Share Email Print

Proceedings Paper

Echo state network prediction method and its application in flue gas turbine condition prediction
Author(s): Shaohong Wang; Tao Chen; Xiaoli Xu
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

On the background of the complex production process of fluid catalytic cracking energy recovery system in large-scale petrochemical refineries, this paper introduced an improved echo state network (ESN) model prediction method which is used to address the condition trend prediction problem of the key power equipment--flue gas turbine. Singular value decomposition method was used to obtain the ESN output weight. Through selecting the appropriate parameters and discarding small singular value, this method overcame the defective solution problem in the prediction by using the linear regression algorithm, improved the prediction performance of echo state network, and gave the network prediction process. In order to solve the problem of noise contained in production data, the translation-invariant wavelet transform analysis method is combined to denoise the noisy time series before prediction. Condition trend prediction results show the effectiveness of the proposed method.

Paper Details

Date Published: 26 May 2011
PDF: 7 pages
Proc. SPIE 7997, Fourth International Seminar on Modern Cutting and Measurement Engineering, 799735 (26 May 2011); doi: 10.1117/12.887330
Show Author Affiliations
Shaohong Wang, Beijing Institute of Technology (China)
Beijing Information Science & Technology Univ. (China)
Tao Chen, Beijing Information Science & Technology Univ. (China)
Xiaoli Xu, Beijing Information Science & Technology Univ. (China)

Published in SPIE Proceedings Vol. 7997:
Fourth International Seminar on Modern Cutting and Measurement Engineering
Jiezhi Xin; Lianqing Zhu; Zhongyu Wang, Editor(s)

© SPIE. Terms of Use
Back to Top