| dc.description.abstract | A water distribution network (WDN) supplies drinking water to homes and businesses, and links water sources to consumers. Such networks are typically complex and dynamic, consisting of thousands of nodes interconnected by thousands of elements. Control and management of a WDN as a whole is challenging as there is no granular information about different parts of the network to support location of issues. Partitioning a WDN into smaller sectors is a strategy to manage its complexity, as advised by the International Water Association. Advantages include (a) enhanced leakage and burst detection and management, (b) improved contamination spread control (associated with water security), (c) a capacity to provide different pressure levels, (d) enhanced demand management, and (e) better rehabilitation and work planning. Water security (i.e., safeguarding the network from intentional or unintentional water quality intrusion/problems), in particular, has become increasingly important, and requires the division of the distribution system into isolated sectors. All water that enters into an isolated sector is metered and is consumed within it. In the event of a contamination incident, such sectors limit the number of people who might be exposed to the threat, and minimise the number of pipes and their lengths that need to be decontaminated. The process of partitioning a WDN into a set of isolated sectors is referred to as water network sectorisation (WNS), and is the focus of this thesis. There are three categories of requirements in WNS: -- Structural requirements, which are related to the structure of the network and include direct access to at least one source for each sector (i.e., the path from the sector to a water source must not contain any nodes in other sectors), sector isolation (i.e., there should be no flow exchange between sectors), sector size limitations, sector size balance, connectedness of the partitioned network, and the requirement that sectors should cross as few mains as possible. The first two requirements are specifically related to sectorisation, while the latter ones are general in WDN partitioning. -- Hydraulic requirements, which are hydraulic objectives and constraints including customer demands satisfaction, pressure requirements, network reliability, energy efficiency, limited water velocity in pipes, minimum nodal elevation differences within sectors, and water quality. These requirements must be satisfied in all types of network design and re-design, including partitioning and sectorisation, and must also be considered during the WNS process. -- Economic requirements, which are related to the cost of sectorisation. A significant challenge with WNS is that closing links in the network to address sector isolation deteriorates hydraulic behaviour of the network. In particular, network reliability, energy efficiency, and water quality are negatively affected. Graph theory can be used to address the structural requirements, but existing graph-theory techniques cannot address direct access to a source for each sector and sector isolation. Current solutions for WDN partitioning do not address all the requirements of sectorisation. In particular, direct access to a source for each sector and sector isolation are not addressed in most of the existing approaches. Those approaches that do address direct access and sector isolation, cannot handle large networks, as the number of sources is a limiting factor for the number of sectors. The only scalable approach that addresses direct access and sector isolation, does not address some other sectorisation requirements, e.g., pressure requirements during different consumption scenarios, energy efficiency, limited water velocity in pipes, minimum nodal elevation differences within the sectors, and the requirement that sectors should cross as few mains as possible are not addressed. This thesis explores WNS and proposes a WDN sectorisation method, called WDN-PARTITION, to address various requirements (specially water security requirements) while minimising the negative impact on the other structural and hydraulic requirements. WDN-PARTITION first satisfies water security-related constraints and generates a collection of feasible solutions (i.e., solutions that address structural requirements of sectorisation) using a novel heuristic graph partitioning algorithm. Then, the best solutions in this initial collection are identified using a many-objective optimisation procedure, based on the hydraulic requirements. The proposed method works well for both small and large networks, as the number of sources is not a limiting factor for the number of sectors. WDN-PARTITION has been implemented and integrated with a hydraulic network simulator (i.e., EPANET). The simulation-based evaluation assesses the results of the proposed method on a series of publicly available real world water distribution networks.
The largest available network has been used as a benchmark to compare the proposed method with three baseline approaches, i.e., a manual engineering approach (Murray et al. 2009) and two automated solutions (Diao et al. 2013 and Ferrari et al. 2014). The results show that WDN-PARTITION is a good alternative for the existing approaches, as it achieves its design objectives to partition a WDN into isolated sectors, satisfying all structural and hydraulic requirements with less than 1% deterioration in network resilience index (a metric for network reliability) and water age (a metric for water quality), comparing to the original network.
A water distribution network (WDN) supplies drinking water to homes and businesses, and links water sources to consumers. Such networks are typically complex and dynamic, consisting of thousands of nodes interconnected by thousands of elements. Control and management of a WDN as a whole is challenging as there is no granular information about different parts of the network to support location of issues. Partitioning a WDN into smaller sectors is a strategy to manage its complexity, as advised by the International Water Association. Advantages include (a) enhanced leakage and burst detection and management, (b) improved contamination spread control (associated with water security), (c) a capacity to provide different pressure levels, (d) enhanced demand management, and (e) better rehabilitation and work planning. Water security (i.e., safeguarding the network from intentional or unintentional water quality intrusion/problems), in particular, has become increasingly important, and requires the division of the distribution system into isolated sectors. All water that enters into an isolated sector is metered and is consumed within it. In the event of a contamination incident, such sectors limit the number of people who might be exposed to the threat, and minimise the number of pipes and their lengths that need to be decontaminated. The process of partitioning a WDN into a set of isolated sectors is referred to as water network sectorisation (WNS), and is the focus of this thesis. There are three categories of requirements in WNS: -- Structural requirements, which are related to the structure of the network and include direct access to at least one source for each sector (i.e., the path from the sector to a water source must not contain any nodes in other sectors), sector isolation (i.e., there should be no flow exchange between sectors), sector size limitations, sector size balance, connectedness of the partitioned network, and the requirement that sectors should cross as few mains as possible. The first two requirements are specifically related to sectorisation, while the latter ones are general in WDN partitioning. -- Hydraulic requirements, which are hydraulic objectives and constraints including customer demands satisfaction, pressure requirements, network reliability, energy efficiency, limited water velocity in pipes, minimum nodal elevation differences within sectors, and water quality. These requirements must be satisfied in all types of network design and re-design, including partitioning and sectorisation, and must also be considered during the WNS process. -- Economic requirements, which are related to the cost of sectorisation. A significant challenge with WNS is that closing links in the network to address sector isolation deteriorates hydraulic behaviour of the network. In particular, network reliability, energy efficiency, and water quality are negatively affected. Graph theory can be used to address the structural requirements, but existing graph-theory techniques cannot address direct access to a source for each sector and sector isolation. Current solutions for WDN partitioning do not address all the requirements of sectorisation. In particular, direct access to a source for each sector and sector isolation are not addressed in most of the existing approaches. Those approaches that do address direct access and sector isolation, cannot handle large networks, as the number of sources is a limiting factor for the number of sectors. The only scalable approach that addresses direct access and sector isolation, does not address some other sectorisation requirements, e.g., pressure requirements during different consumption scenarios, energy efficiency, limited water velocity in pipes, minimum nodal elevation differences within the sectors, and the requirement that sectors should cross as few mains as possible are not addressed. This thesis explores WNS and proposes a WDN sectorisation method, called WDN-PARTITION, to address various requirements (specially water security requirements) while minimising the negative impact on the other structural and hydraulic requirements. WDN-PARTITION first satisfies water security-related constraints and generates a collection of feasible solutions (i.e., solutions that address structural requirements of sectorisation) using a novel heuristic graph partitioning algorithm. Then, the best solutions in this initial collection are identified using a many-objective optimisation procedure, based on the hydraulic requirements. The proposed method works well for both small and large networks, as the number of sources is not a limiting factor for the number of sectors. WDN-PARTITION has been implemented and integrated with a hydraulic network simulator (i.e., EPANET). The simulation-based evaluation assesses the results of the proposed method on a series of publicly available real world water distribution networks.
The largest available network has been used as a benchmark to compare the proposed method with three baseline approaches, i.e., a manual engineering approach (Murray et al. 2009) and two automated solutions (Diao et al. 2013 and Ferrari et al. 2014). The results show that WDN-PARTITION is a good alternative for the existing approaches, as it achieves its design objectives to partition a WDN into isolated sectors, satisfying all structural and hydraulic requirements with less than 1% deterioration in network resilience index (a metric for network reliability) and water age (a metric for water quality), comparing to the original network. | |
