Development of Data-Driven Methods for Performance Prediction of Reverse Osmosis Desalination

Reverse Osmosis (RO) desalination systems account for a significant share of the worldwide desalination technology market. The performance of these systems is affected by a wide range of operational, technical, and environmental factors, making their modeling challenging due to their dependence on multiple parameters. In this respect, recent developments in machine learning and artificial intelligence have greatly enhanced the modeling and optimization of complex energy and water systems, providing powerful data to enhance prediction accuracy. Machine learning approaches offer promising tools for accurate modeling and performance prediction of such complex systems and have been widely applied to different types of desalination systems. This study aims to evaluate different data-driven methods, namely the Group Method of Data Handling (GMDH) and Adaptive Neuro-Fuzzy Inference System (ANFIS), for modeling an RO system based on the data provided by an experimental study. The inputs of the proposed model are feed pressure, feed concentration, and temperature as the key effective factors on the performance of RO desalination systems, while the output is the permeation rate. To enhance the accuracy of the ANFIS-based model, two optimization algorithms, namely Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), were employed to determine the parameters of the membership functions of the ANFIS method. It was found that all the developed models had high precision in modeling the considered systems. Among the developed models, GMDH had the highest prediction accuracy, followed by the PSO-ANFIS and GA-ANFIS, respectively. The values of R۲ for the models based on GMDH, PSO-ANFIS, and GA-ANFIS were ۰.۹۹۹, ۰.۹۹۷, and ۰.۹۹۳, respectively. Average Absolute Relative Deviation (AARD) of the mentioned models for the overall data was ۳.۵۸%, ۴.۵۵%, and ۱۰.۷۸%, respectively.

کلیدواژه ها:

Reverse osmosis desalination ، Machine Learning ، ANFIS ، GMDH