Feature extraction of NWP data for wind power forecasting using 3D-convolutional neural networks

Kazutoshi Higashiyama, Yu Fujimoto, Yasuhiro Hayashi

Research output: Contribution to journalConference article

3 Citations (Scopus)

Abstract

Wind power is one of the most attractive forms of electricity from the viewpoints of cost efficiency and environmental protection. However, the instability of wind power has a serious impact on a grid system. Reliable wind power forecasting will help to utilize storage systems and backup generators effectively for mitigating the instability. This paper proposes a feature extraction procedure for numerical weather prediction (NWP) data based on the three-dimensional convolutional neural networks (3DCNNs). An advantage of 3D-CNNs is to automatically extract the spatio-temporal features from NWP data focusing on the targeted wind farm. Feature extraction based on 3D-CNNs was applied to real-world datasets; the results show significant performance in comparison to several benchmark approaches, and also show that the proposed extraction scheme based on 3DCNNs achieves to derive intrinsic features for prediction of wind power generation from NWP data.

Original languageEnglish
Pages (from-to)350-358
Number of pages9
JournalEnergy Procedia
Volume155
DOIs
Publication statusPublished - 2018 Jan 1
Event12th International Renewable Energy Storage Conference, IRES 2018 - Dusseldorf, Germany
Duration: 2018 Mar 132018 Mar 15

Fingerprint

Wind power
Feature extraction
Neural networks
Environmental protection
Farms
Power generation
Electricity
Costs

Keywords

  • Convolutional neural networks
  • Deep learning
  • Feature extraction
  • Numerical weather prediction
  • Wind power forecasting

ASJC Scopus subject areas

  • Energy(all)

Cite this

Feature extraction of NWP data for wind power forecasting using 3D-convolutional neural networks. / Higashiyama, Kazutoshi; Fujimoto, Yu; Hayashi, Yasuhiro.

In: Energy Procedia, Vol. 155, 01.01.2018, p. 350-358.

Research output: Contribution to journalConference article

@article{bb8d5119b8614352b7ce566114838d89,
title = "Feature extraction of NWP data for wind power forecasting using 3D-convolutional neural networks",
abstract = "Wind power is one of the most attractive forms of electricity from the viewpoints of cost efficiency and environmental protection. However, the instability of wind power has a serious impact on a grid system. Reliable wind power forecasting will help to utilize storage systems and backup generators effectively for mitigating the instability. This paper proposes a feature extraction procedure for numerical weather prediction (NWP) data based on the three-dimensional convolutional neural networks (3DCNNs). An advantage of 3D-CNNs is to automatically extract the spatio-temporal features from NWP data focusing on the targeted wind farm. Feature extraction based on 3D-CNNs was applied to real-world datasets; the results show significant performance in comparison to several benchmark approaches, and also show that the proposed extraction scheme based on 3DCNNs achieves to derive intrinsic features for prediction of wind power generation from NWP data.",
keywords = "Convolutional neural networks, Deep learning, Feature extraction, Numerical weather prediction, Wind power forecasting",
author = "Kazutoshi Higashiyama and Yu Fujimoto and Yasuhiro Hayashi",
year = "2018",
month = "1",
day = "1",
doi = "10.1016/j.egypro.2018.11.043",
language = "English",
volume = "155",
pages = "350--358",
journal = "Energy Procedia",
issn = "1876-6102",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Feature extraction of NWP data for wind power forecasting using 3D-convolutional neural networks

AU - Higashiyama, Kazutoshi

AU - Fujimoto, Yu

AU - Hayashi, Yasuhiro

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Wind power is one of the most attractive forms of electricity from the viewpoints of cost efficiency and environmental protection. However, the instability of wind power has a serious impact on a grid system. Reliable wind power forecasting will help to utilize storage systems and backup generators effectively for mitigating the instability. This paper proposes a feature extraction procedure for numerical weather prediction (NWP) data based on the three-dimensional convolutional neural networks (3DCNNs). An advantage of 3D-CNNs is to automatically extract the spatio-temporal features from NWP data focusing on the targeted wind farm. Feature extraction based on 3D-CNNs was applied to real-world datasets; the results show significant performance in comparison to several benchmark approaches, and also show that the proposed extraction scheme based on 3DCNNs achieves to derive intrinsic features for prediction of wind power generation from NWP data.

AB - Wind power is one of the most attractive forms of electricity from the viewpoints of cost efficiency and environmental protection. However, the instability of wind power has a serious impact on a grid system. Reliable wind power forecasting will help to utilize storage systems and backup generators effectively for mitigating the instability. This paper proposes a feature extraction procedure for numerical weather prediction (NWP) data based on the three-dimensional convolutional neural networks (3DCNNs). An advantage of 3D-CNNs is to automatically extract the spatio-temporal features from NWP data focusing on the targeted wind farm. Feature extraction based on 3D-CNNs was applied to real-world datasets; the results show significant performance in comparison to several benchmark approaches, and also show that the proposed extraction scheme based on 3DCNNs achieves to derive intrinsic features for prediction of wind power generation from NWP data.

KW - Convolutional neural networks

KW - Deep learning

KW - Feature extraction

KW - Numerical weather prediction

KW - Wind power forecasting

UR - http://www.scopus.com/inward/record.url?scp=85058214403&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058214403&partnerID=8YFLogxK

U2 - 10.1016/j.egypro.2018.11.043

DO - 10.1016/j.egypro.2018.11.043

M3 - Conference article

AN - SCOPUS:85058214403

VL - 155

SP - 350

EP - 358

JO - Energy Procedia

JF - Energy Procedia

SN - 1876-6102

ER -