Dynamic modeling of solid oxide fuel cell stack based on local linear model tree algorithm

This paper presents a dynamic model for Solid Oxide Fuel Cell (SOFC) stack using Local Linear Model Tree (LOLIMOT) algorithm which is useful for both dynamic and steady-state studies. Most of conventional SOFC models require a large number of parameters and factors, which are difficult to be determined or estimated. In this paper, a LOLIMOT-based model, which does not require the parameters of fuel cell, is proposed for each operation mode of SOFC. In these models, decision tree-based feature selection approach is exploited to select inputs of the LOLIMOT. The proposed models are trained in a short time and they have little errors.In order to illustrate the effectiveness of the proposed models, they are applied to a simulated model of a 5-kW SOFC stack which show satisfactory results for both steady-state and dynamic studies. Furthermore, the proposed model demonstrates convincing results for real-time simulation studies.

► Local Linear Model Tree algorithm is successful in modeling solid oxide fuel cells.
► The model is useful for both dynamic and steady-state studies.
► The stack voltage and the active power of SOFC are chosen as the model outputs.
► Decision tree feature selection approach is used to select suitable model inputs.
► Current, temperature, pressure and fuel utilization are some of the model inputs.