Implementation of a convenient method in water network simulation to maximize the benefit of pressure management and reduce the leakage

A water distribution network (WDN) is established to meet water demand for human use with appropriate pressure and good quality (DeZuane 1997). Water quality problems and pressure drops are major issues associated with any WDN when it suffers from leakage (Kourbasis et al. 2020).

When pressure increases, the amount of leakage increases accordingly (Shammas & Al-Dhowalia 1993; Kadu & Dighade 2015; Marlim & Kang 2024). Raised pressure also causes the pipes to burst and the frequency of leakage to increase (Makaya & Hensel 2014). Pressure management contributes to reducing the level of leakage and the frequency of new leaks (Farley 2003; Balut & Urbaniak 2011). Reducing the pressure in WDN should be carried out without affecting the required minimum pressure level and desired demand during the day (Vicente et al. 2016). Analysis of water demand and pressure for WDN, especially at night, where the flow is minimal and pressure is high, helps in detection of water loss (Thornton et al. 2008; Berardi et al. 2016). The comparison of water flow based on water demand, billing water, and measured flows and pressures in WDN using the supervisory control and data acquisition (SCADA) system is done using a mathematical model to detect the water loss in the entire WDN (Karadirek et al. 2012; Kowalska et al. 2022).

In most developing countries, the water loss due to WDN leakage is more than 40% of produced water (Mutikanga et al. 2009; Aburawe & Mahmud 2011; Rogers 2014). While rehabilitation of deteriorated WDN pipes offers an effective solution to reduce water loss, it faces significant economic and implementation challenges (Güngör et al. 2017). Leak detection in WDNs is carried out using several techniques, either in the field or through mathematical modeling (Lahlou 2005).

In the field, most of the effective devices depend on sound signal techniques, which provide high accuracy in detecting leaked pipes (Seyoum et al. 2017). There are limits in using acoustic devices, associated with cost and time, affected by buried pipe depth, external noise, and some of these devices need to excavate in part of pipes (Fuad et al. 2019).

Pressure management involves actively controlling pressure within the WDN to maintain an adequate service level, it is a reliable method used in literature to struggle and reduce the leakage (Gonelas & Kanakoudis 2016; Samir et al. 2017; Aboelnga et al. 2018; Garmendia et al. 2019; Keivan et al. 2020). The authors depend on a genetic algorithm pressure management model to minimize the WDNs leakages (Nazif et al. 2010; Korkana et al. 2016). One of the common applicable pressure management tools is installing pressure-reducing valves (PRVs) in smaller segments of the WDNs called district-metered areas (DMAs) (Puust et al. 2010; Kanakoudis et al. 2014; Kanakoudis & Gonelas 2016; Hoko & Chipwaila 2017; Dalton 2019; Marlim & Kang 2024). The formation of the DMA and placement of PRVs in large WDNs require testing various scenarios within a reliable simulation model that accurately reflects the real-world hydraulics of the WDN (Karadirek et al. 2012; Gonelas & Kanakoudis 2016; Bui et al. 2022; Shekofteh et al. 2023).

García-Ávila et al. (2019) used PRVs to optimize pressure in the WDN of Azogues city, Ecuador, with a population of 70,064 capita and a 1,200 km² area to reduce water leakage. By utilizing EPANET software, they achieved a reduction in the leakage by 31.65%. This study demonstrates the efficacy of PRVs in improving WDN efficiency. Gupta et al. (2017) developed a novel algorithm for optimizing the location of PRVs in the WDN to minimize leakage. This solution resulted in a reduction of 26.64% in water leakage, making it a cost-effective method for improving WDN performance. Mahdavi et al. (2010) conducted a similar study to optimize the location and settings of PRVs in the WDN of Mahalat city, Iran, with a population of 160,000 capita and an area of 46 km², Using a genetic algorithm with EPANET, they achieved 40% reduction in water leakage. Gupta & Kulat (2018) implemented a combined approach using variable speed pumps and PRVs in the WDN of Nagpur city, India, and achieved a leakage reduction of 26.3%.

Recently, water leak detection has become prominent in water society worldwide; the leakage is detected using commercial software or mathematical modeling. Most of these models use measured field data (pressures and flows) for DMA or use historical data of pipe characteristics in the network to detect the leakage along the entire WDN, depending on WDN hydraulic analysis (Aburawe & Mahmud 2011). The splitting of WDN to multi-DMAs has been proved as effective in pressure management and leak detection (Kowalska et al. 2022).

Pressure management and leakage detection in WDNs remain important areas of study, as evidenced by numerous scientific publications. Software like WaterGEMS and its associated tools, along with other powerful solutions, are used to analyze WDNs and estimate water loss. These tools are valuable, trusted, and widely used by experts and researchers, as demonstrated in various studies (Samir et al. 2017; Ghosh et al. 2019; Nalatawada et al. 2021; Kowalska et al. 2022; Mohseni et al. 2022; Mukherjee et al. 2022; Sudheer et al. 2020). Samir et al. (2017) adopted a method using WaterCad to minimize the leakage in DMA by simulating different PRV settings and achieved reducing the leakage by up to 37% in DMA. Shekofteh et al. (2023) proposed a new method using graph theory to optimize the number of DMAs in the WDN. These optimized DMAs contribute to maintaining appropriate pressure and flow throughout the WDN, leading to improved network management. Wu et al. (2010) studied the impact of pressure reduction on node leakage using the hydraulic model that utilizes pressure-dependent flow emitters. The model helps in leakage detection after calibration.

Indeed, the detection of leakage by using the field method is costly and difficult spatially in large networks; also, mathematical model methods are not perfectly accurate, and their estimated leakage locations often require verification through field methods. So, the objective of this study is not detecting the leakage but reducing the leakage throughout the WDN by using WaterGEMS software to maximize the benefits of pressure management in the WDN, particularly in the expected leakage region of the network. This technique is considered a tool for leakage reduction that needs to be performed for WDN.

The WaterGEMS model structure includes reservoirs, tanks, junctions, pipes, valves, and pumps, along with the necessary data to create a WDN. For the West Esna model, two types of data were collected: primary and secondary. The primary data, essential for building the WDN in WaterGEMS software, include coordinates and elevation data for junctions, reservoirs, tanks, pumps, and valves, as well as pip start and end junctions, diameters and lengths of pipes, types, and diameters of valves. Secondary data, supplementary for modeling WaterGEMS elements, consist of junction demand, pump head and flows, tanks and reservoirs, water levels, and pipe roughness coefficients and materials (Shushu et al. 2021; Abdulsamad & Abdulrazzaq 2022).

It is noted that both El Sheikh Meskien pumping station produces about 69.2% of total flow, while the other pumping station contributes 30.8%, as follows: 1.1, 2.9, 3.1, 6.0, 8.2, and 9.7% for Gezeiret El Kiman, El Ghareira, Eltraa Naser Nakaly, Esna, El Negou Compact, and El Kiman Compact pumping stations, respectively.

In order to monitor the produced water from the water treatment plants to WDN, in December 2021, a campaign was conducted to install eight ultrasonic flowmeters model InnovaSonic207i from SIERRA, USA with a pressure logger. These flow and pressure devices were installed at the main headers of the eight finished pumping stations and connected with the SCADA system at Luxor Holding Company for water and wastewater to monitor and record the flow and pressure online. The measurement data obtained from SCADA are used to verify the water consumption, derive the demand pattern multiplier, and model calibration.

WaterGEMS connect edition update 2 software, Bentley Systems Inc., USA, is used to simulate and analyze the West Esna WDN. WaterGEMS software has become popular software in analyzing and optimizing the water network (Świtnicka et al. 2017). In addition to using WaterGEMS software for hydraulic analysis of the WDN, it is used in verification of water demand by comparing measurements of produced water with the consumed water through the network.

WaterGEMS software is also used in this study to calibrate the WDN by using the Darwin calibrator. The calibration of the WDN is carried out using a Darwin calibrator to fit the simulated pressure and flow values through the WDN with actual measurement data by modifying the set WDN parameters values (Świtnicka et al. 2017). Operation scenarios are used in WaterGEMS software to maximize the benefits of pressure management in WDN to reduce the leakage along the entire network.

As mentioned in Section 2.2, there are eight chambers distributed at the finished pumping stations in the West Esna WDN. Each chamber is equipped with a flow meter and pressure gauge. These devices continuously measure flow and pressure at their respective locations and transmit the data to the SCADA system. The SCADA system collects and stores this real-time data in a spreadsheet format. The data include hourly measurements of flow and pressure for each of the eight locations. These field data were a crucial component to be used in the Darwin calibrator to calibrate the hydraulic model of WDN.

To optimize the water distribution (WDN) model, the Darwin calibrator utilizes a powerful genetic algorithm search engine to identify the best model parameters. The field data were entered into the Darwin calibrator to optimize the hydraulic model over a 24-h period simulation. The calibrator groups for pipes and nodes are created, then the calibration runs and Darwin automatically adjusts the pipe group roughness and node group demands values to achieve a close match between measured and simulated flow and pressure data. The pipe's roughness and demand node values are modified by multiplying factors to their original values after the calibration run and the Darwin calibrator automatically makes the differences between observed and simulated flow and pressure minimum using genetic algorithm and hydraulic solver. This refined and calibrated model becomes suitable for new scenarios, including identifying water losses, implementing pressure management, and reducing overall leakage (Wu et al. 2004; Aburawe et al. 2013; Abdulsamad & Abdulrazzaq 2022).

The Darwin calibrator employs a genetic algorithm to refine simulated parameters within a WDN using field measurement data. It minimizes discrepancies by calculating the squared differences, absolute differences, or maximum differences between simulated and observed values. The calibration process adjusts pipe roughness coefficients and junction water demands, with variability multipliers ranging from 0.5 to 1.5 (Kowalska et al. 2018).

The calibration process begins with creating a calibration study. This involves assigning field data snapshots, including junction pressures or hydraulic grades and pipe discharge flows. Additionally, roughness and demand groups are formed by selecting specific pipes and junctions. Finally, calibration criteria based on fitness type are established (Wu & Walski 2010).

An optimized run is created within the calibration study. This step involves specifying roughness and demand adjustment groups, activating the selected field data, and setting calibration options such as fitness tolerance, maximum trials, and the number of solutions (Wu et al. 2002; Wu & Walski 2005). The solution includes the adjusted roughness and demand groups, presenting both the original and adjusted roughness (Hazen–Williams coefficient) and emitter coefficients (leakage junctions and quantities). Additionally, the simulated results for observed data are presented alongside the differences between simulated and measured values (Wu et al. 2002; Wu & Walski 2005).

Pipes with the same hydraulic and physical properties can be grouped into one group, and the same roughness coefficient multiplier is applied to the pipes in each group. The junctions that have the same demand patterns and within the same topological area can also be aggregated as one group, but the emitter coefficient for leakage junctions will not be the same (Wu et al. 2002).

A genetic algorithm generates a population of trial solutions for the model parameters. The hydraulic solver then simulates each trial solution and predicts the hydraulic grade and flows at predetermined measurement points within the network. The calibrator evaluates the fitness of the solution by comparing between simulated and observed data. The parameters that minimize the discrepancy between model predictions and field observations are selected as the fittest and considered for subsequent solutions. (Wu & Walski 2005).

District metering is the process of dividing the WDN into zones with boundaries (DMAs) to control pressure and supply demands (Girard & Stewart 2007; Shao et al. 2019). WaterGEMS software offers tools to create DMAs using an algorithm based on graph theory. The DMA creation tool selects the modularity pipes to be the DMA group also DMA analysis determines the boundary of the DMA from elements that can be metered or controlled in the DMA from WDN elements (Kowalska et al. 2022). WaterGEMS software is carried out for dividing of the West Esna distribution network. The dividing of WDN is based on either the target number of nodes or pipe length. The DMAs usually are placed between 500 and 3,000 properties or represent the length of the rural system network (Karadirek et al. 2012; Gomes et al. 2015). The dividing of the WDN is not used only for comparing the pressure and flow in DMA, but it is also used, to identify the main flow path throughout the DMA and to control the flow and pressure within this DMA.

The WaterGEMS DMA tool helps users to create water distribution zones (DMAs) by automatically dividing the network into smaller sections based on user-specified criteria, like the desired number of nodes or pipe length. The tool uses a graph theory-based algorithm that groups pipes with high connectivity into a DMA. Users can manually adjust the suggested DMAs and review their characteristics, including boundary elements like pipes, pumps, and valves (Kowalska et al. 2022).

The goal is to create DMAs of similar size and minimize the number of closed valves. However, in some cases, it may not be possible to achieve these goals due to the network's configuration. In such situations, the tool may create larger DMAs with more closed valves. Users can still manually merge adjacent DMAs to reduce the number of closed valves (Alvisi & Franchini 2014).

The simulated flow is based on actual demand that was driven from billed data. The analysis data of flow were driven from WaterGEMS, the simulated flows were less than the measured flows which referred to existing leakage in the network (Aburawe & Mahmud 2011). The average water consumption is 120 L/C/D based on billing data for the WDN, while it is about 156 L/C/D based on measured data. The total produce water is 55,295 m³/days, while the consumed water is 42,536 m³/days, which means that the percent of leakage is about 23%. The low leakage percent is due to reducing the pumping station outlet pressure via the gate valves partial closure (Samir et al. 2017).

Observed leakage junctions were detected and the number of leakage junctions varied from one to six in each region of 19 residential areas representing the West Esna WDN. The total number of leakage junctions was 71with the emitter coefficient varied between 0.01 and 0.5.

The emitter coefficient (k) represents the characteristics of an orifice-like leakage junction. It describes how much water leaks (Q) from a specific point (junction) based on the water pressure there. This relation is expressed by the formula: Q = k * (P^n), where n is called ‘the emitter exponent’. Typically, n equals 0.5 for leaks behaving like an orifice (Kabaasha et al. 2016).

WaterGEMS software uses this concept to simulate leakage. Its Darwin calibrator tools, like a skilled problem solver, optimizes the amount of water flow through each junction demand using a genetic algorithm to adjust the emitter coefficients in the above formula until the simulated water flow and pressure match actual field observations. This approach is consistent with the method used in previous studies by Desai & Rajapara (2021); and Wu et al. (2004, 2015).

As noted before, the reduction of water loss of 23% was due to reducing the pumping station outlet pressure via partial closure gate valves (Karadirek et al. 2012; Shushu et al. 2021). The Darwin calibrator tools were used to determine the amount of leakage along the entire network in case these gate valves were fully open and no control valve was used. The simulated results, as illustrated in Figure 7, show the leakage percent of 33% compared to 23% when using a control valves scenario (Karadirek et al. 2012; Samir et al. 2017).

The boundary pipes for each DMA were investigated in order to install the PRVs, the results of installed PRVs for each DMA were illustrated in Table 3.

As shown in Figure 7, the installation of PRVs at boundary pipes reduced simulated supplied flows. This decreased the leakage percentage from 23 to 17%, aligning with previous findings in the literature (Karadirek et al. 2012; Gupta & Kulat 2018; Monsef et al. 2018; García-Ávila et al. 2019; Zhang et al. 2021).

• A mathematical model is developed to reduce leakage throughout the WDN by maximizing the benefits of pressure management in the WDN and maintaining the minimum required pressure.

• Field measurements data within the WDN indicated significant leakage, necessitating calibration. Darwin calibrator tools of WaterGEMS were used successfully to calibrate the WDN by fitting simulated parameters with observed data, thereby identifying the amount and location of leakage entire the network.

• The implementation of partial closure gate valves significantly reduced water loss by 23% in the network. Without these control valves, the Darwin calibrator estimated a leakage rate of 33%, highlighting the effectiveness of the implemented measures in mitigating water loss.

• DMAs tool in WaterGEMS is an effective tool to identify feeding flow pipes and boundary pipes in order to help in pressure management.

• Installing PRVs at boundary pipes of DMA is mostly effective in addition to the gate valve at pumping stations.

• The simulated supplied flows after using PRVs at boundary pipes were less than the values before installation of PRVs as illustrated in Figure 7 and the percent of leakage was reduced from 23 to 17%.

• The results demonstrated that the strategy of employing valve closures and PRVs at the boundary pipes of DMAs effectively mitigates water leakage and alleviates the strain on ageing infrastructure in developing countries.

• The study is based on modeling simulations; real-case installations of PRVs could require different configurations to guarantee correct operation and maintenance.

• The actual results in terms of leakage reduction will be measured after DMA implementation and could differ from the model simulation. These results can be used to adjust model parameters to fit the actual leakage reduction.

The authors are thankful to the authorities and management of Luxor Holding Company for water and wastewater, and for providing West Esna WDNdata.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

The authors confirm sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.