Stochastic Gradient Approach for Optimization of Pumping Schedules in Complex Water Supply System

Energy saving in operation of water pumping plants and the minimization of water decit are important issues, which should be considered when optimizing multi-reservoirs and multi-users water supply systems. Undoubtedly, this problem is characterized by a high uncertainty level due to hydrologic variability and water demand behaviour. The aim of this paper is to provide an ecient DSS (Decision Support System) through the optimization of emergency water pumping plants activation schedules. The obtained results allow the water system's authority to get a robust decision policy, minimizing the risk of wrong future decisions. A cost-risk balancing problem has been modelled to manage this problem, in order to balance the damages in terms of shortage water occurrences and energy-cost requirements. These problems are aected by a considerable uncertainty level in input data. Besides, decisions can be aected and modied in response to climatic trends. We develop here a simulation model for this problem, which is optimized using the stochastic gradient methods. We have applied our model for optimization of water pumping in specic area in South-Sardinia (Italy) characterized by Mediterranean climate. The water supply system under consideration includes four pumping plants transferring water between ve storage reservoirs serving dierent types of water demands. By applying our combined simulation and optimization procedure we obtained both the energy-costs saving and the reduction of water decit to admissible level even during heavy drought periods