Fuel cell flow channel multi-objective collaborative optimization design method considering GDL compression deformation

Abstract

The invention discloses a fuel cell flow channel multi-objective collaborative optimization design method considering GDL compression deformation. The method comprises the following steps: step 1, establishing and verifying a PEMFC numerical calculation model considering an assembly force effect; 2, determining the optimal assembly force of the PEMFC; step 3, flow field structure parameter influence mechanism analysis and sensitivity analysis; and step 4, flow field structure optimization based on a neural network and a multi-objective algorithm to obtain a final optimization scheme. According to the method, assembly force optimization and flow field structure optimization are systematically combined for the first time, the optimal assembly force is determined to serve as the basis of flow field optimization, multi-parameter collaborative optimization is carried out on the premise, and the defect that the assembly force optimization and the flow field structure optimization are separated in a traditional method is overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of proton exchange membrane fuel cells, and particularly relates to a fuel cell flow channel multi-objective collaborative optimization design method considering GDL compression deformation. BACKGROUND

[0002] Proton exchange membrane fuel cells (PEMFC) have broad application prospects in the fields of transportation and fixed power stations as a kind of efficient and clean energy conversion device. In PEMFC, the assembly force causes compression deformation of the GDL (Gas Diffusion Layer), which directly affects the physical parameters such as porosity and permeability of the GDL, and changes the contact state between the bipolar plate and the GDL, affecting the contact resistance and contact thermal resistance, and then significantly affecting the performance of the cell. At the same time, the design of the flow field structure is crucial to the performance of the PEMFC. The parameters such as flow channel width, height, rib width and turning angle directly affect the transport of reaction gas, the discharge of liquid water, the uniformity of current density distribution and the flow channel pressure drop.

[0003] Most of the current researches separate the optimization of assembly force and the optimization of flow field structure. The research on assembly force often uses a linear GDL constitutive model or ignores the influence of contact thermal resistance; while the optimization of flow field structure is usually carried out under a certain fixed assembly force which is not necessarily optimal. This fragmented optimization method cannot reveal the coupling influence mechanism between assembly force and flow field structure, and it is difficult to obtain an overall design scheme that optimizes the comprehensive performance of PEMFC.

[0004] The document Shi L, Xu SC, Liu JL. Influences of assembly pressure and flow channel size on performances of proton exchange membrane fuel cells based on a multi-model. Int J of Hydrogen Energy 2022; 47: 7902-7914. optimizes the flow channel width and depth of PEMFC by combining the influence of assembly force in the framework of artificial neural network, and verifies the performance improvement effect of geometric design, but only optimizes the flow channel width and depth, does not involve the rib width and turning angle parameters which significantly affect mass transfer and contact resistance, and does not simultaneously consider the multi-objective optimization requirements of flow channel pressure drop and current density uniformity, so the optimization scheme is difficult to improve the comprehensive performance of PEMFC.

[0005] The document “Yu ZT, Xia L, Xu GP, Wang CJ, Wang DH. Improvement of the three-dimensional fine-mesh flow field of proton exchange membrane fuel cell (PEMFC) using CFD modeling, artificial neural network and genetic algorithm. Int J of Hydrogen Energy 2022; 47: 35038-35054.” performs multi-objective optimization on the three-dimensional fine-mesh flow field by combining Latin hypercube sampling, artificial neural network and non-dominated sorting genetic algorithm II, which makes the power density increase by 6.98%, but does not perform sensitivity analysis on the flow channel structure parameters, cannot quantify the influence weight of each parameter on the performance of the PEMFC, and leads to the problems of large amount of calculation and low efficiency in the optimization process.

[0006] Therefore, there is a need for a design method that collaboratively optimizes assembly force and flow field structure, considering the coupling effects of GDL nonlinear compression deformation, contact resistance and multiple flow field structure parameters, to systematically improve the overall performance of the PEMFC. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application provides a fuel cell flow channel multi-objective collaborative optimization design method considering GDL compression deformation.

[0008] The technical solution of the present application to solve the technical problem is to provide a fuel cell flow channel multi-objective collaborative optimization design method considering GDL compression deformation, characterized in that the method comprises the following steps: Step 1, establishing and verifying the PEMFC numerical calculation model considering the effect of assembly force: obtaining the nonlinear compression deformation relationship of GDL through compression test, obtaining the contact resistance and contact thermal resistance under different assembly forces through contact resistance test; then establishing a PEMFC numerical calculation model coupling the nonlinear compression deformation of GDL, contact resistance and contact thermal resistance, and verifying the accuracy of the model through polarization curve test; Step 2: Determine the optimal assembly force for the PEMFC: Using the PEMFC numerical calculation model obtained in Step 1, simulate the performance of the PEMFC under different assembly forces to obtain the power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index at a specified current density. Then, use the entropy weight method to comprehensively evaluate the power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index under different assembly forces, calculate the comprehensive score for each assembly force, and select the assembly force with the highest comprehensive score as the optimal assembly force for the PEMFC. Step 3: Analysis of the Influence Mechanism and Sensitivity of Flow Field Structural Parameters: Based on the optimal assembly force determined in Step 2, the channel width, channel height, rib width, and turning angle are selected as key flow field structural parameters. Using the PEMFC numerical calculation model obtained in Step 1, the influence of individual changes in the four key flow field structural parameters on power density, pressure drop in the cathode channel, and uniformity index of current density distribution in the proton exchange membrane is analyzed. For combinations of different key flow field structural parameters, the entropy weight method is used to calculate their comprehensive score, and the relative sensitivity factor of each key flow field structural parameter is calculated through sensitivity analysis. Step 4: Flow field structure optimization based on neural network and multi-objective algorithm: Using key flow field structure parameters as input variables and their corresponding power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index as output variables, multiple sets of sample points are generated in the design space using the Latin hypercube sampling method. Then, the output variable values ​​of all sample points are calculated using the PEMFC numerical calculation model obtained in Step 1 to form a sample dataset. The backpropagation neural network is then trained using the sample dataset to establish a performance prediction model. The performance prediction model is used as the objective function and optimized using the multi-objective gray wolf optimization algorithm to obtain the Pareto optimal solution set. Then, the entropy weight method is used to comprehensively score the PEMFC power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index in the Pareto optimal solution set, and the combination of key flow field structure parameters with the highest comprehensive score is selected as the final optimization scheme.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention is the first to systematically combine assembly force optimization and flow field structure optimization. By determining the optimal assembly force as the basis for flow field optimization, and under this premise, multi-parameter collaborative optimization is carried out, overcoming the drawback of the separation of the two in the traditional method.

[0010] (2) The nonlinear compression deformation of GDL, ECR (electrical contact resistance), and TCR (thermal contact resistance) are coupled to the PEMFC flow channel optimization model. In the prior art, it is often assumed that the Young's modulus of GDL is a fixed value. This invention obtains nonlinear compression data of GDL under different assembly forces through experiments, including thickness, porosity, and conductivity. At the same time, it obtains the ECR and TCR between the bipolar plate and GDL. These data are accurately coupled to the numerical calculation model of PEMFC flow channel optimization, so that the model is more in line with the actual operating conditions of PEMFC and the accuracy of flow channel optimization is improved.

[0011] (3) The entropy weight method is used for comprehensive evaluation of multiple indicators, which ensures the balance of optimization results in power, voltage drop and uniformity. The multi-objective optimization combination method, which combines Latin hypercube sampling, backpropagation neural network and multi-objective gray wolf optimization algorithm, significantly improves the optimization efficiency in complex multi-dimensional design space.

[0012] (4) By optimizing four parameters, namely flow channel width, flow channel height, rib width and turning angle, the power density is increased, the voltage drop is reduced and the current uniformity is improved simultaneously, thus making up for the problems of incomplete parameter optimization and single-objective limitations.

[0013] (5) Sensitivity analysis clarified the influence of each structural parameter, providing clear guidance for prioritization in engineering design, reducing redundant iterative calculations for non-critical parameters, lowering design complexity and computational costs, and significantly improving the efficiency and accuracy of flow channel design. Finally, experimental verification ensured the practicality and reliability of the optimization results.

[0014] (6) The optimized flow channel structure can significantly improve the heat and mass transfer efficiency inside PEMFC, significantly increase the oxygen mass fraction at the GDL-catalyst interface in the under-fin region, make the reaction gas distribution more uniform, reduce the overall concentration polarization loss of the battery, make the current density distribution more uniform, and make the bipolar plate temperature distribution more uniform, avoid proton exchange membrane aging caused by local overheating, and extend battery life.

[0015] (7) The optimized flow channel structure enhances the drainage capacity of the flow channel by reasonably adjusting the ratio of flow channel height to rib width, reduces water accumulation in the flow channel, and reduces the problem of increased mass transfer resistance caused by water accumulation. Combined with the porosity change after GDL compression, the water in the flow channel is more easily discharged, further improving the stability of the battery under high current density conditions and avoiding performance degradation caused by water blockage of the flow channel.

[0016] (8) The optimized flow channel structure reduces the pressure drop in the cathode flow channel and reduces the parasitic power consumption of PEMFC; at the same time, it increases the power density and ultimately achieves an overall increase in output power. Attached Figure Description

[0017] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a stress-strain curve diagram of the GDL compression test in Embodiment 1 of the present invention; Figure 3 This is a graph showing the comprehensive PEMFC score under different assembly forces in Embodiment 1 of the present invention. Figure 4 This is a graph showing the relative sensitivity factors of different flow field structural parameters in Embodiment 1 of the present invention. Figure 5 This is a schematic diagram of the backpropagation neural network structure in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the Pareto front for multi-objective optimization in Embodiment 1 of the present invention; Figure 7 This is a performance comparison chart of PEMFC before and after optimization in Embodiment 1 of the present invention. Detailed Implementation

[0018] Specific embodiments of the present invention are given below. These specific embodiments are only used to further illustrate the present invention in detail and do not limit the scope of protection of the present invention.

[0019] This invention provides a multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation (hereinafter referred to as the method), characterized by the following steps: Step 1: Establish and verify the PEMFC numerical calculation model considering the effect of assembly force: Obtain the nonlinear compression deformation relationship of GDL through compression test, and obtain the contact resistance and contact thermal resistance under different assembly forces through contact resistance test; then establish a PEMFC numerical calculation model that couples the nonlinear compression deformation, contact resistance and contact thermal resistance of GDL, and verify the accuracy of the model through polarization curve test. Preferably, step 1 specifically includes the following steps: S11. Conduct a compression test on the GDL using a general compression testing device to obtain the nonlinear compression deformation curve of the GDL; S12. Based on the nonlinear compression deformation curve obtained in S11, the thickness distribution of GDL under different assembly forces under the rib and below the flow channel is obtained by finite element method simulation; then, according to the thickness distribution of GDL after compression, its porosity, electrical conductivity, thermal conductivity and permeability after compression are calculated by empirical formula. Meanwhile, under different assembly forces, the contact resistance between the bipolar plate and the GDL under different contact stresses was measured using a contact resistance tester; the contact thermal resistance under different contact stresses was obtained by fitting existing experimental data. Preferably, in step S12, the empirical formulas include the GDL porosity calculation formula, the GDL electrical conductivity calculation formula, the GDL in-plane thermal conductivity calculation formula, the GDL through-plane thermal conductivity calculation formula, and the GDL permeability calculation formula.

[0020] Preferably, in step S12, the formula for calculating the porosity of GDL is: (1) In equation (1), Porosity after GDL compression; The initial porosity of GDL; This represents the initial thickness of the GDL. This represents the thickness after GDL compression. The formula for calculating the conductivity of GDL is: (2) In equation (2), k eff K represents the conductivity after GDL compression. sol The electrical conductivity of GDL carbon fiber; The formula for calculating the in-plane thermal conductivity of GDL is: (3) In equation (3), Q eff Q is the thermal conductivity of GDL after compression. sol The thermal conductivity of GDL carbon fiber; The formula for calculating the thermal conductivity of a plane using GDL is: (4) The formula for calculating GDL penetration rate is: (5) In equation (5), K is the permeability of GDL after compression; d is the diameter of GDL carbon fiber.

[0021] S13. Based on the data obtained in step S12, a geometric solid model of PEMFC is established using 3D modeling software. The geometric solid model of PEMFC is then meshed into discrete small units for numerical calculation. Then, the solver and boundary conditions are set in the CFD software to obtain the PEMFC numerical calculation model. The PEMFC numerical calculation model couples the non-uniform physical parameters of GDL through a user-defined file and sets the contact resistance and contact thermal resistance at the interface between the bipolar plate and GDL. S14. Perform polarization curve testing and compare the test results with the simulation results of the model in step S13. If the maximum error does not exceed 5%, the model is considered to have passed verification. If the maximum error exceeds 5%, return to S13 and start from mesh generation again.

[0022] Preferably, in step 1, the range of the assembly force is 0.5MPa to 2MPa.

[0023] Step 2: Determine the optimal assembly force for the PEMFC: Using the PEMFC numerical calculation model obtained in Step 1, simulate the performance of the PEMFC under different assembly forces to obtain the power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index at a specified current density. Then, use the entropy weight method to comprehensively evaluate the PEMFC power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index under different assembly forces, calculate the comprehensive score under each assembly force, and select the assembly force with the highest comprehensive score as the optimal assembly force for the PEMFC. Preferably, in step 2, the formula for calculating the uniformity index of current density distribution within the proton exchange membrane is: (6) In equation (6), S uni R is the uniformity index of current density distribution within the proton exchange membrane; R is the root mean square deviation of the distribution; AVE is the mean value of the distribution.

[0024] Step 3: Analysis of the Influence Mechanism and Sensitivity of Flow Field Structural Parameters: Based on the optimal assembly force determined in Step 2, the channel width, channel height, rib width, and turning angle are selected as key flow field structural parameters. Using the PEMFC numerical calculation model obtained in Step 1, the influence of individual changes in the four key flow field structural parameters on power density, pressure drop in the cathode channel, and uniformity index of current density distribution in the proton exchange membrane is analyzed. For combinations of different key flow field structural parameters, the entropy weight method is used to calculate their comprehensive score, and the relative sensitivity factor of each key flow field structural parameter is calculated through sensitivity analysis. Preferably, in step 3, the reference values ​​for the channel width, channel height, rib width, and turning angle are 1.0 mm, 1.0 mm, 1.0 mm, and 0%, respectively; the range of the channel width, channel height, and rib width is 0.6~1.4 mm; the turning angle is used to quantify the geometric sharpness of the channel bend, and the value range is 0%~100%, which physically means the ratio of the depth of the cutting area at the channel corner to the rib width (in this embodiment, the values ​​are 0%, 25%, 50%, 75%, and 100%; where 0% represents a right-angle corner (no cutting), corresponding to the traditional serpentine channel design; 100% represents a corner cutting depth equal to the rib width, forming a semi-circular bend).

[0025] Preferably, in step 3, the specific calculation formula for sensitivity analysis is as follows: (7) In equation (7), S is the relative sensitivity factor; M i+1 and M iThese are the combined scores for the (i+1)th and ith structural parameters, respectively; M a P is the average of the combined scores under the two structural parameters. i+1 and P i These are the values ​​of the (i+1)th and ith structural parameters, respectively; P a This is the average of the two structural parameters.

[0026] Step 4: Flow field structure optimization based on neural network and multi-objective algorithm: Using key flow field structure parameters as input variables and their corresponding power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index as output variables, multiple sets of sample points are generated in the design space using the Latin hypercube sampling method. Then, the output variable values ​​of all sample points are calculated using the PEMFC numerical calculation model obtained in Step 1 to form a sample dataset. The backpropagation neural network is then trained using the sample dataset to establish a performance prediction model. The performance prediction model is used as the objective function and optimized using the multi-objective gray wolf optimization algorithm to obtain the Pareto optimal solution set. Then, the entropy weight method is used to comprehensively score the PEMFC power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index in the Pareto optimal solution set, and the combination of key flow field structure parameters with the highest comprehensive score is selected as the final optimization scheme.

[0027] Preferably, in step 4, the Latin hypercube sampling method is to uniformly divide the value range of each input variable into several intervals equal to the number of samples, and randomly select a sample from each interval.

[0028] Preferably, in step 4, the number of hidden layer neurons in the backpropagation neural network is 5, the learning rate is set to 0.04, and the number of training iterations is 5000. When the regression coefficient R of the training set, validation set, and test set are all greater than 0.95, the model training is considered complete.

[0029] Preferably, in step 4, the optimization objectives of the multi-objective gray wolf optimization algorithm are to maximize the power density, minimize the pressure drop in the cathode channel, and minimize the uniformity index of the current density distribution in the proton exchange membrane.

[0030] Preferably, the method further includes step 5: applying the final optimized scheme obtained in step 4 to the design and manufacturing of the fuel cell stack to achieve a systematic improvement in fuel cell performance.

[0031] Example 1: Step 1: Establish and verify the PEMFC numerical calculation model considering assembly forces: S11, GDL Compression Test: Using a general-purpose compression testing device, select a TGP-H-060 type GDL sample, cut to 1cm×1cm size, and stack five samples for compression testing; the test pressure is uniformly increased from 0N to 1800N, displacement data is recorded, pressure is converted into stress, and displacement is converted into strain, obtaining the nonlinear stress-strain curve of the GDL, such as... Figure 2 As shown; S12, GDL Compression Deformation Simulation: Based on the stress-strain curve obtained from the experiment, a finite element model of PEMFC was established. The thickness distribution of GDL under different assembly forces (0.5, 1.0, 1.5, 2.0 MPa) was simulated by changing the Young's modulus of GDL. Based on the thickness of GDL after compression, its porosity, electrical conductivity, thermal conductivity and permeability after compression were calculated by empirical formula. The ECR between the bipolar plate and GDL under different contact stresses was measured by a contact resistance tester. According to standard GB / T 20042.6-2024, an automatic bipolar plate ECR tester (FT-541SJB-341, ROOKO) was used to measure the ECR between BP and GDL. The equipment is equipped with a copper electrode with a diameter of 6.0 cm. To ensure that the copper electrode is in complete contact with the sample, the diameters of the GDL and graphite plate samples used for testing are 6.5 cm and 7.0 cm, respectively. The ECR test principle is shown in formula (8): (8) In equation (8), r is the ECR between BP and GDL; r1 is the sum of the bulk resistance of the graphite plate, the bulk resistance of GDL, the ECR between the two graphite plates and GDL, the bulk resistance of the two copper electrodes, and the ECR between the two GDLs and copper electrodes; r2 is the sum of the bulk resistance of the two copper electrodes, the bulk resistance of GDL, and the ECR between the two GDLs and copper electrodes; BP r is the volume resistance of the graphite plate. GDL The bulk resistance of the GDL; A con This represents the contact area between the GDL and the graphite plate. During the experiment, a GDL sample was first placed and ensured to completely cover the copper electrode. The pressure was gradually increased so that the contact stress of the GDL sample was 1.0 MPa. At this time, 30 seconds were waited to ensure the test state was stable, and the total resistance value was recorded as r2. Then, GDL, graphite plate, and GDL were placed between the two copper electrodes in the order of GDL, and the pressure was gradually increased so that the contact stress on GDL gradually increased from 0 MPa to 4.5 MPa. During this process, 30 seconds were waited every 0.3 MPa, and the total resistance value was recorded as r1. The ECR between BP and GDL under different contact stresses was calculated using formula (8). The existing TCR test data were fitted to obtain the TCR variation curve with contact stress. S13. PEMFC Numerical Calculation Model Establishment: A three-dimensional numerical calculation model of a single serpentine PEMFC was established based on ANSYS Fluent, with an activation area of ​​2cm×2cm; the physical parameters of the deformed GDL were coupled into the PEMFC numerical calculation model through user-defined files to specify different physical parameters of the GDL in the areas below the rib and the flow channel; ECR and TCR were added to the interface between the bipolar plate and the GDL. S14. Model Validation: Mesh independence validation was performed, with a mesh size of 2,048,000. Polarization curves were obtained through CFD simulation. PEMFC samples were prepared, and polarization curve tests were conducted using a G60 testing system. The simulation results were compared with the experimental results, and the maximum error did not exceed 3%, validating the accuracy of the model.

[0032] Step 2: Determine the optimal assembly force for PEMFC: S21. Performance Index Acquisition: Using the PEMFC numerical calculation model obtained in step 1, simulate the current density of 0.95 A / cm² under assembly forces of 0.5, 1.0, 1.5, and 2.0 MPa. 2 Power density, cathode channel pressure drop, and uniformity index of current density distribution in the PEM at that time; S22. Entropy Weight Method Comprehensive Evaluation: Using power density as a positive indicator and the pressure drop in the cathode channel and the uniformity index of the current density distribution in the proton exchange membrane as negative indicators, the performance of PEMFC under different assembly forces is comprehensively evaluated; the calculation results are as follows: Figure 3 As shown, the PEMFC comprehensive score is the highest under an assembly force of 1.0 MPa, therefore 1.0 MPa is determined to be the optimal assembly force.

[0033] Step 3: Analysis of the influence mechanism and sensitivity analysis of flow field structural parameters: S31. Based on the optimal assembly force of 1.0 MPa, the flow channel width, flow channel height, rib width, and turning angle are selected as key flow field structural parameters, with reference values ​​of 1.0 mm, 1.0 mm, 1.0 mm, and 0%, respectively. S32. Influence Law Analysis: Keeping the other three parameters as baseline values, the system changes one parameter (values ​​of flow channel width, flow channel height, and rib width: 0.6, 0.8, 1.0, 1.2, 1.4 mm; values ​​of turning angle: 0%, 25%, 50%, 75%, 100%), and uses the PEMFC numerical calculation model to analyze the influence of each parameter on the performance indicators. S33. Sensitivity Analysis: For different combinations of parameters, the entropy weight method was used to calculate the comprehensive score, and the relative sensitivity factor was calculated using the sensitivity analysis formula; the results showed that ( Figure 4 The relative sensitivity factor of rib width is the highest (-1.5536), which is at the "extremely sensitive" level and is a parameter that needs to be focused on during optimization.

[0034] Step 4: Flow field structure optimization based on neural networks and multi-objective algorithms: S41. Variable Definition: Using four key flow field structural parameters as input, with a value of 0.95 A / cm². 2 The power density, pressure drop in the cathode channel, and uniformity index of current density distribution in the proton exchange membrane are the outputs. S42, Latin hypercube sampling: 60 sample points are drawn within the design space; S43. Sample Calculation: Calculate the output values ​​of all sample points using the PEMFC numerical calculation model from step 1 to form a sample dataset; S44. Neural Network Training: Constructing a Backpropagation Neural Network ( Figure 5 The input layer has 4 nodes, the hidden layer has 5 nodes, and the output layer has 3 nodes. The 60 sets of data are randomly divided into training set (75%), validation set (15%), and test set (15%). The learning rate is 0.04, and the iteration is 5000 times. After training, the R coefficient of each dataset is greater than 0.95. S45. Multi-objective optimization: The trained neural network is used as the objective function of MOGWO, with the objectives of maximizing power density, minimizing the pressure drop in the cathode channel, and minimizing the uniformity index of the current density distribution in the proton exchange membrane. This yields the Pareto optimal solution set. Figure 6 ); S46. Optimal solution selection: The entropy weight method is used to comprehensively evaluate the solutions in the Pareto solution set, and the solution with the highest comprehensive score is selected as the final optimized solution: channel width 1.1mm, channel height 1.2mm, rib width 0.6mm, turning angle 50%.

[0035] Validation of optimization results: Simulation verification: A new PEMFC numerical calculation model was established based on the optimization scheme and simulation was performed; compared with the original model, the simulation results were obtained at 0.95 A / cm². 2 At that time, the power density increased by 4.51%, the pressure drop in the cathode channel decreased by 37.91%, and the uniformity index of the current density distribution in the proton exchange membrane decreased by 23.43%; compared with the neural network prediction, the errors were all less than 5%. Experimental verification: PEMFC samples were prepared according to the optimized scheme and tested; the optimized polarization curve and power density curve ( Figure 7 Significant improvements were achieved in the high current density region, with the error between the experimental and simulated power density values ​​being less than 2%, and the experimental value of the pressure drop in the cathode flow channel decreasing by 29.22%, verifying the effectiveness of the optimization scheme.

[0036] Any aspects not covered in this invention are applicable to existing technologies.

Claims

1. A multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation, characterized in that, The method includes the following steps: Step 1: Establish and verify the PEMFC numerical calculation model considering the effect of assembly force: Obtain the nonlinear compression deformation relationship of GDL through compression test, and obtain the contact resistance and contact thermal resistance under different assembly forces through contact resistance test; then establish a PEMFC numerical calculation model that couples the nonlinear compression deformation, contact resistance and contact thermal resistance of GDL, and verify the accuracy of the model through polarization curve test. Step 2: Determine the optimal assembly force for the PEMFC: Using the PEMFC numerical calculation model obtained in Step 1, simulate the performance of the PEMFC under different assembly forces to obtain the power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index at a specified current density. Then, use the entropy weight method to comprehensively evaluate the power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index under different assembly forces, calculate the comprehensive score for each assembly force, and select the assembly force with the highest comprehensive score as the optimal assembly force for the PEMFC. Step 3: Analysis of the Influence Mechanism and Sensitivity of Flow Field Structural Parameters: Based on the optimal assembly force determined in Step 2, the channel width, channel height, rib width, and turning angle are selected as key flow field structural parameters. Using the PEMFC numerical calculation model obtained in Step 1, the influence of individual changes in the four key flow field structural parameters on power density, pressure drop in the cathode channel, and uniformity index of current density distribution in the proton exchange membrane is analyzed. For combinations of different key flow field structural parameters, the entropy weight method is used to calculate their comprehensive score, and the relative sensitivity factor of each key flow field structural parameter is calculated through sensitivity analysis. Step 4: Flow field structure optimization based on neural network and multi-objective algorithm: Using key flow field structure parameters as input variables and their corresponding power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index as output variables, multiple sets of sample points are generated in the design space using the Latin hypercube sampling method. Then, the output variable values ​​of all sample points are calculated using the PEMFC numerical calculation model obtained in Step 1 to form a sample dataset. The backpropagation neural network is then trained using the sample dataset to establish a performance prediction model. The performance prediction model is used as the objective function and optimized using the multi-objective gray wolf optimization algorithm to obtain the Pareto optimal solution set. Then, the entropy weight method is used to comprehensively score the PEMFC power density, cathode channel pressure drop, and proton exchange membrane current density distribution uniformity index in the Pareto optimal solution set, and the combination of key flow field structure parameters with the highest comprehensive score is selected as the final optimization scheme.

2. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, Step 1 specifically includes the following steps: S11. Conduct a compression test on the GDL using a general compression testing device to obtain the nonlinear compression deformation curve of the GDL; S12. Based on the nonlinear compression deformation curve obtained in S11, the thickness distribution of GDL under different assembly forces under the rib and below the flow channel is obtained by finite element method simulation; then, according to the thickness distribution of GDL after compression, its porosity, electrical conductivity, thermal conductivity and permeability after compression are calculated by empirical formula. Meanwhile, under different assembly forces, the contact resistance between the bipolar plate and the GDL under different contact stresses was measured using a contact resistance tester; the contact thermal resistance under different contact stresses was obtained by fitting existing experimental data. S13. Based on the data obtained in step S12, a geometric solid model of PEMFC is established using 3D modeling software. The geometric solid model of PEMFC is then meshed into discrete small units for numerical calculation. Then, the solver and boundary conditions are set in the CFD software to obtain the PEMFC numerical calculation model. The PEMFC numerical calculation model couples the non-uniform physical parameters of GDL through a user-defined file and sets the contact resistance and contact thermal resistance at the interface between the bipolar plate and GDL. S14. Perform polarization curve testing and compare the test results with the simulation results of the model in step S13. If the maximum error does not exceed 5%, the model is considered to have passed verification. If the maximum error exceeds 5%, return to S13 and start from mesh generation again.

3. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 2, characterized in that, In step S12, the empirical formulas include the GDL porosity calculation formula, the GDL electrical conductivity calculation formula, the GDL in-plane thermal conductivity calculation formula, the GDL through-plane thermal conductivity calculation formula, and the GDL permeability calculation formula.

4. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 3, characterized in that, In step S12, the formula for calculating the porosity of GDL is: （1） In equation (1), Porosity after GDL compression; The initial porosity of GDL; This represents the initial thickness of the GDL. This refers to the thickness after GDL compression; The formula for calculating the conductivity of GDL is: （2） In equation (2), k eff K represents the conductivity after GDL compression. sol The electrical conductivity of GDL carbon fiber; The formula for calculating the in-plane thermal conductivity of GDL is: （3） In equation (3), Q eff Q represents the thermal conductivity of GDL after compression. sol The thermal conductivity of GDL carbon fiber; The formula for calculating the thermal conductivity of a plane using GDL is: （4） The formula for calculating GDL penetration rate is: （5） In equation (5), K is the permeability of GDL after compression; d is the diameter of GDL carbon fiber.

5. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, In step 1, the range of the assembly force to be examined is 0.5MPa~2MPa.

6. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, In step 2, the formula for calculating the uniformity index of current density distribution within the proton exchange membrane is: （6） In equation (6), S uni R is the uniformity index of current density distribution within the proton exchange membrane; R is the root mean square deviation of the distribution; AVE is the mean value of the distribution.

7. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, In step 3, the baseline values ​​for the flow channel width, flow channel height, rib width, and turning angle are 1.0 mm, 1.0 mm, 1.0 mm, and 0%, respectively; the values ​​for the flow channel width, flow channel height, and rib width are all in the range of 0.6 to 1.4 mm; the turning angle is used to quantify the geometric sharpness of the flow channel bend, and the value range is 0% to 100%.

8. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, In step 3, the specific calculation formula for sensitivity analysis is as follows: （7） In equation (7), S is the relative sensitivity factor; M i+1 and M i These are the combined scores for the (i+1)th and ith structural parameters, respectively; M a P is the average of the combined scores under the two structural parameters. i+1 and P i These are the values ​​of the (i+1)th and ith structural parameters, respectively; P a This is the average of the two structural parameters.

9. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, In step 4, the Latin hypercube sampling method is to uniformly divide the range of values ​​of each input variable into several intervals equal to the number of samples, and randomly select a sample from each interval. In step 4, the number of hidden layer neurons in the backpropagation neural network is 5, the learning rate is set to 0.04, and the number of training iterations is 5000. When the regression coefficient R of the training set, validation set, and test set are all greater than 0.95, the model training is considered complete. In step 4, the optimization objectives of the multi-objective gray wolf optimization algorithm are to maximize the power density, minimize the pressure drop in the cathode channel, and minimize the uniformity index of the current density distribution in the proton exchange membrane.

10. The multi-objective collaborative optimization design method for fuel cell flow channels considering GDL compression deformation according to claim 1, characterized in that, The method also includes step 5: applying the final optimization scheme obtained in step 4 to the design and manufacturing of fuel cell stacks.

Images