Scheduling of Water Distribution System Rehabilitation Using Structured Messy Genetic Algorithms Driss Halhal Godfi-ey A. Walters zyx Water and Electricity Distribution Co. School of Engineering (RAID) and Computer Science 5 Rue Okuba Ibn Naffy University of Exeter BP 286, Tangier, Morocco Exeter EX4 4QF, UK G.A.Walters@exeter.ac.uk Dragan A. Savic School of Engineering and Computer Science University of Exeter Exeter EX4 4QF, UK D.Savic@exeter.ac.uk Driss Ouazar Hydraulic Systems Analysis Laboratory Mohammadia School of Engineers (EMI) BP 765, Agdal, Rabat, Morocco ouazar@emi.ac.ma Abstract A methodology is presented for the optimal design and scheduling of investment for the rehabilitation of water distribution networks. Based on the evolutionary programming technique known as Structured Messy Genetic Algorithms, the methodology utilizes a multi-objective formulation which improves the evolutionary process and provides non- dominated optimal solutions over a range of costs and benefits. The model is applied to an example-a small artificial network of fifteen pipes. The effects on the optimal solutions of varying parameters such as interest rate and inflation rate are also investigated. Keywords Water distribution, genetic algorithms, scheduling, multi-objective, optimization, rehabil- itation, networks. 1 Introduction Drinkmg water distribution networks are essential and expensive components of the infras- tructure of all urbanized areas. Most systems have been developed over a period of time, and much of the pipework is of considerable age. Pipes and fittings gradually deteriorate, with internal corrosion and depositions causing loss of carrying capacity and a consequent in- crease in pumping pressures and energy costs, pressure fluctuations and inadequate pressure at customers’ taps. The high cost involved in remedial works, together with budget restrictions applied by water utilities, make phasing the only financially practicable rehabilitation strategy in most cases. It is therefore sensible that the money available be optimally invested over a period of time taking into account both the physical response of the system with time and the influence of financial factors such as inflation and interest rate. 01999 by the Massachusetts Institute of Technology Evolutionary Computation zyxw 7(3): 31 zyxw 1-329