Stochastic Design of Hydraulic Structures in Irrigation Canal Networks

Abstract

A methodology is presented for optimal design of hydraulic structures in a branched irrigation canal network subject to parametric and objective uncertainty. Monte Carlo simulation of steady spatially varied flow is incorporated into the Hooke and Jeeves direct‐search method to determine a high‐performance, low‐cost solution. Fuzzy membership functions address subjectivity associated with interpreting expected values of performance measures in light of water delivery objectives. Solution of an example problem yields an optimal design for 10 pipe diversion structures in a lined canal network. Water supply level, potential crop evapotranspiration, canal hydraulic resistance, and irrigation application efficiency were modeled as random variables. Results indicated that predicted performance for a given design can be subject to significant variability. Coefficients of variation in system performance measures for adequacy, efficiency, dependability, and equity of water delivery were 0.13, 0.04, 0.17, and 0.20, respectively, at the optimal design. The method describes variability in system performance and relative cost for non‐optimal designs as well. Such information is valuable for assessing the anticipated impact of alternative designs on overall system performance.