Analysis And Prediction Of Contact Resistance Of Single Proton Exchange Membrane Fuel Cell

As a non-pollution power generation device,Proton Exchange Membrane(PEM)fuel cell has the advantages of high reliability,high energy efficiency,low temperature and quick start.It has been widely applied in the new energy automobile industry.The output power is an important performance index of PEM fuel cell.The loss of output power is mainly caused by the contact resistance between the metal plate and the gas diffusion layer in PEM fuel cell.Although a series of techniques have been developed to keep the contact resistance at low levels,the assembly errors of PEM fuel cell still cause the increase of the contact resistance.Decreasing all the assembly errors can realize acceptable contact resistance,but will lead to a severe increase in manufacturing costs.Therefore,it is necessary to find out key assembly errors that have significant effect on contact resistance,and reveal the variation of contact resistance caused by different assembly error values,in order to guide the assembly process optimization of PEM fuel cell from the design stage.To fulfill these engineering requirements,single PEM fuel cell as the main target,after developing the finite element simulation model of single PEM fuel cell,this paper analyzes the influence of assembly errors on contact resistance,and predicts the distribution of contact resistance with different assembly error values:1)Based on assembly errors in single PEM fuel cell,the finite element simulation model between the metal plate and the gas diffusion layer is developed,and the contact resistance calculation formula based on the contact stress is design.The case study of a single PEM fuel cell is provided.The data samples composed of assembly errors and contact resistance are obtained.2)Considering the simultaneous change of assembly errors and the complex relationship between assembly errors and contact resistance,an improved global sensitivity analysis method is proposed to evaluate influence degrees of assembly errors on contact resistance.In the proposed method,the assembly error is randomly sampled based on the adaptive Sobol sequence,and the Monte Carlo algorithm is employed to estimate the variation variance of contact resistance to calculate the sensitivity indices of contact resistance to assembly errors accurately.Based on sample data obtained by the finite element simulation,the proposed method is compared with a traditional sensitivity analysis method in terms of computational efficiency,and the influence degrees of assembly errors on contact resistance is analyzed.The results demonstrate that the proposed method improves the computational efficiency,and provides key assembly errors that have significant impacts on contact resistance.3)Considering the characteristics of small sample size and high dimensional inputs of contact resistance prediction problem,the contact resistance prediction model based on Least Squares Support Vector Machine(LS-SVM)is designed by taking key assembly errors as input variables.The kernel function selection between linear function,polynomial function and radial basis function,and the genetic algorithm-based optimization of hyperparameters are employed to improve the prediction accuracy of contact resistance in the LS-SVM.Based on historical data collected from the finite element simulation model,the contact resistance prediction accuracy of LS-SVM with different kernel functions and hyperparameters is compared.The results demonstrate that the LS-SVM with on radial basis kernel function realizes accurate prediction of contact resistance,and provides contact resistance distribution with different assembly error values.Finally,the prototype system of single PEM fuel cell contact resistance analysis and prediction method is developed.The system outputs provide scientific suggestions for the reasonable decrease of assembly errors in a specific single PEM fuel cell.