Author(s): Mahendra Kadu; Rajesh Gupta; Promod R. Bhave
Linked Author(s):
Keywords: Optimization; Genetic algorithm; Water distribution network; Water transmission systems
Abstract: Genetic algorithm (GA), in a simple or improved and modified form, has been used for providing a practically-favored single-size pipe optimal solution for the basic nonlinear water distribution network (WDN) optimization problem. The decision variables in design and expansion of water distribution networks are link diameters which are discrete in nature. The application of methodology can be extended to: design singe-level urban gravity water transmission system (WTS) in which water from the main balancing reservoir (MBR) is supplied to several other elevated service reservoirs (ESRs) located in different parts of a city; and also for regional rural water supply systems (RRWSSs) in which ESRs are located in different villages. However, in large RRWSSs, multi-level gravity WTS may be used which consists of a network from MBR to various village-ESRs with zonal balancing reservoirs (ZBRs) in between located at strategic locations. Herein, heights of the different ZBRs become another decision variable in the design. In this paper a methodology for design of multi-level WTSs is developed. Link diameters are considered as discrete variables and height of ZBRs are considered as continuous variables. The modified Genetic Algorithm with reduction in search space using critical path method (Kadu, Gupta and Bhave 2007) and additional GA operators to take care of continuous variables is suggested for optimal design of multi-level WTSs. The methodology is applied to an illustrative example and the singlesize GA solution is compared with LP split-pipe solution. The comparison shows that the modified GA methodology coupled with critical path method is effective and efficient in identifying optimal solution along with a set of near-optimal solutions. The GA provides a wide choice to the field engineers to select an appropriate design by easily comparing the near-optimal solutions.
Year: 2007