Hydraulic performance and optimal booster chlorination in an existing urban water distribution network: a case study of IIT Roorkee, India Open Access

Water supply and distribution systems are critical infrastructures that provide communities with reliable access to clean and safe water. These systems include water sources, pump stations, treatment plants, service reservoirs, overhead tanks (OHTs), and pipe networks. Their primary function is to supply and distribute potable water from the source to the community for various purposes (Mehta et al. 2017). In addition to meeting the needs of human consumption, effective water supply and distribution systems are essential for sustaining economic activities, ensuring public health, and promoting overall societal well-being (Kadhim et al. 2021). Efficient management of these systems plays a pivotal role in fostering sustainable development and improving the quality of life for populations (Beker & Kansal 2023).

Maintaining water quality in urban water distribution systems (WDS) is crucial, as these systems represent the final before human consumption. Even treated water can become contaminated during distribution, posing significant risks to public health (Islam et al. 2017). Water intended for drinking and food preparations must meet established quality standards and remain free of organic and inorganic contaminants that could cause diseases. Among the various methods used for water disinfection, chlorine is the most commonly employed disinfectant (Mazhar et al. 2020). The efficiency of the disinfections by chlorine depends on residual chlorine concentration ensuring continuous protection of microbial contaminants throughout the distribution networks. In WDS, residual chlorine levels of between 0.2 and 0.3 mg/L are frequently maintained to act as a sentinel for the entry of pollutants and to stop the growth of unwanted bacteria and other organisms (Galal-Gorchev 1996).

The chlorine dosage and the location of injections during disinfections are factors that significantly affect residual chlorine levels in WDS (Javadinejad et al. 2019). However, excessive chlorine concentrations in WDS can lead to several challenges, including pipe corrosion and unpleasant taste and odor for consumers near the entry points of the systems. Moreover, elevated levels of chlorine can result in the formation of harmful disinfection by-products, such as trihalomethanes, which are carcinogenic at high concentrations (Goyal & Patel 2015). For this reason, many national regulations impose strict limits on the maximum allowable residual chlorine concentration (C_max). Reducing the maximum residual concentration (C_max) at the treatment plant may alleviate these concerns. Still, it could result in inadequate chlorine levels at distant junctions with longer residence times in large water distribution networks (WDN) (Kansal et al. 2005).

The proper location of booster chlorination (BC) stations emerges as a critical solution to maintain the recommended residual chlorine levels throughout WDN, ensuring both the safety and effectiveness of disinfection. To address the optimal placement, operation, scheduling, number, and chlorine dosage of the BC stations in WDNs, several researchers have developed optimization models using various approaches. While many studies discuss life cycle costs, they typically focus only on capital costs, chlorine dosage, and operation/maintenance costs, without accounting for the replacement costs associated with the BC stations. This study addresses this gap by presenting a novel methodology that integrates LCC analysis, with a particular emphasis on replacement costs, into the optimization of BC station placement and dosage, ensuring the economic feasibility and sustainability of the proposed solutions. Cost and time overruns are common challenges in construction projects worldwide, impacting both efficiency and sustainability (Ugural & Burgan 2021). Similarly, Rahman et al. (2012) emphasize that time and cost efficiency are key indicators of a project's overall success.

Recently, Sime & Kansal (2024) investigated an optimization problem for BC stations using a mixed integer linear programming problem. The aim is to minimize chlorine dosage and life cycle costs by increasing the number of BC stations. However, during cost analysis, the study does not consider the life span of BC stations with the design period of WDN. Similarly, Sert & Altan-Sakarya (2017) proposed a linear programming approach to determine the optimal injection rate of the BC stations by considering two scenarios, with and without initial concentrations as unknown variables. However, this methodology focuses on optimizing disinfectant mass injection and maintaining residual chlorine concentration throughout the WDN without addressing the economic aspects like capital costs, operations/maintenance costs, and replacement costs of BC stations over their lifespan. This omission results in a deficiency in comprehending the long-term economic viability and sustainability of establishing booster disinfection stations.

In addition to this gap, several studies are investigating the hydraulic performance of existing WDN using EPANET, WaterCAD, and WaterGEMS software. For example, Mekonnen (2023) and Jeong & Kang (2020)) evaluate the hydraulic performance of existing WDS. However, most studies aim to determine the pressure and velocity in WDN without effectively addressing how to allocate node water demand. Similarly, Agunwamba et al. (2018) and Nasrollahi et al. (2021) have focused on hydraulic modeling and pressure management but often rely on simplified demand allocation methods. Proper demand allocation is critical to accurately representing the hydraulic and chlorine dynamics in the network, and its neglect can compromise the accuracy of simulation results and optimization outcomes. This study employs the Thiessen polygon approach to spatially allocate junction water demand and integrates it with WaterGEMS hydraulic modeling to simulate pressure, velocity, and chlorine distribution across the WDN.

The methodology uses mixed-integer linear programming and water quality modeling WaterGEMS software to determine the optimal BC stations for effective disinfection and the present worth (PW) method was applied to evaluate the LCC of BC stations. Minimizing the chlorine dosage and number of BC is the objective function, and maintaining residual chlorine in WDN is a constraint during the optimization of BC.

The following information parameters are needed to model and simulate any urban area's existing WDN. (1) Pipe information (length, diameter, and roughness); (2) OHT data (elevation and diameter); (3) pump data (discharge rate and pump head); and (4) junction data (demand, pressure, and water quality), and population data. These datasets were collected through various methods, including field observations, experimental tests, historical records, and interviews with field workers. These data were also collected using a pressure gauge, magnetic flow meter, and plastic jerrycan.

Junction water demand allocation refers to the demand at a specific point in a WDN, typically at junctions. WDN models simplify by assuming that water is drawn only from a limited number of junctions, ignoring the numerous actual service connections along the pipelines. As suggested by Mulatu (2017), the water demand at a particular junction can be estimated by dividing the total population by the number of junctions and multiplying by a standard per capita water demand rate. However, this method does not accurately represent the actual allocation of junction water demand, as it assumes an even distribution of users across junctions without considering the spatial variation of users. Similarly, grouping existing water users solely based on demand requirements overlooks the geographical proximity of users to specific junctions, potentially leading to an inaccurate distribution of localized demand patterns. Therefore, the Thiessen polygon approach is more suitable, as it delineates the service area of each junction spatially, providing an equitable distribution of water demand based on the proximity of users to the nearest junction, which is particularly useful in areas with high population density.

According to Mekonnen (2023), it is essential to calibrate and validate a model to ensure that its results accurately reflect the values observed in the field. Similarly, Kassahun & Dargie (2024) state that model calibration continuously evaluates and adjusts parameters by comparing the simulated outcomes with the observed values. This study uses WaterGEMS software for two purposes:

• (1) Hydraulic analysis in the WDN to assess pressures at various nodes and velocities in various pipelines during the various hourly demands.

• (2) To assess the residual chlorine concentration at various nodes; identify the nodes with less than min. Chlorine concentration, and identify/suggest the booster chlorine concentration stations.

However, the model is calibrated and validated through the observed and simulated pressures and residual chlorine concentration at several nodes of the WDN to ensure the accuracy of the model. According to Tufa & Abate (2022), 2–10% of the samples are sufficient for calibration and validation. In this study, four dead-end locations (J-56, J-82, J-84, and J-91) were selected (out of the total 93 locations) (4% of the total nodes) for calibration and validation of pressures at the peak hourly demand and pressure is measured by the use of a pressure gauge. The residual chlorine is calibrated and validated through the three overhead tanks (OHT-1, OHT-2, OHT-3, and OHT-7) by assessing residual chlorine using the 4500-Cl B Iodometric Method. Therefore, the statistical correlation coefficient (R²) was used to evaluate the model's accuracy.

The calibrated and validated WaterGEMS model was run to simulate pressure, velocity, and residual chlorine levels. To evaluate the pressure and velocity of the existing WDN, the WaterGEMS model was run under peak and MHC scenarios. For residual chlorine levels, the model was run for 24 h starting from the time of chlorine dosage injection.

The primary objectives of optimizing BC stations are to minimize the chlorine dosage required for disinfection, reduce the number of BC stations, and ensure safe and efficient water distribution. This study proposed an optimization approach to address the operational and economic challenges associated with disinfection by chlorine, as outlined below.

Maintaining chlorine residuals within a WDN is crucial to protecting users from microbial contamination (Avvedimento et al. 2022). The regulatory requirement for residual chlorine at the consumer end in India is a minimum of 0.2 mg/l to prevent microorganism contamination at various points in the water distribution (Goyal & Patel 2015). However, many countries practice chlorination at the source or service reservoirs to meet these regulatory requirements without considering the size of the network and chlorine decay in WDN.

According to Avvedimento et al. (2022), installing BC stations is the best solution to address issues of under-dosing and over-dosing effectively and maintain residual chlorine levels across the WDN. This study analyzed five scenarios to determine the ideal location and the number of BC stations, as well as the chlorine dosage needed for effective disinfection. In scenario 1, only the chlorine dosage was increased from 0.5 to 4 mg/l to assess its impacts on residual chlorine levels across the WDN before proposing BC stations. Scenario 2 proposed installing one BC station, followed by scenario 3 with two BC stations. Scenario 4 analyzed the performance with three BC stations, and finally, scenario 5 evaluated the residual chlorine across the full network with four BC stations.

Scenario 3 implemented two booster stations (BC1 at P-124 and BC2 at P-67), with chlorine dosages of 3 and 2 mg/l, respectively, reducing the total dosage requirement to 4,302 kg/year. Although chlorine distribution improved, J-49 and J-50 still exhibited concentrations below 0 mg/l (Figure 13(c)). Scenario 4 further optimized chlorine distribution by introducing three booster stations (BC1, BC2, and BC3), each with a chlorine dosage of 2 mg/l, reducing the total chlorine mass dosage to 4,226 kg/year. Most junctions showed effective chlorine maintenance, except for J-49, where levels remained unchanged (Figure 13(d)). Finally, scenario 5 implemented four booster stations (BC1, BC2, BC3, and BC4) strategically placed at P-124, P-67, P-9, and P-119, respectively, each with a chlorine dosage of 1 mg/l, significantly lowering the total chlorine dosage to 3,075 kg/year. This scenario ensured that residual chlorine concentrations across all observed junctions exceeded the minimum threshold of 0.2 mg/l, demonstrating the effectiveness of optimized BC in maintaining chlorine levels while minimizing overall chlorine usage (Figure 13(e)).

The cost analysis for all scenarios evaluates BC stations' PW for the base year 2025, including annual O&M costs, replacement costs over 20 years, and fixed costs (E). In scenario 1, the analysis focuses solely on the cost of the existing chlorination method by increasing the chlorine dosage from 0.5 to 4 mg/l without installing BC stations. As no capital and replacement costs are involved, the assessment centered on the annual chlorine costs, which amount to 103,660 INR/year (Table 3), with a PW of 1.17 million INR. Additionally, the annual O&M cost for this scenario is 0.06 million INR/year, contributing a PW of 0.01 million INR.

For model calibration and validation purposes, samples were taken during chlorine injection and throughout 24 hours for laboratory tests. Table 4 compares the simulated residual chlorine levels and the experimental measured in the laboratory. This comparison ensures the accuracy and reliability of the WaterGEMS model in predicting chlorine decay over time.

The statistical analysis and performance evaluation of the WaterGEMS model for the IIT Roorkee WDN revealed strong agreement between observed and simulated values for both pressure and residual chlorine. Key statistical metrics, such as minimum, maximum, mean, median, standard deviation, and coefficient of variation, were calculated to assess the data distribution and variability. The model performance was evaluated using metrics like root mean squared error (RMSE), mean absolute error (MAE), Nash–Sutcliffe efficiency, and R-squared (R²). For pressure, the RMSE was 1.90, MAE was 1.43, and R² was 0.86, indicating reliable hydraulic predictions (Figure 17). Figure 18 presents the calibration results for chlorine residuals at selected overhead tanks: (a) OHT-1, (b) OHT-2, (c) OHT-3, and (d) OHT-7. The simulated results closely align with observed values, demonstrating the model's calibration accuracy across different locations. The RMSE, MAE, and R² were 0.04, 0.03, 0.96 respectively (Figure 19), indicating exceptional accuracy in water quality simulations (Table 5). These results highlight the model's robustness and its potential for optimizing network operations and ensuring compliance with safety standards.

The optimization of BC placement and dosage demonstrated the potential for significant improvements in maintaining residual chlorine levels across the WDN. In the IIT Roorkee WDN, the current system applies chlorine to the OHTs at a concentration of 0.5 mg/l over 24 h, with a total chlorine dosage of 1.775 kg/day across seven OHTs. However, this application method fails to maintain adequate residual chlorine levels throughout the network, highlighting a critical limitation of the existing system. This issue underscores the need for a more strategic approach to chlorine dosing and distribution. The proposed methodology for optimizing BC stations was developed using data from the existing system. By simulating various scenarios in WaterGEMS, we compared the existing system's performance with the optimized model. Key metrics chlorine concentration, water pressure, and velocity were evaluated. The results indicate that the optimized approach significantly enhances chlorine distribution, ensuring compliance with safety standards and improving overall system efficiency.

The five scenarios analyzed revealed that increasing chlorine dosage alone (scenario 1) is insufficient to maintain adequate chlorine concentrations throughout the network. However, the strategic placement of BC stations (scenarios 2–5) significantly improved chlorine distribution, with scenario 5 emerging as the most effective approach. By implementing four booster stations with optimized chlorine dosages, scenario 5 ensured that residual chlorine concentrations exceeded the minimum threshold of 0.2 mg/l across all observed junctions while minimizing overall chlorine usage. This approach not only enhances water quality but also reduces operational costs and environmental impacts associated with excessive chlorine use.

The LCC analysis of the BC scenarios provided valuable insights into the economic feasibility of different strategies. While scenario 1 (increased chlorine dosage without booster stations) had the lowest initial cost, its inability to maintain adequate chlorine levels makes it an unsustainable long-term solution. In contrast, scenarios 2–4, which involved the installation of booster stations, offered better performance but at higher costs. Scenario 5, despite having the highest initial investment, emerged as the most cost-effective option over 30 years due to its lower chlorine dosage requirements and improved system performance. The LCC analysis highlights the importance of considering both capital and operational costs when evaluating water distribution system improvements.

The findings underscore the importance of adopting a systematic approach to BC, considering factors such as network topology, flow patterns, and chlorine decay rates. Additionally, practical recommendations for implementing BC stations were provided, considering factors such as cost, feasibility, and population coverage. These recommendations are grounded in the analysis of the existing system, ensuring that the proposed solutions are both technically sound and practically viable.

The main objective of this study was to evaluate the existing chlorination approach and determine the optimal placement and number of BC stations, along with assessing the hydraulic performance of the WDN at IIT Roorkee, India. This was achieved using mixed-integer linear programming and WaterGEMS V8i software. The analysis identified that the existing chlorination approach was unable to maintain the recommended residual chlorine levels throughout the networks. The findings demonstrated that implementing four strategically placed BC stations (BC₁, BC₂, BC₃, and BC₄) at links P-124, P-67, P-9, and P-119, with a chlorine dosage of 1 mg/L, effectively addressed the challenges of maintaining residual chlorine levels above the minimum recommended threshold of 0.2 mg/l across the WDN. This configuration reduced the total chlorine mass dosage to 3.075 kg/year and minimized the LCC to INR 101.90 million over 30 years, making it the most efficient and cost-effective solution. Hydraulic performance analysis revealed that most junctions maintained acceptable pressure levels during peak and minimum demand. However, 16.13% of junctions experienced suboptimal pressures (<15 m) during peak demand. Velocity analysis indicated that 14.29% of pipes fell below the minimum threshold, with a minor fraction exceeding the maximum desirable velocity. Despite these deviations, the optimized system improved water quality and operational efficiency significantly. This study provides a replicable framework for addressing disinfection and hydraulic challenges in urban WDNs, promoting robust, cost-efficient, and environmentally sustainable systems. Future research should consider the impacts of seasonal and long-term variations in water demand on pressure, velocity, and residual chlorine levels to refine BC station placement and system performance further.

We thank the IIT Roorkee water supply office for their invaluable support through this study. Their provision of essential data and allocation of labor resources greatly contributed to the successful completion of our research.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.