Online implementation of SVM based fault diagnosis strategy for PEMFC systems

•A classification based fault diagnosis approach is proposed for Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems.•Individual fuel cell voltages are considered as the variables for diagnosis.•A high-compacted embedded system is designed and fabricated to realize the approach.•Online implementa...

Full description

Saved in:
Bibliographic Details
Published in:Applied energy 2016-02, Vol.164, p.284-293
Main Authors: Li, Zhongliang, Outbib, Rachid, Giurgea, Stefan, Hissel, Daniel, Jemei, Samir, Giraud, Alain, Rosini, Sebastien
Format: Article
Language:English
Subjects:
Automatic
Engineering Sciences
Fault diagnosis
Fluid mechanics
Mechanics
Online implementation
PEMFC system
Physics
SVM classification
System in Package
Thermics
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•A classification based fault diagnosis approach is proposed for Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems.•Individual fuel cell voltages are considered as the variables for diagnosis.•A high-compacted embedded system is designed and fabricated to realize the approach.•Online implementation is carried out to validate the strategy.•A diagnosis rule is proposed to improve the robustness and accuracy. In this paper, the topic of online diagnosis for Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems is addressed. In the diagnosis approach, individual cell voltages are used as the variables for diagnosis. The pattern classification tool Support Vector Machine (SVM) combined with designed diagnosis rule is used to achieve fault detection and isolation (FDI). A highly-compacted embedded system of the System in Package (SiP) type is designed and fabricated to monitor individual cell voltages and to perform the diagnosis algorithms. For validation, the diagnosis approach is implemented online on PEMFC experimental platform. Four concerned faults can be detected and isolated in real-time.
ISSN:0306-2619
1872-9118
DOI:10.1016/j.apenergy.2015.11.060