Computational-Based Framework for Optimizing Dynamic Processes with Plant-Model Mismatch

Abstract

A general computational sequence in optimizing the operation of a dynamic process is firstly highlighted in this paper. However, in most cases these dynamic processes include process-model mismatch, which shifts the optimal operation of the process. To overcome this, a model-mismatch estimator such as the neural network technique has been implemented in the optimization strategy. A modified general computational framework to incorporate these mismatches is developed for this purpose. The framework also allows the use of discrete process data in a continuous model to predict discrete and/or continuous mismatch profiles. The strategy is applied on a batch distillation system and the optimal operation using model mismatches is found to be comparable to that using the actual process model.