Optimal weekly releases from a multireservoir hydropower system

Abstract

The operation of a multi-unit electric energy generating system is studied under certain and uncertain future inflow conditions. The generating units include thermoplants, hydroplants with regulating reservoir and run-of-river hydroplants. The objective is to minimize the expected cost of the operation of the system while meeting a previously defined energy demand. A case study is formulated based on the electric energy generating system of the South of Brazil. The system is composed of 6 hydroplants with regulating reservoirs, 2 run-of-river hydroplants, and 8 thermoplants. In order to obtain a better insight into the nature and peculiarities of the system's operation it is initially studied considering the future to be deterministic. An aggregation-optimization-disaggregation procedure is proposed to identify a near optimal solution while reducing substantially the computational effort. This consists of the development of an aggregated representation of the system composed of a hypothetical and unique reservoir with overall energy inflows and releases. Optimal operation of the aggregated system is determined by a new and efficient optimization technique specifically developed for this problem. A disaggregation procedure defines the operation of each system's unit given the operation of the aggregated system. The procedure is based on a heuristic approach that has as a main objective to minimize water spills. An aggregated representation of the system is again adopted for the definition of optimal strategy of operation when the future inflows are uncertain. The characteristics of operation of each reservoir are introduced into the aggregated formulation utilizing the peculiarities of the optimal deterministic operation. A modification of Massé's Chain of Marginal Expectations is used in the computations. The resultant strategy of operation can be presented as a function of aggregated values of energy storage and inflow. The strategy explicitly considers the autocorrelation of aggregated energy inflows. The strategy also implicitly accounts for the cross-correlations among the energy storages and inflows to each reservoir. Finally, a substantial part of the autocorrelation of the energy inflows and storages in each reservoir is indirectly considered in the strategy. Theoretical significance of the strategy is obtained without burdensome computational effort.