Chance-Constrained Genetic Algorithm for Water Supply and Irrigation Canal Systems Management

Large and complex water supply and irrigation canal networks exist in arid regions of the United States, as well as in many arid locations in other countries. It is necessary to establish management strategies for the operation of a water supply and irrigation canal network to meet the water demands at various locations in the network over time without causing problems such as water shortage or flooding in the areas served by the network. A previous study by El Gamel and Harrell demonstrated that by incorporating an unsteady flow model into a genetic algorithm-(GA-) based framework, efficient strategies for scheduling gate operations and allocating crops can be determined to address the various objectives (such as profits from crops and costs of damages due to flooding and water shortages) and constraints (such as maintaining stability of gates and adequate water levels at water supply intakes) over a period of time using deterministic models. However, the actual values of many of the input parameters for this type of problem are not known with certainty due to variability over space and time, as well as lack of information. The GA framework is flexible to incorporate estimates of reliability to search for efficient solutions that satisfy the constraints with a specified likelihood under conditions of uncertainty. Because a GA evaluates the fitness of each potential solution, it is a straightforward extension to evaluate each solution repeatedly using a set of Monte Carlo (MC) realizations of uncertain parameters generated based upon their probability distributions.