Evaluating utilization of structures' settings of one reach in the others

Controlling structures in irrigation canals to accurately deliver and distribute the water, and to keep it needs the appropriate control techniques. Sarsa reinforcement learning, as a branch of artificial intelligence, has recently been used to control the structures and improve water delivery and distribution in irrigation canals. To improve Sarsa efficiency and reduce the required time of operational pattern learning, the Sarsa algorithm in the E1R1 canal was developed and linked to a non-linear model of the canal to learn the operational pattern of one reach of the canal and apply the results to the other reaches. Operational scenarios were defined in this regard, and standard performance indicators was used for assessment. The results showed that Sarsa can be used successfully with the proposed idea, maintaining water depth within a dead band of 5% in the learning step and that of 10% while utilizing the learning results. The efficiency and adequacy indicators were close to the desired value.