Adaptive Learning of Hybrid Models for Nonlinear Model Predictive Control of Distillation Columns

Abstract Our previous work has shown that replacing parts of the classical compartmentalization model reduction approach for distillation columns by offline-trained artificial neural networks (ANNs) improves computational performance. In real-life applications, absence a high-fidelity data generation can, however, prevent deployment this approach. Therefore, we propose method utilizes solely plant measurement data, starting from small initial set and then continuously adapting to newly measured data. We demonstrate in closed-loop simulations compare benchmarks using either or an offline trained reduced control.