太陽能是未來主要的能源供給，但是其供電受天氣影響並不穩定。預測發電量是可行解決之道，配上適當的電力調度可使電力供給穩定，但是如何使其預測準確又是問題。本論文嘗試以支援向量回歸來進行預測，因為支援向回歸原理簡單，發展完整，容易改良預測結果。以台電跟中央氣象局過往資料進行預測，最後我們結合k-平均演算法和支援向量回歸可使預測平均絕對值誤差縮小為總裝置容量的7 % ，並與決策樹、類神經網路、k-鄰近法進行比較。之後進行延伸，嘗試預測台電即時機組發電量，也證實此方法實作是可行的。之後以插值的方式加上天氣預報來預測近日發電模型，不過其準確度受限於天氣預報。

Photovolatic will be the major power supply in the future, but it is not stable due to the weather condtion. Forecasting power output of PV system and optimal power dispatach can solve this problem, but how to accurate prediction? This thesis tries to predition based on support vector regression(SVR) and improve this method. The power data is collected from Taiwan Power Company and the weather data is collected from Taiwan Central Weather Bureau(TCWB). We use those historical data to forecast the PV output. Finally, we propose algorithm that is combined by K-Means Algorithm and SVR.It’s mean relative error is reduced to 7 %. This algorithm has better prediction accuracy than regression tree, K nearest neighbors regression and neural network. We extend this method to the online forecast. It still works, but needs to improve. Use the interpolation and weather forecast to predict the receant PV power output. Beacause of the inaccurate weather forecast, but its accuracy is limited by weather forecast.

致謝 III
摘要 IV
ABSTRACT V
目錄 VI
圖目錄 VIII
表目錄 XI
第一章 緒論 1
1-1 研究背景 1
1-2 研究目標 3
1-3 文獻回顧討論 3
1-4 章節摘要 3
第二章 支援向量回歸 5
2-1 前言 5
2-2 支援向量機 5
2-2-1 線性支援向量機 5
2-2-2 軟限制支援向量機 11
2-2-3 非線性支援向量機 13
2-3 支援向量回歸 17
2-3-1 Tube Regression 17
2-3-2 二次規劃 20
第三章 支援向量回歸預測 27
3-1 太陽能資料 27
3-2 時間序列 29
3-3 天氣資料 39
3-4 依天氣分類預測 42
3-4-1 k-平均演算法(K-Means Algorithm) 44
第四章 其他預測方法 48
4-1 回歸樹 48
4-2 K-鄰近法 (K Nearest Neighbors) Regression 51
4-3 Neural Network 53
4-3-1 Feedforward Neural Network 54
4-3-2 Recurrent Neural Network 54
4-3-3 Deep Recurrent Neural Network(DRNN) 56
第五章 更短期即時預測 58
5-1 前言 58
5-2 核三生水池光電更短期預測 60
5-3 天氣預報加入預測 63
第六章 結論與未來研究方向 67
6-1 結論 67
6-2 未來研究方向 67
參考文獻 68

[1]B. Peder, M. Henrik, and N. H. Aalborg, “Online short-term solar power forecasting,”Solar Energy, vol. 83, pp. 1772-1783, May. 2009.
[2]Y. Li, Y. Su, and L. Shu, “An ARMAX model for forecasting the power output of a grid connected photovoltaic system,” Reneweable Energy, No. 66, pp. 78-89, 2014
[3] H.-T. Yang, C.-M.Huang, and Y.-S,Huang, “A weather-based hybrid method for 1-day ahead hourly forecasting of PV power output,” IEEE Trans. Sustain. Energy, vol. 5, no. 3, pp. 917-926, Jul. 2014
[4]U.K. Das, K.S. Tey, M.Seyedmahmoudian, M.Y.I. Idris, S. Mekhilef, B. Horan, and A. Stojcevski, “SVR-based model to forecast PV power generation under different weather conditions,” Energies, vol. 10, no. 7, Article number:876, June. 2017
[5]J. Zeng and W. Qiao “Shout-term solar power prediction using support vector machine, ” Renewable Energy, vol. 52,pp. 118-127, 2013
[6]J. Shi, W. J. Lee, Y. Liu, Y. Yang, and P. Wang, “Forecasting power output of photovoltaic systems based on weather classification and support vector machine,” IEEE Trans. Industry Application, vol. 48, no. 3, pp. 1064-1069, May. 2012
[7]M.W. Ahmad, M. Mourshed, and Y. Rezgui, “Tree-based ensemble methods for predicting PV power generation and their comparison with support vector regression, ” Energy, vol. 164, pp. 465-474, 2018
[8]B. Jing, Z. Qian, Y. Pei, and J. Wang, “Ultra short-term PV power forecasting based on ELM segmentation,” J. Eng, vol. 2017,lss. 13, pp. 2564-2568, 2017
[9]A. Gensler, J. Henze, and B. Sick, “Deep learning for solar power forecasting an approach using autoencoder and LSTM neural network,” IEEE International Conference on System, Oct. 2016
[10]K.Y. Bae, H.S. Jang, and D.K. Sung, “Hourly solar irradiance prediction based on support vector machine and its error analysis, ” IEEE Trans. Power System, vol. 32, no. 2, pp. 935-945, Mar. 2017
[11]A. Mellit, and A. M. Pavan, “A 24-h forecast of solar irradiance using artificial neural network:application for performance prediction of a grid-connected PV plant at Trieste,Italy, ” Solar Energy, vol. 84, pp. 807-821, 2010
[12]A. Alzahrani, P. Shamsi, C. Dagli, and M. Ferdowsi, “Solar irradiance forecasting using deep neural networks,” Procedia Computer Science, No. 114, 2017
[13]J.R. Andrade, and R.J. Bessa, “Improving renewable energy forecasting with a grid of numerical weather prediction,” IEEE Tran. Sustainable Energy, vol. 8, NO. 4, pp. 1571-1580, Oct. 2017
[14]C. Voyant, G. Notton, S. Kalogirou, M. Nivet, C. Paoli, F. Mottem, and A. Fouilloy, “Machine learning methods for solar radiation forecasting: A review,” Renewable Energy, vol. 105, pp. 569-582, 2017
[15]A. J. Smola, and B. Scholkopf , “A tutorial on support vector regression,” vol. 14, Statistics and Computing, vol. 14, pp. 199-222, 2004
[16]林軒田,機器學習基石[Online]
https://zh-tw.coursera.org/learn/ntumlone-mathematicalfoundations
[17]C.-C. Chang, and C.-J. Lin, “LibSVM: A library for support vector machines,” Acm Trans. Inyell. Syst. Techil., vol. 2, no. 3, 2011,
[18] D. Cournapeau , Python Scikit-learn [Online]:
https://scikit-learn.org/stable/
[19] J. Perktold, S. Seabold and J. Taylor, Python Stastmodels[Online]:
https://www.statsmodels.org/stable/index.html
[20]中央氣象局觀測資料查詢[Online]:
http://e-ervice.cwb.gov.tw/HistoryDataQuery/index.jsp
[21] Python Keras[Online] : https://keras.io/
[22]L. Richardson, Python BeautifulSoup[Online]:
https://www.crummy.com/software/BeautifulSoup/bs4/doc/
[23]台灣電力公司機組發電量[Online]:
https://www.taipower.com.tw/d006/loadGraph/loadGraph/genshx_.html
[24]中央氣象局即時觀測資料[Online]:
https://www.cwb.gov.tw/V7/observe/
[25]中央氣象局未來天氣預報[Online]:
https://www.cwb.gov.tw/V7/forecast/
[26]台大大氣系，台大天氣資料[Omline]:
http://www.as.ntu.edu.tw/
[27]P. Ineichen, “A broadband simplified of the Solis clear sky model,” Solar Energy, vol. 82, pp. 758-762, 2008
[28]R. Marquez, and C. F.M. Coimbra, “Intra-hour DNI forecasting based on cloud tracking image analysis,” Solar Energy, vol. 91, pp. 327-336, 2013
[29]H. T.C. Pedro and C. F.M. Coimbra “Nearst-neighbor methodology for prediction of intra-hour global horizontal and direct normal irradiances,” Reneweable Energy, No. 80, pp. 770-782, 2015