An artificial neural network‐based optimization of reverse electrodialysis power generating cells using CFD and genetic algorithm

Reverse electrodialysis (RED) is a renewable energy production method that employs salinity gradient to produce electricity. The salinity gradient between the rejected brine of desalination process and river water/seawater is a reliable source of energy, particularly for desalination plants located in susceptible areas. In this study, the performance of RED is predicted using computational fluid dynamics and an artificial neural network. This approach reduces the computational costs of optimization, and more importantly, networks can be updated by more data in the future. Since geometric, hydrodynamic, and electrochemical variables affect the performance of these cells, ignoring any of them will influence the final design. We can consider all of these factors through deep learning. Performance parameters such as Sherwood number, Power number, and concentration polarization coefficient are evaluated in this study. Mass transport and pressure drop are optimized using genetic algorithm, and accessible electrical power is obtained for the optimized cases that help designers make final decisions. Using predictors and a set of optimized cases provide an efficient tool for the design. Based on our results, RED cells can produce net power density of 2.4 W m−2 by using rejected brine of desalination and river water as the two solutions. In addition, Sherwood number of 80 and Power number of 5248 show a good balance between the amount of mass transfer and pressure drop in RED cells.