Reservoir management: A reliability programming approach

A reliability programming model is developed to determine the optimum monthly releases from a multipurpose reservoir. The model is based on chance‐constrained linear programming (CCLP) and dynamic programming (DP). The flood and drought reliabilities are represented in the CCLP in the form of chance constraints. These reliabilities are varied parametrically from their minimum required levels to their maximum possible values. The trade‐offs between flood and drought reliabilities are made using a forward two‐dimensional DP routine. The model does not require equal flood and/or drought reliabilities during the year. Folsom reservoir of the California Central Valley Project is modeled to demonstrate the use of this reliability programming approach. The resulting operation policy shows high reliabilities of flood and drought during winter and summer months, respectively. The variable risk levels avoid unnecessary releases during summer (for flood protection) and unnecessary storage of water during winter (for drought protection).