TRADING OFF RELAIBILITY AND COST IN OPTIMAL WATER DISTRIBUTION SYSTEM DESIGN

摘要

Most design engineers would list “reliability” as the primary objective in water system design while most optimization models focus on “least cost” as the primary objective. Of course, both objectives are important and the key to successful design lies in effectively trading off both objectives. rnLeast-cost design methods, using a single (or handful of) design scenario(s) to evaluate the hydraulic performance of a solution do not work for evaluation of reliability because there are numerous modes of failure for each solution trial. Within reason, each of these solutions needs to be evaluated for multiple loadings and important elements being taken out of service. rnFor any of the solution methods investigated, there are tradeoffs between computational speed, range of alternatives considered and thoroughness of design. The key is to not evaluate every possible failure in the system but to select a reasonable set of failure scenarios corresponding to various problems and objectives in the system. rnIt is not reasonable to expect the same piping project will solve problems with rapid growth in the south of town, poor fire flows near school in the west, lack of storage in the north, need for a new pressure zone to the east and serious tuberculation in the old central part of town. Similarly, it is not reasonable to expect a single model run should solve all five problems. Instead, design engineers need to pose the optimization model in a way that reasonably manages the dimensionality of the design problem and allows the optimization to find low cost and reliable design solutions. rnThis paper evaluates several alternative approaches for trading off reliability and costs in a genetic algorithm optimization model for a real system.