Advanced hybrid approaches based on graph theory decomposition, modified evolutionary algorithms and deterministic optimisation techniques for the design of water distribution systems.

Abstract

The cost of water distribution system (WDS) design or rehabilitation is normally expensive. Over the past 40 years, a number of optimization¹ techniques have therefore been developed to find optimal designs for WDSs in order to save costs, while satisfying the specified design criteria. Often there are a large number of decision variables involved. The majority of currently available optimization techniques exhibit limitations when dealing with large WDSs. Two limitations include (i) finding only local optimal solutions and/or (ii) exhibiting computational inefficiency. The research undertaken in this dissertation has focused on developing advanced optimization techniques that are able to find good quality solutions for real-world sized or large WDS design or rehabilitation strategies with great efficiency. There were three objectives for the research: (i) the modification and improvement of currently available optimization techniques; (ii) the development of advanced hybrid optimization techniques (evolutionary algorithms combined with traditional deterministic optimization techniques) and (iii) the proposal of novel optimization methods with the incorporation of graph decomposition techniques. The most novel feature of this research is that graph decomposition techniques have been successfully incorporated to facilitate the optimization for WDS design. A number of decomposition techniques have been developed to decompose WDSs by the use of graph theory in this research. Real-world sized or large WDSs are used to demonstrate the effectiveness of the proposed advanced optimization techniques described in this thesis. Results show that these advanced methods are capable of obtaining sound optimal solutions with significantly improved efficiency compared to currently available optimization techniques. The main contribution of this thesis is the provision of effective and efficient optimization techniques for real-world sized or large WDS designs or rehabilitation problems. ¹American spelling has been used in this thesis as all the publications included in this thesis have been submitted to or published in American journals.