Stack Temperature Estimation in System Environment by Utilizing the Design of Experiments Methodology

Abstract

The behavior of the maximum temperature measured inside a SOFC stack with respect to three independent input variables (stack current, air flow and air inlet temperature) is examined by using a full factorial screening experiment, following the design of experiments methodology. The experiments were carried out with a complete 10 kW e SOFC system. Multivariate regression models are developed to estimate said temperature and a statistical analysis is carried out on the model parameters.

Solid oxide fuel cell stack temperature estimation with data-based modeling – Designed experiments and parameter identification

“…The experiment design used here extends that of the previous work [9], where it was found that while the three first inputs (I, _ V air and T air,in ) as such have a statistically significant effect on the stack temperature, their interactions do not. This enabled omitting the interaction effects of these inputs from the current experimental design.…”

“…To meet these aims, the experiments were designed to fit within a pre-defined system operating space, and in particular, around two representative "central" operating conditions (Section 2). Thus, in comparison to the previous work [9,18], the data series collected and used in this study is larger in terms of the number of inputs as well as their range of variation. In addition, the number of data points collected is larger than before, which enables splitting the data set to a part used for model development and another part used solely for model validation.…”

“…In Refs. [9,16,17] DoE is utilized in particular for the analysis and the development of linear regression models of SOFCs.…”

“…The dependencies between internal SOFC stack temperatures and three system inputs, namely, load current, air flow and air temperature at stack module inlet, were experimentally evaluated in a preceding work [9,18]. The experiments were carried out around a single specified nominal operating point and the models were created for the steady state estimation of the stack temperature by utilizing multivariable linear regression (MLR).…”

“…Note also that the central operating point of FFD1 is also the system design point ("nominal operating condition" or NOC). The FFD structure was decided upon earlier experimental results [9]. Both FFDs are 2 4e1 fractional factorial designs obtained with the design generator _ V NG ¼ I$ _ V ir $T air;in (with the defining relation…”