Fuzzy Rule-Based Modeling of Reservoir Operation

Abstract

A fuzzy rule-based model is constructed to derive operation rules for a multipurpose reservoir. The case study of the Tenkiller Lake in Oklahoma illustrates the methodology. Operation rules are generated on the basis of economic development criteria such as hydropower; municipal; industrial and irrigation demands; flood control and navigation; and environmental criteria such as water quality for fish and wildlife preservation, recreational needs, and downstream flow regulation. The fuzzy rule-based model operates on an “if-then” principle, where the “if” is a vector of fuzzy explanatory variables or premises and “then,” of fuzzy consequences. The reservoir storage level, estimated inflows, and demands are used as the premises and release from the reservoir is taken as the consequence. Split sampling of historical data (mean daily time series of flow, lake level, demands, and releases) is used to train and then validate the rules. Different performance indices are calculated and two figures of merit, namely, engineering sustainability and engineering risk are developed for evaluating the rules generated by the model, which appears to be easy to construct, apply, and extend to a complex system of reservoirs.