An efficient data-driven desalination approach for the element-scale forward osmosis (FO)-reverse osmosis (RO) hybrid systems.

Freshwater shortages are a consequence of the rapid increase in population, and desalination of saltwater has gained popularity as an alternative water treatment method in recent years. To date, the forward osmosis-reverse osmosis (FO-RO) hybrid technology has been proposed as a low-energy and environmentally friendly next-generation seawater desalination process. Scaling up the FO-RO hybrid system significantly affects the success of a commercial-scale process. However, neither the ideal structure nor the membrane components for plate-and-frame FO (PFFO) and spiral-wound FO (SWFO) are known. This study aims to explore and optimize the performance of SWFO-RO and PFFO-RO hybrid element-scale systems in the desalination of seawater. The results showed that both hybrid systems could yield high water recovery under optimal operating conditions. The prediction of the system performance (water flux and reverse salt flux) by artificial intelligence was considerably better (R > 0.99, root mean square error <5%) than that of conventional mass balance models. A Markov-based decision tree successfully classified the water flux level in hybrid systems. An optimal set of operational conditions for each membrane system was proposed. For example, in RO, a combination of the feed solution (FS) flow rate (≥17.5 L/min), FS concentration (<17,500 ppm), and operation pressure (<35 bar) would result in high water permeability (>40 LMH). In addition, five SWFO elements and four PFFO elements should be the optimal numbers of FO membranes in the hybrid FO-RO system for effective seawater desalination, especially for long-term operation.