成功大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	陳世昌 Chen, Shih-Chang
論文名稱：	水泥製程參數對污染排放影響之研究應用類神經網路與分群技術 Setting operation parameters for NOx control in cement kilns using neural networks and cluster technology
指導教授：	吳植森 Wu, Chih-Sen
學位類別：	碩士 Master
系所名稱：	管理學院 - 工業與資訊管理學系碩士在職專班 Department of Industrial and Information Management (on the job class)
論文出版年：	2007
畢業學年度：	95
語文別：	中文
論文頁數：	82
中文關鍵詞：	氮氧化物、燃煤水泥旋窯
外文關鍵詞：	Coal-fired cement kilns, Nitrogen oxides
相關次數：	點閱：54 下載：2
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

摘要
水泥工業乃國家重要的基礎工業，與國防及經濟發展息息相關，是各國不可或缺的基礎工業，惟水泥製造屬高耗能與高污染的產業，尤其水泥旋窯製程中需要燃燒大量煤炭，燃燒過程中產生了大量的空氣污染物，如二氧化碳、硫氧化物、氮氧化物及懸浮微粒，當它釋放於自然大氣中時，經光化學反應後，與大氣中的水分、氧氣及其他污染物結合形成硫酸與硝酸等酸性物質，是酸雨主要酸性成分的來源，酸雨造成土壤日益的酸化、農作物的死亡及威脅食物的安全等危害。
現行去除氮氧化物的技術，主要是燃燒程序的修正與排氣煙道後處理設備兩項技術，燃燒程序的修正為控制燃燒溫度、滯留時間、降低空氣預熱溫度及降低燃燒室熱負荷等方法，來減少氮氧化物的生成；或是設置排氣煙道後處理設備，將氮氧化物轉換成氮氣和水的媒介，如選擇性觸媒還原法或選擇性非觸媒還原法。燃燒程序的修正可能導致主燃燒區燃燒不完全、增加煙塵、一氧化碳排放及降低旋窯燃燒效率等缺點；排氣煙道後處理設備，選擇性觸媒還原法系統中的觸媒老化、廢棄處理、氨氣外洩及還原劑的高成本支出等相關問題，尋求燃燒效率與減少氮氧化物排放之間的最佳平衡點，是本研究的主要目的。
本研究應用類神經網路與分群規則技術，從連續排放監測系統與操作日報表資料庫中，挖掘出氮氧化物排放與水泥製造程序之間的相關規則，研究結果發現：
(1) 監督式分群模式較非監督式分群模式佳。
(2) 篩選合適的輸入因子，可有效地降低非監督式分群模式的誤判率。
(3) 可控因子的輸入分群模式較量測因子的輸入分群模式佳。
(4) 氮氧化物排放濃度與旋窯產量成正相關；與旋窯一氧化碳含量成負相關。
本研究發現之分群規則，可以完全解釋製程參數對氮氧化物排放的影響，有效降低氮氧化物的生成，達成低污染製程與節約能源的研究目的。

關鍵字：燃煤水泥旋窯、氮氧化物、選擇性非觸媒還原法、連續排放監測系統、規則、類神經網路

Abstract
Cement is essential material to national defense construction, the livelihood of people and industrial development. As cement production requires considerable input of coal and petroleum. The cement industry has placed a high priority on energy savings as a primary means of achieving cost reductions. However air pollution from cement kilns is responsible for some of the most pressing environmental problems today. These pollutants are extremely harmful to both public health and to the environment. For example, nitrogen oxides and sulfate oxides emission contributes to the increase or growth of photochemical smog and acid rain, to the enhancement of greenhouse effects. When released into the atmosphere, NOx with water, oxygen, and other cement kilns pollutants, such as NOx, will form acidic compounds. These acidic compounds return to the earth in the form of acid deposition, the resulting acid rain affects a lot of land, damaging crops and threatening food security.
The first action to control NOx emissions, also known as primary measures, normally takes the form of combustion modifications. Where the limits on NOx emissions cannot be met by combustion control, flue gas treatment must be installed. The dominant methods in use are SCR, selective non-catalytic reduction (SNCR) and combined processes for nitrogen oxide removal.
In this paper, the neural network (NN) and rules models are used to estimate the NOx emission of coal-fired cement kilns. The coal burned and the operation parameters can be changed toward the optimal low NOx combustion condition using the NN and rules technology.
Because the wide use of the continuous emission monitoring system (CEMS) and the a lot of operation variation in emission monitor in the cement kilns , a lot of input and output data can be used to train the NN. With the accuracy and robustness of an NN model is strongly influenced by the availability of training data, it is important that the training data must cover a wide range of the operation conditions.
The results generated show that the NN and rules model is able to consider the trade-offs between environmental requirement and economic objective. Thus, a balanced operation condition between environments protects and economic growth can be obtaining through our approach.

Keywords：Coal-fired cement kilns, Nitrogen oxides (NOx),
Selective non-catalytic reduction (SNCR),
Continuous emission monitoring system (CEMS),
Rules, Neural network(NN).

目    錄
摘要.................................................I
Abstract................................................II
目錄....................................................IV
表目錄................................................VIII
圖目錄..................................................IX
第一章  緒  論...........................................1
1-1  研究背景與動機......................................2
1-2  研究目的............................................4
1-3  研究流程............................................6
第二章  文獻探討.........................................8
2-1  氮氧化物特性........................................8
2-1-1  氮氧化物來源......................................8
2-2  氮氧化物生成機制....................................9
2-2-1  熱生成機制........................................9
2-2-2  燃料生成機制.....................................10
2-2-3  瞬時生成機制.....................................11
2-3  氮氧化物控制技術...................................11
2-3-1  燃燒前處理.......................................12
2-3-2  燃燒程序的修正...................................12
2-3-3  選擇性觸媒還原法.................................12
2-3-4  選擇性非觸媒還原法...............................12
2-4  小 結..............................................13
2-5  資料探勘...........................................15
2-5-1  資料探勘工具的選擇...............................16
2-5-2  資料探勘的技術...................................17
2-6  統計分析...........................................19
2-6-1  最近芳鄰法.......................................19
2-7  決策樹及規則.......................................20
2-7-1  類神經網路.......................................21
2-7-2  網路模式與訓練範例之選擇.........................22
2-8  自我組織映射圖網路.................................24
2-8-1  自我組織映射神經網路的相關應用...................26
2-9  類神經網路相關研究.................................27
第三章  研究方法........................................29
3-1  研究架構及步驟.....................................29
3-1-1  資料前置處理.....................................29
3-1-2  第一階段預測模式之建立...........................32
3-1-3  第二階段分群規則模式建立.........................33
3-1-4  結果彙整與分析...................................33
3-2  污染排放分類模式建置...............................34
3-2-1  類神經倒傳遞網路的學習法則.......................37
3-2-2  倒傳遞網路的學習.................................37
3-2-3  倒傳遞網路的學習演算步驟.........................38
3-3  自我組織映射圖網路.................................40
3-3-1  自我組織映射圖網路演算步驟.......................41
3-3-2  網路訓練及測試...................................42
3-4  數值型輸出的評估技術...............................43
3-5  運用規則產生器挖掘出污染排放之相關規則.............45
3-5-1  規則產生器之基本架構.............................45
3-5-2  非監督模式的規則解釋.............................46
3-5-3  規則解釋與評估...................................47
第四章  實例分析........................................48
4-1  研究公司背景簡介...................................48
4-1-1  連續排放監測系統的資料收集單元...................48
4-1-2  資料來源.........................................48
4-2  輸入資料屬性定義...................................49
4-3  倒傳遞網路預測模式驗證.............................51
4-3-1  倒傳遞網路分類模式架構...........................51
4-3-2  倒傳遞網路分類模式參數設定.......................53
4-3-3  倒傳遞網路模式預測結果...........................54
4-3-4  倒傳遞網路氮氧化物預測模式架構...................55
4-3-5  倒傳遞網路氮氧化物預測結果.......................56
4-4  自我組織映射圖網路分群模式驗證.....................58
4-4-1  自我組織映射圖網路分群參數設定...................58
4-4-2  自我組織映射圖網路分群結果.......................59
4-5  非監督式分群模式之驗證.............................61
4-5-1  非監督式分群模式之結果...........................62
4-5-2  非監督式分群模式之修正...........................66
4-6  小 結..............................................68
第五章  結論與建議......................................70
5-1  研究結論...........................................70
5-2  分群模式穩健度不佳的原因...........................71
5-3  研究發現與討論.....................................72
5-4  研究建議...........................................78
參考文獻................................................79

表 目 錄
表1-1：環保署公告固定污染源公私場所及應監測之項目........3
表2-1：旋窯燃料氮氧化物生成機制.........................10
表2-2：可行之燃燒控制改善方式...........................14
表2-3：資料探勘技術工具與任務的建議表...................17
表2-4：資料探勘技術分類預測準確率比較...................28
表4-1：輸入屬性資料統計表...............................50
表4-2：類神經網路訓練樣本集準確率之比較.................55
表4-3：類神經網路測試樣本集準確率之比較.................55
表4-4：類神經網路訓練組NOx預測值準確率之比較............56
表4-5：類神經網路測試組NOx預測值準確率之比較............57
表4-6-A：訓練資料子集的交叉驗證.........................59
表4-6-B：測試資料子集的交叉驗證.........................60
表4-7：高誤判率資料子集2進行不同學習率測試..............60
表4-8：修正分群模式之輸入因子...........................67

圖 目 錄
圖1-1：研究流程..........................................7
圖2-1：溫度與氮氧化物生成關係............................9
圖2-2：知識發現流程圖...................................15
圖2-3：資料探勘技術選擇.................................18
圖2-4：二元決策樹.......................................21
圖2-5：隱藏層神經元數目與預測精確度.....................23
圖2-6：自我組織映射圖網路模式...........................24
圖2-7：鄰近區域觀念模式.................................25
圖2-8：廢水申報資料之應用分析...........................26
圖3-1：研究架構圖.......................................30
圖3-2：資料前置處理.....................................32
圖3-3：三層類神經倒傳遞網路.............................34
圖3-4：類神經倒傳遞網路模式流程圖.......................35
圖3-5：概念樹狀架構.....................................46
圖4-1：連續排放監測系統(CEMS)架構圖.....................49
圖4-2：倒傳遞網路(BPN)架構圖............................52
圖4-3：隱藏層神經元數目之預測精確率.....................53
圖4-4：倒傳遞網路(BPN)預測模式參數設定..................54
圖4-5：類神經網路訓練組NOx預測值準確率之比較............57
圖4-6：類神經網路測試組NOx預測值準確率之比較............58
圖4-7：高誤判率資料集與交叉驗證資料集準確率比較.........61
                                    

參考文獻
中文部分：
Richard J. Roiger ; Michael W. Geatz 著;曾新穆、李建億譯：資料探勘 Data Mining，台灣培生教育出版股份有限公司，2003。
空氣污染防制專責人員訓練教材，行政院環境保護署，2002。
環球水泥CEMS系統操作手冊，2000。
黃怡華，應用類神經網路與關聯法則於銀行消費性貸款，成功大學碩士論文，2004。
盧瑞山，漫談資料探勘技術於環境資料庫之應用，環保訓練園地，第68期，2003。
網站部分：
行政院環境保護署網站：http://www.epa.gov.tw , 2006.02。
英文部分：
Adriaans, P. and Zantinge.,”Data Mining”, Addison Wesley, 1996.
Alsabti, K. Ranka, S. and Singh, V.,”An Efficient K-Means Clustering Algorithm”, PPS/SPDP Workshop on High performance Data Mining, 1997.
Akgun, F.,“Investigation of energy saving and NOx reduction possibilities in a rotary cement kiln”, International Journal of Energy Research 27, 455–465, 2003.
Anand, S. Keren, N. Tretter, M. J. Wang, Y. O’Connor, T. M. and Mannan, M. S.,“Harnessing data mining to explore incident databases”, Journal of Hazardous Materials 130 , pp.33–41,2006.
Berry, M. J. A. and Linoff, G.,“Data Mining Techniques”, John Wiley & Sons Inc, 1997.
Berson, A. Smith, S. and Thearling, K.,“Building Data Mining Application for CRM”, McGraw Hill, pp. 2–13, 1999.
Berry, M. J. A. and Linoff, G.,“Mining the Web: transforming customer data into customer value”,John Wiley & Sons, 2001.
Chelani, A. B. Gajghate, D. G. and Hasan, M. Z., “Prediction of ambient PM10 and toxic metals using artificial neural networks”, Journal of the Air and Waste Management Association 52, pp.805–810,2002.
Fayyad, U. M. Piatetsky-Shapiro, G. Smyth, P. and Uthurusamy, R.,“Advances in knowledge discovery and data mining”, Menlo Park, CA:AAAI/MIT Press, 1996.
Freme, F. L. and Hong, B. D.,“US coal supply and demand: 1999 review”, Energy information administration, 1999.
Haykin, S.,“Neural Networks: A Comprehensive Foundation”, 2nd edn. Prentice Hall,Englewood Cliffs, NJ, 1999.
Huang, Y. M. Hung, C. M. and Jiau, H. C.,“Evaluation of neural networks and data mining methods on a credit assessment task for class imbalance problem”,Nonlinear Analysis: Real World Applications , 7 ,pp.720–747,2006.
Jain, L. C. and Vemuri, V. R.,“Industrial applications of neural networks”,Boca Raton：The CRC press,1999.
Jiang, D. Zhang, Y. Hu, X. Zeng, Y. Tan, J. and Shao, D., “Progress in developing an ANN model for air pollution index forecast”,Atmospheric Environment,38,pp.7055-7064, 2004.
Kohonen, T.,”Self-organization and associative memory”, 3rd edn. Springverlag, Berlin Heidelberg,1989.
Katharina, D. C. S. and Dirk, U. P., “The application of non-parametric techniques to solve classification problem in complex data sets in veterinary epidemiology an example”, Intelligent Data Analysis,3(1),pp.23-35,1999.
Kusiak, A.and Kurasek, C.,“Data Mining of Printed-Circuit Board Defects”,IEEE Transactions on Robotics and Automation, Vol.17,NO.2, April,2001.
Kolehmainen, M. Martikainen,H.and Ruuskanen,J.,“Neural networks and periodic components used in air quality forecasting”, Atmospheric Environment 35, pp.815–825, 2001.
Kuo, R. J. Wang, H. S. Hu,T. L., and Chou, S. H., “Application of Ant K-Means on Clustering Analysis”,Computers and Mathematics with Applications,Vol.50 pp.709–1724,2005.
Lemieux, P. Stewart, E. Realff, M. and Mulholland, J. A., “Emissions study of co-firing waste carpet in a rotary kiln”,Journal of Environmental Management ,Vol.70,pp. 27– 33,2004.
Li, S. T.and Shue, L.Y.,“Data mining to aid policy making in air pollution management”, Expert Systems with Applications 27, pp.331–340,2004.
Medsker, L.,“Neural networks with application. Arlington”,VA：The Institute for Professional Education, 1997.
Mak, B. and Munakata, T.,“Rule extraction from expert heuristics：A comparative study of rough sets with neural networks and ID3”,European Journal of Operational Research,136(1),pp.212–229,2002.
Negnevitsky, M.,”Artificial Intelligence:A Guide to Intelligent Systems”, 2nd edn. Addison Wesley, 2005.
Wood, S. C.,“Select the Right NOx Control Technology”,Chemical Engineering Progress, Jan,pp.32–38, 1994.
Wey, W. M.,“An integrated expert system/operations research approach for the optimization of waste incinerator sitting problems”, Knowledge-Based Systems Vol.18,pp.267–278,2005.
Yi, J. C. Anandalingam, G. A.and Sorrell, L. A.,“An expert system to physician-detailing planning”, Expert Systems with Applications 25,pp.533–544,2003.
Zhou, Q. Huang, G. H.and Chan, C. W., “Development of an intelligent decision support system for air pollution control at coal-fired power plants”,Expert Systems with Applications Vol,26,pp.335–356,2004.
Zhou, H. Cen, K.and Fan, J.,“Modeling and optimization of the NOx emission characteristics of a tangentially fired-boiler with artificial neural networks”, Energy ,Vol.29,pp.167–183,2004.

校內：2009-08-28公開
校外：2012-08-28公開

簡易檢索 / 詳目顯示

相關論文