Fault diagnosis of thermal management system in a polymer electrolyte membrane fuel cell
Fault diagnosis of thermal management system in a polymer electrolyte membrane fuel cell
复制标题
DOI：
10.1016/j.energy.2020.119062
复制
发表时间：
2021
期刊：
影响因子：
9
通讯作者：
J. Park;In Seop Lim;E. Choi;M. S. Kim
中科院分区：
文献类型：
--
作者：
J. Park;In Seop Lim;E. Choi;M. S. Kim
聚合物电解质膜燃料电池系统商业化的障碍之一是其可靠性。提高系统可靠性的一个很好的方法是建立故障诊断模型。提出了一种聚合物电解质膜燃料电池热管理系统的诊断方法。随着燃料电池电堆功率的增加，来自电堆的废热也随之增加，从而热管理系统也受到关注。在这项研究中，热管理系统故障实验反复进行各种负载和堆栈退化条件下。随着电堆的退化，电堆的废热增加，故障响应及其对系统的影响也发生变化。在这项研究中开发的诊断模型检测故障的组件级和诊断的严重性的故障组件。该诊断模型适用于电堆退化的燃料电池系统。在模型开发过程中，将单任务学习技术应用于神经网络诊断模型，以提高诊断精度，并与传统的多任务学习技术进行了比较。
One of the barriers for commercialization of polymer electrolyte membrane fuel cell system is its reliability. One good solution for reliability improvement is to develop fault diagnosis model. This study suggests a diagnostic method for polymer electrolyte membrane fuel cell thermal management system. With the increase of fuel cell stack power these days, waste heat from the stack also increased and thereby thermal management system is also receiving attention. In this study, thermal management system fault experiments are repeatedly performed under various loads and stack degradation conditions. With the degradation of stack, waste heat from the stack increases and fault response as well as its impact on the system varies. The diagnosis model developed in this study detects faults on component-level and diagnoses severity of the faulty components. Furthermore, the diagnosis model is applicable to fuel cell system with degraded stack. In the process of model development, single-task learning technique is applied to neural network diagnosis model for higher diagnostic accuracy and compared with conventional multi-task learning technique.