LAUSR

A data‐driven strategy considering sensor reliability and cost is presented for polymer electrolyte membrane fuel cell (PEMFC) fault diagnosis. An increase of pressure drop in the flow field is a reliable indicator of PEMFC flooding fault. Still, it highly depends on the pressure measurement accuracy of the stack inlet and outlet. The abnormal sensor will mislead the diagnosis results, and an extra pressure sensor in the outlet means higher system cost. Thus, the related behavior is investigated. The throttle opening set, one of the control signals, is chosen to replace the cathode pressure drop. The throttle opening set, the stack voltage, and cell voltage monitoring (CVM) standard deviation are employed as diagnostic features, and the support vector machine (SVM) is used as the classifier. Compared to the approach using single‐cell voltages, the strategy mentioned above minimizes extra sensors and avoids the computation load of dimensionality reduction. High accuracy in diagnosing flooding and dehydration via identifying experimental data of different fault severity verifies the effectiveness of the presented strategy.