Journal of Water Resources Planning and Management

This study applied a generalized likelihood uncertainty estimation (GLUE) approach to a Hydrological Simulation Program—Fortran (HSPF) model used for the simulation of hydrology and transport of fecal coliform bacteria at Moore’s Creek, Va. The procedures of the GLUE application to Moore’s Creek HSPF model include: (1) a Latin hypercube sampling approach to generate multiple parameter sets from a prior specified parameter space; (2) utilization of multiple calibration measures to evaluate the performance of HSPF predictions associated with each parameter set and definition of acceptance criteria to select the acceptable parameter sets; (3) a fuzzy logic model to calculate the likelihood values of the acceptable parameter sets; and (4) estimation of the distributions and construction of the uncertainty bounds for HSPF flow and fecal coliform predictions. Among 50,000 randomly generated parameter sets, 381 parameter sets are accepted. These parameter sets can lead to various HSPF flow and fecal coliform simulations closely matching the observed data given the observation errors. The results suggest that the multiple acceptable parameter sets identified by GLUE provide a sound basis for decision makers to evaluate uncertain hydrologic and water quality responses to a potential watershed management option.

Developed and demonstrated in this paper is an ant colony methodology extending previous work on ant colony optimization for least-cost design of gravitational water distribution systems with a single loading case, to the conjunctive least-cost design and operation of multiple loading pumping water distribution systems. Ant colony optimization is a relatively new meta-heuristic stochastic combinatorial computational discipline inspired by the behavior of ant colonies: ants deposit a certain amount of pheromone while moving, with each ant probabilistically following a direction rich in pheromone. This behavior has been used to explain how ants can find the shortest path between their nest and a food source, and inspired the development of ant colony optimization. The optimization problem solved herein is through linking an ant colony scheme with EPANET for the minimization of the systems design and operation costs, while delivering the consumers required water quantities at acceptable pressures. The decision variables for the design are the pipe diameters, the pumping stations maximum power, and the tanks storage, while for the operation—the pumping stations pressure heads and the water levels at the tanks for each of the loadings. The constraints are domain pressures at the consumer nodes, maximum allowable amounts of water withdrawals from the sources, and tanks storage closure. The proposed scheme is explored through base runs and sensitivity analysis using two pumping water distribution systems examples.

Peak daily water demand forecasts are required for the cost-effective and sustainable management and expansion of urban water supply infrastructure. This paper compares multiple linear regression, time series analysis, and artificial neural networks (ANNs) as techniques for peak daily summer water demand forecast modeling. Analysis was performed on 10 years of peak daily water demand data and meteorological variables (maximum daily temperature and daily rainfall) for the summer months of May to August of each year for an area of high outdoor water usage in the city of Ottawa, Canada. Thirty-nine multiple linear regression models, nine time series models, and 39 ANN models were developed and their relative performance was compared. The artificial neural network approach is shown to provide a better prediction of peak daily summer water demand than multiple linear regression and time series analysis. The best results were obtained when peak water demand from the previous day, maximum temperature from the current and previous day, and the occurrence/nonoccurrence of rainfall from five days before, were used as input data. It was also found that the peak daily summer water demand is better correlated with the rainfall occurrence rather than the amount of rainfall itself, and that assigning a weighting system to the antecedent days of no rainfall does not result in more accurate models.

A model of pipe wall and hydraulic profile effects on free chlorine decay in a water distribution system is developed using a two-step parameter assignment method. The model employs a Bayesian statistical method and Monte Carlo Markov chain to reflect wall decay coefficients as used in EPANET and is demonstrated to provide an efficient approach for EPANET water quality model calibration. The findings indicate that the wall decay coefficients and overall decay rate are largely influenced by flow velocity, pipe diameter, and pipe roughness. The largest impacts are demonstrated on unlined cast iron pipe, and lesser impacts progressively in ductile pipe, and then polyvinylchloride pipe.

This paper presents the development of a dynamic artificial neural network model (DAN2) for comprehensive urban water demand forecasting. Accurate short-, medium-, and long-term demand forecasting provides water distribution companies with information for capacity planning, maintenance activities, system improvements, pumping operations optimization, and the development of purchasing strategies. We examine the effects of including weather information in the forecasting models and show that such inclusion can improve accuracy. However, we demonstrate that by using time series water demand data, DAN2 models can provide excellent fit and forecasts without reliance upon the explicit inclusion of weather factors. All models are validated using data from an actual water distribution system. The monthly, weekly, and daily models produce forecasting accuracies above 99%, and the hourly models above 97%. The excellent model accuracy demonstrates the effectiveness of DAN2 in forecasting urban water demand across all time horizons. Finally, we compare our results with those of an autoregressive integrated moving average model and a traditional artificial neural network model.

The efficient and effective search for the optimum design solution of a water distribution network using genetic algorithms (GAs) is governed by several factors such as representation scheme, population size, hydraulic simulation model, fitness function, penalty method, GA operators, number of generations, and more importantly the size of the search space. This paper proposes a modified GA that uses basic operators along with their derivatives randomly. Further, a methodology based on critical path method is suggested to reduce the search space. A software tool, GA-WAT, based on the proposed methodology is developed and first tested and verified for its efficiency and effectiveness on two previously published single source networks. Later, it is applied to the optimal design of a larger, two-source hypothetical network. The results obtained indicate that the modified GA with reduction in search space proposed herein is more effective, especially for large practical networks.

This work presents a novel approach for solving groundwater management problems with reduced computational effort. We replace a groundwater flow model governed by a partial differential equation with a simple model governed by an ordinary differential equation. Model reduction is achieved with empirical orthogonal functions, i.e., principal components. Replacement of the full-scale model by a reduced model allows implementation of the embedding approach for optimal groundwater management. Comparing the results obtained with the full-scale simulation model, preliminary analyses show that the reduced model is able to reproduce head variations in the flow domain with good accuracy and, to a certain degree, the sensitivities of head with respect to pumping. A key advantage of the reduced model is that it is simple and easy to solve, and in many instances captures the dominating characteristics of the original model. In view of the many sources of uncertainty influencing groundwater simulation, the accuracy provided by a reduced model may be sufficient for planning purposes. As with other examples of model reduction presented in recent research efforts, the methodology shows promise in presenting general trends, but does not eliminate the need for the original model when more detailed analyses are needed.

A calibrated simulation model is linked with two optimization models to investigate alternatives for enhancing seawater intrusion barrier operations for the Alamitos Barrier Project in Los Angeles. Two types of management problems are analyzed, the optimal scheduling problem (OSP) and the optimal well location problem. The OSP objective is to minimize the total injected water subject to constraints on the state variables: Hydraulic head and chloride concentration at target locations. Two OSP formulations are considered, a pure hydraulic gradient formulation, and a combined hydraulic and transport formulation. When considering all 43 injection wells over a five-year planning horizon, the simulation-optimization model could not significantly improve upon the assigned initial injection rates. However, if a subset of the injection wells is exclusively considered, more favorable injection policies are obtained where less water is injected, compared with either the mean or annual mean derived from the historical record. Next, a genetic algorithm (GA) is linked with the calibrated simulation model to determine the locations of new injection wells that maximize one of two alternative fitness functions, which quantify barrier improvement. Parallel processing is implemented to accelerate the convergence of the GA.

A method for near-optimal real-time on-line operation of urban water distribution system is presented and demonstrated. It uses a reduced model (RM) of the network which reproduces its performance over time with high fidelity with optimization by a genetic algorithm (GA). The RM-GA software uses network data, forecasted demands for an operational planning time-horizon ( $24\mathrm{h}$ ahead), field data from a SCADA on the current status of the network, time-of-day energy cost data, and operator-imposed constraints on tank water levels and demand junctions pressures. The GA looks for the near-least-cost operation plan for the whole time horizon, using the reduced network model, and activates the first hour of the plan through the SCADA. At the end of the hour, the status of the network is updated from field data, the time horizon is rolled ahead by $1\mathrm{h}$ , and the process repeated. The software was developed for and demonstrated on the distribution system of Haifa. Experiments were conducted, in which the operation which simulates the operators' set-point operation is compared with that obtained by the RM-GA, and energy cost savings of 8 and 10% were calculated for a summer and winter month, respectively.

A water distribution system was analyzed for optimal sensor placement based on different intrusion scenarios and optimization criteria. The spatial distributions of the selected sensor networks were analyzed and compared using a geographic information system to determine spatial trends in sensor placement. Frequency, average nearest neighbor, and spatial autocorrelation analyses indicated the different intrusion scenarios and optimization criteria created networks with sensors in similar locations and with the same placement order, particularly for the first few sensors placed. Thus, sensor networks based on different optimization criteria and attack scenarios are expected to offer similar protection to the water distribution system. The relationship between sensor location and water demand was also analyzed using a geographic information system and a chi-square analysis. Sensor locations selected by minimizing the volume of consumed contaminated water or minimizing the population affected are likely to coincide with network nodes with a high reachable average demand. Alternatively, sensor locations selected by maximizing the detection likelihood are likely to coincide with network nodes with a low reachable average demand.

In the present technical note, a simple methodology for identifying the source location of an accidental contamination in a water distribution network is formulated. Starting from measured solute concentration data and using the pollution matrix concept, the methodology first selects a group of candidate nodes, which are possible intrusion points. Among all candidate nodes, the source location is then identified, minimizing the differences between simulated and measured concentrations. The source location procedure, which incorporates a hydraulic network simulation model, is formulated as an optimization problem, linearized using the water fraction matrix concept. An example illustrates methodology effectiveness. An uncertainty analysis is performed in order to demonstrate the methodology robustness with respect to uncertainties in some input data, like concentration measurements and water demands.