A combined stochastic dynamic programming-statistical disaggregation approach applied to multiple reservoir systems

The stochastic dynamic programming (SDP) method faces computational difficulties when used to determine the optimal operation of multiple storages. A new approach, a combined SDP-statistical disaggregation approach is introduced to determine releases for a special situation relating to multiple reservoir systems, that is, for a system of multiple storages where operational data are available. The approach consists of defining an equivalent single reservoir which represents the system of multiple reservoirs. The optimal releases from the equivalent single reservoir are derived by the use of SDP. Disaggregation of the optimal releases from the equivalent single reservoir, to produce the releases from the individual storages is based on historical operational data. The Melbourne (Australia) water supply system is considered as the example. The releases derived from the combined SDP-statistical disaggregation approach are tested by operating a simulation model, and the conclusion is made that the approach produces satisfactory releases for a system of multiple reservoirs where operational data are available. The method cannot be applied to existing systems where insufficient or no operational data are available, or to proposed systems where operational data are not available. The method uses a small amount of computer time.