Shankar Narasimhan S

Leakage in water distribution systems is a growing concern throughout the world. Huge amounts of water and energy is wasted through undetected leaks and unauthorized usage in water distribution networks alongside potential health hazards caused by the leaky pipes. As most of the pipelines are buried underground, detecting a leak becomes a complex task. This work focuses on application of graph partitioning technique to Intermittent Water Distribution Networks for identification of leaks without use of a hydraulic model. The network structure is exploited to find the minimum number of valves to be closed or pipes to be isolated, and volume balance is carried out at each stage of partitioning to isolate the leak to a desired level of network resolution. This methodology is applied to a custom built, unique laboratory experimental setup that reflects intermittent water distribution network. The results demonstrate that the leaks can be localized to a smaller part of large network with the application of repeated graph partitioning technique.

Water Distribution Networks in many rural areas supply water from storage reservoirs at the source of water to a group of villages through gravity or by pumping. Due to poor operational policies, they often fail to meet the demand of the beneficiary villages. This work proposes a rational technique for scheduling the supply in such networks to maximize the water delivered to the villages in an equitable manner. A secondary objective is to minimize the number of valve operations. The problem is formulated as a LP followed by an ILP. The parameters in the optimization problem are obtained by simulations of a well calibrated hydraulic model or experimental data if such a model is not available.

Optimal operation of water distribution networks can be posed as a scheduling problem where the objective is to meet the time varying demand while meeting constraints on supply, pressure etc. In the present work, we propose a robust optimization problem to address uncertainty in the parameters of the model used for optimization. The resulting problem is a second order cone program that can be solved efficiently. The formulation ensures a high probability of meeting the demands, adding to the practical significance. Further, we provide the results of applying this technique on a laboratory scale water distribution network.