In this paper an adaptive expert generalized predictive multivariable controller (AEGPMC) is developed for control of a seawater reverse osmosis desalination plant, which supplies freshwater for a mineral processing facility. Based on the use of subspace system identification algorithms a dynamic multivariable model, which considered the permeate flow rate and the permeate conductivity as the controlled variables, is derived and its parameters are computed from experimental data. This model is used in the design of the AEGPMC. The obtained simulation results demonstrate that the performance of the control system with the designed AEGPMC is much more appropriate and robust than with the other multivariable GPC.