Distribution System Optimization Becoming More Powerful, More Popular
The use of optimization analysis to identify superior low-cost water distribution planning and operating solutions is becoming more common. Several water utilities in Texas and California have gone so far as to specify in their requests for proposals that a formal optimization technique be used by the consultant team in preparing comprehensive water distribution master plans.
As interest in optimization analysis increases, the developers of genetic algorithm (GA) pipe network optimization at Optimatics Pty Ltd. in Adelaide, Australia are applying innovative advances that greatly improve the efficiency of the GA search algorithm. The firm's ongoing research is aimed at advancing the methods by which the desirability or fitness of network solutions is evaluated, the way network solutions are represented as individuals in the artificial evolution, and the way new and better populations of solutions are generated from old populations.
The use of various "elitist" concepts is one such advance. By maintaining a population of elite solutions in parallel with the working population, successive working populations have more freedom to explore and the GA search is able to narrow in on near-optimal solutions much faster. This innovation can also help to identify solutions having even lower costs than previously found for certain types of problems, as illustrated in the accompanying plot.
The GA optimization analysis technique has been used by Optimatics and Frey Water Engineering, Inc. of Arlington Heights, Illinois since 1993 to identify viable low-cost solution alternatives for water distribution planning, design and operations problems. It has been proved on numerous studies in the U.S. and Australia, including an expansion plan review for the Fort Collins-Loveland Water District in Colorado. In that study, use of the procedure eliminated 10.6 miles of proposed new pipe from the recommended year 2015 expansion solution, reducing projected capital costs from nearly $6 million to just under $3 million.
In a project employing the technique, the analysis starts by reading in data from an existing hydraulic simulation model, along with the allowable locations, sizes, settings and costs for new or replacement pipes, tanks, pumps and valves. The program then automatically generates several hundred thousand trial solutions as it searches for the optimal combination of components that achieve the specified hydraulic performance at lowest possible cost. These trial solutions are generated in a very efficient manner using the rules of natural selection. The preferred solution alternatives can then be verified by inputting them into the original hydraulic simulation model.
The figure below illustrates solution search progress as the GA narrows in on the optimal solution using three different GA search methodologies.
Editor's Note: This article was provided by Jeffery Frey, president of Frey Water Engineering, Inc., Arlington Heights, Illinois, USA. He can be contacted at: Tel. 847-670-7970; Fax. 847-670-7973.