1. Introduction
Disinfection is one of the most important water treatment methods applied along the drinking water supply chain as it provides safe, free of pathogens, water to the end-users [
1]. Water distribution networks (WDNs) are complex systems comprised of several tanks, valves, several kilometers of pipes, and other assets. To manage such complex systems is actually a very hard and challenging task, as there are conflicting decisions to be made, in order to effectively achieve the ultimate goal of a WDN, which is to provide water of good quality at adequate quantity and pressure to the consumers. Usually, a WDN experiences high water losses levels due to leaks and breaks in pipes, fittings, and connecting points. The most common way to reduce water losses and better manage a WDN is to divide it into district metered areas (DMAs) and apply pressure management techniques [
2,
3]. Small areas within a WDN that are hydraulically isolated with one inlet point and one outlet point form a DMA. The division of a WDN into several DMAs has several benefits such as the reduction of accidental or intentional contamination impacts, limiting the contaminated area and the optimal placement of quality sensors to monitor water quality parameters [
3]. Usually, pressure reduction valves (PRVs) are installed at the entering node (inlet point) of a DMA to manage pressure. It is known that lowering the pressures in a WDN results in increased water age levels (being the time the water remains in the network), meaning that the quality of the water is deteriorating [
1]. It is commonly accepted that efficient chlorination in WDNs means that residual chlorine low concentrations should be present in all pipes and nodes in the network. The drinking water quality guidelines launched by the World Health Organization (WHO) (and also the national legislation in Greece) sets a down limit of 0.2 mg/L for residual chlorine concentration in the most remote parts/nodes of the WDN. The WHO provides guidelines for residual chlorine, setting the target of 0.2 to 1.0 mg/L in a WDN [
4]. Usually, chlorination takes place in water reservoirs or tanks, resulting in higher concentrations of residual chlorine near the chlorination points where the consumers report odor problems. At the same time, this chlorination practice (i.e., at the entering points of the network) results in decreased chlorine levels at remote parts/nodes of the network and pipe ends (dead-ends). It is also well known that excessive chlorination causes the formation of disinfection by-products (DBPs) in WDNs, especially at the remote parts. In Greece, the national legislation sets the maximum acceptable level of total trihalomethanes (THMs) to 0.1 mg/L [
5], complying with the former EU Drinking Water Directive 98/83/EC and the revised one 2020/2184 [
6]. Water utilities comply with this level without further investigation of chlorination level and its impact on DBPs’ formation. Thus, it is important to find the balance between efficient chlorination and DBPs’ formation.
The present paper presents a case study network of a small town in Greece, where the impact of DMAs’ formation is assessed on several scenarios where conventional disinfection processes and inline chlorination boosters are used. Initially, the water quality model of the WDN is developed on the basis of the hydraulic simulation model set. The paper aims to (1) investigate the impact of inline boosters on the chlorination’s efficiency, especially at remote parts/nodes of the WDN; and (2) investigate the impact of DMAs’ formation on chlorination scenarios.
2. Chlorination Boosters, Chlorine Residual Modeling, and THMs’ Formation
In 1999, Tryby et al. [
7] proposed the use of chlorination boosters to manage disinfectant residuals in WDNs. They arrived at the conclusion that applying inline booster chlorination results in reduced total chlorine mass used for disinfection and at the same time ensures effective chlorination at the remote parts/nodes of the network. Since then, several studies dealt with the optimization of boosters’ locations in WDNs [
7,
8,
9,
10,
11,
12]. Research teams have also addressed the issue of adequate residual chlorine concentrations and hydraulic performance of the WDN [
13,
14]. However, these studies, except for optimizing residual chlorine concentrations, took into consideration the maximization of water supply mass [
13] and operational functioning of pumps [
14]. Many studies used hydraulic and water quality analyses to obtain residual chlorine concentration results using re-chlorination. Several studies showed how pressure-driven analysis affects water quality analysis [
15,
16,
17]. Studies related to the optimization of DMAs’ formation took into consideration water age and water pressure using advance modeling techniques [
18,
19,
20]. These studies resulted in forming smaller DMAs, achieving a balance between pressure and water age.
A water quality analysis is based on hydraulic analysis of a WDN, as concentration changes of a chemical compound, water age, etc. are based on the results of the hydraulic analysis. Chlorine reacts with natural organic matter and inorganic substances in water, with the pipe material and the biofilm on pipes’ walls. Many kinetic models are used to describe chlorine decay in WDNs. These models are first-order or second or higher-order kinetic models. The most commonly used ones are the first-order kinetic models for chlorine decay [
12,
21] and for wall reaction. Researchers use these models for their simplicity and availability and because they represent chlorine decay in the WDN in a reasonable way [
21].
Equation (1) represents the first-order model describing chlorine decay for the bulk fluid [
22,
23]:
where C is the free chlorine concentration (mg Cl/L); t is the time (days or hours); and k
b is the bulk decay coefficient (days
−1 or hour
−1). There are several studies in the literature reporting first-order chlorine decay values ranging from 0.12 to 17.7 L∙mg
–1∙day
–1 at temperatures from 14 to 28 °C [
24]. The first-order chlorine decay model for pipe wall reactions is given by Equation (2) [
21]:
where k
w is the wall decay coefficient (m/day); r
h is the hydraulic radius (m); and C
w is the chlorine concentration at the pipe wall (mg/L). The concentration of chlorine at a pipe’s wall is a function of the bulk chlorine concentration, as Rossman et al. [
25] indicated. They also reported that the model describing chlorine decay at a pipe’s wall incorporates variations in pipe diameters and non-steady flow under turbulent and laminar flow conditions [
22]. The reaction coefficient k
w is the difference of the overall decay coefficient k
T and the bulk decay coefficient k
b [
21]:
Many researchers showed that THMs’ formation can be modeled using chlorine decay [
21] and developed a linear relationship:
where THM is the total THMs concentration (μg/L); Y is a yield parameter (μg of THMs formed per mg of chlorine consumed); and M is the intercept from linear regression analysis of experimental data [
18]. In their study, Carrico and Singer [
21] assumed that Y is equal to 40 μg of THMs formed per mg of chlorine consumed. This parameter depends on the chemical composition of water including the organic matter, temperature, and pH. There is a range of values in the literature; however, it can be estimated only per case using laboratory experiments.
3. Methods and Study Area
The hydraulic simulation software used in this study is Watergems V8i (Bentley Systems, Incorporated, Exton, PA, USA), which was used for hydraulic simulation and for water quality modeling. Specifically, the software provides water age analysis, trace analysis (calculating the water percentage coming from a specific node) and constituent analysis, which can be used for many different water quality parameters such as chlorine residual, total dissolved solids, etc. In order to model chlorine residual concentrations within a WDN, the value of diffusivity and the reaction order and rate for bulk reaction and wall reaction are necessary. The diffusivity value for chlorine suggested by Watergems is 1.208 × 10−9 m2/s. First-order kinetic models are assumed for bulk and wall reactions. The bulk reaction rate is −0.3 (mg/L)/day and the wall reaction rate is −0.305 m/day.
The study area is Eani district in the Kozani municipality in Greece, serving 2006 people. The average water volume entering the network is 1566.96 m
3 per day. The system consists of one reservoir, three tanks, 333 pipes of total length of 29,211 m, and 259 pipe junctions (
Figure 1). Most of the pipes in the system are polyvinyl chloride (PVC) pipes, and only two pipes of 102 m are made of high-density polyethylene (HDPE). The pipes’ diameters range from 57 to 203.4 mm. The study is based on the digital twin of the water distribution network already developed using Watergems. Domestic water demand follows a 24-h demand pattern, based on the literature (
Figure 2a). Non-revenue water components are simulated in the hydraulic model. Specifically, the apparent losses are added to the junctions’ demand proportionally, since their time distribution is similar to the domestic one. However, as real losses’ time distribution is similar to pressure distribution, real losses are allocated in the junctions as a separate demand with its own time pattern, which is inversely proportional to pressure (
Figure 2b). To allocate the demand to the network’s junctions, the spatial allocation of water demand at the street level method (SAWDSL) [
26] has been adopted. The digital twin has been calibrated and validated.
Water supplied to the network comes from 2 different water sources: (a) water from boreholes (modeled as a reservoir) is stored in tank T-2 and then enters the network with gravity, supplying the high zone and through the tank T-1 the lower zone and (b) water pumped from springs supplies the network through the tank T-1. There is no water treatment, except for chlorination at the entry nodes.
Table 1 shows the water volume abstracted.
The paper analyzes chlorine residual concentrations in the water distribution system when it is operating as one district metered area (DMA) (scenario A) and when it is divided in three DMAs (scenario B) (
Figure 1b). There is a pressure reduction valve (PRV) installed at the inlet point of DMA2, setting pressure to 300 KPa, in scenario B. For each scenario (A and B), conventional and booster chlorination processes are thoroughly analyzed. Conventional chlorination processes refer to the supply of chlorine in the water supply nodes (reservoir and springs), and booster chlorination processes refer to the installation of boosters inline the WDN, as supplementary chlorination devices. Initially, only the water sources are chosen as chlorine inlet/entering points, forming the first five scenarios for different chlorine concentrations. Then, using trial and error, the locations of boosters are chosen, and simulations are made for different chlorine concentrations. The inline chlorination boosters provide chlorination in WDN’s branches and in remote parts/nodes of the network. Watergems is used for the calculation of the chlorine concentration at each pipe and node.
Several nodes and pipes are chosen in the water distribution network to illustrate the chlorine behavior. Nodes and pipes near the water supply sources and nodes and pipes with high water age values are chosen. Pipes P-85, P-87, and P-146 are close to tank T-2, P-49 and P-51 are near tank T-1, and, pipes-10, pipes-32, P-64, and P-15 have increased water age ranging from 20.36 to 63.17 h (scenario A) and from 20.89 to 63.39 h (scenario B) (
Table 2 and
Figure 3). Nodes J-54, J-55, and J-77 are close to tank T-2, nodes J-35 and J-36 are near tank T-1, and nodes J-42, nodes-20, nodes-110, manholes-87, nodes-56, manholes-15, J-1, J-34, J-16, nodes-47, manholes-55, 495, J-7, and 473 are located close to remote parts/nodes of the network experiencing high water ages.
Table 2 shows the water ages for the pipes and nodes selected in both scenarios, A and B. The nodes and pipes close to the water sources are chosen as the whole network is fed from these points. It is evident from
Table 2 that when the network is divided in DMAs (scenario B), the water age gets higher values.
Figure 4a,b show the calculated water age of the pipes over a 288-h period (12 days). Water age values do not vary a lot for pipes P-49, P-51, pipes-10, P-64, pipes-32, and P-15. For pipes P-85, P-87, and P-146, water age values show a downwards trend after having reached their maximum values at about 15–20 h. This is because the initial water age in the tank T-2 is considered to be zero. Then, the tank supplies the network with water, and the water age values are lower. The same trend is evident from
Figure 4c,d, except for the initial water age in the tank T-2 considered to be 3.5 h.
5. Conclusions
The primary aim of the present study is to address (a) the impact of boosters’ chlorination on critical pipes in a WDN and (b) the impact of DMAs’ formation on critical pipes’ chlorination. A thorough analysis took place, including several chlorination scenarios, where chlorine is injected only at water sources or both at water sources and inline using chlorination boosters. The impact of chlorination boosters and DMAs is studied in critical pipes selected in the WDN. These pipes are located near the water sources and at remote parts in the network with high water age values. The model simulations revealed that the use of additional chlorination boosters provide a more uniform chlorination throughout the WDN, achieving effective chlorination of 0.2 mg/L without exceeding the maximum value of 0.5 mg/L, avoiding the formation of DBPs. The use of additional inline chlorination boosters allows for more effective chlorination compared to conventional chlorination taking place at the water sources of the network. Using chlorination boosters, there is a significant reduction of total chlorine dose compared to conventional chlorination. The formation of DMAs (and installation of PRVs) showed that although the operating pressure is lower and the WDN does not suffer from high water losses, water age values are increased, especially in remote pipes of the network. This fact results in delayed chlorination exposing the system to pathogens. However, the excess time needed for effective chlorination is not proportional to reduced pressure. Therefore, when water operators make decisions to improve the network’s performance level (such as DMAs’ formation and pressure reduction), it is important to achieve the proper balance between water pressure and effective chlorination. When water utilities want to optimize their water pressure, they should use residual chlorine as an additional criterion in optimization. Water utility managers should not underestimate both water age and residual chlorine during the decision-making process to reduce water losses, as water quality is important for consumers’ health.