We present a method for determining optimal modes of operation for autonomously oscillating systems with uncertain parameters. In a typical application of the method, a nonlinear dynamical system is optimized with respect to an economic objective function with nonlinear programming methods, and stability is guaranteed for all points in a robustness region around the optimal point. The stability constraints are implemented by imposing a lower bound on the distance between the optimal point and all stability boundaries in its vicinity, where stability boundaries are described with notions from bifurcation theory. We derive the required constraints for a general class of periodically operated processes and show how these bounds can be integrated into standard nonlinear programming methods. We present results of the optimization of two chemical reaction systems for illustration. 2 process performance has been investigated for decades. For example, Douglas and Rippin 3 (1966) demonstrate that the performance of an isothermal continuous stirred-4 tank reactor (CSTR) may be improved by periodic forcing of the feed. The 5 authors also consider a first order irreversible exothermic reaction in a non-6 isothermal CSTR. For this case, they show that autonomous oscillations may 7 lead to increased average product concentration compared to steady state op-8 eration. Similar investigations have been carried out later by other authors.9 Jianquiang and Ray (2000) use autonomous oscillations to improve the per-10 formance of a bioreactor used for sludge water treatment. Stowers et al. 11 (2009) show that oscillations can increase the product yield in yeast fermen-12 tation. Parulekar (2003) demonstrate that the performance of series-parallel 13 reactions can be improved by forced periodic operation. The authors also 14 discuss the benefit of forced periodic operation compared to steady state op-15 eration in recombinant cell culture processes. Abashar and Elnashaie (2010) 16show that periodically forced fermentors provide higher average bioethanol 17 concentrations than fermentors operated in steady state.
18Whenever models of the production process of interest and its economics 19 are available, it is an option to use linear or nonlinear programming methods 20 to find an optimal mode of operation. It is known, however, that optimizing a 21 dynamical system in this way may result in a steady state or periodic mode 22 of operation that, while optimal with respect to the economic objective, 23 is unstable (Mönnigmann and Marquardt, 2002). In general, optimal but 24 unstable solutions are not useful in practice. 25 2 Approaches inspired by applied bifurcation theory have been used to state 26 constraints on stability properties in optimization problems. Since these 27 methods are based on normal vectors to manifolds of critical points such as 28 bifurcations points, they are jointly referred to as the normal vector approach 29 for short. Originally, the normal vector approach was developed to guarantee 30 stability of optimal equili...