Optimization of water supply system operation

An optimization model has been developed to generate pump schedules in real-time operation for a complex water supply system by considering major difficulties posed by a complicated tariff; especially demand charges, discrete pump discharges, and other physical constraints. The model decomposes in space and time the system into several subsystems, and planning periods into operational periods, Progressive optimality is applied to solve a dynamic programming model. The pump discharges are discretized and arranged by heuristic methods in order to reduce the number of times pumps are switched on. In real-time operation, billing monthly demand is estimated by a long-term model and then incorporated in a short-term model to obtain an optimal daily schedule in a relatively short computational time. Application is made to the city of Pittsburgh's water supply system. To implement the model in making operational decisions, a computer software package is also coded. A simulation run on part of the system in a test year shows that the optimal solution could have saved 20% of the actual cost.