Near real-time topological adaptation in looped water distribution networks: enhancing resilience and water quality Open Access

Water distribution networks (WDNs) are essential infrastructure systems responsible for supplying potable water to communities. Ageing water infrastructure in the developed world concerns engineers and policymakers for multiple reasons (Makropoulos et al. 2008a; Baird 2010; Butler et al. 2018). These critical networks are now challenged by ageing urbanization and climate change (Makropoulos & Butler 2010; Nikolopoulos et al. 2021; Goliopoulos et al. 2022).

Water losses detected in the WDNs during the past years, along with problems related to serving the demand, have been a priority for the water utility companies for environmental, sustainability, financial, and many other reasons (Makropoulos et al. 2008b; Latinopoulos 2014; Blocher et al. 2020). In 2018, Thames Water was ordered to pay out £120 million to compensate customers for poor leakage management (BBC 2018). Infrastructure resilience, including WDNs, is also at the core of the post-pandemic recovery funding packages in the US (NYT 2021) and the EU (EU 2021). WDN's resilience is defined as the ability of the WDN to meet the customer's demand without interruptions and recover from failures (Makropoulos et al. 2018; Rokstad et al. 2023). Furthermore, the treatment of water is more developed, but ageing water networks cannot always ensure the distribution of water with the same quality as the treatment plants (Mohammed et al. 2021).

As new challenges are presented, water utilities are trying to improve their operation by offering high levels of resilience and water quality (Sakellari et al. 2005; Baki et al. 2018; Makropoulos & Savíc 2019). Several strategies have been proposed to improve the resilience and serviceability of WDNs, as well as to improve the water quality. These studies have been focusing on dynamic network topologies to improve the network's operation (Wright et al. 2015a, b, c; Marsili et al. 2023; Anchieta et al. 2024) or water quality network optimization (Brentan et al. 2021; Shmaya & Ostfeld 2022). Serviceability is the ability of the water networks to continue offering reliable services to customers (Ofwat 2012). Moreover, throughout the past decades, several metrics have been developed to assess the performance of WDNs (Makropoulos & Butler 2004; Behzadian et al. 2014). One of the main ways to tackle these challenges is to intervene in the topology of the networks.

WDN's resilience can be achieved by improving connectivity and redundancy of the network (Wright et al. 2014; Ulusoy et al. 2020). Sometimes, minimizing leakage with pressure management and reaching high resilience in the network can be contradictory objectives (Wright et al. 2015a, b, c). For wealthier countries, the resilience of the WDNs is one of the priorities, because of the deterioration of the infrastructure without sufficient investment for replacing components (Hoult et al. 2009; Makropoulos 2017; Nikolopoulos et al. 2020). Folkman (2018) stated that pipe breaks increased by 27% in the US between 2012 and 2018, and 16% of the total installed water mains are beyond their useful life.

Problems of water quality deterioration and water age are rapidly increasing globally (Barros et al. 2023), and it is the time taken for the water to travel from the source to the consumption locations within the distribution system (Monteiro et al. 2021). Water quality is affected by several parameters, such as treatment process, pipe material and age, hydraulic conditions, environmental conditions, and residence time (Armand et al. 2018). Water age is a common metric for assessing water quality (Mabrok et al. 2022), and the optimal values of water age to achieve the best water quality performance are less than 10 h (Monteiro et al. 2021). In the past few years, pressure control and district metering areas design are used to control water losses, but both are reducing water velocities in the WDNs (Giustolisi et al. 2023). Also, the problems of water quality are developing in countries where the WDN becomes deteriorated due to various physical, environmental, and operational factors, including aging infrastructure (Kim et al. 2022).

Two primary configurations for WDNs, looped and branched, have been widely employed while trying to support the new problems of the WDNs (Gibson & Karney 2023). Branched networks are simpler in terms of design and maintenance due to the straightforward layout, and the construction cost is lower (Martínez 2010).

A significant category of branched networks is the self-cleaning networks and is getting more popular, as water quality is becoming a more important concern for water utilities (Jenks et al. 2023a). Self-cleaning networks use cutting-edge technology, smart sensors, and software to smartly manage and ensure water quality (Jenks et al. 2023b). On the other hand, branched networks have reduced redundancy and a higher risk in serviceability, especially in disruptive scenarios like pipe bursts or even for maintenance reasons.

Looped networks have been used more widely in the past few decades since the looped topology made the water utilities more confident about supplying consumers with enough water and ensuring redundancy (Martínez-Rodríguez et al. 2011).

Looped networks are characterized by interconnected pipelines that form closed loops, offer several advantages (Saleh & Tanyimboh 2013). Looped networks provide multiple pathways for water flow, reducing the risk of service interruption due to disruptive events like pipe breaks or maintenance activities (Martínez-Rodríguez et al. 2011). Thus, looped network offers higher levels of resilience and redundancy in the WDNs, which leads to higher levels of serviceability (Vertommen et al. 2022). The resilience of the infrastructure is a relatively new concept, but many definitions and extensive research can be found (Timashev 2015; Doorn et al. 2019; Quitana et al. 2020), which mainly focus on the absorption, adaptation, and recovery from a failure. For wealthier countries, the resilience of the WDNs is one of the priorities, because of the deterioration of the infrastructure without sufficient investment for replacing components (Hoult et al. 2009). Depending on the network topology and other characteristics, looping promotes better water mixing, reducing the likelihood of stagnant zones and improving water quality.

However, building looped networks can be more expensive due to the need for additional pipelines and infrastructure (Zischg et al. 2018). Moreover, maintaining and repairing looped networks can be more challenging due to the interconnected nature of the system (Berardi et al. 2022). These networks are considered more complicated (Banõs et al. 2010).

Taking all of the above into consideration, it would be accurate to say that branched networks have the advantages of lower construction cost and better water quality, whereas looped networks have the advantage of increased resilience and redundancy of the network, which makes them easier to operate during disruptive events.

Furthermore, technology has been developed rapidly over the past two decades (Ciliberti et al. 2023) and is now offering water utilities new tools for WDN's management in real time (Loureiro et al. 2014; Tsiami & Makropoulos 2021). The water utilities are now able to interfere in the WDN's topology during its operation to achieve the optimal balance between serviceability, water quality, and facing disruptive events efficiently (Abraham et al. 2017).

The novelty of this study is based on the real-time approach to the network's management with topology adaptation, based on the branched and looped topology. To the best of our knowledge, this is the first study to propose real-time network operation based on water quality, for adjusting the WDN's operation and proposing a planning tool for the WDN's operators, and the first study that focuses on combining the advantages of both branched and looped networks.

Some studies, which are using the real-time approach, focus on cost and energy reduction (Shamir & Salomons 2008) or improving pressure and operation during disruptive events like bursts and leaks (Allen et al. 2011; Wang et al. 2015; Abu-Mahfouz et al. 2019; Creaco et al. 2019; Kalyanapu et al. 2023). None of these studies focuses on real-time operation to improve water quality while maintaining the same serviceability.

The present study aims to optimize the WDN's operation by adapting its topology in real time. The study includes making changes in the network's topology by using remote-controlled valves to make branched topologies in looped networks to combine the advantages of both topologies. Both serviceability and water quality are assessed for each situation. Finally, a conceptual operational plan is proposed to the water utility professionals to improve each WDN's performance.

The EPANET benchmark network Net3 was chosen to evaluate the applicability of the proposed method (Rossman 2016). The Net3 water network is based on the North Marin Water District in Novato, CA. The system has an average demand of 4.9 MGD, serves a population of around 64,000 people over an area of about 100 mi². The network has 2 raw water sources, 2 pump stations, 3 elevated storage tanks, 92 nodes, and 117 pipes. The network was first published by Clark et al. (1994), since then it has been frequently used for water quality, chlorine residual, and disinfection byproduct formation modeling in the literature.

The software used for this study was EPANET, for modifications and simulations on the WDN, along with Python Spyder, for specific orders and the presentation of the results. Libraries used in Python included ‘numpy’ for mathematical operations, ‘pandas’ for data analysis, ‘matplotlib’ for the presentation of the analysis, and ‘wntr’ and ‘network’ for the network's simulation. EPANET was chosen to model the water distribution system, and the use of Python was selected to make specific functionalities as described below and improve the presentation of the results. Using the relevant libraries in Python improved the simulation duration and made the criticality analysis easier.

Moreover, the total network demand of the first day is presented in Figure 2. The maximum total demand is presented at 14 h, whereas the minimum total demand is presented at 4 h.

For each area, critical pipes were shown during this study. Pressure-dependent analysis was chosen for the simulations because it simulates the real condition of a WDN, and the relationship between nodal pressure and discharge is considered (Shirzad 2020). The pressure threshold for the analysis was set at 30 psi.

The first step is to show critical pipes of the network, by testing their impact in disruptive events. Pipe criticality is evaluated based on a critical link analysis (CLA), which performs pressure-dependent analysis for each pipe closure of the network. CLA shows which pipes should not be closed to achieve maximum serviceability. The CLA is commonly used by water companies to identify which links are crucial for the operation of the network. The implementation of CLA involves closing each link of the network and assessing the impact on the network. For links (pipes) that are preferred to stay closed or close occasionally, valves should be introduced. The CLA was implemented on the network with a loop of all pipes removed one by one, then simulating the network and assessing the serviceability and the water quality.

At this point, it is important to mention that since the CLA shows the result of a pipe closure, it is considered to be a very accurate metric for the impact during the repair period of the pipe (Wright et al. 2015a, b, c), when the pipe needs to be isolated to repair it and a very accurate metric for the impact of significant bursts which lead to large water losses. The main limitation of the methodology is that the results are not accurate for smaller leaks or bursts, which are also important disruptions but are minor.

At the same time, during the implementation of the CLA, for each pipe closure, the average quality of the network is calculated and compared with the average quality of the network without any pipe closure. The water age, which is a major factor for quality deterioration in water distribution systems, is used to assess water quality in our case.

This methodology enables the WDN operator to spot which pipes can be closed with remote-controlled boundary valves to improve.

Finally, the results of the simulations should be considered to decide whether some loops of the network can be closed to create a branched network and include the advantages of both network topologies. An operational planning tool is created for network operators to improve the network's resilience and water quality by creating branched sections in looped networks. This study evaluates the results of the simulation to create an operational action plan for the WDN operators to optimize the resilience and the quality of the network by adapting the network's topology in real time.

In this section, the results of the simulations are presented and interpreted. The first part includes the investigation of the advantages and disadvantages of creating branched topologies in looped networks. The second part uses these results to implement a more structured methodology to add remote-controlled boundary valves in a WDN to improve its water quality without reducing its resilience and redundancy.

The result of the analysis shows that the pipes are more critical and would create more problems during a disruptive event. This figure (Figure 5) also shows that the river is a much more important water source than the lake, some pipes are critical for the serviceability of the network, and some points where loops are improving the network's resilience. Moreover, Figure 5 indicates that the pipes that lead to serviceability issues when they are removed are not suitable for closure to improve water quality. On the other hand, the dark blue pipes show that their closure would not lead to problems with the serviceability of the network, making them suitable for closing them.

The network's operator could add some links by extending the network to improve resilience, especially near the pipes, which are very critical for the network. Moreover, the network's pressure varies between different moments.

Thus, the pipe removal was efficient in creating a branched network with improved water quality, but closing the pipe could lead to problems with the network's serviceability. The following step is to test the impact of pipe removal at pipe ‘204’, which is the alternative for creating a branched network at Area 3. Pipe ‘204’ is the pipe that has a valve icon added on the simulations after the pipe removal.

These results prove that the pipe could be closed with a remote-controlled valve and open only when pressure issues occurred, or during disruptive events. However, supplying the section of the network only with one pipe when a boundary valve is installed on pipe ‘204’ is risky, because in case of a burst or a similar disruptive scenario, the whole section could face serviceability issues until the valve opens and redundancy is achieved again. This means that the pipe can improve the resilience of this section of the network, but the data availability and communication, which are related to disruptive events (bursts, fire hydrants, etc.), should be developed to an acceptable level.

The network's quality is measured in all cases as the average water age. When both pipes are open and operating, the average water age in the network is 33.6 h. When pipe ‘229’ is removed, the average water age in the network is 16.7 h, and when pipe ‘204’ is removed, the average water age in the network is 23.4 h.

This comparison highlighted that there should be a balance between serviceability and quality improvement. This study focused on keeping high levels of the network's resilience and serviceability as a priority for the operator and trying to create a branched topology to improve the quality.

The pipes that are in the blue area are the ones that create serious serviceability issues when they are closed, and their closure is not considered for quality improvement. The pipes that are in the yellow area are not critical for serviceability, but their closure does not improve the water quality significantly. The pipes in the green area can be closed, since their closure is not critical for serviceability, but when they are closed, the network's quality is greatly improved. The separation of Figure 10 in these sections is chosen as a proposed operational strategy in this study and could be modified based on the WDN operator's priorities.

The proposed methodology significantly reduces the water age, achieves high performance, and keeps the network's serviceability and resilience. However, installing remote-controlled boundary valves on the network increases the risk of not meeting demand, especially during disruptive scenarios. This is the reason why remote-controlled boundary valves are chosen instead of simple boundary valves, so that in case of a disruptive event, the valves can open to increase the redundancy of the network.

The network's operator should focus on preventing supply interruptions. The risk of supply interruptions can only be minimized if loops are added on the network at the parts of the network that present a branched topology.

• (2) Evaluating the network's criticality and water quality

A CLA of the network should be performed by successive pipe removals of all pipes, and assess the serviceability and the water quality of the WDN. The serviceability is measured with the number of nodes that occur low-pressure conditions during a pipe closure. The water quality is measured using the average water age of the network.

• (3) Creating branched topologies

Branched topologies should be created in looped networks to improve water quality. The results of the CLA performed in step 2 should be evaluated, considering the network's special characteristics and the operator's priorities. At this stage, potential critical customers should also be considered. Remote-controlled boundary valves should be added at the pipes chosen by the WDN's operator to create branched topologies. The boundary valves should be able to open immediately in case of a disruptive event. The cost of the implementation should be feasible, taking into account that in Net3, which, as mentioned before, serves a population of around 64,000 people, only three remote-controlled boundary valves should be installed. However, the cost would definitely vary depending on the network's topology. However, if more valves were to be installed, the water operator could still prioritize the most important ones based on the analysis (Figure 10) and choose to install as many valves as it would be affordable for the water company.

Finally, the efficiency and usefulness of the planning tool can be maximized along with the existence of a credible digital twin of the network, which will be updated with continuous data from the network's performance in real time.

Digital twins are emerging technologies that include topological data, along with data from SCADA, sensors, meters, and other sources (Brasil et al. 2022) to reproduce real-time models of the water networks (Ramos et al. 2022, 2023; Mücke et al. 2023; Sinagra et al. 2023). These models, along with other technologies, can optimize the WDN's operation and assist the operator's decision-making ability (Brahmbhatt et al. 2023).

Digital twins can be useful tools for the implementation of the near real-time operation, which – until now – has been achieved in very few studies and remains very simplistic (Creaco et al. 2019). In more detail, digital twins enable the operator to make the simulations presented in this study, to spot the best pipes for remote-controlled boundary valve installation, and then, optimize the network's operation by simulating the benefits and challenges of every intervention on the network.

The digital twins can be efficient – especially if the resilience of the network is dependent on them – only if the data of the network can be provided with very short time steps. For example, the 1-h time step, which is provided in this study of Net3, is not efficient because it could lead to 1 h of reduced serviceability during a disruptive scenario, which is not considered acceptable in many cases. A time step of a few minutes should be used in such cases, so that the operator can detect the problem early enough and take action to open relevant closed valves and increase the redundancy of the network until the disruptive event is fixed.

The planning tool for enhancing WDN's topology and quality, by keeping the maximum serviceability, is a holistic approach for improving the network's operation. Supply interruptions can be minimized because the loops can always offer redundancy and resilience in the network. Moreover, a branched topology created with boundary valves installed improves the water quality of the WDN. The planning tool's details will define which pipes should have remotely controlled boundary valves and sensors for spotting disruptive events.

WDN operators can use the planning tool to maximize the redundancy of the network and, at the same time, the water quality, with simple steps. The practical implementation of the tool is efficient because, with simple steps, the operator can focus on the topological weaknesses of the WDN and improve them by adding loops and remote-controlled valves. The remote-controlled boundary valves are crucial for the optimal implementation of the tool, because in case of a disruptive scenario (burst, fire hydrant, etc.), the relevant remote-controlled valve can be immediately opened to add redundancy on the network and minimize the serviceability issues. Of course, as can be understood from the above, the optimal implementation of the toolbox is significantly influenced by the existence of warning systems of disruptive events, especially bursts, which is an important issue for future research. However, the rapid development and installation of smart water meters, instruments, and monitoring equipment on water networks offer the WDN's operators several ways for early detection of disruptive events. Thus, the toolbox should be used in combination with the metering and automation systems used by each operator, while following the steps presented above.

It is also important to mention that the implementation of the planning tool can be very beneficial to the WDN operators, but only if implemented appropriately for each case. To achieve this, expert support is required for each case. The development of the planning tool would include steps to improve transferability and reduce the need for a lot of customization from experts. This would include automation tools that could be used by WDN operators without expert support. These improvements are considered to be the major target for future research.

The discussion around looped and branched WDNs has been growing over the past few years. Looped networks are considered better for WDN's redundancy and resilience. On the other hand, branched networks have better water quality and reduced construction costs. Climate change, aging infrastructure, urbanization, regulatory pressures, and financial constraints are posing a risk to WDN's resilience and water quality.

This paper investigated the ability to create branched topologies in looped networks. Net3 was chosen for the simulations, and the results showed that adding boundary valves to create branched topologies can significantly reduce the water age of the network. A CLA analysis was performed to assess the results of each pipe removal. For each pipe removal, the serviceability and the water quality of the network were measured by counting the number of nodes that had low-pressure conditions and the average water age of the network. The pipes that showed that the serviceability of the network was disrupted were excluded from the valve placement assessment. The pipes, which improved the water quality greatly when they were removed, without water supply interruptions, were the optimal points for adding remote-controlled boundary valves. The proposed methodology, applied to the Net3 model, reduced water age from 33.6 to 16.7 or even 8.6 h in critical sections, significantly improving water quality while maintaining serviceability and considered a very high performance according to research. The results also showed that only 7 out of 117 pipes can improve the water quality when boundary valves are used on them.

Furthermore, a planning tool was created for the WDN's operators to choose the optimal points of adding remote-controlled boundary valves to create branched topologies in looped networks. The planning tool for enhancing WDN's topology and quality, by keeping the maximum serviceability and high resilience of the network, is a holistic tool for the network's operation. The planning tool's details should be used in combination with the operator's experience to define which pipes should have remote-controlled boundary valves and sensors. The remote-controlled boundary valves can transform the network's topology in real time, based on its needs, especially during disruptive events. Their placement enables the WDN's operators to combine the advantages of branched and looped network topologies. Details about the planning tool, such as data availability, simulation time steps, and especially making it simpler for WDN operators in order to use it without expert support, should be a matter of future research and depends on the goals and special characteristics (human resources, consumer behavior, topology, etc.) of the operator. However, regardless of the details, the tool's methodology should remain the same.

Future work may include a more extended evaluation of the resilience risks associated with quality improvements, or the possibility of adding loops on a branched network, along with remote-controlled boundary valves to maximize both quality and serviceability. A significant next step is also testing the methodology developed in this paper in real-world systems or integrating it with predictive maintenance algorithms. Finally, a critical part of future research is to implement the planning tool with water operators to assess its efficiency and potential improvements, which can be beneficial for them.

It is hoped that the above tool can support the WDN's operators to evaluate the resilience of their network and add remote-controlled valves and sensors to improve the water network and increase its adaptability to disruptive events, especially along with a reliable digital twin of the network.

All relevant data are available from an online repository at https://github.com/OpenWaterAnalytics/EPANET/tree/dev.

The authors declare there is no conflict.