Online optimization of water desalination by synthesizing a flexible RO system

Abstract

Fresh water is indispensable to survive, but her distribution around a word is not uniform in result of that 40% of the world’s population is suffering from a serious shortage of water. As nearly 97,5 % of water is salty, is a natural step to try to extract drinkable water from salt water. This process is known as desalinization. Many industrial ways exist to desalinate sea or brackish water. The process that seems to be the best way of doing that is application of reverse osmosis. It is verified that this process takes less energy, occupies less space, is more flexible to modifications ( is just to add new modules, depends of needs – scaling up is direct ) and does not involve vaporizing the water, which gives better operation efficiency than other desalination technologies. Membrane performance can be difficult to model from the beginning, so it is important to create models that predict the performance of membranes under various operating conditions. Also dynamic simulators are particularly useful in studying plant controllability. Create such model is objective of this thesis. The model is based on mixed-integer nonlinear programming (MINLP) whose objective is to minimize the total annual cost, while incorporating thermodynamic, technical and flexibility constraints. The cost includes cost of equipments operating ( energy needs ) and cleaning cost of membrane fouling. Acquisitions’ cost is not included because the model is made to optimize the already existing desalinization plant. The software used is Matlab. This choice has been made because Matlab is a wide-spread, user-friendly, reliable software that embeds a lot of methods, including optimization and fitting methods. The construction of the model was successful. On case study, by using a model, the cost was reduced in 13 %. To check the behaviour of the model in multiple conditions various sensibility studies were done. The model is pretty flexible and adaptable. The great strength of the model is that it works on real data, thanks to data acquisition, and simulate on most possible way the real process, which includes an analyse of membrane fouling and concentration polarization problems. Also in the end, user receives very important information about optimal parameters for plant operating and when to clean the membranes.