A general approach to develop reduced order models for simulation of solid oxide fuel cell stacks

Abstract

Numerical models for solid oxide fuel cells (SOFCs) are needed in system modeling studies of fuel cell-based power generation systems. A reduced order modeling approach based on response surface techniques is developed for SOFC stacks. This approach creates a numerical model that can quickly compute desired performance variables of interest based on the stack's input parameter state. The developed method first carefully samples the multidimensional design space based on the input parameter ranges, automatically evaluates an existing detailed stack model at each of the sampled points, and performs regression for selected performance variables of interest to determine the response surfaces. After error analysis to ensure that sufficient accuracy is established for the response surfaces, they are then implemented in a calculator module for use by the system-level software. The benefit of this modeling approach is that it is sufficiently fast for integration with system modeling software while still providing high fidelity information about the internal distributions of key variables in the fuel cell. This paper describes the sampling, regression, sensitivity, error, and principal component analysis to identify the most appropriate methods for simulating a planar SOFC stack.