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：	施昊沅 Shih, Hau-Yuan
論文名稱：	運用希爾柏特黃轉換進行彰濱場址海氣象分析及預測 Sea Meteorological Analysis and Prediction at Chanbin Sites by Using the Hilbert Huang Transform
指導教授：	林大惠 Lin, Ta-Hui
共同指導教授:	謝志敏 Hsieh, Chih-Min
學位類別：	碩士 Master
系所名稱：	工學院 - 機械工程學系 Department of Mechanical Engineering
論文出版年：	2023
畢業學年度：	111
語文別：	英文
論文頁數：	117
中文關鍵詞：	希爾伯特黃轉換、總體經驗模態分解、風速預測、離岸風電、長短期記憶神經網路
外文關鍵詞：	Hilbert-Huang Transform, Ensemble Empirical Mode Decomposition, Wind speed prediction, Offshore wind energy, Long Short-Term Memory Neural Network
相關次數：	點閱：100 下載：9
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

台灣缺乏自然能源，能源供給高度仰賴進口，也因此經濟發展非常容易受到國際局勢影響。為了達成能源永續此一政策目標，政府以離岸風電做為主力且規劃能達到5.7GW的發電量。然而離岸風力發電也存在一些困難，風機會受到颱風、季節性季風、風速無順序性大變化、海洋風浪等因素影響，造成風力發電機壽命減短、供應電網穩定性不佳等問題，因此有必要對風場的海氣象數據進行分析和預測。
本研究採用總體經驗模態分解 (EEMD) 方法，分析台電一期旁由台電桅杆蒐集到的海氣象資料。在有限條件下海氣象資料，首先藉由已知訊號的分解實驗驗證了分解之正確性，並通過分析缺失風速資料對分解結果的影響，更好地了解資料完整對於分析海氣象的必要性。本文分析了彰濱外海海氣象的特徵，結果顯示IMF1-IMF5具有代表特徵意義。以其中2019年風資料作為研究主軸，將彰濱外海區分為四個季節，以更好地理解海洋氣象狀況，且更進一步分析了四個颱風對海氣象和波浪的影響，驗證了EEMD方法在解析信號方面的能力，並說明風和浪是密不可分的自然現象。
早期的作法直接以長短期記憶神經網路進行預測，本文結合希爾伯特黃轉換法中的總體經驗模態分解方法分解海氣象訊號並進行單步預測與多步預測，對於風速急升、急降做出進一步探討，結果顯示長短期記憶模型 (LSTM) 模型的效果比 EEMD-LSTM 模型預測更準確，但 EEMD 對於起風點以及落風點的預測具有一定的成效且提供更多的訊息，此外對多步預測做出測試，預測更多未來風資料。

Taiwan lacks natural resources and thus mainly relies on imported energy by making its economic development highly vulnerable to the international situations. In order to achieve the policy goal of energy sustainability, the government has made offshore wind power mainstay for generating the capacity of 5.7GW. However, this study revealed that the offshore wind power would also face the challenges, such as the impact of typhoons, seasonal monsoons, large changes in wind speed without order ocean winds, and waves, which can reduce the lifespan of wind turbines and cause instability in power supply to the grid. Therefore, it is necessary to analyze and predict the sea meteorological data of wind farms.
This study employed the Ensemble Empirical Mode Decomposition (EEMD) method to analyze the sea meteorological data under limited conditions. Firstly, the correctness of the decomposition was verified by decomposing the known signals. The impact of missing wind speed data on the decomposition results was analyzed to better understand the necessity of complete data for analyzing the sea meteorology. The characteristics of the sea meteorology in the offshore area of Changhua were analyzed by decomposing the signal of five years of data, and the importance of IMF1-IMF5 was demonstrated. By using the wind data of 2019 as the main focus of the study, the offshore area of Changhua was divided into four seasons to better understand the sea meteorological conditions. The impact of four typhoons on sea meteorology and waves were further analyzed, by verifying the ability of the EEMD method to decompose signals by demonstrating the wind and waves are inseparable natural phenomena.
This study combined the Empirical Mode Decomposition (EMD) method of the Hilbert-Huang Transform with a Long Short-Term Memory (LSTM) neural network for wind speed forecasting. The traditional approach involved the application of LSTM to directly predict the wind speed. In this study, the EEMD method was first applied to decompose the sea meteorological signal, and then the LSTM model performed one-step and multi-step predictions. The study further investigated the wind speed spikes and drops, and the results showed that the LSTM model has better prediction accuracy than that of the EEMD-LSTM model. However, it was suggested that the EEMD was effective in predicting the starting and ending points of wind events and also provided an additional information. Additionally, the study performed the tests on multi-step prediction to forecast the future wind data.

摘要	i
Abstract	ii
致謝	iv
Contents	v
List of Tables	vii
List of Figures	viii
1.	Introduction	1
1.1	Development of Wind Power in Taiwan	1
1.2	Wind Prediction	3
1.3	Research Motivation	6
2.	Data Source	8
2.1	Taipower Wind Farm	8
2.2	Taipower Meteorological Mast	8
2.3	Sea Meteorological at Chanbin Sites	9
3.	Methodology	11
3.1	Hilbert Huang Transform	11
3.1.1	Empirical Mode Decomposition	11
3.1.2	Ensemble Empirical Mode Decomposition	13
3.2	Long Short-Term Memory	14
3.3	Research Procedure	17
4.	Data Preprocessing	19
4.1	Missing Data Analysis	19
4.2	Time Scales for Sampling Analysis	22
4.3	Wind Data Decomposition	23
5.	Analysis of Wind Data in 2019	25
5.1	Wind Direction Analysis	25
5.2	Wind Speed Analysis	26
6.	Analysis of Typhoons in 2019	28
7.	Wind Prediction	31
7.1	Annual Wind Speed Forecast	31
7.2	The Effect of IMF Features on the Wind Ramp Prediction	32
7.3	Composite Signal Analysis	34
7.4	Multi-Step Forecasting	35
8.	Conclusions	37
9.	References	40
Tables and Figures	43
Appendix	77
A.1 Comparison of missing data decomposition	77
A.2 EEMD of wind speed data	81
A.3 Prediction of each IMF layers	97
A.4 Prediction of the peak and valley for each IMF	104

 
List of Tables    
Table 4.1 The time scales of the IMF at different sampling rates.	43
Table 6.1 Typhoons in 2019.	44
Table 7.1 Predicted results of each model.	44
Table 7.2 Predicted performance of each layer of IMF.	45
Table 7.3 Prediction of soaring and plunging wind speed.	45
Table 7.4 Predicted results of each model.	46

 
List of Figures
Figure 1.1 Global energy consumption [1].	47
Figure 1.2 A total of 36 potential offshore wind farm sites in the coastal areas of Taiwan [6].	47
Figure 2.1 Wind Farm of Taipower Phase I [18].	48
Figure 2.2 Taipower Offshore meteorological mast.	48
Figure 2.3 Definition of wind direction.	49
Figure 2.4 Time series of wind direction and wind speed for (a) 2017, (b) 2018, (c) 2019, (d) 2020, (e) 2021.	51
Figure 3.1 Architecture of LSTM Unit [22].	52
Figure 3.2 Upper and lower envelopes of the EMD algorithm.	52
Figure 3.3 Prediction architecture for non-simplified models.	53
Figure 3.4 Prediction architecture for simplified models.	53
Figure 3.5 Research process.	53
Figure 4.1 Original Signal.	54
Figure 4.2 Composite Signal.	54
Figure 4.3 Decomposition of composite signals.	55
Figure 4.4 FFT verification of IMF 1-3 components of the composite signal.	55
Figure 4.5 Percentage of missing data for each year.	56
Figure 4.6 Comparison of missing data decomposition for IMF1、IMF2.	56
Figure 4.7 Comparison of missing data decomposition for IMF6、IMF7.	57
Figure 4.8 Comparison of missing data decomposition for IMF11、IMF14.	57
Figure 4.9 Correlation coefficient between complete data and missing data decomposition.	58
Figure 4.10 Energy distribution between complete data and missing data.	58
Figure 4.11 Yearly power distribution.	59
Figure 5.1 Monsoon in Taiwan [30].	59
Figure 5.2 Wind rose for the year of 2019.	60
Figure 5.3 Time series of wind direction.	61
Figure 5.4 EEMD of wind speed data.	62
Figure 5.5 Seasonal power distribution.	63
Figure 6.1 Typhoon of Danas.	64
Figure 6.2 Typhoon of Lekima.	64
Figure 6.3 Typhoon of Bailu.	65
Figure 6.4 Typhoon of Mitga.	65
Figure 6.5 EEMD of wind speed data.	66
Figure 6.6 EEMD of wave data.	67
Figure 7.1 Prediction by LSTM (a) Time series, (b) Correlation plot.	68
Figure 7.2 Prediction by EEMD-LSTM (a)Time series, (b) Correlation plot.	69
Figure 7.3 Prediction by EMD-LSTM (a) Time series, (b) Correlation plot.	70
Figure 7.4 Prediction by EMD-LSTM (a) Time series, (b) Correlation plot.	71
Figure 7.5 Prediction by EMD-LSTM (a) Time series, (b) Correlation plot.	72
Figure 7.6 Prediction compared of three model.	73
Figure 7.7 Peak and valley of predicted IMF6.	74
Figure 7.8 Peak and valley of predicted IMF7.	75
Figure 7.9 Prediction of (a) soaring, (b) plunging wind speed.	76
Figure A.1 Comparison of missing data decomposition for IMF1、IMF2.	77
Figure A.2 Comparison of missing data decomposition for IMF3、IMF4.	77
Figure A.3 Comparison of missing data decomposition for IMF5、IMF6.	78
Figure A.4 Comparison of missing data decomposition for IMF7、IMF8.	78
Figure A.5 Comparison of missing data decomposition for IMF9、IMF10.	79
Figure A.6 Comparison of missing data decomposition for IMF11、IMF12.	79
Figure A.7 Comparison of missing data decomposition for IMF13、IMF14.	80
Figure A.8 Comparison of missing data decomposition for Residual.	80
Figure A.9 Wind speed data from 2017 to 2021.	81
Figure A.10 IMF1 of wind speed data from 2017 to 2021.	82
Figure A.11 IMF2 of wind speed data from 2017 to 2021.	83
Figure A.12 IMF3 of wind speed data from 2017 to 2021.	84
Figure A.13 IMF4 of wind speed data from 2017 to 2021.	85
Figure A.14 IMF5 of wind speed data from 2017 to 2021.	86
Figure A.15 IMF6 of wind speed data from 2017 to 2021.	87
Figure A.16 IMF7 of wind speed data from 2017 to 2021.	88
Figure A.17 IMF8 of wind speed data from 2017 to 2021.	89
Figure A.18 IMF9 of wind speed data from 2017 to 2021.	90
Figure A.19 IMF10 of wind speed data from 2017 to 2021.	91
Figure A.20 IMF11 of wind speed data from 2017 to 2021.	92
Figure A.21 IMF12 of wind speed data from 2017 to 2021.	93
Figure A.22 IMF13 of wind speed data from 2017 to 2021.	94
Figure A.23 IMF14 of wind speed data from 2017 to 2021.	95
Figure A.24 Residual of wind speed data from 2017 to 2021.	96
Figure A.25 Predicted IMF 1.	97
Figure A.26 Predicted IMF 2.	97
Figure A.27 Predicted IMF 3.	98
Figure A.28 Predicted IMF 4.	98
Figure A.29 Predicted IMF 5.	99
Figure A.30 Predicted IMF 6.	99
Figure A.31 Predicted IMF 7.	100
Figure A.32 Predicted IMF 8.	100
Figure A.33 Predicted IMF 9.	101
Figure A.34 Predicted IMF 10.	101
Figure A.35 Predicted IMF 11.	102
Figure A.36 Predicted IMF 12.	102
Figure A.37 Predicted IMF 13.	103
Figure A.38 Predicted IMF 14.	103
Figure A.39 Peaks and valleys of predicted IMF1.	104
Figure A.40 Peaks and valleys of predicted IMF2.	105
Figure A.41 Peaks and valleys of predicted IMF3.	106
Figure A.42 Peaks and valleys of predicted IMF4.	107
Figure A.43 Peaks and valleys of predicted IMF5.	108
Figure A.44 Peaks and valleys of predicted IMF6.	109
Figure A.45 Peaks and valleys of predicted IMF7.	110
Figure A.46 Peaks and valleys of predicted IMF8.	111
Figure A.47 Peaks and valleys of predicted IMF9.	112
Figure A.48 Peaks and valleys of predicted IMF10.	113
Figure A.49 Peaks and valleys of predicted IMF11.	114
Figure A.50 Peaks and valleys of predicted IMF12.	115
Figure A.51 Peaks and valleys of predicted IMF13.	116
Figure A.52 Peaks and valleys of predicted IMF14.	117
                                    

[1] Enerdata, “Total energy consumption, ” The Enerdata Yearbook, https://yearbook.enerdata.net/total-energy/world-consumption-statistics.html, 2021 (accessed Feb 1, 2023).
[2] United Nations Department of Economic and Social Affairs, “Sustainable Development Goals, ” The 17 GOALS, https://sdgs.un.org/goals,2015 (accessed Feb 1, 2023).
[3] Wikipedia contributors, “Energy in Taiwan,” Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/w/index.php?title=Energy_in_Taiwan&oldid=1152901054 (accessed Feb 2, 2023).
[4] 行政院新聞傳播處, “全力推動離岸風電—打造台灣成為亞洲離岸風電技術產業聚落, ” 行政院全球資訊網, https://www.ey.gov.tw/Page/5A8A0CB5B41DA11E/9eebb9b8-490b-4357-963f-a48a981852a7, 2019 (accesse May 1, 2023).
[5] 4C Offshore, “Global Offshore Win Speeds Rankings, ” 4C Offshore, https://www.4coffshore.com/windfarms/windspeeds.aspx, 2014 (accesse April 29, 2023).
[6] 工業技術研究院, “潛力場址, ” 風力發電單一服務窗口, https://www.twtpo.org.tw/offshore_show.aspx?id=963, 2015 (accesse Feb 2, 2023).
[7] Wikipedia contributors, “Wind power,” Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/w/index.php?title=Wind_power&oldid=1156496382 (accessed Feb 3, 2023).
[8] S. S. Soman, H. Zareipour, O. Malik and P. Mandal, “A review of wind power and wind speed forecasting methods with different time horizons,” North American Power Symposium 2010, Arlington, TX, USA, 2010, pp. 1-8.
[9] Y. K. Wu and J. S. Hong, “A literature review of wind forecasting technology in the world,” 2007 IEEE Lausanne Power Tech, Lausanne, Switzerland, 2007, pp. 504-509.
[10] C. W. Potter and M. Negnevitsky, “Very short-term wind forecasting for Tasmanian power generation,” IEEE Transactions on Power Systems, vol. 21, no. 2, pp. 965-972, 2006.
[11] B. Candy, S. J. English, and S. J. Keogh, “A Comparison of the Impact of QuikScat and WindSat Wind Vector Products on Met Office Analyses and Forecasts,” IEEE Transactions on Geoscience and Remote Sensing, vol. 47, no. 6, pp. 1632-1640, 2009.
[12] A.Brendan, “Time Series Forecasting with ARIMA , SARIMA and SARIMAX,” Medium, https://towardsdatascience.com/time-series-forecasting-with-arima-sarima-and-sarimax-ee61099e78f6 , 2022, (accessed Feb 6, 2023).
[13] Q. Cao, B. T. Ewing, and M. A. Thompson, “Forecasting wind speed with recurrent neural networks,” European Journal of Operational Research, vol. 221, no. 1, pp. 148-154, 2012/08/16/ 2012.
[14] E. Cadenas and W. Rivera, “Rivera, W.: Wind speed forecasting in the South Coast of Oaxaca, México. Renew. Energy 32(12), 2116-2128,” Renewable Energy, vol. 32, pp. 2116-2128, 2007.
[15] H.C. Fang, “Wind Power Potential Evaluation and Sea Meteorological Prediction at Chanbin Sites,” Master’s thesis, Department of Mechanical Engineering, National Cheng Kung University, 2021.
[16] W. Hu, Q. Yang, H.-P. Chen, Z. Yuan, C. Li, S. Shao and J. Zhang, “New hybrid approach for short-term wind speed predictions based on preprocessing algorithm and optimization theory,” Renewable Energy, vol. 179, pp. 2174-2186, 2021.
[17] M. Majidi Nezhad, A. Heydari, M. Neshat, F. Keynia, G. Piras, and D. A. Garcia, “A Mediterranean Sea Offshore Wind classification using MERRA-2 and machine learning models,” Renewable Energy, vol. 190, pp. 156-166, 2022.
[18] Taipower, “Phase 1 of the Offshore Wind Power Project - Powering Up Green Energy with Sea Breeze, “ Taipower Sustainability Section, https://csr.taipower.com.tw/en/news/kZ/detail, 2022 (accessed March 28, 2023).
[19] 交通部中央氣象局, “臺灣的四季怎麼劃分？,” 交通部中央氣象局, https://www.cwb.gov.tw/V8/C/K/Encyclopedia/climate/climate2_list.html#climate2-02 , 2023 (accessed March 31, 2023).
[20] Norden E. Huang, Zheng Shen, Steven R. Long, Manli C. Wu, Hsing H. Shih, Quanan Zheng, Nai-Chyuan Yen, Chi Chao Tung and Henry H. Liu, “The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis,” Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, vol. 454, no. 1971, pp. 903-995, 1998.
[21] Z. WU and N. E. Huang, “Ensemble Empirical Mode Decomposition: A Noise-Assisted Data Analysis Method,” Advances in Adaptive Data Analysis, vol. 01, no. 01, pp. 1-41, 2009.
[22] 黃日鉦, “長短期記憶模型 (Long Short Term Memory, LSTM) 模型,” 人工智慧與深度學習:理論與Python實踐, 臺北市: 碁峰資訊, 2020, pp. 146-151.
[23] S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Computation, vol. 9, no. 8, pp. 1735-1780, 1997.
[24] Wikipedia contributors, “Logistic function,” Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/w/index.php?title=Logistic_function&oldid=1148617753 (accessed April 30, 2023).
[25] E. O. Brigham and R. E. Morrow, “The fast Fourier transform,” IEEE Spectrum, vol. 4, no. 12, pp. 63-70, 1967.
[26] R. Yang and B. Xing, “A Comparison of the Performance of Different Interpolation Methods in Replicating Rainfall Magnitudes under Different Climatic Conditions in Chongqing Province (China),” Atmosphere, vol. 12, no. 10, pp. 1318, 2021.
[27] T. Liu, H. Wei, and K. Zhang, “Wind power prediction with missing data using Gaussian process regression and multiple imputation,” Applied Soft Computing, vol. 71, pp. 905-916, 2018.
[28] P. Schober, C. Boer, and L. A. Schwarte, “Correlation Coefficients: Appropriate Use and Interpretation,” Anesthesia & Analgesia, vol. 126, no. 5, pp. 1763-1768, 2018.
[29] Ccjou, “向量範數,” 線代啟示錄, https://ccjou.wordpress.com/2013/08/13/%E5%90%91%E9%87%8F%E7%AF%84%E6%95%B8/, 2013 (accessed April 16, 2023).
[30] Y. Sangheon, “Holocene Vegetation Responses to East Asian Monsoonal Changes in South Korea,” in: Juan A. Blanco & Houshang Kheradmand (ed.), Climate Change - Geophysical Foundations and Ecological Effects, IntechOpen, 2011.
[31] 颱風資料庫, “有發警報颱風列表,” 颱風資料庫, https://rdc28.cwb.gov.tw/TDB/public/warning_typhoon_list/ , 2019 (accessed April 28) .

校內：立即公開
校外：立即公開

簡易檢索 / 詳目顯示

相關論文