MODELING, OPTIMIZATION, AND CONTROL OF REVERSE OSMOSIS WATER TREATMENT IN KAZEROON POWER PLANT USING NEURAL NETWORK

ABSTRACT

Neural network modeling and the back-propagation concept were utilized to develop data-driven models for predicting reverse osmosis (RO) plant performance and finding control strategies. Considering different commissioning times, the process of three RO plants was successfully modeled using an artificial neural network (ANN). Moreover, long-term forecasting of performance degradation was developed. Time (h), transmembrane pressure (TMP; bar), conductivity (µs/cm), and flow rate (m³/h) were utilized as ANN inputs. The effects of operating time and TMP on performance at mean values of feed conductivity and flow rate were investigated using three-dimensional figures. Genetic algorithm (GA) was employed to find optimum paths of TMP, feed flow rate, and control strategies during a specific period of time. The RO plant was monitored for 5000 h corresponding to the results generated by GA (optimum paths), and experimental results were compared to the prediction made by the model. The differences strongly implied the robustness of the ANN model.

Keywords:

• Artificial neural network
• Control strategies
• Genetic algorithm
• Modeling
• Reverse osmosis

Notes

Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/gcec.