Quantitative diagnosis of PEMFC membrane humidity with a vector-distance based characteristic mapping approach

Abstract

Membrane dehydration or flooding fault is one of the main causes of performance degradation for Proton Exchange Membrane Fuel Cell (PEMFC). Effective and accurate diagnosis of membrane humidity faults is necessary to ensure the optimal operation of PEMFC. However, the research gap lies in the quantitative diagnosis of membrane humidity values under complex working conditions of vehicle application. In this paper, characteristic spaces describing membrane humidity are established as multi-matrices consisting of the amplitude spectrum extracted from output voltage and pressure drop in PEMFC under different vehicle working conditions. In the multi-matrix, the relationship between the membrane humidity fault levels and the amplitude spectrum can be described as characteristic vectors. Then, a vector-distance based mapping approach under multi-information characteristic vectors is developed to describe the relationship between characteristic vectors and humidity values. Consequently, the membrane humidity of the PEMFC can be quantitatively evaluated with the projection length between characteristic vector and humidity-change vector, with which the humidity value of PEMFC membrane can be obtained under different working conditions in real-time. The performance of the proposed a vector-distance based characteristic mapping approach is verified by New European Driving Cycle (NEDC) and World Light Vehicle Test Procedure (WLTP) working conditions. The diagnostic results show that the quantitative humidity diagnosis strategy can achieve diagnostic accuracy of 98.16% and 92.24% under NEDC and WLTP working conditions, compared with the single information of output voltage-based and pressure drop-based diagnosis strategy, the accuracy is improved by 4.37% and 4.08% respectively.