臺灣博碩士論文加值系統

本研究主要探討基於機器學習進行校園用電量預測之應用。對現代社會來說，能源有如空氣一樣的重要，電能更是現代人們不可或缺的一部分。而電費的契約更是五花八門，延伸發展出的電費計算方式其中之一就是契約容量。主要應用在學校與大用電戶。
本研究使用國立宜蘭大學數位電表的資料來進行所謂的機器學習，訓練資料分別有2017年以及2018年電表數據資料，測試資料則包含2018一整年與2019年一月至五月電表資料，經前處理後以每小時一筆數據共12312筆數據，接著以不同特徵時段與演算法，並以平均絕對百分比誤差(Mean absolute percentage error)做為模型的效能指標，並以最佳模型做出預估電表值，進而做出預估的契約容量值。最後再以2018年、2019年的實際用電數值做出比較，從數據來看平均每個月可以節省22萬元電費，說明本研究的方法可以達到節能減碳愛地球的目的。

This study focuses on the application of campus power consumption forecasting based on machine learning. For modern society, energy is as important as air, and electricity is an indispensable part of modern people. The contract for electricity bills is even more varied. One of the ways to calculate the electricity bill is using the contract capacity, which mainly used in schools and heavy electricity users.
In National Ilan University, we use digital Watt-hour meters to hourly collect power consumption. The collected data is used to predict power consumption by using machine learning techniques. The training data includes the data collected in year 2017 and 2018. The test data includes the data collected in year 2018 and the January to May in 2019. After pre-processing, a total of 12,312 data records are obtained. Then, we apply different machine learning algorithms, include decision tree, support vector machine (SVM), k-nearest neighbors (KNN), subspace algorithms, to predict hourly power consumption. We adopt the mean absolute percentage error (MAPE) as the performance matric to choose the best model from these machine learning algorithms. Based on the predicted power consumption, we are able to estimate a better value of contract capacity. Numerical evaluation results have shown that our proposed method can save up to 220,000 NTD per month on average, which indicates the study can achieve the purpose of energy efficiency and carbon reduction.

摘要 I
Abstract II
致謝 III
目錄 IV
圖目錄 VI
表目錄 VIII
第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 1
1.3 研究方法 2
1.4 論文架構 3
第二章 文獻探討 4
2.1 機器學習演算法 4
2.1.1 決策樹演算法 4
2.1.2 SVM演算法 5
2.1.3 KNN演算法 7
2.1.4 子空間判別演算法 8
2.2 台電用電 9
2.2.1 不同時間用電 10
2.2.2 最高需量對電費成本影響 11
第三章 用電量預測演算法 17
第四章 電腦模擬結果 22
4.1 資料來源 22
4.2 模擬結果 23
4.2.1 使用訓練一整年的資料(8760筆)測試一整年資料(8760筆) 23
4.2.2 使用訓練一整年的資料(8760筆)測試5個月資料(3552筆) 24
4.2.2.1 以2017一整年的訓練資料測試2019年5個月的資料為例 25
4.2.2.2 以2018一整年的訓練資料測試2019年5個月的資料為例 26
4.2.3 使用訓練二年的資料(17520筆)測試5個月資料(3552筆) 26
4.3 模擬結果延伸 27
4.3.1 預估2018年1月至12月節省量 28
4.3.2 預估2019年1月至5月節省量 29
4.4 實際應用狀況 31
第五章 結論與未來展望 32
5.1 結論 32
5.2 未來展望 32
附錄1 33
附錄2 38
附錄3 48
參考資料 53

[1] 郭千瑜。2013。智慧電網中以戶為單位之用電特徵分析。碩士論文。國立臺灣師範大學資訊工程學系。
[2] 陳束弘、林政廷、蔡宗成、張語軒。電力耗能負載預測與節能應用。Journal of Taiwan Energy Volume 1, No. 5, December 2014, pp. 601-612
[3] S. S. Gavankar and S. D. Sawarkar, "Eager decision tree," 2017 2nd International Conference for Convergence in Technology (I2CT), Mumbai, 2017, pp. 837-840.
[4] P. Sutheebanjard and W. Premchaiswadi, "Fast convert OR-decision table to decision tree," 2010 Eighth International Conference on ICT and Knowledge Engineering, Bangkok, 2010, pp. 37-40.
[5] H. Xie and F. Shang, "The study of methods for post-pruning decision trees based on comprehensive evaluation standard," 2014 11th International Conference on Fuzzy Systems and Knowledge Discovery (FSKD), Xiamen, 2014, pp. 903-908.
[6] W. Dawoud, I. Takouna and C. Meinel, "Elastic VM for rapid and optimum virtualized resources' allocation," 2011 5th International DMTF Academic Alliance Workshop on Systems and Virtualization Management: Standards and the Cloud (SVM), Paris, 2011, pp. 1-4.
[7] B. Sanjaa and E. Chuluun, "Malware detection using linear SVM," IFOST, Ulaanbaatar, 2013, pp. 136-138.
[8] Chang, C.-C. & Lin, C.-J. (2001). LIBSVM: a library for support vector machines. Software available from http://www.csie.ntu.edu.tw/~cjlin/libsvm.
[9] 林巧玲，2000，「短期電力負載需求之整合模型建構與預測」，碩士論文，國立臺灣大學工業 工程學研究所。 
[10] Mohammad M. Hoque, Mohammad A. Hannan, Azah Mohamed, "Optimal algorithms for the charge equalisation controller of series connected lithium-ion battery cells in electric vehicle applications", IET Electrical Systems in Transportation, vol. 7, no. 4, pp. 267-277, 2017.
[11] Wen-Jing Shen, Han-Xiong Li, "A Sensitivity-Based Group-Wise Parameter Identification Algorithm for the Electric Model of Li-Ion Battery", IEEE Access, vol. 5, pp. 4377-4387, 2017.
[12] Rui Xiong, Jiayi Cao, Quanqing Yu, Hongwen He, Fengchun Sun, "Critical Review on the Battery State of Charge Estimation Methods for Electric Vehicles", IEEE Access, vol. 6, pp. 1832-1843, 2018.