Scenario-based approaches to flood control operations support

Flood control reservoir operations typically involve forecasts of reservoir inflows and uncontrolled tributary flows at points downstream of the dam. Since such forecasts often are accompanied by appreciable uncertainty, operators may be interested in the results of scenario, or what if, analysis when reservoir operating rules allow for some discretion. One approach to scenario analysis is to determine an optimal release schedule for each flow scenario considered. While intuitively appealing, this approach may do little to help the operator select the particular schedule that must be implemented. A better approach is to determine a release schedule that optimally hedges against the identified contingencies. To this end, a two-stage decision model of the flood control operating problem is developed. It is assumed that flow data for the first stage are known with certainty, corresponding to a short-term forecast, while data for the second stage comprise the scenarios of interest, corresponding to probabilistic forecasts for periods further in the future. Scenario generation, objective function formulation, and computational issues are discussed. Copyright 2004 ASCE