Evolutionary calibration of downstream canal control algorithms

Increasing the delivery efficiency of an irrigation system is becoming more important everyday. The limited water resources and increasing demand require us to be more scrupulous with every drop of water. Automatic control is often a viable alternative to complete rehabilitation of water delivery systems. This paper introduces an automatic control algorithm that restores the canal conditions by compensating for both flow changes and volume changes in the system. The control algorithm is a multi input multi output downstream canal controller. The controller is tested using the guidelines suggested by the ASCE task committee on canal automation. The test cases were simulated using CANALMAN, which is a one-dimensional dynamic model. The controller performed better in canals with milder slopes compared to steeper canals. Not only is the controller still adequate for most canals, but also it introduces a more rule-based logic. The rule-based logic is more intuitive and easier to understand by water managers. The controller introduced here was calibrated using evolutionary algorithms, which are robust search engines that permit handling more complex optimization problems. The results indicate the potential for more rule-based algorithms. The paper conclude by recommending further exploitation of evolutionary methods through applying learning classifier systems in canal control problems.