Hits:

Indexed by:期刊论文

Date of Publication:2010-08-01

Journal:ICIC Express Letters

Included Journals:EI、Scopus

Volume:4

Issue:4

Page Number:1373-1378

ISSN No.:1881803X

Abstract:Coke oven gas is one of the most important byproduct energy in steel industry. The real-time prediction for the amount of generation and consumption of COG is a very significant problem. A prediction method based on time series is proposed in this paper, where an improved echo state network combined with Gaussian process, instead of using linear regression in traditional echo state network, is established. This method makes inferences about the relationship between the reservoir states and networks output using Gaussian process in order to eliminate the ill condition of the model. The experimental results using practical data demonstrate the proposed approach exhibits high prediction precision and can provide effective probability estimation for the prediction. ICIC International ? 2010 ISSN 1881-803X.