ABSTRACT

In the proton exchange membrane fuel cell (PEMFC), the uneven distribution of contact pressure on the gas diffusion layer (GDL) will lead to the increase in local contact voltage, thus increasing the ohmic loss of the stack, and will affect the porosity to change the reaction gas concentration. Therefore, it is essential to optimize the consistency and uniformity of contact pressure. In this paper, the bolt-encapsulated different scale stack models are established. The influence of the cell number on the distribution characteristics of the contact pressure on the outermost and innermost GDLs in different stacks is discussed. According to the GDL contact pressure distribution characteristics in the 6-stack, it is found that it can be used as the data extraction model for optimizing the clamping condition of the 80-stack. Subsequently, the optimal Latin hypercube design is used to randomly generate 280 data sets with different clamping conditions, and a data-driven surrogate model based on the Support Vector Regression is used to predict the contact pressure distribution characteristics of GDLs in the 6-stack according to changes in clamping condition. Then, the surrogate model is taken as the fitness evaluation function in Gray Wolf Optimizer, and the clamping condition of the 80-stack is optimized, which improves the uniformity and consistency of the contact pressure distribution on the GDL by 66.67%. This paper proposes a strategy to optimize the clamping condition of large-scale stacks based on the calculation and prediction of the contact pressure distribution on the GDLs in small-scale stacks.

摘要

在质子交换膜燃料电池（PEMFC）中，气体扩散层（GDL）上的接触压力分布不均会导致局部接触电压升高，从而增加电堆的欧姆损耗，并影响孔隙率改变反应气体浓度。因此，必须优化接触压力的一致性和均匀性。本文建立了螺栓封装的不同尺度堆栈模型。讨论了单元数对不同电堆中最外层和最内层GDLs接触压力分布特性的影响。根据GDL在6叠中的接触压力分布特征，发现它可以作为优化80叠夹持条件的数据提取模型。随后，采用最优拉丁超立方体设计随机生成280个不同钳位条件的数据集，并采用基于支持向量回归的数据驱动代理模型根据变化预测6-stack中GDL的接触压力分布特征在夹紧状态。然后，将代理模型作为灰狼优化器中的适应度评估函数，对80-stack的夹紧条件进行优化，使GDL上接触压力分布的均匀性和一致性提高了66.67%。本文提出了一种基于小型电堆中 GDL 接触压力分布的计算和预测来优化大型电堆夹紧条件的策略。