Improved Directional Derivatives for Modifier-Adaptation Schemes

MARTAND SINGHAL; ALEJANDRO G. MARCHETTI; DOMINIQUE BONVIN; TIMM FAULWASSER

Toulouse

Congreso; IFAC World Congress 2017; 2017

International Federation of Automatic Control (IFAC)

Modier adaptation enables the real-time optimization (RTO) of plant operation in the presence of considerable plant-model mismatch. For this, modier adaptation requires the estimation of plant gradients, which is experimentally expensive as this might involve several online experiments. Recently, a directional modier-adaptation approach has been proposed, which uses the process model to compute offline a subset of input directions that are critical for plant optimization. This allows estimating directional derivatives only in the critical directions instead of full gradients, thereby reducing the burden of gradient estimation. However, in certain cases such as change of active constraints and large parametric uncertainties, directional modier adaptation may lead to signicant suboptimality. Here, we propose an extension to directional modier adaptation, whereby, at each RTO iteration, we compute a set of critical directions that are robust to large parametric perturbations. We draw upon a simulation study of the run-to-run optimization of the Williams-Otto semi-batch reactor to illustrate the performance of the proposed extension.