Stochastic Approximation in Online Steady State Optimization Under Noisy Measurements

We present a novel approach to online steady state optimization (Real-time Optimization) of processes subject to model uncertainties and noisy measurements. The idea behind of the method is to exploit the inherent stochastic properties of the problem by means of Stochastic Approximation. In combination with Modifier Adaptation, the proposed approach is able to provide convergence towards the actual optimum, despite the presence of plant-model mismatch and highly noisy measurements. The performance of the method is illustrated by simulation studies for a benchmark problem.