隨著風力發電的普及化，人們越來越關注風力發電機主動控制，並且其中一個目標是會以參與公用電網的頻率調節為目的，產出滿足風電場所要達到之功率設定點。在本研究中建立一個由12台風力發電機所組成的風電場做為模擬環境，以過去歷史的風速資料作為深度學習的訓練資料，在每台風力發電機皆有理想之最大功率點控制的情況下，控制其風場針對各種風速變化狀況的每台風力發電機功率設定點，最後成功達成風場之最大功率曲線追蹤。本研究提出一種新的有效功率控制策略結合深度學習之方法，在滿足追蹤功率要求之功率並達成風場功率最大化儲備，在功率設定點控制下減少由尾流效應引起的風速不足，此方法優先考慮最下游風力發電機的產生功率，從而衰減尾流干擾。從結果可看出與傳統的功率分配相比，控制策略能夠增加功率儲備，其中每台風力發電機的功率設定點與其可用功率成正比。

With the popularization of wind power generation, people are paying more and more attention to the active control of wind turbines. One of the goals is to participate in the frequency adjustment of the public grid, and the output meets the power set point to be reached by the wind power station. Establish a wind farm consisting of 12 wind turbines, using historical wind speed data as training materials for deep learning. Under the condition that each wind turbine has the ideal maximum power point control, control its wind field. Each wind turbine power set point for various wind speed changes. Finally, the maximum power curve tracking of the wind field was successfully achieved. This research proposes a new effective power control strategy combined with deep learning method to meet the power requirements of tracking power and achieve maximum wind farm power reserve, and reduce the wind speed shortage caused by wake effect under power set point control. This method prioritizes the generated power of the most downstream wind turbines, thereby attenuating wake disturbances. From the results, the control strategy can increase the power reserve compared to the conventional power distribution, where the power set point of each wind turbine is proportional to its available power.

Acknowledgement I
Abstract in Chinese I
Abstract in English III
Contents IV
List of Figures VI
List of Tables VIII
1 Introduction 1
1.1 Motivation 1
1.2 Background 2
1.2.1 Wind power development 2
1.2.2 Deep Learning 5
1.3 Problem Statement 9
2 Wind Farm Model 10
2.1 Wind Energy Conversion Systems 11
2.1.1 DC Bus and Battery Bank Equivalent Circuit 13
2.1.2 Wind power reserve 14
2.2 Wind Power Plants 15
2.2.1 Wake-effect Model 15
2.2.2 Wind farm power modeling 16
3 Control Stategies 19
3.1 Active Power Control 20
3.2 Deep Learning 20
3.2.1 Deep Neural Networks 20
3.2.2 DNNs Training 24
4 Simulation Result 26
4.1 Matlab SimWindFarm Module 26
4.2 Wind farm layout 27
4.3 Constant wind speed wake effect 28
4.4 Random wind speed power tracking 37
5 Conclusion 48
References 49
List of Figures
1.1 Charles Brush’s windmill of 1888, used for generating electricity 3
1.2 Global annual installed wind capacity [1] 5
1.3 Global cumulative installed wind capacity [1] 5
1.4 (a)Single-Layer Perceptron (b)Two-Input/Single-Neuron Perceptron 6
1.5 Types of machine learning 7
1.6 Supervised Learning 7
1.7 Unsupervised Learning 8
1.8 Reinforcement Learning 8
2.1 Wind energy conversion system 11
2.2 Per-phase circuit and phasor diagram of the PMSG 12
2.3 DC Bus Equivalent Circuit 13
2.4 Wake effect on the Jensen model 15
2.5 Wind farm distributed control system 16
2.6 Induction factor a is related to wind speed v and power set point Pr 17
2.7 Induction factor a is related to wind speed v and power set point Pr 18
3.1 Control structure of the active power control of a wind farm 20
3.2 Single neuron[2] 21
3.3 Sigmoid function curve 22
3.4 ReLU function curve 22
3.5 Neural network architecture 23
4.1 SimWindFarm Overview 26
4.2 Wind farm layout corresponding to the case study 27
4.3 First column wind speed 29
4.4 Second column wind speed 30
4.5 Third column wind speed 31
4.6 Fourth column wind speed 32
4.7 First column wind power 33
4.8 Second column wind power 34
4.9 Third column wind power 35
4.10 Fourth column wind power 36
4.11 Random wind speed 38
4.12 Total generated power 38
4.13 First column wind generated power 39
4.14 Second column wind generated power 40
4.15 Third column wind generated power 41
4.16 Fourth column wind generated power 42
4.17 Total generated power 43
4.18 First column wind generated power 44
4.19 Second column wind generated power 45
4.20 Third column wind generated power 46
4.21 Fourth column wind generated power 47
List of Tables
4.1 Wind farm layout specification 28
4.2 Wind farm Simulation data 37

[1] “Global wind report.” April 2018.
[2] H. B. Demuth, M. H. Beale, O. De Jess, and M. T. Hagan, Neural Network Design, 2nd ed. USA: Martin Hagan, 2014.
[3] “A wind energy pioneer: Charles f. brush,” Tech. Rep. 1, September 2008.
[4] R. W. Righter, Wind Energy in America: A History, March 2018.
[5] L. Gapay, Hawaiians get Boeing’s Last Wind Machine Makani Ho’Olapa will Bring Power to 1,140 Residences, 1987.
[6] “Btm forecasts 340-gw of wind energy by 2013,” 2009.
[7] S. Hewitt, L. Margetts, and A. Revell, “Building a digital wind farm,” Archives of Computational Methods in Engineering, vol. 25, no. 4, pp. 879–899, Nov 2018. [Online]. Available: https://doi.org/10.1007/s11831-017-9222-7
[8] S. S. Lauha Fried, Shruti Shukla, Wind Energy Engineering: Growth Trends and the Future of Wind Energy, 2017.
[9] D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. Huang, A. Guez, T. Hubert, L. Baker, M. Lai, A. Bolton, Y. Chen, T. Lillicrap, F. Hui, L. Sifre, G. van den Driessche, T. Graepel, and D. Hassabis, “Mastering the game of go without human knowledge,” Nature, vol. 550, pp. 354–, Oct. 2017. [Online]. Available: http://dx.doi.org/10.1038/nature24270
[10] R.REN, “Renewable energy policy network for the 21st century: Paris: Ren21 secretariat.” 2013.
[11] M. B. Christiansen and C. B. Hasager, “Wake effects of large offshore wind farms identified from satellite sar,” Remote Sensing of Environment, vol. 98, no. 2, pp. 251 – 268, 2005. [Online]. Available: http://www.sciencedirect.com/science/ article/pii/S0034425705002476
[12] B. P. Hayes and S. Z. Djokic, “Modelling of wind generation at all scales for transmission system analysis,” IET Generation, Transmission & Distribution, vol. 7, pp. 1144–1154(10), October 2013. [Online]. Available: https://digital library.theiet.org/content/journals/10.1049/iet-gtd.2012.0271
[13] J. Thongam and M. Ouhrouche, “Mppt control methods in wind energy conversion systems.” ResearchGate, June 2011.
[14] Y. Li, Z. Xu, and K. P. Wong, “Advanced control strategies of pmsg based wind turbines for system inertia support,” IEEE Transactions on Power Systems, vol. 32, no. 4, pp. 3027–3037, July 2017.
[15] R. Doherty, A. Mullane, G. Nolan, D. J. Burke, A. Bryson, and M. O’Malley, “An assessment of the impact of wind generation on system frequency control,” IEEE Transactions on Power Systems, vol. 25, no. 1, pp. 452–460, Feb 2010.
[16] H. Zhao, Q. Wu, Q. Guo, H. Sun, and Y. Xue, “Distributed model predictive control of a wind farm for optimal active power controlpart ii: Implementation with clustering-based piece-wise affine wind turbine model,” IEEE Transactions on Sustainable Energy, vol. 6, no. 3, pp. 840–849, July 2015.
[17] P. Liu, W.-T. Yang, C.-E. Yang, and C.-L. Hsu, “Sensorless wind energy conversion system maximum power point tracking using takagi-sugeno fuzzy cerebellar model articulation control,” Appl. Soft Comput., vol. 29, no. C, pp. 450–460, Apr. 2015. [Online]. Available: http://dx.doi.org/10.1016/j.asoc.2015.01.019
[18] E. Muljadi, S. Drouilhet, R. Holz, and V. Gevorgian, “Analysis of permanent magnet generator for wind power battery charging,” in IAS ’96. Conference Record of the 1996 IEEE Industry Applications Conference Thirty-First IAS Annual Meeting, vol. 1, Oct 1996, pp. 541–548 vol.1.
[19] B. K. Bose, Modern power electronics and AC drives. Prentice Hall, c2005.
[20] F. Valenciaga, P. F. Puleston, and P. E. Battaiotto, “Power control of a solar/ wind generation system without wind measurement: a passivity/sliding mode approach,” IEEE Transactions on Energy Conversion, vol. 18, no. 4, pp. 501–507, Dec 2003.
[21] B. S. Borowy and Z. M. Salameh, “Dynamic response of a stand-alone wind energy conversion system with battery energy storage to a wind gust,” IEEE Transactions on Energy Conversion, vol. 12, no. 1, pp. 73–78, March 1997.
[22] R. Billinton and A. A. Chowdhury, “Incorporation of wind energy conversion systems in conventional generating capacity adequacy assessment,” IEE Proceedings C - Generation, Transmission and Distribution, vol. 139, no. 1, pp. 47–56, Jan 1992.
[23] N. Jensen, A note on wind generator interaction. Risø National Laboratory, 1983.
[24] H. Kim, C. Singh, and A. Sprintson, “Simulation and estimation of reliability in a wind farm considering the wake effect,” IEEE Transactions on Sustainable Energy, vol. 3, no. 2, pp. 274–282, April 2012.
[25] K. Xie, H. Yang, B. Hu, and C. Li, “Optimal layout of a wind farm considering multiple wind directions,” in 2014 International Conference on Probabilistic Methods Applied to Power Systems (PMAPS), July 2014, pp. 1–6.
[26] A. Zertek, G. Verbic, and M. Pantos, “A novel strategy for variable-speed wind turbines’ participation in primary frequency control,” IEEE Transactions on Sustainable Energy, vol. 3, no. 4, pp. 791–799, Oct 2012.
[27] G. S. Janaka EkanayakeNick, Nick JenkinsG, “Power system frequency response from fixed speed and doubly fed induction generator-based wind turbines,” 2004.
[28] H. Badihi, Y. Zhang, and H. Hong, “Design of a pole placement active power control system for supporting grid frequency regulation and fault tolerance in wind farms,” IFAC Proceedings Volumes, vol. 47, no. 3, pp. 4328 – 4333, 2014, 19th IFAC World Congress. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1474667016422792
[29] K. Shimizu, S. Ito, and S. Suzuki, “Tracking control of general nonlinear systems by direct gradient descent method,” IFAC Proceedings Volumes, vol. 31, no. 17, pp. 183 – 188, 1998, 4th IFAC Symposium on Nonlinear Control Systems Design 1998 (NOLCOS’98), Enschede, The Netherlands, 1-3 July. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1474667017403326
[30] J. L. B. Diederik P. Kingma, “Adam: A method for stochastic optimization,” 2015.
[31] J. Grunnet, M. Soltani, T. Knudsen, M. Kragelund, and T. Bak, “Aeolus toolbox for dynamics wind farm model, simulation and control,” in European Wind Energy Conference and Exhibition, EWEC 2010, 2010.