應用深度學習與給水管戶籍化演算法統計分析漏水潛能
| dc.contributor | 張國楨 | zh_TW |
| dc.contributor | Chang, Kuo-Chen | en_US |
| dc.contributor.author | 陳佳瑩 | zh_TW |
| dc.contributor.author | Chen, Chia-Ying | en_US |
| dc.date.accessioned | 2022-06-08T02:52:14Z | |
| dc.date.available | 9999-12-31 | |
| dc.date.available | 2022-06-08T02:52:14Z | |
| dc.date.issued | 2022 | |
| dc.description.abstract | 臺灣降雨分布在地理位置與地形等時間及空間因素影響下，水資源難以保存，每逢枯水期「水」尤顯珍貴，除在平時呼籲民眾節約用水外，如何有效降低漏水率亦是一項重要的課題，本研究旨在探討如何規劃管線汰換區域與預測漏水位置以降低管網漏水率(研究範圍：臺北自來水事業處供水轄區)。本研究提出以邏輯斯迴歸(Logistic regression)與深度神經網路(Deep Neural Networks，DNN)演算法，針對給水管過去16年來修漏資料與現行給水管線空間資料分析漏水潛能影響特徵，透過演算法學習產生模型計算每支給水管漏水潛能，再搭配「給水管戶籍化演算法(Service-pipe Grouping Algorithm，SGA)」，使管線從「點」整合成「群」並結合圖面，另以街廓作為區域劃分，更加彈性且貼近給水管線埋設位置，可視化呈現漏水潛能。本研究結果顯示，針對給水管以深度神經網路演算法，透過分處、口徑、管材及管齡等特徵計算漏水潛能，可有效預測漏水位置，並結合給水管戶籍化演算法以街廓或分群呈現漏水潛能空間分布，協助規劃管線汰換優先順序。 | zh_TW |
| dc.description.abstract | The distribution of rainfall in Taiwan is affected by time and space factors such as geographic location and topography, making it difficult to preserve water resources."Water" is especially precious during dry seasons. In addition to appealing to the public to save water in peacetime, how to effectively reduce the water leakage rate is also an important issue. This study aims to explore how to plan pipeline replacement areas and predict the location of water leakage to reduce the leakage rate of the pipeline network. (The scope of research: Jurisdiction covers of Taipei Water Department) We use the method with Logistic regression and Deep Neural Networks (DNN)to find out the characteristics of the leakage cases data recorded during the last 16 years. We train the model, and then use the"Service-pipe Grouping Algorithm (SGA)" to integrate the leakage locations from "points" into "groups" and mapping with the pipeline and the street-division map to identify the location of the service-pipe with the high leakage potential visually. The result of this research shows that the Deep Neural Networks algorithm is used for the Service-pipe to calculate the water leakage potential through the characteristics of office, caliber, stuff and age, it can effectively predict the location of water leakage. And combined with the Service-pipe Grouping Algorithm, the spatial distribution of water leakage potential is presented by street profile or grouping, assist in planning pipeline replacement priorities. | en_US |
| dc.description.sponsorship | 地理學系空間資訊碩士在職專班 | zh_TW |
| dc.identifier | 008233110-40830 | |
| dc.identifier.uri | https://etds.lib.ntnu.edu.tw/thesis/detail/db335f1bde2919e87520a7986c3ad297/ | |
| dc.identifier.uri | http://rportal.lib.ntnu.edu.tw/handle/20.500.12235/117796 | |
| dc.language | 中文 | |
| dc.subject | 漏水潛能 | zh_TW |
| dc.subject | 管網分析 | zh_TW |
| dc.subject | 深度學習 | zh_TW |
| dc.subject | water leakage potential | en_US |
| dc.subject | Pipe Network Analysis | en_US |
| dc.subject | deep learning | en_US |
| dc.title | 應用深度學習與給水管戶籍化演算法統計分析漏水潛能 | zh_TW |
| dc.title | Application of deep learning and Service-pipe Grouping Algorithm to statistical analysis of water leakage potential | en_US |
| dc.type | 學術論文 |