STATISTICAL PREDICTION OF TIME SERIES OF WIND SPEED AND WAVE HEIGHT BY CONVOLUTIONAL NEURAL NETWORK

DOI Web Site Web Site 2 References Open Access

Bibliographic Information

Other Title
  • 畳み込みニューラルネットワークCNNを用いた風速・波高の時系列の統計的予測
  • タタミコミ ニューラルネットワーク CNN オ モチイタ フウソク ・ ナミ ダカ ノ ジケイレツ ノ トウケイテキ ヨソク

Search this article

Description

<p> Although researches of impact assessment of climate change are studied energetically and needs of projection of oceanographic phenomenon have been increasing, simulating oceanographic phenomenon using numerical models requires computational cost. This study predicts time series of oceanographic phenomenon such as wind speeds and wave heights from spatial atmospheric phenomenon using Deep Learning (CNN: Convolutional Neural Network). This study also examines difference of the effects of physical factors of atmospheric phenomenon and hyper parameter of CNN. The physical factors mean combination of explanatory variables, input ranges of atmospheric phenomenon, and so on. The hyper parameters of CNN mean mini-batch size and number of epochs. Inputting the instantatneous value of pressure field could predict wind speeds for Ise Bay and a point on the Pacific Ocean. Inputting the time history of wind speeds could predict wave heights for Ise Bay and Tottori with high precision because it enables to consider the effect of swell. Developed CNN can predict high waves by inputting the time series of past atmospheric phenomenon and increasing training periods (number of training data).</p>

Journal

References(2)*help

See more

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top