Distributed model predictive control based on the alternating direction method of multipliers for branching open canal irrigation systems

Abstract

Most studies of open canal automation concentrated on the control of single canal pool or cascaded multiple canal pools, neglecting the hydraulic coupling effects between the main canal and lateral canals. Low transmission efficiency and poor equity of agriculture water delivery may be caused in the practical application of large-scale irrigation networks. Considering the high computation and communication requirements for large-scale operating systems, a distributed model predictive control (DMPC) algorithm based on the alternating direction method of multipliers (ADMM) is developed for the branching open canal irrigation systems (BOCISs). A simplified BOCIS example, originating from Test case 2 proposed by the American Society of Civil Engineers, is used for control tests. Moreover, for control performance comparisons and evaluations, centralized model predictive control (CMPC) is employed by modifying the state space of the integrator delay (ID) model. The results show that the control performance of the proposed DMPC converges to the global optimal solution of the CMPC under the normal scenario, and the degradation is less than 0.35 %. The communication and coordination process of DMPC contributes to more powerful robustness and interference immunity than CMPC. Furthermore, in case of sudden accidents occurring in some subsystems, the proposed DMPC is more convenient and timely to execute fault isolation. The improvements on water level oscillations and overshoots can reach 20.3 % and 52.1 % in the other subsystems which need normal water delivery, respectively. Equipped with outstanding control characteristics, the proposed ADMM-based DMPC framework enables the water authorities of large-scale BOCISs to promote surface water distribution in a practicable, flexible, and secure way, showing great potential in developing intelligent irrigation districts.