運用資料探勘模型預測工業用能源最佳化輸出：以燃煤蒸汽鍋爐之產量為例;Predicting and optimizing industrial energy output with a data mining model: A case study of coal-fired steam boilers’ output

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/77590

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/77590

Title:	運用資料探勘模型預測工業用能源最佳化輸出：以燃煤蒸汽鍋爐之產量為例;Predicting and optimizing industrial energy output with a data mining model: A case study of coal-fired steam boilers’ output
Authors:	彭裕少;Peng, Yu-Hsiao
Contributors:	資訊管理學系
Keywords:	燃煤蒸汽鍋爐;資料探勘;最佳化;支援向量迴歸;粒子分群法;Coal-fired Steam Boiler;Data Mining;Optimization;Support Vector Regression;Particle Swarm Optimization
Date:	2018-07-10
Issue Date:	2018-08-31 14:49:12 (UTC+8)
Publisher:	國立中央大學
Abstract:	紡織工業因應市場需求與環保法規的限制，需要對能源使用與污染物排放等議題做出進一步的控管與改善。提供紡織工業製程使用之熱源多由蒸汽鍋爐所產生。此種燃燒控制系統可透過不斷校正鍋爐操作變數而達到燃燒的最佳化，進而降低汙染物排放以及提高能源輸出效率等效果。新興之資料探勘方法因資料蒐集成本降低、計算機計算能力提高等而逐漸發展蓬勃，並於工業能源領域之預測與優化擁有良好的研究成果。因此，本研究將使用支援向量迴歸與粒子分群法等資料探勘之技術，針對工業能源產出建立工業產量預測模型框架，並以燃煤蒸汽鍋爐為實際研究案例。研究所提出之預測模型將進一步部屬至Web-based資訊系統以呈現預測結果。由預測模型的建模實驗顯示，支援向量迴歸對於本研究預測之蒸汽壓與排氣含氧量具有良好的預測準確度與資料解釋力。最佳化實驗也顯示粒子分群法演算之鍋爐操作值可以有效增加蒸汽壓並降低排氣含氧量。本研究另外建立類神經網路預測模型與研究提出之支援向量迴歸進行準確度與資料解釋力的比較實驗。實驗結果顯示支援向量迴歸於準確度與資料解釋力皆顯著地大於類神經網路。粒子分群法同樣透過驗證實驗與基因演算法比較其優化效果。實驗結果呈現在有限時間內，粒子分群法尋找出的鍋爐操作變數較基因演算法而言更能增加蒸汽壓產出並降低排氣含氧量。;The textile industry needs to make progress in the improvement on energy usage and pollution control due to the requirement of global markets and the environmental regulations from the government. Heat sources for textile industrial processes are mostly provided by steam boilers. Combusting control systems for boilers can achieve the optimization of combustion though adjusting boilers’ operating parameters continuously. Recently, approaches of data mining receive more attention in the field of energy and widely employed to model for combusting control systems due to the low cost of collecting data and the improvement of computing power. Therefore, the purpose of this study is to use support vector regression (SVR) and particle swarm optimization (PSO), the methods of data mining, to establish a framework for predicting and optimizing the output of industrial machines. A case study of coal-fired steam boiler was conducted to construct the predicting and optimizing models, and then the models were deployed to a web-based information system to present better predicting results. In this study, two comparing experiments using artificial neural network (ANN) and genetic algorithm (GA) were additionally conducted to prove the accuracy of SVR and the capability of PSO. The experiment results show that the accuracy of SVR for predicting steam pressure and oxygen content of boiler is significantly better than ANN. PSO, in the given computing time, also has the significantly better performance than GA in improving steam pressure and reducing oxygen content of boiler.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	158	View/Open

社群 sharing

Loading...