查詢結果分析
來源資料
頁籤選單縮合
題 名 | 以類神經網絡模式分析颱風降雨與半分布並聯式水庫概念模式模擬颱洪歷線之串聯應用=Cascaded Applications of the Artificial Neural Network Model for Typhoon Rainfall Analysis and the Semi-Distributed Parallel-Typed Reservoir Conceptual Model for Flood Simulation |
---|---|
作 者 | 王如意; 謝龍生; 嚴玉書; | 書刊名 | 農業工程學報 |
卷 期 | 44:2 1998.06[民87.06] |
頁 次 | 頁1-25 |
分類號 | 328.55 |
關鍵詞 | 類神經網絡模式; 颱風降雨統計預測模式; 半分布並聯型線性水庫逕流模式; 倒傳遞演算法; 洪災消減; Artificial neural network model; Statistical forecasting model of typhoon rainfall; Semi-distributed parallel-typed reservoir conceptual model; Back propagation network; Flood mitigation; |
語 文 | 中文(Chinese) |
中文摘要 | 本研究之目的乃建立颱風降雨類神經網絡模式,並搭配架構完備且演算過程精簡 之半分布並聯型線性水庫逕流模式,組成一模擬效果良好、計算時間迅速之颱風降雨-逕流 預報體系。 在颱風降雨推估方面,本研究採用颱風降雨統計預測模式之概念,結合影響颱風降雨之 各項物理因子,並以具函數間非線性映射能力之類神經網絡倒傳遞演算法作為模式架構,據 以建立颱風降雨類神經網絡模式。影響颱風降雨之因子包括:颱風位置、暴風半徑、中心氣 壓、中心最大風速、平均移動速度、移動路徑等颱風本體因子與當地氣壓、風向、風速、比 溼度、相對溼度等區域氣象因子,以及雨量站高程等地形特性因子。由上述各因子組成模式 輸入向量,而模式之輸出值則為降雨量,再經範例訓練找出最佳之隱藏層連結架構,即可建 立颱風降雨類神經網絡模式。 於逕流模擬時,本文採用半分布並聯型線性水庫逕流模式。其係根據集水區河川網絡與 土地覆蓋、利用等資料,將集水區劃分成多個核胞,則區內之降雨傳輸型態將遵循數條流路 流達出口。模式中將每一流路中各核胞之漫地流與河渠流視為不同之狀態,而每一流路則比 擬為由數個不同狀態所串聯而成者,其中,每一狀態係利用一線性水庫予以比擬。如此,每 一流路之反應即為若干不同線性水庫之串聯結果,而該集水區出口處之反應乃由所有流路之 反應並聯且線性疊加而形成者。 為驗證文中所研擬颱洪預報系統之合適性,本研究以新店溪秀朗橋上游集水區為研究對 象。由各雨量站之颱風降雨推估結果以及秀朗橋水文站逕流模擬結果顯示,本研究所建立之 颱洪預報體系具有良好之精確性。因此,本研究所建立之颱風降雨類神經網絡模式確有其實 用性,兼以與半分布並聯型線性水庫逕流模式串聯,可及時模擬出精度尚佳之洪水歷線,臻 以提供本省颱洪預報與預警工作之參考應用。 |
英文摘要 | Two adequate models are included in this study to build up a flood forecasting system. They are the artificial neural network (ANN) model for typhoon rainfall analysis and the semi- distributed parallel-typed linear reservoir conceptual model for flood simulation. The first part of the flood forecasting system is to estimate typhoon rainfall with the ANN model. In this study, the back propagation (BP) network, one of the ANN models, is used because of its capability of dealing with nonlinear mapping. The input vector of the ANN model of typhoon rainfall contains the following three kinds of components. The first kind of components, classified as the characteristic factors of the typhoon, includes the position, radius, central pressure, central maximum wind speed, average velocity and path of the typhoon. The second category, classified as the local atmospheric factors, consists of the local pressure, wind direction, wind speed, specific humidity and relative humidity. The third type of component, classified as the geomorphologic factor, includes only the local elevation. The output of the ANN model is typhoon rainfall. The optimum structure of the hidden layer(s) of the ANN model can be obtained by repeated training of the case studies. The second part of the flood forecasting system is to simulate flood with the semi-distributed parallel-typed linear reservoir conceptual model. Based on the pattern of the stream network and the landuse, a project watershed can be divided into several cells. The overland flow and the channel flow of each cell are considered as a specific path to the exit of the watershed. Each path is regarded as the convolutional integral of several linear reservoirs of different states. The total runoff at the exit of the watershed is the sum of the runoff of all paths. To verify the appropriateness of flood forecasting system, data records of several typhoon events in the upstream watershed at Hsiu-Lang Bridge of Hsin-Tien Creek are used. By comparing the estimated results of rainfall and runoff with the observed records, the flood forecasting system created by this study shows good capability for practical use. Consequently, this study can be further adopted to the estimations of rainfall and runoff during typhoon periods in project watersheds and play a referential role on the planning of flood mitigation in Taiwan. |
本系統中英文摘要資訊取自各篇刊載內容。