Comparison of simplified methods for the optimisation of water distribution systems

Abstract

Evolutionary algorithms have been applied for the last 20 years to find near-optimal designs of water distribution systems (WDS). However, they suffer the disadvantage that they require a significant amount of computer time, especially for water distribution systems of a realistic size. In this paper various simplified methods to optimise the design of WDSs are compared. The methods start by reducing the full water distribution network to a tree network using the shortest path or bulk flow methods. Pipes are then sized using non-linear programming. The optimal power use surface methodology is also tested. These methods are compared on two case studies, namely the Hanoi and Balerma problems. The two networks are characterized by different numbers of pipes and reservoirs. Results show that the simplified methods are relatively fast in finding solutions with a cost 5-10% larger than the best known solution. Although it is likely that further improvements are possible, this result offers an acceptable approximation for real problems.