University of Tehran Press Iranian Journal of Soil and Water Research 2008-479X 46 2 2015 06 22 Application of Reinforcement Learning Algorithm for Determining the Operational Instructions of the On-Request Method for Optimal Water Distribution and Delivery Application of Reinforcement Learning Algorithm for Determining the Operational Instructions of the On-Request Method for Optimal Water Distribution and Delivery 283 291 55933 10.22059/ijswr.2015.55933 FA Kazem Shahverdi Ph. D. Candidate, Dept. of Water Structure Engineering, University of Tarbiat Modares 0000-0001-8098-0931 Mohammad Javad Monem Associate Professor, Dept. of Water Structure Engineering, University of Tarbiat Modarres 0000-0001-7735-9976 Majid Nili Professor, Electrical Engineering, University of Tehran Journal Article 2014 09 15 The on-request system is considered as one of the effective water distribution and delivery systems. It can be applied to currently available irrigation networks but, the main challenge for its application is the extraction and provision of appropriate operational instructions. The main objective followed in the research the development of Fuzzy Sarsa reinforcement Learning (FSL) model for extracting operation al scheduling for the on request irrigation systems. The FSL is to be evaluated in the E1R1 canal of Dez network. Requested discharges are the input of the algorithm and the output comprised of the optimum operational instructions. Water depth and flow performance indicators were made use of for an evaluation of the two performed scenarios. In scenario No. 1, as an exemplary sample, in which turnouts No. 5 and 6 demands increase from 0.1 to 0.2 m3/s while the other turnouts are closed, the minimum values of efficiency and adequacy indicators were recorded as 0.989 and 0.994; and while maximum and average values of water depth deviations being  obtained 8.4% and 7.4%, respectively. Considering the results, FSL can be applied as manual adjustment of the structures available on the present irrigation networks for a determination of the operational instructions. The on-request system is considered as one of the effective water distribution and delivery systems. It can be applied to currently available irrigation networks but, the main challenge for its application is the extraction and provision of appropriate operational instructions. The main objective followed in the research the development of Fuzzy Sarsa reinforcement Learning (FSL) model for extracting operation al scheduling for the on request irrigation systems. The FSL is to be evaluated in the E1R1 canal of Dez network. Requested discharges are the input of the algorithm and the output comprised of the optimum operational instructions. Water depth and flow performance indicators were made use of for an evaluation of the two performed scenarios. In scenario No. 1, as an exemplary sample, in which turnouts No. 5 and 6 demands increase from 0.1 to 0.2 m3/s while the other turnouts are closed, the minimum values of efficiency and adequacy indicators were recorded as 0.989 and 0.994; and while maximum and average values of water depth deviations being  obtained 8.4% and 7.4%, respectively. Considering the results, FSL can be applied as manual adjustment of the structures available on the present irrigation networks for a determination of the operational instructions. https://ijswr.ut.ac.ir/article_55933_d4f0886cfb50e0fbf8f0eebddf36b27d.pdf