Optimal control for a simplified 1D fuel cell model

Abstract

Molten carbonate fuel cells are a promising technology for the operation of future stationary power plants. To enhance service life, a detailed knowledge of their dynamical behaviour is essential. The possibility of fast and save load changes is important for daily operation of these power plants. To predict the dynamical behaviour of fuel cells a hierachy of mathematical models has been developed in the past. Recently a systematic model reduction was applied to a 2D crossflow model. We present here the new 1D counterflow model and discuss a suitable discretization method. Accordingly we set up a method of optimal control following the first-discretize-then-optimize approach. Results are shown for simulation and optimal control in the case of load changes.

Optimal control of coupled multiphysics problems: Guidelines for real‐life applications demonstrated for a complex fuel cell model

“…Due to the complexity of the system and the long process time (up to 3 h), but mainly due to the non-efficient application of numerical differentiation for the computation of the gradient of the objective function with respect to the discretized control variables only coarse discretizations can be handled. Nevertheless, the numerical results were sufficiently trustworthy thanks 6 Adaptivity in space is not an issue here, since equidistant grids perform sufficiently well; see [24]. Thereby, one avoids the book-keeping of handling adaptive grids for each of the 21 space-dependent space variables at several time instances.…”

“…1. These models have been developed by Heidebrecht [10] and later simplified in [24]. A major task in operating such MCFC plants is to control appropriate technical devices in an optimal way in order to perform a change in the external load.…”

“…The spatially 1D counter flow configuration of Fig. 2 forms the basis of the simplified 1D model of [24]. In contrast, Fig.…”

“…Complete models with up to 28 state variables can be found for the 2D cross flow model in [10] and [6] and for the 1D counter flow model in [24] and [23].…”

“…al. [26], Heidebrecht [10], and [6] and [24]. All these phenomena have to be simultaneously taken into account in order to describe the reallity as precise as possible and necessary.…”