Neural-Network Approach to Dynamic Optimization of Batch Distillation: Application to a Middle-vessel Column

Abstract

A framework is proposed to optimize the operation of batch columns with substantial reduction of the computational power needed to carry out the optimization calculations. The proposed framework relies on the use of an artificial neural network (ANN) based process model to be employed by the optimizer. To test the viability of the framework, the optimization of a pilot-plant middle-vessel batch column (MVBC) is considered. The maximum-product problem is formulated and solved by optimizing the column operating parameters, such as the reflux and reboil ratios and the batch time. It is shown that the ANN based model is capable of reproducing the actual plant dynamics with good accuracy, and that the proposed framework allows a large number of optimization studies to be carried out with little computational effort.