2022-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A review of proton exchange membrane fuel cell water management: Membrane electrode assembly | |
| Maier et al. | Mass transport in PEM water electrolysers: A review | |
| Miao et al. | Parameter estimation of PEM fuel cells employing the hybrid grey wolf optimization method | |
| Salomov et al. | Pore-scale modeling of fluid flow through gas diffusion and catalyst layers for high temperature proton exchange membrane (HT-PEM) fuel cells | |
| Kulkarni et al. | The effect of non-uniform compression and flow-field arrangements on membrane electrode assemblies-X-ray computed tomography characterisation and effective parameter determination | |
| Rama et al. | An X-ray tomography based lattice Boltzmann simulation study on gas diffusion layers of polymer electrolyte fuel cells | |
| Simaafrookhteh et al. | Microstructure reconstruction and characterization of the porous GDLs for PEMFC based on fibers orientation distribution | |
| Ma et al. | Impact of gas diffusion layer spatial variation properties on water management and performance of PEM fuel cells | |
| Tranter et al. | Pore network modelling of capillary transport and relative diffusivity in gas diffusion layers with patterned wettability | |
| Fazeli et al. | Incorporating embedded microporous layers into topologically equivalent pore network models for oxygen diffusivity calculations in polymer electrolyte membrane fuel cell gas diffusion layers | |
| Cawte et al. | A 3D convolutional neural network accurately predicts the permeability of gas diffusion layer materials directly from image data | |
| Almodfer et al. | Improving parameter estimation of fuel cell using honey badger optimization algorithm | |
| Pourrahmani et al. | The impacts of the gas diffusion layer contact angle on the water management of the proton exchange membrane fuel cells: Three‐dimensional simulation and optimization | |
| Khatoonabadi et al. | Insights on the interaction of serpentine channels and gas diffusion layer in an operating polymer electrolyte fuel cell: Numerical modeling across scales | |
| Csoklich et al. | Laser structured gas diffusion layers for improved water transport and fuel cell performance | |
| Xiao et al. | Microstructure reconstruction using fiber tracking technique and pore-scale simulations of heterogeneous gas diffusion layer | |
| Zhang et al. | Advancing next-generation proton-exchange membrane fuel cell development in multi-physics transfer | |
| CN115017803B (en) | Method for designing local and global directional regulation and control of thermal mass resistance in porous structure of fuel cell stack | |
| Sim et al. | Effects of operating conditions, various properties of the gas diffusion layer, and shape of endplate on the open-cathode proton exchange membrane fuel cell performance | |
| Shin et al. | Numerical analysis on transport properties of self-humidifying dual catalyst layer via 3-D reconstruction technique | |
| Wang et al. | Lattice Boltzmann simulation of cathode catalyst layer degradation on transport reaction process within a proton exchange membrane fuel cell | |
| Satjaritanun et al. | Numerical Study of Mass Transport and Electrochemical Kinetics inside Porous Structure Layers of PEMFC Using Direct Simulation Approach | |
| Sepe et al. | Multiscale three-dimensional modeling of two-phase transport inside porous transport layers | |
| Baz et al. | A review of water management methods in proton exchange membrane fuel cells | |
| Liao et al. | Pore-scale study of effects of cathode catalytic layer component gradient distribution and perforation on reaction transport processes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |