Optimisation of irrigation infrastructure rehabilitation using genetic algorithms

Abstract

Techniques that optimise the design and operation of water distribution systems have taken a long time to make the transition from theory to practice. There are several reasons for this including the size and complexity of existing systems, the vast number of possible solutions and the wide range of system-specific design considerations, performance requirements and operational limitations encountered in practice. This paper describes the application of a class of powerful search and optimisation techniques called genetic algorithms to the optimisation of the design of piped irrigation delivery systems. The genetic algorithm (GA) search is a computer-simulated evolution process that imitates nature's ongoing search for better solutions. In the application to pipe network optimisation, the GA searches for near-optimal cost networks or network expansions which meet specified design criteria and performance requirements for a system subject to defined future demand patterns. The GA computes the actual cost of proposed designs based on commercially available component sizes and uses an efficient, reliable hydraulic simulation model to check the hydraulic feasibility of trial network solutions. The GA search method is robust and is well-suited to the design of complex engineering systems. The formulation of the GA model is very flexible and the GA can be used to optimise any system variable that is input to the hydraulic simulation model. This may include pipe network layout, pipe sizes, tank locations and pump sizes. It is possible to incorporate any type of cost or objective function and any performance constraint that may be imposed on the system (and evaluated by hydraulic simulation), no matter how specific. A case study is presented to demonstrate the versatility of the GA technique and the significant cost savings that may be achieved by the application of this technique to a practical irrigation network design problem. The case study involves the rehabilitation of an aged pressure pipe system (constructed in the 1920s) in the Loveday Division of the Cobdogla Irrigation Area in the South Australian Riverland. The design achieved by the GA search saved 11.0% of the estimated cost for the supply and construction of pipelines, compared to the design determined by the conventional design approach based on experience and the trial-and-error application of a hydraulic simulation package.