智慧化動床推估模式之發展與應用-以濁水溪為例

Development and Application of Intelligent Mobile-bed Estimation Model - A Case Study on Zhuoshui River

楊琬茹(Wan-Ju Yang)

指導教授 ： 葉克家 ； 吳祥禎

國立交通大學/工學院/土木工程系所/碩士(2020年)

摘要

臺灣河川具坡陡流急之特性，集水區地質多屬砂、頁、板岩，質地脆弱，易崩塌，導致河流土砂生產量大，加上近年受到氣候變遷與極端降雨之影響，河道沖淤情形加劇，造成河道沖淤失衡、流路變遷。以往為了降低易致災河段發生災害的風險，電腦數值模式（如CCHE1D）被大量應用於河川輸砂之研究，提供防洪設施之河道穩定策略，如在危險河段中建置水工結構物，評估堤防高度是否足夠因應颱洪事件等。但在氣候變遷的衝擊下，降雨量往往出乎預料，預先進行的災害風險分析也無法確切描述災害程度，故即時評估流域水文及地文現況對河道沖淤變化之影響顯得格外重要。
本研究以濁水溪流域寶石橋至河口段為研究範圍，採用AI技術之類神經網路模式建置智慧化動床推估模式。因實測資料有限，故蒐集多變量蒙地卡羅法及隨機降雨序列機制衍生的多組事件帶入動床數值模式CCHE1D的模擬結果做為本模式資料來源。根據建置完成的ANN模式模擬結果可知，本研究發展的智慧化動床推估模式可短時間量化在不同水文、地文及輸砂因子的條件下，河道各斷面的最終底床沖淤狀況。未來應用方面可結合降雨預報系統，提前預警可能發生淤積過量或沖刷嚴重的河段，讓政府單位及沿岸居民做好完善的災前整備工作及災中應變作為。

關鍵字

CCHE1D ； AI ； 類神經網路模式 ； 災害預警 ； 河川輸砂

並列摘要

Fluvial sedimentation and riverbed scour are common issues in Taiwan’s river because of steep terrain, fragile geology and torrents of rain that caused by typhoon. In general, in order to reduce the occurrence of fluvial sedimentation hazards, the numerical mobile-bed simulation model is widely applied in evaluating the riverbed change under various hydrological and geographical variables that provides governing agency to conduct river stability plan, such as reinforcement of spur dikes and groynes, etc. However, in recent years, Taiwan has influenced by climate change, and the rainfall intensity and magnitude of heavy rain has increased. It is uncertain to guarantee that the critical hydraulic structures could ensure the flood discharge capability under the changing climate. Accordingly, it is necessary to predict the real time riverbed change for early warning in order to reduce the impacts of extremely upcoming rainfall.
In this study, Baoshi Bridge to the Zhuoshui River estuary is chosen as the study area. The purpose is to establish an intelligent bed-elevation estimation (IBEE), which is simple and quick model to predict the riverbed change by AI technique (ANN). The data used in IBEE model, including multiple rainfall events and relevant factors which are intended to predict the riverbed change are generated by multivariates Monte Carlo simulation approach. In addition, due to the limited measurable riverbed elevation data, the corresponding riverbed changes are estimated through CCHE1D model, a 1D numerical mobile-bed simulation model, with the generated rainfall events. The result shows that IBEE model can quantify the riverbed change under consideration of factors, including rainfall factors (i.e., average rainfall depth and maximum rainfall intensity), physiographical factors (i.e., initial riverbed elevation and roughness coefficient) and sediment factors (i.e., coefficients of discharge-sediment rating curve and parameters of CCHE1D model). For further application, it could combine the rainfall forecast system to provide the early warning for reaching with serious deposition or scour, and the sophisticated pre-disaster preparation and disaster response.

並列關鍵字

CCHE1D ； early warning ； artificial intelligence (AI) ； ANN ； fluvial sedimentation

參考文獻

43. 呂宜瑾(2019)，「淺析台灣人工智慧醫療之發展」，取自https://portal.stpi.narl.org.tw/index/article/10499

1. Armanini, A. and di Silvio, G. (1988). “A One-dimensional Model for the Transport of a Sediment Mixture in Non-Equilibrium Conditions.” Journal of Hydraulic Research, IAHR, 26(3), 275-292.

2. Bagnold, R.A. (1966). “An Approach to Sediment Transport Problem from General Physis.” Geological Survey Professional Paper.

3. Basser, H., Karami, H., Shamshirband, S., Jahangirzadeh, A., Shatirah, A., and Saboohi, H. (2014). “Predicting Optimum Parameters of a Protective Spur Dike Using Soft Computing Methodologies – A Comparative Study.”Computers & Fluids, 97, 168–176.

4. Bouzeria, H., Ghenim, A., and Khanchoul, K. (2017).“Using Artificial Neural Network (ANN) for Prediction of Sediment Loads, Application to the Mellah Catchment, Northeast Algeria.”Journal of Water and Land Development, 33,47-55.