Application of extreme learning machine for estimation of wind speed distribution

• Published in: Climate dynamics Vol. 46; no. 5-6; pp. 1893 - 1907
• Main Authors: Shamshirband, Shahaboddin, Mohammadi, Kasra, Tong, Chong Wen, Petkovic, Dalibor, Porcu, Emilio, Mostafaeipour, Ali, Ch, Sudheer, Sedaghat, Ahmad
• Format: Journal Article
• Language: English
• Published: Springer Berlin Heidelberg 01.03.2016
• ISSN: 0930-7575; 1432-0894
• DOI: 10.1007/s00382-015-2682-2

The knowledge of the probabilistic wind speed distribution is of particular significance in reliable evaluation of the wind energy potential and effective adoption of site specific wind turbines. Among all proposed probability density functions, the two-parameter Weibull function has been extensively endorsed and utilized to model wind speeds and express wind speed distribution in various locations. In this research work, extreme learning machine (ELM) is employed to compute the shape (k) and scale (c) factors of Weibull distribution function. The developed ELM model is trained and tested based upon two widely successful methods used to estimate k and c parameters. The efficiency and accuracy of ELM is compared against support vector machine, artificial neural network and genetic programming for estimating the same Weibull parameters. The survey results reveal that applying ELM approach is eventuated in attaining further precision for estimation of both Weibull parameters compared to other methods evaluated. Mean absolute percentage error, mean absolute bias error and root mean square error for k are 8.4600 %, 0.1783 and 0.2371, while for c are 0.2143 %, 0.0118 and 0.0192 m/s, respectively. In conclusion, it is conclusively found that application of ELM is particularly promising as an alternative method to estimate Weibull k and c factors.