Determining the Contamination Source in Water Distribution Networks Using Genetic Algorithm

Contamination of drinking water is known as a major threat of water security around the world. As contamination enters a water distribution network, it spreads rapidly into the network and poses health and safety risks to the community. Using a set of sensors to report the concentration of chlorine or any other chemical, useful observations can be made to detect, identify and manage pollution. Based on these observations, location, concentration and start time of contamination can be determined and decision makers can be informed. In this paper, a simulation-optimization approach is used to solve the problem of contamination source characterization in which the EPANET software is used as a simulator and the Genetic Algorithm is used as an optimizer. The model developed in this paper is implemented on EPANET example 3. Modeling of water distribution networks uses information as input data which can cause error in model simulation. Pipe roughness and chlorine deterioration rate are among these inputs. The model has been implemented to find the location, start time and concentration of inlet pollution and the effect of pipe roughness and chlorine deterioration rate on the model responses have been investigated. The pollution entry scenario is applied to the network and the model presented is accurate in finding the location and time of the contamination. As the variables increase, the model accurately estimates the location and time of entry of the contamination but does not have complete accuracy in estimating the concentration of contamination, which is calculated with standard deviation of σ = 4.8% -8.1%.