A Novel Fault Diagnosis Method of PEMFC System Based on Data Space Feature Decision Tree Group and Extreme Learning Machine

For the operation process of proton exchange membrane fuel cells (PEMFCs) system, fault diagnosis plays a crucial role in ensuring the safety and reliability of system. To improve the accuracy and rapidity of fault diagnosis, a fault diagnosis method of PEMFC based on data space feature decision tree group (DTS) and extreme learning machine (ELM) is proposed. Firstly, principal component analysis (PCA) is used to reduce the dimension of sensor data such as current, temperature and gas flow rate of system, which reflects the operating states of fuel cell. Besides, spatial features of system can be extracted and processed by random rotation matrix. Furthermore, multiple sets of spatial rotation feature data are set to diagnose the health state of fuel cell system by combining decision tree group and extreme learning machine (DTS-ELM). Finally, the proposed method was verified and analyzed, and the experimental results indicate that this method can quickly identify four operating states such as normal state, flooding, membrane dying and hydrogen leakage. The diagnostic accuracy and operating time of this method are 99.54% 0.261s, respectively. Therefore, on-line rapid fault diagnosis of proton exchange membrane fuel cell system can be realized by the proposed method.