Reservoir Flood Season Segmentation and Optimal Operation of Flood-Limiting Water Levels

Abstract

A flood season segmentation-based reservoir optimal operation framework was proposed to maximize the use of the flood control storage in flood seasons. The proposed framework consists of an ensemble-method approach to segment the flood season for improved reservoir operation, a stochastic simulation model to generate abundant inflow hydrographs, and a Monte Carlo technique to calculate the failure probabilities of flood regulation under different given flood-limiting water levels (FLWLs). The benefits of the proposed framework were illustrated through the operation of Chengbihe Reservoir in China. Results indicate that an ensemble-method approach of a fuzzy clustering method, a probability change-point analysis, and a statistical graphical technique provide more reasonable segmentations than that of an individual method alone. The flood season of the Chengbihe Reservoir is segmented into an early flood season (April 1–May 31), a main flood season (June 1–August 20), and a late flood season (August 21–October 31). The FLWL in the main flood season remains to be the value currently in use (185.00 m), whereas the FLWL in the late flood season can be lifted maximally up to 188.50 m, which is equivalent to an increase of the active storage by