Abstract
A full-scale Pressure Retarded Osmosis process (PRO) is optimized in non-ideal operating conditions using Grey Wolf Optimization (GWO) algorithms. Optimization process included the classical parameters that previous studies recommended such as operating pressure, and feed and draw fractions in the mixture solution. The study has revealed that the recommended operating pressure ΔP = Δπ/2 and the ratio of feed or draw solution to the total mixture solution, ̴ 0.5, in a laboratory scale unit or in an ideal PRO process are not valid in a non-ideal full-scale PRO module. The optimization suggested that the optimum operating pressure is less than the previously recommended value of ΔP = Δπ/2. The optimization of hydraulic pressure resulted in 4.4% increase of the energy output in the PRO process. Conversely, optimization of feed fraction in the mixture has resulted in 28%–70% higher energy yield in a single-module PRO process and 9%–54% higher energy yield in a four-modules PRO process. The net energy generated in the optimized PRO process is higher than that in the unoptimized (normal) PRO process. The findings of this study reveal the significance of incorporating machine-learning algorithms in the optimization of PRO process and identifying the preferable operating conditions.