Title: Electric power communication bandwidth prediction based on adaptive extreme learning machine

Authors: Li Di; Zheng Zheng; Song Wang; Ruidong Zhang; Min Xia; Kai Hu

Addresses: State Grid Henan Economics Research Institute, Zheng Zhou, 450052, China ' State Grid Henan Economics Research Institute, Zheng Zhou, 450052, China ' State Grid Henan Electric Power Company, Zheng Zhou, 450052, China ' Jiangsu Collaborative Innovation Canter on Atmospheric Environment and Equipment Technology, Nanjing University of Information Science and Technology, Nanjing, 210044, China ' Jiangsu Collaborative Innovation Canter on Atmospheric Environment and Equipment Technology, Nanjing University of Information Science and Technology, Nanjing, 210044, China ' Jiangsu Collaborative Innovation Canter on Atmospheric Environment and Equipment Technology, Nanjing University of Information Science and Technology, Nanjing, 210044, China

Abstract: Bandwidth demand forecasting is the basis and foundation of the power communication network planning. For the traditional neural network learning, there are many problems, such as slow convergence speed, more iterative times, and easy to fall into local optimum. An adaptive extreme learning machine model based on the theory of extreme learning machine and K nearest neighbour theory is proposed to predict the bandwidth of electric power communication. The adaptive metrics of inputs can solve the problems of amplitude changing and trend determination, and reduce the effect of the over-fitting of networks. The proposed algorithms are validated using real data of a province in China. The results show that this method is better than the traditional neural network, autoregressive models, self organisation models, and single extreme learning machine model. It can be used in electric power communication bandwidth prediction.

Keywords: electric power communication; bandwidth prediction; extreme learning machine; ELM; K nearest neighbours.

DOI: 10.1504/IJES.2018.091786

International Journal of Embedded Systems, 2018 Vol.10 No.3, pp.233 - 240

Received: 22 Jul 2016
Accepted: 09 Jan 2017

Published online: 16 May 2018 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article