Genetic algorithm optimisation

A recent project to replace inefficient open channels in the Murrumbidgee irrigation area near Leeton and Griffith in NSW with a pressurised pipe network is expected to save up to 50,000 ML of water per annum when completed.

Optimatics was engaged by Murrumbidgee Irrigation (MI) to plan, design and optimise the system which involves more than 450 km of pipe covering 30,000 hectares.

During the project, hydraulic simulation models of 22 horticultural zones were developed and pipeline and pump station options determined. Data from MI's GIS database was used to identify boundaries and coordinates of each farm and develop demands for each customer based on farm areas.

Genetic algorithm optimisation was used to determine the best combination of options for each of the 22 horticultural zones in the project. Options included pipe locations and sizes, pump station locations and capacities and optimal farm outlet locations.

GA uses a 'survival of the fittest' search technique, to find true or approximate solutions to optimisation and search problems. The solutions are represented as genomes (or chromosomes) and the GA then creates a population of solution options represented by strings of numbers analogous to chromosomes.

Genetic operators such as mutation, selection and crossover/recombination are applied to evolve the solutions in order to find the best one(s). The process is similar to natural selection with the fittest solutions being the most likely to survive and breed to 'evolve' better and better solutions. Fitness is a measure of each solution's cost and hydraulic performance.

GA optimisation was first developed in the 1970s by Professor John Holland at the University of Michigan. It has since been applied to many industries to identify near-optimal solutions to complex engineering problems and can replace the simulation trial-and-error step in a traditional distribution system design project. GA optimisation was first applied to water systems by researchers at Adelaide University in the 1990s. Optimatics has been developing and using its own optimisation software since 1995.

The GA technique enabled Optimatics to evaluate the cost and hydraulic performance of hundreds of thousands of potential solutions and establish the lowest cost viable solution that meets the design criteria. Used for a project like this, it has the potential to save millions of dollars in capital and operating costs. Whole life cycles were taken into consideration in each optimisation run, which simultaneously considered hundreds of options for pipes, pumps and outlets for each zone.

The Murrumbidgee area is large and there were many variables and potential solutions. "The main challenge was the size of what we had to do," says Optimatics project manager, Alana Duncker. "Most clients have a model and are just looking for optimal system augmentations or a master plan design.

"With the Murrumbidgee project we had to start from scratch, gathering physical data to build their models and there were a lot of options to consider with hundreds of possible pipe routes." The location of where the farmer draws water from the system was also a decision variable which gave the optimisation an extra level of flexibility and scope to find better solutions.

Dozens of optimisation runs were carried out for each model area, producing many low-cost viable solutions for each zone. At least two optimised designs for each model area were selected and presented to MI.

Optimatics was recognised for the Murrumbidgee project at the Association of Consulting Engineers Australia (ACEA) 2006 Awards for Excellence, held in Sydney in December. The company was awarded a Gold Award of Merit in the Small Firm category and a Certificate of Recognition in the Water category.