Advanced control of a reverse osmosis desalination unit

Abstract An experimental investigation of constrained model predictive control (CMPC) for a reverse osmosis (RO) desalination unit has been conducted. For comparison purposes, results with traditional PID-type control have also been obtained. The experimental unit consists of a series of four cellulose acetate membranes. A 486-PC is used as the data acquisition and control computer. It is interfaced to the experimental unit via analogue-to-digital and digital-to-analogue converter boards. The models required for CMPC and PID-type controls are obtained by step testing. The RO system has four outputs and two inputs. The outputs are (1) permeate flow rate, (2) permeate conductivity, indicative of the salt content in the product, (3) trans-membrane pressure, and (4) inlet pH. The inputs are (1) flow rate of reject water and (2) inlet acid flow rate. The production objectives are to produce the specified flow rate of permeate, having the desired salt content, subject to the constraints that the inlet pH and the trans-membrane pressure are within specified bounds. It is shown that CMPC can achieve these goals. It is also demonstrated that CMPC can maximize the throughput subject to the constraints on the other three outputs. A comparison of the results with CMPC and PI control reveals the excellent capability of CMPC for RO desalination plant operations.