Studies On The Optimization Of Reverse Osmosis Networks For Seawater Desalination

The reverse osmosis (RO) has been widely used for seawater and brackish waterdesalination. The main advantages of RO over other desalination processes are its nophase change, energy saving, modularity, etc. Therefore, RO developed rapidly andits market share increased recently. However, it is still necessary to develop asystematic optimization method for RO process. In this paper, studies on optimizationof RO networks for desalination of seawater and brackish water to reduce cost andpower consumption of desalted water are carried out.First, an optimization model of RO networks with pressure exchanger (PX) isestablished based on finite difference membrane transport model, and salinity increasecaused by volumetric mixing in PX is considered. The optimum design problem canbe solved as a mixed-integer nonlinear problem (MINLP) using outer-approximationalgorithm. Discrete numbers constraints are added to make sure the number ofmembrane elements in each PV and PVs employed in the RO stage are integers. Thevariation of salinity of feed seawater is studied using the RO super-structure model,and the maximum product concentration is500mg/L, the results show that when thefeed water salinity increases, the desalted water cost increases correspondingly, andsystem processes also change. For the feed seawater concentration higher than32000mg/L, single-stage RO system is favored. When the feed seawater concentration isbelow28000mg/L, two-stage RO system with inter-stage pump is the better choice.Different types of membrane modules are selected for different feed seawater salinity.The desalted water cost increases with the decreases of permeate concentrationrequirement. For the looser permeate concentration requirement range form300~500mg/L, single-pass configuration can meet the required quality of desalted water. Whenthe lower permeate quality requirement of concentration from50~100mg/L, two-passsystem is selected.Secondly, multi-objective optimization (MOO) of RO networks based on finitedifference membrane transport model is solved by ε-constraint algorithm, and non-inferior solutions (Pareto optimal set) are obtained. Decision-makers can chooseproper RO design configuration process for different design requirement. RO systems with or without ERD are optimized and compared as single objective optimizationusing desalted water cost and unit energy cost as the optimization objective,respectively. Two-stage RO system with inter-stage pump is favored for RO systemswith turbine or without energy recovery device (ERD). For RO system with PX, asimple single-stage system is favored. The desalted water cost of the RO system withturbine dose not decline much comparing with the system without ERD, and thedesalted water cost of the RO system with PX is the lowest. For a specific RO projectusing PX as the ERD, the MOO optimal process configurations achieve to producepotable water with a desalted water cost between US$0.52—0.69/m3and powerconsumption between2.30—2.79kW h/m3. The variation of the power consumptionis contrary to the desalted water cost. One-stage layout is proved to minimizesimultaneously the power consumption and the desalted water cost. With increasingthe feed water salinity, the pareto curve shift to the upper right, and the lowestdesalted water cost and the lowest power consumption of the RO system are bothincrease.Finally, a generalized disjunctive programming (GDP) approach is established forthe optimization of RO networks. The nonlinear model based on GDP is proposed thatrelies on the identification and application of mass balances and equilibrium equationsfor existence membrane modules in PV. For non-existing or inactive modules, theequations considered are simply input-output relations in which no mass transfer takesplace. RO stage takes similar treatment. A variant of the logic-based outerapproximation algorithm is proposed to solve the problem. It would be desirable notonly to reduce the size of the NLP sub-problems, but also to avoid infeasibilities thatare due to the non-existing RO stage and linearization at zero flows in the ROsuperstructure. RO seawater projects are studied using the proposed logic-based outerapproximation algorithm, and several factors such as electricity price, membranemodule price, average seawater temperature, etc, are concerned. The desalted watercost is very sensitive to electricity price. When the electricity price increases, thedesalted water cost can be decrease through increasing operating pressure orincreasing membrane area. The desalted water cost is more sensitive to changes inelectricity price than membrane price. If the seawater temperature increases, theoperation pressure required by the RO system decreases and the total membrane areareduces, so the desalted water cost decreases. Study on the optimization of RO networks is carried out, taking into account thesalinity increase caused by volumetric mixing in PX, and discrete numbers constraintsare added to make sure the number of membrane elements in each PV and PVsemployed in the RO stage are integers. Optimization researches show that the resultsin this paper could be utilized in the actual engineering design of RO process orsystem evaluation.