UC Irvine

This thesis investigates the challenge of designing a sufficiently reliable solid oxide fuel cell electrical system for use in a data center. Such a system will lower data center emissions and increase data center efficiency, without sacrificing operational performance.

In order to accomplish this goal, the reliability and availability of such a system must be analyzed and understood. Data centers desire 99.9999% uptime. Currently, extensive backup equipment is used to ensure the availability of services. The proposed design alternative was analyzed by several methods to ensure sufficient reliability and availability.

First, potential system designs were explored. Designs for both system as a whole and for individual fuel cells were considered. Once potential designs were identified, a reliability block diagram analysis of the potential fuel cell system was completed. This analysis was done to understand the reliability of the system without repair and without additional redundant technologies. From this analysis, it was apparent that redundant components would be necessary.

Building on this knowledge, a program was written in MATLAB to show that the desired system reliability could be achieved by a combination of parallel components, regardless of the number of additional components needed. Having shown that the desired reliability was achievable through some combination of components, a dynamic programming analysis was undertaken to assess the ideal allocation of parallel components.

A Markov analysis was then conducted. The Markov analysis allowed for the consideration of a repairable system. This allowed for an understanding of the reliability and availability of the system.