Application of Artificial Neural Network to Allocate Regional Water Resources

• Main Authors: Hwung, Yih-ming, 黃義銘
• Other Authors: FREDERICK N.-F. Chou
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/40194305348006761350

碩士 === 國立成功大學 === 水利及海洋工程學系 === 85 === The regional water resources system is getting complex day by day. In the mean time, water management personnel feels more complicate to distribute water to desired users favorably in the system. Better water management poli-cies may be analyzed by mathematical programming algorithm (MPA). However, thelarge- scale characteristics of regional water resources system prevents MPA''s from being applied efficiently to develop the optimal operation policies. Net-work flow model can properly represent all the essential elementof a reservoir-river basin water distribution system. In addition, applying the network fow programming (NFP) model to simulate the water distribution over a large- scale system has the advantage of light computation burden. A dynamic network flow model (DNFM) can optimally allocate the water of a regional water resources system. The optimal allocation obtained by DNFM considered future stream flow of a system. Due to forecasting error of river flow and no simple form of allocation policy, the DNFM is difficult to be ap- plied in site operation. A atificial neural network (ANN) has the capability of learning. It was applied to allocate regional water resource in this study. First, the DNFM was applied to analyze the optimal transbasin water transport between river basins of Kaoping Chi and Tsengwen Chi. The water allocation optimal of DNFM was then learned by ANN model. A well trained ANN model was validated to allocate the water for this regional system. The results showed most water supplies were correctly allo-cated.