應用類神經網路於集水區與水庫之水質預測分析

Application of Artificial Neural Network for Forecasting the Water Quality in the Watershed and Reservoir

林信宏(Sin-Hong Lin)

指導教授 ： 徐年盛

國立臺灣大學/工學院/土木工程學研究所/碩士(2008年)

https://doi.org/10.6342/NTU.2008.02280

摘要

台灣地區水庫淤積及優養化問題嚴重，每經颱風豪雨過後，水庫集水區總會有土壤沖蝕或土石崩塌的情形，而大量的汙染物及營養鹽(磷、氮等)隨著暴雨逕流及泥砂沖刷一起進入水庫庫區內，使之產生嚴重的淤積現象，並造成原水的混濁及水體之透明度降低，以及水質惡化，而可能讓水庫呈現優養化的狀態，並且水庫庫容減少縮短水庫的壽命。本研究之目的為利用類神經網路(Artificial Neutral Network, ANN)模擬降雨時在水庫與其集水區內，所造成之泥砂、營養鹽(氮、磷等)及藻類之生長情形。
本研究以石門水庫為例，先收集前人研究集水區模式BASINS的資料，再結合水庫模式CE-QUAL-W2模擬水庫內藻類及營養鹽的生長情形，最後利用倒傳遞類神經網路取代複雜的集水區BASINS及水庫CE-QUAL-W2模式，以便可即時推估水庫及集水區泥砂、營養鹽產量及藻類生長的情形。
在集水區模式BASINS方面，類神經網路之輸出值相關系數皆大於0.67，且峰值並不會相差太大，趨勢也能夠符合；在水庫模式CE-QUAL-W2方面，其類神經網路訓練的結果相關係數大都有0.80以上，而驗證的結果也大多有0.65以上的相關性，可見若有連續資料，可應用倒傳遞類神經網路進行對水庫水質的連續預測。

關鍵字

優養化 ； 水質 ； CE-QUAL-W2 ； 類神經網路 ； 石門水庫

並列摘要

Eutrophication and soil deposition problem can affect water quantity and quality in the reservoir. In a heavy rainfall of typhoon, a river with a large number of pollution and nutrient flow into the reservoir, it can reduce the reservoir’s volume and cause eutrophication problem. It will affect water chemistry’s properties and quantity. Certain species of algae cause taste and odor problems in drinking water. These problems become prominent as the water body becomes more eutrophic.
The purpose of thesis is using ANN (Artificial Neural Network) to predict the suspended solid、nutrient (nitrogen and phosphorous) and algae in the watershed and reservoir.
ANN is used in this study to take the place of the BASINS and CE-QUAL-W2 model to real time forecast the water quality in Shi-men watershed and reservoir. In the BASINS, ANN trains Discharge、SS、PO4、NH3-N and NO3-N. The correlation coefficients are all over 0.67. In the CE-QUAL-W2, ANN trains Chl-a、PO4、NH3-N、NO3-N and DO. The correlation coefficients are all over 0.65. This shows ANN can use the continuous data to forecast the water quality at the moment in the watershed and reservoir.

並列關鍵字

eutrophication ； water quality ； CE-QUAL-W2 ； ANN ； Shi-men reservoir

參考文獻

1. Brion et al., “Using neural networks tp predict peak Cryptosporidium concentrations,” Journal AWWA, pp.99-105, 2001.

3. Chen, C. W. and Orlob, G. T., “Ecological Simulation of Aquatic Environmental,” Chapter 12 in: Systems Analysis and Simulation in Ecology, Vol.3, B.C. Patten, Academic Press, New York, 1975.

4. Di Toro, D. M., O’Connor, D. J., and Thomann, R. V., “A dynamic model of the phytoplankton population in the Sacramento-San Joaquin Delta,” Advances in Chemistry Series, American Chemical Society, Vol.106, pp.132-180, 1971.

7. Garvey, E., Tabiason, J.E., Hayes, M., Wolfram, E., Reckhow, D.A. and Male, J.W., “Coliform Transport in a Pristine Reservoir: Modeling and Field Studies, “Water Science and Technology, Vol.37, No.2, pp.137-144,1998.

8. Gunduz, O., Syoupak, S., and Yurteri, C., ”Development of Water Quality Management Strategies for the Proposed Isikli Reservoir,” Water Science and Technology, Vol.37, No.2, pp.369-376, 1998.