On-line optimal control of ethanol production in a fed-batch reactor by using neural network estimator

Abstract

This research presents the implementation of an on-line optimal control with neural network estimator to control a fed-batch reactor for the production of ethanol. Due to the presence of unknown disturbances and model-plant mismatches, an open-loop optimal control may not give the optimal performance when applied to the actual process. To improve the control performance, an on-line optimal control is developed to modify the optimal feed profile of a fed-batch reactor whenever feedback information of the system is available. In this work, the formulated optimal control problem is solved by a sequential method in which the control profile is parameterized by using a piecewise constant function. Artificial neural network is used to estimate unmeasured state variables which are employed as feedback information of the system. The ethanol fermentation process by Saccharomyces cerevisiae in a fed-batch reactor is chosen as a case study to demonstrate the proposed control strategy. The simulation results have shown that the on-line optimal control with neural network estimator gives a better control performance in terms of the amount of the desired ethanol product, compared with the off-line optimal control