Short-Term Wind Speed Forecasting Based on Optimizated Support Vector Machine

Abstract:

To avoid the impact which is caused by the characteristics of the random fluctuations of the wind speed to grid-connected wind power generation system, accurately prediction of short-term wind speed is needed. This paper designed a combination prediction model which used the theories of wavelet transformation and support vector machine (SVM). This improved the model’s prediction accuracy through the method of achiving change character in wind speed sequences in different scales by wavelet transform and optimizing the parameters of support vector machines through the improved particle swarm algorithm. The model showed great generalization ability and high prediction accuracy through the experiment. The lowest root-mean-square error of 200 samples was up to 0.0932 and the model’s effect was much stronger than the BP neural network prediction model. It provided an effective method for predicting wind speed.

• Periodical: Applied Mechanics and Materials (Volumes 300-301)
• Pages: 189-194
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.300-301.189
• Online since: February 2013
• Authors: Yu Sun, Ling Ling Li, Xiao Song Huang, Chao Ying Duan
• Keywords: Parameter Optimization, Particle Swarm Algorithm, Support Vector Machine (SVM), Wavelet Transformation, Wind Speed Forecasting

References

[1] Chongqing Kang, Qing Xia, Zuohao Hu. New Connotation of Forecasting Issues in ElectricityMarket [J]. Automation of Electric Power Systems, 2004，28(18): 1-6(In Chinese).

[2] Yazhou Lei, Weisheng Wang, Huizhu Dai, et al. Analysis of windpower value to power system operation[J]. Power SystemTechnology, 2002，26(5): 10-14(In Chinese).

[3] Jingxu Liu. Research on model selection for support vector regression and application of IT [D]. Changsha: Colloge of In-formation System and Management, National University of De-fense Technology, 2006, (In Chinese).

[4] Chunzheng Ye, Xianrong Chang, Weiguo Gu. Hort-term load forecasting based on wavelet transformation and support vector machines [J]. Power System Protection and Control, 2009，37(14): 42-45, (In Chinese).

[5] Stephane GM. A theory for multiresolution signal decomposition: The wavelet representation [J]. IEEE Tran on Pattern Analysis andMachine Intelligence, 1989, 11(7): 674-693.

DOI: 10.1109/34.192463

[6] Yishan Xu, Bi Zeng, Xiuwen Yin, et al. BP neural net-work and its applications based on improved PSO[J]. Computer Engineering and Application, 2009, 45(35): 233－235.

[7] Junfang Li, Buhan Zhang, Guanglong Xie, et al. Grey predictor models for wind speed-wind power prediction[J]. Power System Protection and Control, 2010，38(19): 152-159.

[8] Xiuyuan Yang, Yang Xiao, Shuyong Chen. Wind speed and generatedpower forecasting in wind farm[J]. Proceedings of the CSEEm2005m11(6): 1-5(In Chinese).

[9] Xiaolan Wang, Xanhong Zhang,　Huizhong Wang. Non-stationarytime series prediction based on wavelet analysis and AR-LSSVM[J]. Control and Decision, 2008, 23(3): 357-360 (In Chinese).

[10] Feng Pan, Haozhong Cheng, Jingfei Yang, et al. Power System Short-term Load Forecasting Based onSupport Vector Machines [J]. Power SystemTechnology, 2004, 28(21).

[11] Ran Li, Siqing Sheng, Ying Wang. Weak voltage areas recovery based on improved particle swarm optimizational gorithm[J]. Power System Protection and Control, 2009, 37(24): 109－113.