Dynamic Modification of Super Short Term Numerical Wind Forecast Based on Neural Networks at Wind Farm
WU Xi1, WANG Binbin1, ZHOU Hai2, YU Jiang1, CUI Fang2
1. Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science & Technology, Nanjing 210044, China;2. Institute of Clean Energy Generation, State Grid Electric Power Research Institute, Nanjing 210003, China
Abstract：For effective planning and scheduling and for Wind Power Prediction (WPP) at 70 meters above the ground and 0-4h super short term wind speed forecasting, this paper uses the NWP wind speed of MM5 grids from the National Meteorological Center to analyze the prediction error at the wind tower height in a wind farm which is located off the coast. Based on the meteorological data from the wind tower and after data statistical analysis, it is found that the numerical forecast wind speed errors have correlations with themselves and the prediction errors are caused by the elements of sustainability. A method using earlier observation errors and turbulent index to revise the wind speed forecasting of MM5 is discussed and an ANN dynamic modification model for super short term forecasting is set up. The results show that after correction of the forecast wind speed, the mean absolute error is reduced and the prediction accuracy is improved effectively. It is also shown that the error index decreases about 40%, and the prediction curve can better reflect the high frequency of wind speed fluctuations, which better agrees with the measured wind speed curve. Update can be done once every four hours, satisfying the requirements of power grid dispatching. The method is simple and economic and can be used widely in small and medium-sized wind farms. It will help effective use of wind power as well as safe operation of power companies.
吴息;王彬滨;周海;余江;崔方. 基于神经网络的风电场超短期风速数值预报的动态修订[J]. , 2013, 31(34): 39-44.
WU Xi;WANG Binbin;ZHOU Hai;YU Jiang;CUI Fang. Dynamic Modification of Super Short Term Numerical Wind Forecast Based on Neural Networks at Wind Farm. , 2013, 31(34): 39-44.
2013, Vol. 31 
