چكيده به لاتين
In this study, dynamic consequence management in water distribution networks (WDNs) due to pollution injection is considered using different approaches. For this reason, EPANET simulation model is embedded in several optimization algorithms such as Dynamic Programming (DP), single objective Genetic Algorithm (GA), and multi-objective Non-dominated Sorting Genetic Algorithm-II (NSGA-II). To calculate the spatial and temporal extent of contamination in WDNs, the EPANET 2 simulation model is employed. EPANET is a well-established open-source software that can be easily linked with any optimization approaches through its toolkit. This study considers three different objective functions in its modeling scheme including minimization of response actions, consumed contamination mass and total number of polluted nodes. The best operation of time-varying consequence management should be selected among several potential locations of valves and hydrants. In this study, three approaches are developed for dynamic concequence management in WDNs following contamination detection. In the first approach, consequence management in WDN is evaluated based on Pressure Driven Analysis (PDA) and compared with Demand Driven Analysis (DDA). Implementation of a consequence management strategy by changing modes of operation for nominated valves and hydrants may modify the topology of the network. Any change in topology may cause pressure-deficient condition in the network, thereby reducing the actual water withdraw. Therefore, a consequence management plans achieved by PDA may evidently be more realistic and reliable in pressure-deficient cases and may readily be used for operational conditions in WDNs. In the second approach, the optimization procedure allows each valve and hydrant to change its operation mode at each time step considering the adverse effects on human health. It was revealed that in dynamic cases, the decision variables may vary significantly from one stage to another, resulting in a flexible operating strategy that would reduce the total contaminant mass consumption compared to the static case. And finally in the second approach, applicability of clustering in large-scale WDNs for recognizing contaminated zone and reducing computational time is evaluated. The clustering method is based on flow direction in the network. The contaminant distribution within the cluster is controlled by reoperation of the connecting pipes and nominated hydrants inside the cluster. Results show the appropriate performance of clustering in consequence management problems for large scale WDNs.
