Use of Convex Model Approximations for Real-Time Optimization via Modifier Adaptation Grégory François and Dominique Bonvin ∗ Laboratoire d’Automatique, École Polytechnique Fédérale de Lausanne CH-1015 Lausanne, Switzerland E-mail: dominique.bonvin@epfl.ch Abstract Real-Time Optimization (RTO) via modifier adaptation is a class of methods for which measurements are used to iteratively modify a model-based optimization problem with the particularity that the plant model remains unchanged. The modifier terms correspond to the deviations between the measured and predicted constraints and between the measured and pre- dicted cost and constraint gradients. If the iterative scheme converges, these modifiers guaran- tee that the converged point is a KKT point for the plant. If, furthermore, the plant model is adequate, i.e. if it is able to predict the correct curvature of the cost function at the converged inputs, convergence to a (local) plant optimum is obtained. The main advantage of modifier adaptation lies in the fact that these properties do not rely on specific assumptions regarding the nature of the uncertainty. In other words, modifier adaptation is able to reject the effect of not only parametric uncertainty, like most RTO methods, but also of process disturbances and structural plant-model mismatch. This paper revisits RTO via modifier adaptation and shows that the modifier terms can be seen as linear approximations of the errors in the predicted cost and constraints. As a second contribution, this paper proposes to construct convex approxi- mations of the plant model to be used in the modifier-adaptation framework. It is shown that ∗ To whom correspondence should be addressed 1