Journal of Water Resources Planning and Management

This paper investigates the application of multiobjective evolutionary algorithms to the identification of the payoff characteristic between total cost and reliability of a water distribution system using the well-known “Anytown” network as an example. An expanded rehabilitation problem is considered where the design variables are the pipe rehabilitation decisions, tank sizing, tank siting, and pump operation schedules. To provide flexibility, the network is designed and operated under multiple loading conditions. Inclusion of pump operation schedules requires consideration of water system operation over an extended period. The cost of the solution includes the capital costs of pipes and tanks as well as the present value of the energy consumed during a specified period. Optimization tends to reduce costs by reducing the diameter of, or completely eliminating, some pipes, thus leaving the system with insufficient capacity to respond to pipe breaks or demands that exceed design values without violating required performance levels. A resilience index is considered as a second objective to increase the hydraulic reliability and availability of water during pipe failures. Sensitivity analysis of solutions on the payoff curve generated by twin-objective optimization shows poor performance of these networks under random pipe failure or a pump being out of service. The minimum surplus head is added as a third objective to overcome the shortcomings of the resilience index. Results are presented for the payoff characteristics between total cost and reliability for $24\mathrm{h}$ design and five loading conditions.

Genetic algorithms (GAs) have been shown to apply well to optimizing the design and operations of water distribution systems (WDSs). The objective has usually been to minimize cost, subject to hydraulic constraints such as satisfying minimum pressure. More recently, the focus of optimization has expanded to include water quality concerns. This added complexity significantly increases the computational requirements of optimization. Considerable savings in computer time can be achieved by using a technique known as metamodeling. A metamodel is a surrogate or substitute for a complex simulation model. This research uses a metamodeling approach to optimize a water distribution design problem that includes water quality. The type of metamodels used are artificial neural networks (ANNs), as they are capable of approximating the nonlinear functions that govern flow and chlorine decay in a WDS. The ANNs were calibrated to provide a good approximation to the simulation model. In addition, two techniques are presented to improve the ability of metamodels to find the same optimal solution as the simulation model. Large savings in computer time occurred from training the ANNs to approximate chlorine concentrations (approximately 700 times faster than the simulation model) while still finding the optimal solution.

Optimal design and rehabilitation of a water distribution system is a constrained nonlinear optimization problem. A penalty function is often employed to transform a constrained into a nonconstrained optimization problem within the framework of a genetic algorithm search. A penalty factor is used for defining the penalty function and calculating the penalty cost for the solutions with constraint violation. Effective penalty factors vary from one optimization model to another. This paper introduces a self-adaptive penalty approach to artificially evolve both the penalty factor and the design solutions. Solution robustness is proposed and quantified, along with solution fitness, for representing the boundary between the infeasible and feasible solutions of a high-dimension optimization problem. The solution space of multiple dimensions is mapped onto a 2D space, providing significant insight into the self-adaptive penalty approach, which is compared with other constraint-handling techniques tested on a benchmark example. The results show that this approach is more effective than other penalty methods used for searching for optimal and near-optimal solutions. The self-adaptive penalty approach relieves modelers from tuning the penalty function and facilitates a practical optimization modeling for water distribution design.

The Metropolitan Water District of Salt Lake and Sandy recently began construction of the Point of the Mountain Aqueduct. This 1,524 mm (60 in.) diameter pipeline will convey finished water approximately 9 km ( $12\mathrm{mi}$ ) through mostly developed areas of two communities. Construction of a large-diameter transmission pipeline through heavily developed cities creates many engineering, construction, and public relations challenges. A geographic information system (GIS)-based route selection process was used to provide a rational basis for narrowing hundreds of potential alternatives into one final alignment corridor. The route selection process was based on construction costs as well as important noncost issues. The use of GIS data and GIS analysis software was critical to the success of this project. The GIS software allowed large amounts of pipeline cost-related data to be collected, stored, and documented for each alignment alternative. It was also used to analyze the network of possible alternatives to quickly determine the optimum route between two points based upon construction costs. This allowed for a logical selection and ranking of alternatives, resulting in one final alignment corridor that was acceptable to all of the stakeholders in the project.

The Massachusetts Water Resources Authority (MWRA) supplies wholesale unfiltered surface water to 2.2 million people in 43 communities, including Boston. Current MWRA treatment consists of adding hypochlorite for primary disinfection, soda ash and carbon dioxide for corrosion control, hydrofluosilicic (hexafluosilicic) acid for fluoride treatment, and hypochlorite and ammonia for maintaining distribution system residuals. The new Walnut Hill water treatment plant, scheduled to come on line in early 2005, will consolidate current treatment facilities under one roof and use ozone instead of hypochlorite as the primary disinfectant. It was observed from distribution system monitoring and bench-test results that the decay of chloramines was affected by prechlorination. The decay of chloramines released ammonia for subsequent nitrification. Complete nitrification and subsequent denitrification at current target alkalinity and combined chlorine levels affected pH. Increased microbiological activity associated with nitrification also decreased haloacetic acid concentrations. Bench and pilot tests indicate that the use of ozone will increase the rate of chloramine decay and will affect pH stability. Ozonation of natural organic matter produces organic acids, which typically decreases the pH in bottle tests. Results from a pilot plant, set up to simulate the new treatment plant with old cast-iron pipes simulating the distribution system, indicate that corrosion and biological activity inside the pipe have major impacts on water quality and that pipe surface reactions often cause pH to increase.

It has become generally accepted that water quality can deteriorate in a distribution system through reactions in the bulk phase and at the pipe wall. These reactions may be physical, chemical, or microbiological in nature. Perhaps one of the most serious aspects of water-quality deterioration in a network is the loss of disinfectant residual that can weaken the barrier against microbial contamination. Recent studies have suggested that one factor contributing to the loss of disinfectant residuals is internal corrosion of the pipe wall material. Recent studies have suggested that in older unlined metal pipes, the loss of chlorine residual may increase with increasing flow rates. To systematically assess the effect of free chlorine loss in corroded metal pipes, subject to changes in velocity, the authors conducted a study under controlled conditions in a specially constructed pipe loop located at the U.S Environmental Protection Agency’s (U.S. EPA’s) Test and Evaluation (T&E) Facility in Cincinnati, Ohio. Results from the pipe-loop study supported the concept that the rate of free chlorine residual loss increased with velocity.

Recent amendments to the Safe Drinking Water Act emphasize efforts toward safeguarding our nation’s water supplies against attack and contamination. Specifically, the Public Health Security and Bioterrorism Preparedness and Response Act of 2002 established requirements for each community water system serving more than 3,300 people to conduct an assessment of the vulnerability of its system to a terrorist attack or other intentional acts. Integral to evaluating system vulnerability is the threat assessment, which is the process by which the credibility of a threat is quantified. Unfortunately, full probabilistic assessment is generally not feasible, since there is insufficient experience and∕or data to quantify the associated probabilities. For this reason, an alternative approach is proposed predicated on Markov latent effects (MLE) modeling, which provides a framework for quantifying imprecise subjective metrics through possibilistic or fuzzy mathematics. Here, the MLE approach is introduced and demonstrated within the context of water supply system threat assessment.

A well-known network solver (EPANET) and a novel water-use generator (PRPsym) were linked in Monte Carlo computer experiments to simulate a deliberate biochemical assault on a municipal drinking-water distribution system. The attack was modeled as a steady 6-hour injection delivering 3,600 g of a soluble conservative contaminant to a single node on the main line in a small town. Migration of the contaminant plume was tracked for 55 hours throughout the pipe network, and the cumulative mass loading was computed at four target nodes strategically located on looping links and dead-end branches. This exercise was repeated for 1,000 independent trials to establish a baseline distribution of consumer dose exposures at the target nodes. A battery of simulation experiments was then performed to examine the sensitivity of the nodal load distributions to various system characteristics and water-use patterns. Results show that variability in the total mass load received at a node can be apportioned between the variability in the water-use volume and variability in the mean delivered concentration. Overall, however, the operation of the network storage tank had the greatest influence on the nodal mass loadings. This study demonstrates that Monte Carlo techniques are a useful tool for simulating the dynamic performance of a municipal drinking-water supply system, provided that a calibrated model of realistic network operations is available.

We present a model for optimizing the placement of sensors in municipal water networks to detect maliciously injected contaminants. An optimal sensor configuration minimizes the expected fraction of the population at risk. We formulate this problem as a mixed-integer program, which can be solved with generally available solvers. We find optimal sensor placements for three test networks with synthetic risk and population data. Our experiments illustrate that this formulation can be solved relatively quickly and that the predicted sensor configuration is relatively insensitive to uncertainties in the data used for prediction.

The effects of a partial blockage on pipeline transients are investigated analytically. A partial blockage is simulated using an orifice equation, and the influence of the blockage on the unsteady pipe flow is considered in the governing equations using a Dirac delta function. A simplified, linear dimensionless governing equation has been derived, and an analytical solution expressed in terms of a Fourier series has been developed under nonvarying boundary conditions. The linear analysis indicates that pipe friction and a partial blockage both introduce damping on fluid transients. The friction damping and blockage damping are exponential for each of the individual harmonic components. For each individual harmonic component, the blockage-induced damping depends on the blockage magnitude and position and is also independent of measurement location and the transient event. A new blockage detection method using the blockage-induced transient damping is developed based on the analytical solution. The magnitude of the blockage-induced damping rate indicates the size of the blockage, and the ratios of different damping rates can be used to locate the blockage. The proposed blockage detection method has been successfully used in detecting, locating, and quantifying a pipe blockage based on laboratory experiments.