(625c) A New Approach for Robust Online Pefc Sensors Using Symbolic Dynamics

In fuel cell stacks and systems, there is an urgent need for non-intrusive, robust, rapid and inexpensive sensors to enable both stable performance and extended lifetime operation. Achieving this goal is complicated by 1) the desire to achieve active control with a minimal number of sensors, 2) a limited number of basic physical signals available (e.g. voltage, current, temperature, pressure), 3) the highly coupled nature of the thermophysical and electrochemical processes involved, and 4) the impracticality of applying many of the diagnostic tools used to study single cells in a laboratory environment to a full-sized stack.

Carbon monoxide, even in ppm levels, can dramatically reduce the performance of a fuel cell stack if not remediated. Remediation generally requires bleeding a small fraction of air into the fuel stream to promote oxidation of the CO. However, this process is inefficient and must be controlled based on real-time feedback of feed-stream CO.

In this work, a novel robust online electrochemical sensor [1] capable of discerning the continually changing carbon monoxide levels in a reformed fuel stream delivered to a stack is described. Methods of symbolic dynamics [2,3] and pattern recognition have been developed for other electromechanical systems, but never fuel cells. In this work, this approach has been applied to discern minute levels of CO poisoning so that proper control can be achieved. Figure 1 shows a plot of the anomaly measure, a numerical indication of the deviation from a normal CO-free state, versus the CO level for a laboratory-based sensor. The sensor sensitivity CO at the ppm level can be tailored for a particular application. A prototype sensor is planned to be installed as a prototype on a full size stack, and will be discussed as time permits.

A similar approach can be used to develop online sensors for a variety of other important fuel cell phenomena. Time permitting, the talk will also discuss the status of a new membrane health sensor based on this approach.

1. M. M. Mench, A. Ray, S. Chin, and E. C. Kumbur, Provisional Patent Application for Invention Disclosure 2004-2921 ?Rapid-response sensor for low-level carbon monoxide,? January 2006.

2. A. Ray, ?Symbolic Dynamic Analysis of Complex Systems for Anomaly Detection,? Signal Processing Vol. 84, No. 7, pp. 1115-1130, 2004.

3. S. Chin, A. Ray and V. Rajagopalan, ?Symbolic Time Series Analysis for Anomaly Detection: A Comparative Evaluation,? Signal Processing, Vol. 85, No. 9, pp. 1859-1868, 2005.