Application of the Sensor Selection Approach in Polymer Electrolyte Membrane Fuel Cell Prognostics and Health Management

Abstract

Abstract:In this paper, the sensor selection approach is investigated with the aim of using fewer sensors to provide reliable fuel cell diagnostic and prognostic results. The sensitivity of sensors is firstly calculated with a developed fuel cell model. With sensor sensitivities to different fuel cell failure modes, the available sensors can be ranked. A sensor selection algorithm is used in the analysis, which considers both sensor sensitivity to fuel cell performance and resistance to noise. The performance of the selected sensors in polymer electrolyte membrane (PEM) fuel cell prognostics is also evaluated with an adaptive neuro-fuzzy inference system (ANFIS), and results show that the fuel cell voltage can be predicted with good quality using the selected sensors. Furthermore, a fuel cell test is performed to investigate the effectiveness of selected sensors in fuel cell fault diagnosis. From the results, different fuel cell states can be distinguished with good quality using the selected sensors.

Effect of Sensor Set Size on Polymer Electrolyte Membrane Fuel Cell Fault Diagnosis

“…One possible reason for this is that several sensors used in the analysis are not sensitive to the PEM fuel cell performance variation, and will be more easily affected by measurement/environment noise—hence, the inclusion of these sensors will make a negative contribution to the PEM fuel cell fault diagnosis. It should be noted that the above results are comparable with those of previous studies [ 19 , 33 ], where the performance of all available sensors and selected sensors in identifying fuel cell flooding was investigated and results also demonstrate that the selected sensors can provide better diagnostic performance.…”

“…Compared to the fault diagnosis using fuel cell voltage, more studies have been performed using multiple sensors information [ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ]. Both single [ 7 ] and multiple PEM fuel cell faults [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ] can be identified accurately with information from multiple sensors, using either model-based or data-driven approaches.…”

“…Both single [ 7 ] and multiple PEM fuel cell faults [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ] can be identified accurately with information from multiple sensors, using either model-based or data-driven approaches. Furthermore, in order to increase computational efficiency when dealing with data from multiple sensors, dimension reduction techniques such as principal component analysis have been selected to reduce the dimension of the original dataset, while retaining useful information [ 17 , 18 , 19 , 20 ].…”

Hydrogen Leakage Diagnosis for Proton Exchange Membrane Fuel Cell Systems: Methods and Suggestions on Its Application in Fuel Cell Vehicles

“…In recent years, with the rapid development of data mining technologies, the data-driven gas leakage diagnosis methods have been applied widely [44]. The data-driven methods realize the diagnosis of hydrogen leakage from hydrogen supply system by fault features, which are extracted from the state data of system in case of normal and hydrogen leakage (as shown in Figure 6b).…”

“…In the existing research, there are many diagnosis methods, either for the hydrogen leakage from the hydrogen supply system or for the hydrogen leakage in the stack. As shown in Figure 1, these methods used different strategies to diagnose hydrogen leakage, mainly including environmental hydrogen concentration diagnosis [24][25][26][27][28][29][30], hydrogen pressure decay diagnosis [9,10,[32][33][34]100,[112][113][114], gas leakage diagnosis by model-based methods or data-driven methods [36][37][38][39][40][41][42][43][44][45][46][47][48][49][50]115,123], cross-current diagnosis [10,[100][101][102][103], and diagnosis by stack (or cells) output voltage [10,100,[104][105][106][107][108]…”