In this study, a Nonlinear AutoRegressive
eXogenous (NARX) neural network is used to estimate the wind speed on three
monthly data sets taken from the wind central in Zonguldak province in Turkey.
In the estimation study, the first and second order curve fitting coefficients
of the measured temperature, pressure, humidity and solar radiation parameters
together with the wind speed are used. In the estimation process, before these
coefficients are applied directly to the NARX network structure, the most
suitable features are selected with ReliefF method to minimize the MSE value.
The number of delay steps in the NARX network structure is varied from 3 to 15
and the number of hidden neurons is varied from 3 to 15 to obtain model
parameters that give the least estimation error. In order to determine the
performance of the obtained model, the model is evaluated in terms of
statistical error criteria such as MAE, MSE and RMSE. The model parameters and
features matrix giving the least estimation error for the wind speed estimation
of the NARX network structure are determined. It has been observed that this
approach provides a high performance for estimating the wind speed with related
to other measured parameters.
Subjects | Electrical Engineering |
---|---|
Journal Section | Research Article |
Authors | |
Publication Date | October 25, 2017 |
Submission Date | July 6, 2017 |
Acceptance Date | October 17, 2017 |
Published in Issue | Year 2017 Volume: 4 Issue: 3 |