Optimization of water quality sensor locations in water distribution systems considering imperfect mixing

Nowadays, the quality of water becomes an important concern as the demand of better life increases. The accidental or deliberate intrusion of pollutants which may cause serious damage to community health into water distribution systems is a common operational problem. Several different algorithms and objective functions have been proposed in previous studies regarding the optimal locations of water quality sensors. However, accurate water quality modeling is also an important factor when simulating contaminant propagation in water distribution systems as it may affect the optimal location of sensors. Therefore, in this study, two modeling approaches are evaluated and compared to determine the optimal sensor locations. The first model is the perfect mixing behavior at cross junctions embedded in the EPANET 2.0 program which is popularly employed in previous studies. Then, AZRED 1.0 program using the imperfect mixing behavior - a new program modified from the free source code of EPANET 2.0 by the University of Arizona - is used to compare the results with EPANET. Multi-Objective Genetic Algorithm (MOGA) using Pareto-ranking approach for evaluating fitness is also applied on a water distribution system having a small number of cross-junctions. As a result, the optimal water quality sensor locations are not considerably different in the applied system because the number of cross-junctions is not large enough to cause substantial variations on the pollutant propagations using EPANET and AZRED. The looped type network having large number of cross-junctions has to be applied in further study.