딥러닝을 활용한 저압막여과 공정의 적정 플럭스 도출

Abstract

The low-pressure membrane filtration process includes microfiltration and ultrafiltration, which are mainly operated at pressures within 200 KPa. Membrane fouling reduces the operating efficiency of the low pressure membrane filtration process, so it is important to determine an proper flux to minimize that. The best way to determine the proper flux is to go through a long-term pilot test with one or two membrane modules applied in the field, which can take at least 6 months to over a year and pilot. There is a disadvantage that it is difficult to apply if the quality of the raw water in the water changes. In this study, by using the characteristic that fouling increases or decreases when the flux is increased or decreased, an algorithm is developed to lower the flux when the fouling index exceeds the limit value through artificial intelligence automatic control, and to increase the flux when the value is not reached. It was attempted to derive an proper flux. The reference value of the fouling index was calculated through the low-pressure membrane filtration model, and the flux fluctuation was made to find an appropriate value by using deep learning. The membrane filtration process operation result when the flux fluctuates was scored through a function of the fouling index, and the subsequent operation result when the artificial intelligence fluctuated the flux was feedback, and through this feedback result, the appropriate flux fluctuation value was found by itself. Finally, an proper flux suitable for operating conditions such as raw water quality was derived.