Application of Achieve-Based Genetic Algorithm for Consequence Management of Contaminant Entering in Water Distribution Networks

In this research, for the first time, finding the optimal operation actions in WDN to decrease the optimization time is taken into consideration. Valve(s) and hydrant(s) are also employed for isolating and flushing the contamination out of the network. The proposed embedded simulation-optimization approach for consequence management in this study is compromised EPANET simulation model and archive-based non-dominated sorting genetic algorithm-II (NSGA-II). Two objective functions are considered in this paper. The first objective function, minimized numbers of field operational actions related to expenses of the optimal solutions, whereas the other one minimized “consumed contamination mass” take into account for public health and safety. 20 valves and 31 hydrants are designed to insulate the network and discharge pollution, respectively. Without a follow-up management program, the total amount of contamination consumed in the event of network contamination would be 81.3 kg. Using 15 reactive activities, the mass of contamination consumed has reached 60.6 kg. For extracting the Pareto front between these objective functions with general NSGA-II which is a constraint to a maximum of 15 operational actions, approximately 73 minutes is required. To decrease this optimization time, archive-based NSGA-II is taken into account. With an archiving concept, it is possible to not implement a simulation model for similar chromosomes. Sensitivity analysis on the archive capacity of 0, 100, 200, 500, 1,000, 2,000, 3,000, 4,000, and 5,000 chromosomes has been investigated. As an example, with an increase in archive capacity from 0 to 5000, the required time for extracting the optimal Pareto front is reduced from 73 to about 35 minutes, indicating a decrease of more than 50%. The results showed that if a small amount is selected for the archive capacity, for example, 50 or 100, the time required to extract optimal activities increases slightly relative to the base state. The reason for this is that if a small amount is selected for the archive capacity, part of the implementation time of the simulation-optimization model will be spent on finding similar chromosomes, and due to the low capacity of the archive, t is time to use the archive capacity. Using the archive, it is possible to reduce the time optimization and consequence management of the network in real-time operation.