以類神經網路建構石門水庫集水區泥砂濃度推估模式

Constructing the suspended sediment concentration model in Shihmen Reservoir watershed by artificial neural networks

陳逸鴻(Yi-Hung Chen)

指導教授 ： 張斐章

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.02319

摘要

河川高濃度的泥砂於颱風期間造成水庫淤積現象一直為集水區經營管理與水庫操作過程中重要之議題，因集水區泥砂之量測無法持續且穩定進行觀測，而傳統之回歸分析與泥砂率定曲線方法並無法提供有效而準確之模擬，故建立一精確之集水區泥砂濃度推估模式乃有迫切之必要性。本研究就此議題分為兩部份進行研究，首先，乃以調適性網路模糊推論系統建立石門水庫集水區霞雲水文站泥砂推估模式，並蒐集1982至2009年霞雲水文站流量、泥砂濃度以及上游16個雨量測站之日雨量觀測資料以分析及探討不同輸入變數對泥砂濃度之影響。結果顯示最佳之輸入變數組合為同時刻流量與前兩日累積雨量，而成果亦顯示調適性模糊推論系於泥砂濃度推估能力則較傳統方法準確。
第二部份之研究則著眼於調適性網路模糊推論系統建構石門水庫集水區羅浮水文站颱風期間之泥砂濃度預測模式。模式建構所需之資料採用石門水庫集水區之平均雨量、羅浮水文站入流量以及泥砂濃度，亦測試不同輸入變數組合並逐步找出最佳模式。結果顯示最佳之變數組合為前一小時入流量、前一小時與前兩小時泥砂濃度以及前七小時平均雨量。此結果可做為颱風期間水庫操作及水資源管理之參考，藉由精確之泥砂濃度預測可預先排放含有高濃度的流量，以減緩水庫之淤積速率並延長水庫的使用壽命。

關鍵字

調適性網路模糊推論系統 ； 類神經網路 ； 泥砂濃度 ； 推估 ； 預測 ； 颱風

並列摘要

High concentration of sediment in upstream river is one of the important issues that affect the effectiveness of water resources management and reservoir operations in watersheds during typhoon periods. The measurement of sediment in the river is difficult to continuously and effectively achieve because of time- and human- consuming. Furthermore, traditional methods such as regression analysis and sediment rating curve are not able to provide effective simulation of sediments. As a result, there is a necessity of establishing a precise model for suspended sediment estimation and prediction. The study can be divided into two topics, in which the first topic tries to construct the estimation of sediment concentration at Hsia-Yun gauging station by applying the Adaptive Network-Based Fuzzy Inference System (ANFIS). To analyze the relationship between hydrological variables and suspended sediment concentration, the daily streamflow, precipitation, and sediment concentration data recorded in the years of 1982-2009 were collected. The results not only showed that the best input combination of a model was consisted of current streamflow and the accumulated rainfall but also revealed that the performance obtained from ANFIS outperformed conventional methods in terms of model accuracy, when predicting daily suspended sediment concentrations.
The second topic focuses on the prediction of event-based suspended sediment concentration at Lo-Fu gauging station during typhoon periods using the hourly average rainfall, inflow rate, and suspended sediment concentration. By taking various input variables into account, the study successfully constructed a suspended sediment concentration prediction model by using ANFIS with the input dimensions consisting of antecedent one-hour inflow rate, antecedent one- and two-hour suspended sediment concentration, and antecedent seven-hour rainfall information. Overall, the study demonstrates that the performance obtained from ANFIS can be used as a reference to the management of reservoir operation and water discharge because typhoons always result in a high concentration of suspended sediment in both river and reservoir.

並列關鍵字

Adapted Network-Based Fuzzy Inference System ； Artificial Neural Network ； Suspended Sediment Concentration ； Estimation ； Prediction ； Typhoons

參考文獻

1. Atiya, A.F., S.M. El-Shoura, S.I. Shaheen, and M.S. El-Sherif. 1999. A comparison between neural-network forecasting techniques - Case study: River flow forecasting. Ieee Transactions on Neural Networks. 10:402-409.

2. Chang, F.J., and Y.T. Chang. 2006. Adaptive neuro-fuzzy inference system for prediction of water level in reservoir. Advances in Water Resources. 29:1-10.

3. Chang, F.J., and Y.C. Chen. 2001. A counterpropagation fuzzy-neural network modeling approach to real time streamflow prediction. J. Hydrol. 245:153-164.

4. Chang, F.J., and Y.C. Chen. 2003. Estuary water-stage forecasting by using radial basis function neural network. J. Hydrol. 270:158-166.

5. Chang, Y.T., L.C. Chang, and F.J. Chang. 2005. Intelligent control for modeling of real-time reservoir operation, part II: artificial neural network with operating rule curves. Hydrological Processes. 19:1431-1444.