Prediction of reservoir water quality by using artificial neural network: a case study of Chlorophyll-a in Shihmen Reservoir

Abstract

本研究採用水庫月資料(從2004至2008年)預測石門水庫水體水質因子，預測時利用類神經網路及傳統回歸分析兩者方法，以影響因子視為模式之輸入參數，預測項目視為輸出因子，其中類神經網路係採用典型之倒傳遞類神經模式。本研究以葉綠素a為水質主要預測目標，找出相關影響因子建置適用於石門水庫優養化之預測模式，最後探討模式預測能力。結果顯示類神經網路預測當月份及下一個月份之效果優於線性回歸分析。 Abstract: This study analyzed monthly records of water quality from Shihmen Reservoir from 2004 to 2008 and predicted the reservoir water quality parameter by using both artificial neural network (ANN) and traditional linear regression method. The reaction behavior models were adopted to construct the relationships between the input and output variables. The main water quality parameter, chlorophyll-a, was analyzed to be used in our predictive model. These models then applied to predict the chlorophyll-a in Shihmen R