Constrained Model Predictive Control Algorithm for Cascaded Irrigation Canals

Abstract

Agricultural irrigation accounts for the largest proportion of freshwater use worldwide, and canal automation potentially improves conveyance efficiency in irrigation canal systems. In this paper, model predictive control (MPC) for a cascaded irrigation canal system was formulated using the integrator-delay model. Magnitude and variation amplitude constraints on input and output imposed on the canal operation were identified along with proposed handling methods, and optimal control actions were achieved by quadratic objective function optimization. The MPC, as well as classical proportional-integral (PI) and centralized linear quadratic (LQ) for comparison, were developed for the Changma South Irrigation District cascaded irrigation canals in Gansu Province (China) and numerically tested via SOBEK software. In contrast to the poor performance of PI and LQ in controlling the studied canal, the results show that MPC can efficiently control the canal system under known demand changes and maintain water levels at control points within the operating range. The control performance improves if normal input and output constraints are incorporated in optimization. However, deadband constraints, which is the minimum variation amplitude of input, cause controlled water levels to oscillate around the reference value and degrade the control performance. In summary, MPC can cope with time delays, coupling effects, and constraints inherent in a cascaded irrigation canals system. Furthermore, it is suggested to evaluate more advanced methods for handling the output and deadband constraints in future studies.