Paper Abstract and Keywords |
Presentation |
2015-03-02 11:00
Prediction of temperature distribution by gaussian process dynamical model for green data center Koji Suganuma (NAIST), Yuya Tarutani, Go Hasegawa, Yutaka Nakamura (Osaka Univ.), Norimichi Ukita (NAIST), Kazuhiro Matsuda (NTT - AT), Morito Matsuoka (Osaka Univ.) NS2014-204 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The prediction of temperature distribution is required for reducing the power consumption of the data center. In this paper, we propose a method to predict the temperature distribution in data center by machine learning with the past operation data of the data center. In the proposed method, we use Gaussian process dynamical model (GPDM), which the method applies Gaussian process for dynamical system, as a learning model for reflecting the nonlinearity and dynamics of operation data of the data center. In this paper, we evaluate the performance of proposed method by comparing the prediction of temperature distribution results with the actual data at our experimental data center. As a result, we show that the proposed method can predict temperature distribution within a range of 0.932 $+/-$ 1.51 (root mean square error $+/-$ 1 standard deviation) using Gaussian process dynamical model in this paper. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
data center / power consumption reduction / temperature distribution prediction / gaussian process dynamical model / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 114, no. 477, NS2014-204, pp. 155-160, March 2015. |
Paper # |
NS2014-204 |
Date of Issue |
2015-02-23 (NS) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
NS2014-204 |
|