PEMFC Fault Diagnosis Based on an Equivalent Circuit and OS-ELM Framework

Most current proton exchange membrane fuel cell (PEMFC) fault diagnosis models are not updated after offline training is completed, so they cannot learn new data features. For this problem, a PEMFC fault diagnosis model based on an equivalent circuit and an online sequential extreme learning machine (OS-ELM) framework is proposed. Under different operating conditions, the range of equivalent circuit element values is different, which is conducive to fault diagnosis. Element values constitute a dataset, from which OS-ELM is used to learn data features, and PEMFC fault diagnosis and isolation can be accomplished. OS-ELM has online learning capability. During the operation of PEMFC, equivalent circuit fitting is performed on multiple sets of sample data obtained using electrical impedance spectroscopy (EIS). Optimized fuzzy C-means (FCM) is used to cluster and partition sample data for online updating of OS-ELM models. The experimental results show that the proposed PEMFC fault diagnosis model not only has the speed and accuracy of fault diagnosis but also has the ability to learn fault characteristics online.