TY - JOUR
T1 - Prediction of Wind Power Model Using Hybrid Method Based on WD-SVM Algorithm
T2 - 2nd International Conference on Applied Sciences Mathematics and Informatics, ICASMI 2018
AU - Prasetyowati, A.
AU - Sudiana, D.
AU - Sudibyo, H.
N1 - Funding Information:
Our thanks to research funding provided by the PITTA (HIBAH PUBLIKASI INTERNASIONAL TERINDEKS UNTUK TUGAS AKHIR MAHASISWA UI) grant and the New Renewable Energy Society (METI) BPPT for allowing us to use Pandansimo wind farm data to develop into a wind power prediction model.
Publisher Copyright:
© Published under licence by IOP Publishing Ltd.
PY - 2019/10/23
Y1 - 2019/10/23
N2 - Wind power prediction with original data that has a nonstationary pattern and randomness becomes a major problem when the data is used for the preparation of the generation. Especially the wind power directly connected to the network. In this research, we proposed wavelet decomposition model and support vector machine (WD-SVM) to predict the power scale of wind power in Pandansimo wind farm. Time series data that is built in interval of 1 hour mean in 24 formed day data. The data is parsed using WD that generates the IMF component. The output of the WD model is reprocessed using the SVM model to do the clustering process which the outputs are the scale of selected wind power strength in 1 month approaching the historical data of measurement. Finally after going through the experiment, obtained the data prediction scale of wind power strength that has a high accuracy near the actual conditions. The WD-SVM hybrid model provides a smaller error than the predicted model of NN and WD-NN.
AB - Wind power prediction with original data that has a nonstationary pattern and randomness becomes a major problem when the data is used for the preparation of the generation. Especially the wind power directly connected to the network. In this research, we proposed wavelet decomposition model and support vector machine (WD-SVM) to predict the power scale of wind power in Pandansimo wind farm. Time series data that is built in interval of 1 hour mean in 24 formed day data. The data is parsed using WD that generates the IMF component. The output of the WD model is reprocessed using the SVM model to do the clustering process which the outputs are the scale of selected wind power strength in 1 month approaching the historical data of measurement. Finally after going through the experiment, obtained the data prediction scale of wind power strength that has a high accuracy near the actual conditions. The WD-SVM hybrid model provides a smaller error than the predicted model of NN and WD-NN.
UR - http://www.scopus.com/inward/record.url?scp=85075013025&partnerID=8YFLogxK
U2 - 10.1088/1742-6596/1338/1/012048
DO - 10.1088/1742-6596/1338/1/012048
M3 - Conference article
AN - SCOPUS:85075013025
VL - 1338
JO - Journal of Physics: Conference Series
JF - Journal of Physics: Conference Series
SN - 1742-6588
IS - 1
M1 - 12048
Y2 - 9 August 2018 through 11 August 2018
ER -