ニューラルネットワークによる風速の予測能力の検討

概要

論文の詳細を見る
Effective prediction of wind speed and direction is the basis for a wind energy conversion system aiming at economical fuel control of a stand alone wind energy conversion system. There are several methods for prediction of time serial signal such as statistical approach, linear estimation and Kalman filter approach. However, it is difficult to treat such non-linear prediction problems as wind speed estimation with conventional methods. Hence, the author studied a prediction method using neural networks, which are considered to be powerful for non-linear prediction method. The data base used in our study is composed of time serial points recorded at TAPPI wind park. The first part of the data was used for the training the networks, the other was used for the validation. The types of the neural networks used in this paper are feed forward and recurrent network. It was found that neural networks perform better short-term prediction of wind speed.
八戸工業大学の論文
1998-03-31