Efficient Water Distribution Using Ant Colony Optimization

Choosing optimal parameters for efficient water distribution in regional water supply systems can be a challenge. There are many tunable factors that affect the efficiency and effectiveness of water supply, based on various objective functions: minimizing shortage, maximizing availability, minimizing spillage, or equalizing shortages among all demand nodes. The decision variables are the flow volumes in each of the arcs of the water distribution network. The number of combinations of parameters can grow exponentially so that a brute-force approach to find the optimal variables is prohibitively expensive and impractical. In this project, we propose an Ant Colony Optimization (ACO) solution. ACO has been successfully applied in many applications, especially those related to finding shortest paths, optimal schedules, best vehicle routing, etc. In this work, we use ACO in two benchmark case studies: the Hanoi water network and the New York City water supply tunnel system. The models are highly non-linear and the problem is formulated as discrete optimization. The success of ACO hinges on a well-tuned strategy to incrementally promote good solutions and penalize poor ones in the search process. We show how the performance of the ACO varies with different search strategies and different rules to incrementally update the quality of the path. In particular, we experiment with ACO methods that updates the quality of the current path visited as well as its neighboring paths. We compare our results to other approaches reported in the literature for water supply in Hanoi and New York City.