Title: Short-term forecasting of wind power generation using extreme learning machine and its variants
Authors: Sthita Prajna Mishra
Addresses: MDRC, EE Department, ITER, Siksha 'O' Anusandhan University, Khandagiri, Bhubaneswar, Odisha, 751030, India
Abstract: This paper proposes three efficient and accurate wind power prediction algorithms, i.e., online sequential extreme learning machine (OS-ELM), ridge ELM (R-ELM) and hybrid linear and nonlinear neural network (HNN) trained by Levenberg-Marquardt algorithm. Learning speed rate and computational scalability are essential attributes upon which the accuracy of the prediction algorithm is decided. In this regard, the accuracy of some of the conventional algorithms, such as support vector regression or neural network-based algorithms is very frail. This is due to the fact these algorithms are computed in an iterative manner in which the hidden layers are being updated in each iteration. On the contrary, the proposed ELM-based prediction algorithm computes the output weight vector in a chunk, where the hidden layer is not being updated. Hence, the essential features such as the learning speed rate and computational scalability have been significantly improved that allows a faster response for the proposed algorithm, which is distinguished from the response of the conventional algorithms in the MATLAB/Editor environment, as has been illustrated in the simulation and result section.
Keywords: correlation coefficient; linear neural networks; nonlinear neural networks; hybrid neural networks; HNN; online sequential ELM; extreme learning machine; OSELM; ridge ELM; R-ELM; short-term forecasting; wind power generation; wind energy; simulation.
International Journal of Power and Energy Conversion, 2017 Vol.8 No.1, pp.68 - 89
Available online: 16 Nov 2016 *Full-text access for editors Access for subscribers Purchase this article Comment on this article