Iterative learning predictive control for urban drainage systems

Urban drainage system is important for a modern city. This paper considers the control problem of an urban drainage system when a storm happens. By analyzing a storm, we see that the procedure of every storm is similar and its approximate procedure can be divided into three phases according to the change trend of rainfall. Therefore, we introduce the iterative learning control (ILC) into the urban drainage system and present a method of ILC based model predictive control (MPC). A 2D model based on a simplified model of drainage system is adopted. The procedure of a storm is divided into three phases and the ILC based MPC is designed for each phase by regarding it as a similar batch process. The handle of different length of phases is also considered. Since the learning based on the historic results is utilized, the proposed control method can improve the control result storm by storm. The numerical examples verify the effectiveness of the proposed design.