Abstract
This paper proposes a systematic approach to explore the search space of control parameters on an existing water network in a locality within Mysore, India, to see if continuous water supply can be provided to meet a nominal diurnal demand pattern. The control parameters are the initial states of the reservoirs at some reference hour and hourly maximum flow rates across control valves. The system state should return to its initial state after twenty four hours to follow a diurnal rhythm without overflow or underflow of reservoirs. The number of configurational changes should also be small. The paper proposes an optimisation framework that relaxes the constraints via convex penalty functions, performs approximate gradient descent to minimise the net penalty, and thus identifies a suitable configuration. The proposed framework is ideal for integration into a cyber physical system that can actuate the controls based on real-time demand and flow information.