Modifier adaptation with quadratic approximation in iterative optimizing control

In this paper we combine the quadratic approximation approach used in derivative-free optimization (DFO) with iterative gradient-modification optimization (IGMO) to formulate an efficient scheme for iterative real-time optimization (RTO) under model uncertainty. By combining the robustness of the DFO approach to noisy data with the convergence to the true optimum of the IGMO using empirical gradients, the novel scheme is able to reliably and efficiently optimize the operation of a system based on inaccurate process models and noisy measurements, i.e. for realistic scenarios. Simulation studies for the optimization of the Otto-Williams reactor are used to demonstrate the performance of the new scheme.