Model predictive control of a reverse osmosis desalination unit

Abstract

Model predictive control (MPC) has established itself as the leading form of advanced control in the process industry. One of the most popular MPC strategies is the dynamic matrix control (DMC) algorithm. In this paper, the DMC algorithm with/without constraints is used for the control of a simulated reverse osmosis water desalination unit based on a hollow-fiber membrane. The controller was tuned using a recently presented simple method. Two cases were considered. In the first, the permeate flux was controlled by manipulating feed pressure. The second was a multivariable case in which both the permeate flow rate and conductivity (a measure of quality) are controlled by adjusting the feed pressure and pH. The performances of the DMC were analyzed and compared to those obtained from a control system based on classical PI controllers. The sensitivity of the DMC to changes in a key plant parameter, namely the gain-relating changes in the permeate flow rate to changes in the feed pressure, was also investigated. The results show that the DMC offers a substantially better performance than classical PI controllers. Also, it has good robustness characteristics. It continued to produce good control for even large changes, ±30%, in the gain relating the product flow rate to feed pressure.