Submitted to: United States Committee on Irrigation and Drainage Conference
Publication Type: Proceedings
Publication Acceptance Date: 3/1/2003
Publication Date: 5/15/2003
Citation: Wahlin, B.T., Bautista, E. 2003. Feedforward control with anticipation: volume compensation versus model predictive control. United States Committee On Irrigation And Drainage Conference. p. 487-496. Interpretive Summary: Water is becoming a scarce resource, and irrigation districts are under pressure to use water more effectively. Improved canal operational procedures are needed by districts to reduce their spills and to deliver water more accurately to farmers. It is difficult for farmers to adopt modern irrigation technologies and management strategies and to use the water resource efficiently when deliveries to the farm do not match their water demands. This study compares two methods for scheduling the operation of canal structures with the objective of delivering known water demands through a network of canals. Results show that both schemes are able to deliver the requested flows while maintaining water level fluctuations within safe levels. These results will be of use to irrigation districts and consultants.
Technical Abstract: Anticipatory feedforward control of irrigation water delivery systems refers to control strategies designed to route known demand changes through a canal system. While canal operators in many parts of the world use feedforward control, strategies in use generally have been developed based on experience and not on hydraulic principles. Heuristic approaches can be effective, but only when dealing with delivery systems subject to few flow changes. More sophisticated approaches are needed to deal with systems with many customers and in which very flexible water deliveries are required. Over the past few years, researchers at the U.S. Water Conservation Laboratory (USWCL) have developed an anticipatory feedforward routine based on volume compensation principles. This open-loop controller is based on two simple concepts: (1) determining the change in pool volumes needed to take the canal system from the initial steady-state to the final steady-state and (2) determining the time needed by the volume changes to travel down the canal (i.e., the pool delays). In the process control industry, Model Predictive Control (MPC) has rapidly gained popularity. Although MPC is typically thought of as a feedback controller, it has feedforward capabilities built into it. To utilize MPC¿s anticipatory feedforward control mode, the known flow changes are incorporated into MPC¿s predictions of the system output. In this way, MPC anticipates the flow changes coming and moves the gates to route the flow through the canal system. This paper compares the volume compensation and MPC feedforward control methods. The performance of these open-loop controllers was examined through hydraulic simulations of the Upper Arizona Canal operated by the Salt River Project (SRP). Both feedforward routines performed well on the proposed test. The volume compensation method slightly out performed the feedforward component of MPC. The performance of MPC is more a function of the underlying process model than of MPC itself.