Forecasting wind speed using support vector regression and feature selection

Botha, Gerda Nvan der Walt, Christiaan2018-08-222018-08-222017-12Botha, G.N. and van der Walt, C. 2017. Forecasting wind speed using support vector regression and feature selection. Pattern Recognition Association of South Africa and Robotics and Mechatronics (PRASA-RobMech), 2017, Bloemfontein, South Africa, 30 Nov -1 Dec 2017, pp. 181-186978-1-5386-2315-2DOI: 10.1109/RoboMech.2017.8261144http://ieeexplore.ieee.org/document/8261144/http://hdl.handle.net/10204/10383© 2017 IEEE. Due to copyright restrictions, the attached PDF file contains the accepted version of the published paper. For access to the published item, please consult the publisher's website.Since the wind power generated by a wind farm is entirely dependent on meteorological conditions (such as wind speed, wind direction, humidity etc.), accurately forecasting wind speed based on these conditions over a 1 to 24 hour time horizon is crucial to predict the potential short term energy supply of a wind farm. These short term predictions in turn are crucial in assisting with wind farm planning so that the required base load provided to the electricity grid is always guaranteed (even in the presence of highly variable wind power outputs). In this work we show that the relative prediction performance of a short-term Support Vector Regression (SVR) wind forecasting system can be improved by up to 11.12% by systematically selecting and combining relevant input features that influence short term wind speed. We illustrate our results on meteorological data collected in Alexander Bay, South Africa over a three year period from 2011-2013.enForecastingSupport vector machinesTraining dataWind ForecastingWind speedForecasting wind speed using support vector regression and feature selectionConference PresentationBotha, G. N., & van der Walt, C. (2017). Forecasting wind speed using support vector regression and feature selection. IEEE. http://hdl.handle.net/10204/10383Botha, Gerda N, and Christiaan van der Walt. "Forecasting wind speed using support vector regression and feature selection." (2017): http://hdl.handle.net/10204/10383Botha GN, van der Walt C, Forecasting wind speed using support vector regression and feature selection; IEEE; 2017. http://hdl.handle.net/10204/10383 .TY - Conference Presentation AU - Botha, Gerda N AU - van der Walt, Christiaan AB - Since the wind power generated by a wind farm is entirely dependent on meteorological conditions (such as wind speed, wind direction, humidity etc.), accurately forecasting wind speed based on these conditions over a 1 to 24 hour time horizon is crucial to predict the potential short term energy supply of a wind farm. These short term predictions in turn are crucial in assisting with wind farm planning so that the required base load provided to the electricity grid is always guaranteed (even in the presence of highly variable wind power outputs). In this work we show that the relative prediction performance of a short-term Support Vector Regression (SVR) wind forecasting system can be improved by up to 11.12% by systematically selecting and combining relevant input features that influence short term wind speed. We illustrate our results on meteorological data collected in Alexander Bay, South Africa over a three year period from 2011-2013. DA - 2017-12 DB - ResearchSpace DP - CSIR KW - Forecasting KW - Support vector machines KW - Training data KW - Wind Forecasting KW - Wind speed LK - https://researchspace.csir.co.za PY - 2017 SM - 978-1-5386-2315-2 T1 - Forecasting wind speed using support vector regression and feature selection TI - Forecasting wind speed using support vector regression and feature selection UR - http://hdl.handle.net/10204/10383 ER -