Considerations on multiobjective optimization regarding municipal water supply systems

Date

2020-9

Author

Bektaş, Hayrettin Onur

Metadata

Show full item record

Item Usage Stats

17
views

18
downloads

Optimal water supply system design has always been a challenge for researchers. While most optimization studies focus mainly on pipe diameter sizing, recent works are being done for optimizing other elements of water supply systems. The design of water supply system can be discussed from either single or multiobjective optimization point of view. Multiobjective optimization does not only consider the costs of the system but also some performance measures such as resilience. In this study, a multiobjective optimization methodology is developed by using non-dominated sorting genetic algorithms (NSGA-II). Consequently, Pareto trade-off curves for cost vs. network resilience (In) are obtained. Different sets of solutions are formed with different population sizes; then performances are measured by hypervolume calculations to determine optimum population size for the system considered. Finally, one solution is selected within the pareto curve by compromise programming to offer a design to the decision maker. Shamir network, Hanoi network and a real distribution network from Ankara are studied as case studies. It is concluded that constraints play a decisive role on pareto curves in the methodology. When number of constraints are increased, final pareto curve shifts away from the true pareto curve. Moreover, it is observed that, placing storage tanks in the system have less effect on the network strength when compared to diameter changes. It is also concluded that a storage tank count limitation is present in Shamir and Hanoi networks. After this case specific limit, it becomes very hard to place another feasible tank to the network and final pareto curves consist of very few points.

Subject Keywords

Multiobjective Optimization, NSGA-II, Network Resilience, Optimum Water Distribution System Design, Ankara

URI

https://hdl.handle.net/11511/69040

Collections

Graduate School of Natural and Applied Sciences, Thesis