Algorithm Design Based on Derived Operation Rules for a System of Reservoirs in Parallel

It is challenging to obtain optimal operation rules for a system of reservoirs in parallel. This study derives generic optimal operation rules from a multistage nonlinear programming (NLP) model established for a system of reservoirs in parallel with a single demand. The optimal operation depends on the system's capability in coordinating individual reservoir storages to regulate inflows to those reservoirs and it is interfered by the full-empty cycles of reservoirs due to the variability of the inflows and/or the limitation of storage capacities. In general, small reservoirs are often made empty to leave space for future inflow or full to reserve water for the future, while large reservoirs are operated within their capacities. Based on the derived optimality conditions, a computationally efficient algorithm is developed for the NLP solution, which determines the system level releases and identifies all full and empty states and stages with no release for individual reservoirs in the system. The algorithm is demonstrated through a case study, in which the solution accuracy and computation time are compared with dynamic programming. This paper and subsequent studies on developing algorithms for optimal reservoir operation can potentially relieve the stress on water supply when new emerging sectors such as biomass and biofuel production increase water use.