CAREER: Structure and Robustness in Nonlinear Control: Challenges from Fuel Cell Technology

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

The research objective of this proposal is to develop systematic design tools that exploit, rather than override, nonlinearities and structural properties in feedback systems. Encouraged by our preliminary results, we address the problems of observer design and robustness to unmodeled dynamics, which are crucial for practical applicability of nonlinear control techniques. The variety of dynamic phenomena in nonlinear systems have hampered constructive solutions to these problems. In contrast, we pursue structure-specific designs which are tailored to inherent system properties and nonlinear phenomena. Our robust design tools exploit types of unmodeled dynamics and multivariable structures to recover stability and performance achievable with nominal designs. Our observers take advantage of nonlinearities to enhance observer convergence and to recover stability of full state-feedback designs in certainty-equivalence implementations. In collaboration with United Technologies Research Center, the PI will apply these design tools to emerging feedback problems in fuel cell power systems. Fuel cell technology is particularly suit- able for United States, because the wide geography makes centralized power networks prone to failure, and the strict environmental regulations accentuate the necessity of clean power genera- tion.A major challenge for potential vehicular and portable power applications is the high-order, interconnected, system structure, and severe nonlinearities, which arise from tight integration of system components and wide variations in the electric load. Robustness to unmodeled dynamics is crucial for reliability of low-order designs. Nonlinear observers are of great importance because commercially available sensor technologies are not designed to operate in a fuel cell environment. The study of fuel cells will set the stage for abstractions and broad design methodologies which will also impact other technologies, such as computer networks, chemical processes, and fluid control problems in aeroengines, where high-order system models and severe nonlinearities constitute an obstacle to efficient and reliable feedback designs. The educational objective of the proposal is to make nonlinear concepts accessible to a wider spectrum of students. The PI will introduce a new course for undergraduates, and graduate stu- dents who wish to study nonlinear systems at an introductory level. This course will enhance the appreciation of nonlinear systems among practitioners of the future, and attract more students to our graduate research program. In collaboration with the New Visions Program of local school districts and Rensselaer Polytechnic Institute, we will include high-school students in project teams with our undergraduates. Several pedagogical experiments and computer animations, detailed in the proposal, will give them a motivating real-life exposure to dynamic phenomena and control concepts.