Stabilization of neural network by combination with AR model in FastADR control of building air‐conditioner facilities. Issue 1 (29th October 2015)

• Record Type: Journal Article
• Authors: Fukazawa, Taro; Iwata, Yuji; Morikawa, Junji; Ninagawa, Chuzo
• Is Part Of: IEEJ transactions on electrical and electronic engineering. Volume 11:Issue 1(2016:Jan.)
• Page Start: 124
• Page End: 125
• Publication Date: 2015-10-29
• Subjects: smart grid -- demand response -- AR model -- neural network; Electrical engineering -- Periodicals; Electronics -- Periodicals; 621.3
• DOI: 10.1002/tee.22196
• Languages: English
• ISSNs: 1931-4973

Abstract

Abstract : Fast automated demand response (FastADR) will be one of possible technologies to realize smart grid ancillary services in future. For FastADR control, a statistical prediction model on the nonlinear response of facility loads is necessary. Although neural networks (NNs) can be utilized for nonlinear prediction, they can have harmful exceptional prediction due to bias of the NN on the training data. In this letter, we propose a combination of an autoregressive (AR) model and an NN to avoid the harmful behavior, and discuss the theoretical basis of the method. © 2015 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.