Fuzzy Inference Decision Rule for Optimal Reservoir Operation

Reservoir management is a key component in effective water supply management. In real systems, reservoir operation is typically governed by a decision rule, an input-output relationship that prescribes reservoir release as a function of inflows, storage and other inputs. In this study, fuzzy inference decision rules (FIDRs) for reservoir operation are developed. Due to their highly flexible nature that enables their capturing of highly nonlinear interactions, the FIDRs are expected to perform better than conventional rules. The parameters of the FIDRs are optimized by applying differential evolution to a reservoir simulation model. A comparison of the FIDRs against linear rules, nonlinear rules and artificial neural networks, applied to two scenarios of a hypothetical reservoir on the Ganges River, show the FIDRs to perform best.