Optimization of water distribution networks using metaheuristic methods

In our world, where the population is increasing rapidly and the existing water resources are damaged for various reasons, the need for water is gradually increasing. In this master thesis study, water distribution networks are optimized using metaheuristic optimization methods to reduce the cost of delivering water to consumers, which is of vital importance. The presence of constraints such as pressure, speed and flow in the network makes the optimization process quite complex. However, the Epanet-MATLAB toolkit developed by Marios Kyriakou and Demetrios Eliades enables an interactive bilateral data transfer between Epanet environment and MATLAB programming language, which makes the network optimization to be done in a relatively easier way. In the scope of this thesis, the following optimization processes were used to minimize the costs of two benchmark water distribution networks; genetic algorithm (GA), particle swarm optimization (PSO), ant colony optimization (ACO), harmony search (HS), differential evolution (DE), annealing simulation (SA), artificial bee colony (ABC), bio-geography based optimization (BBO), and invasive weed optimization (IWO). While the ABC algorithm was used in a few studies before, the last two optimization algorithms, namely, BBO and IWO, are used for the first time in this thesis study for the optimization of water distribution networks. For this reason, the results of the bio-geography-based optimization (BBO), invasive weed optimization (IWO), and the artificial bee colony (ABC) algorithms are mainly compared with the results of the other commonly used optimization algorithms in this thesis study. The benchmark networks Alperovits and Shamir, and Hanoi city networks were optimized under the same constraints to compare the performances of the optimization methods used. When compared to several previous studies, the previously reported lowest 419 000-unit cost for the Alperovits and Shamir network in the literature was achieved with lower numbers of evaluations using BBO and IWO algorithms in this study. Similarly, for the Hanoi city grid, a lower cost of $6.081x106 was achieved with lower numbers of evaluations with BBO and IWO algorithms used in this study when compared to some of the previous studies. It is concluded that bio-geography based optimization (BBO), invasive weed optimization (IWO), and artificial bee colony (ABC) optimization algorithms can perform better results than many algorithms that are widely used. Also, the Epanet-MATLAB toolkit was successfully used in the network optimization. Nevertheless, it was determined that the optimization methods could not give equally successful results without changing the control parameters in different networks. Therefore, it is concluded that the algorithm control parameters should be adjusted to the given specific problem. Keywords: Epanet-MATLAB toolkit, metaheuristic optimization algorithms, optimization of water distribution networks, invasive weed optimization, bio-geographical based optimization