Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/14101
標題: 應用倒傳遞類神經模式預測山坡地地下水位
Hillslope Area Groundwater Level estimate - a BPNN approach
作者: 鍾芸菁
Chung, Yun-Jing
關鍵字: Back-Propapation Neural Network;倒傳遞類神經模式;Geographic Information System;Hillslope Area Groundwater Level;地理資訊系統;山坡地地下水位
出版社: 土木工程學系
摘要: 
中文摘要
本研究選定在台中縣梨山及台北市五處山坡地,使用地理資訊系統建置影響因子資料庫,經因素分析選定影響地下水流發生之參數為累積降雨量、降雨強度、主流長度、平均高程等項。同時,本研究採用倒傳遞網路分析影響地下水水位變化因素之相關程度,並決定類神經網路結構之最佳參數,以針對集水區降雨與地下水水位之模擬學習、輸入影響因素個數及加入地文因子學習表現進行探討。
本研究於單一集水區使用轉換函數之參數作為類神經之輸入參數,可以使學習表現達到極高的相關性,以瞭解地下水位與降雨之互動關係,此顯示所選定之六個集水區分別皆可適用此方式進行模擬預測。另一方面,於加入地文因子後之地下水變化學習模擬中雖然相關性並不高,但其中三個集水區的誤差量較低,此亦顯示所選用影響參數仍具其代表性,且此模式在地下水位的變化模擬有較佳的學習表現。

Abstract
This study uses six hillside watersheds in northern and central Taiwan based on Geographic Information Systems technique to establish an impact factor database for groundwater fluctuation analysis. From the analysis, the parameters which affect the groundwater level can be summarized as the accumulate rainfall, rainfall strength, length of main stream and average watershed elevation. Also, through the Back-Propagation Neural Network (BPNN) analysis, factors which affect the groundwater level are analyzed and thus the most significant Neural network parameter can then be determined.
This study uses the transfer function in a single water basin through
Neural Network Propagation method to find out the interactive relationship between the groundwater level and the rainfall event. Moreover, this study reveals that similar prediction is also quite effective for Li-San watershed.
Through the entire analysis, the errors of simulation for three of the
selected watersheds are low enough although the groundwater levels do not show much relevance with the physiographic factors which are chosen. The results also show that the chosen impact factors are somewhat well representative and the method herewith is effective in the simulation of the groundwater level fluctuation.
URI: http://hdl.handle.net/11455/14101
Appears in Collections:土木工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.