A Risk-Based Design of Water Distribution Systems under Uncertainty

Lead Research Organisation: University of Exeter
Department Name: Engineering Computer Science and Maths

Abstract

In real world water distribution systems (WDS) uncertainty can arise in a number of different ways. Variations in the performance of parts (for example pipe roughness) can affect the performance of the system. Uncertainty in the requirements the system must satisfy (such as demand at a node) will affect the ability of the system to meet those requirements. An algorithm which can reduce the number of fitness evaluations required to find performance probabilities for systems operating under uncertainty has the potential to significantly reduce computation times required for optimisation. Furthermore when system uncertainties include mechanical failures such as pipe bursts, blockages and leaks, costs can be associated with underperformance allowing such an algorithm to offer risk-based optimisations of systems by assigning an expected consequence of failure to each design. Such optimisations will find a family of solutions offering a trade-off between the cost of the system and the expected future costs or consequences due to failures and other uncertainties.The need for an optimisation technique which is not only capable of optimising systems under uncertainty, but is also scalable to large WDS is at the heart of the proposed research.This research project brings mathematical techniques for statistical sampling and evolutionary optimisation together with an engineering knowledge of the design of water distribution systems under uncertainty.
 
Description Water distribution systems (WDS) deteriorate as they age and so does their hydraulic and other performance. At the same time, customers and industry regulators often have increasing expectations in terms of the level of service provided. As a consequence, water companies have to periodically (re)design, i.e. rehabilitate their systems. The long-term (re)design of these systems is a complex task that requires minimising total capital and other costs involved whilst delivering the target system performance required. This task is further complicated by the fact that, due to urbanisation, climate and other future changes, a number of uncertainties (such as predicting future water consumption) exist in the long-term WDS planning process. Driven by the above, the overall aim of project was to explore how best to formulate and solve the optimal long-term WDS (re)design problem under future demand and other uncertainties. The research work explored several alternative approaches and resulted in the development of four new methods for solving the above problem in effective and computationally efficient way. Details of these methods can be found in the final project report. All new methodologies were tested and verified on both literature and real-life water distribution systems. The new knowledge gained is summarised in a series of papers published in leading, high impact, water engineering journals and presented at prestigious international conferences.
Exploitation Route The methods developed are generic hence could be potentially applied to solve different engineering and other problems, as long as these can be formulated as optimisation under uncertainty problems. Contact Prof. Zoran Kapelan (z.kapelan@exeter.ac.uk).
Sectors Environment

URL http://www.people.ex.ac.uk/zkapelan
 
Description The findings from this work have been used in my future work on various other projects. This has ultimately led to development of new technologies for water resources management under uncertainty and flood risk management under uncertainty, to name the few. For details please see www.people.ex.ac.uk/zkapelan.
First Year Of Impact 2012
Sector Education
Impact Types Societal