Fault classification of PEM fuel cell systems based on CS-SVM method

Considering the problem that the support vector machine (SVM) algorithm, used in the fault classification of fuel cell systems, is prone to fall into local optimal solution and cannot accurately identify fault categories, a fault classification method based on SVM optimized by cuckoo search (CS) algorithm is proposed. Through the analysis of measured data, the proposed method can effectively identify five fault states, including the leakage of hydrogen, the low pressure of deionized water humidification pumps, deionized ethylene glycol inlet high temperature, signal voltage over-range of deionized ethylene glycol outlet temperature and the low pressure of input air, with classification accuracy up to 93.6%.The advantages of this method are verified by comparing with SVM optimized by traditional grid search (GS) method.