Design of irrigation system of canals using stochastic optimization approach.

Abstract

Design capacities of irrigation system canals determine the size and the cost of the system. Decisions to define the canal capacities depend on many uncertain parameters such as those describing the system operation and crop return. The most economical design can be obtained from an optimization model that accounts for these uncertainties while considering both the system design and operation. A two-stage stochastic problem with recourse is formulated in which the canal capacities are the first stage decisions and the water allocations are made in the second stage. This dissertation examines the two-state regularized stochastic decomposition algorithm and modifies it to better handle general engineering applications. It was used to solve for the optimal canal capacities under uncertain parameters present in the right hand side (RHS) of the second-stage constraints. The L-shaped method was then used to solve for these capacities under uncertain objective coefficients as well as RHS parameters. Obtained capacities from the stochastic approach were compared with results obtained from a traditional design procedure and from a deterministic optimization model in different conditions. The comparison demonstrates the merit of the proposed approach and points out the necessity to consider the parameters uncertainties when designing under certain conditions.