Online Diagnosis of PEMFC by Combining Support Vector Machine and Fluidic Model

This paper deals with the online fault diagnosis of polymer electrolyte membrane fuel cell (PEMFC) stack. In the proposed approach, support vector machine (SVM) and a fluidic model are correlated to realize the diagnosis of the faults on water management. With the help of the fluidic model, health states of the stack can be identified. This procedure is then dedicated to labeling the training data, which is used to train the dimension reduction model, named Fisher discriminant analysis (FDA), and the SVM classifier. The online diagnosis can then be realized by using the trained FDA and SVM models. The proposed approach is illustrated by using the experimental data of a 20-cell PEMFC stack. The feasibility of the approach for online implementation is also affirmed.