This content is not included in your SAE MOBILUS subscription, or you are not logged in.

Online Flooding and Dehydration Diagnosis for PEM Fuel Cell Stacks via Generalized Residual Multiple Model Adaptive Estimation-Based Methodology

Journal Article
2019-01-0373
ISSN: 2641-9637, e-ISSN: 2641-9645
Published April 02, 2019 by SAE International in United States
Online Flooding and Dehydration Diagnosis for PEM Fuel Cell Stacks via Generalized Residual Multiple Model Adaptive Estimation-Based Methodology
Sector:
Citation: Zhou, S., Zhou, S., Jin, J., and Wen, Z., "Online Flooding and Dehydration Diagnosis for PEM Fuel Cell Stacks via Generalized Residual Multiple Model Adaptive Estimation-Based Methodology," SAE Int. J. Adv. & Curr. Prac. in Mobility 1(2):787-794, 2019, https://doi.org/10.4271/2019-01-0373.
Language: English

Abstract:

For proton exchange membrane fuel cell (PEMFC) stack, critical issues such as flooding and dehydration, are caused by improper water management. With respect to the water management failure, PEMFC stack outputs power and efficiency decreased. Therefore, proper water management with diagnosis contributes to the reliability and durability. Existing researches establish Electrochemical Impedance Spectroscopy (EIS) measurement to detect and identify different faults, whereas the sophistication, overwhelmed computational consumption of EIS and unaffordable dedicated instrumentation make it’s unsuitable for commercial application. Therefore, EIS is not considered to be a viable solution to online and real-time diagnostic scheme. In this paper, an innovative method based on EIS, is further developed to identify some critical PEMFC fault conditions online, wherein generalized residual multiple model adaptive estimation (GRMMAE) methodology is applied. The diagnosis process consists of multiple equivalent circuit models that represent signature faults separately, such as flooding and dehydration, causing significant variation of model parameters. Applying a small sinusoidal alternating current (AC) as a perturbation signal to various parallel models and measuring the potential response, whilst the noise from perturbation signal and measurement result in utilizing standard Kalman filters (KF) to generate residual signals. The residual signals are used in the GRMMAE methodology to evaluate probabilities in real-time that determine the type of fault. Simulation result shows that the fault conditions can be detected and identified accurately, hence the effectiveness of the proposed method would be indicated.