Simultaneous optimization of size and operation for the seawater reverse osmosis (RO) system is proposed through a superstructurebased mixed-integer differential-algebraic programming approach. The influences of operating conditions on the performance of pumps, RO membrane, and the properties of seawater are considered. The permeate split (PS) design, interstage permeate split (ISPS) design, and on−off model could give more operability and cost saving. For constant feed conditions, nearly constant profiles of control variables are obtained with a larger water tank. The maximum level in a water tank is an indicator to measure its buffer capacity. For the variation of temperature and time-of-use electricity price, although energy consumption can be effectively saved by one time shut off at a large water demand time period with high electricity price, not much water cost is reduced due to the increase of the plant size. Low load operation is an alternative solution. The ISPS design could balance the flux distributions inside the pressure vessel and extract more permeates with high quality; thus, the water cost and energy consumption could be reduced, compared with the interstage design (ISD) and the PS design. The uncertainties of the maximum concentration of product water and the membrane fouling factor are considered in a twostage stochastic program. It is useful for identifying a robust optimal design and operation solutions for membrane-based seawater desalination.