Multi-objective Storage Reservoir Operation under Uncertainty

Abstract

水量・水質は，貯水池操作において考慮されるべき最も重要な要素である。本研究では，貯水池操作の新たな方法を提案し，Barra Bonita貯水池(ブラジル)に適用を行う。最適化手法と人工知能を用いることが特徴で，ファジイ確率的動的プログラミングを用い，多数のファジイ目的を最適化することにより適切な操作手順を算出する。水質解析は，ニューラルネットワークモデルにより行う。また，有機体や栄養物は，ファジイ回帰モデルにより河川流量の関数として表現する。

Water quantity and quality are considered to be the main driving forces the reservoir operation.Barra Bonita reservoir, located in the southeast region of Brazil, is chosen as the case study forthe application of the proposed methodology. Herein, optimization and artificial intelligence(AI) techniques are applied in the simulation and operation of the reservoir. A fuzzy stochasticdynamic programming model (FSDP) is developed for calculating the optimal operationprocedures. Optimization is applied to achieve multiple fuzzy objectives. Markov chaintechnique is applied to handle the stochastic characteristics of river flow. Water quality analysisis carried out using an artificial neural network model. Organic matter and nutrient loads aremodeled as a function of river discharge through the application of a fuzzy regression modelbased on fuzzy performance functions. The obtained results show that the proposedmethodology provides an effective and useful tool for reservoir operation.