location: Current position: Home >> Scientific Research >> Paper Publications

Online Parameter Optimization-Based Prediction for Converter Gas System by Parallel Strategies

Hits:

Indexed by:期刊论文

Date of Publication:2012-05-01

Journal:IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY

Included Journals:SCIE、EI

Volume:20

Issue:3

Page Number:835-845

ISSN No.:1063-6536

Key Words:Graphic processing unit (GPU) acceleration; Linz Donawitz converter gas (LDG) system; least square support vector machine (LS-SVM); online parameter optimization; parallel particle swarm optimization (PSO)

Abstract:Linz Donawitz converter gas (LDG) is one of the most important sources of fuel energy in steel industry, whose reasonable use plays a crucial role in energy saving and environment protection. In practice, online prediction of variation of gas holder level and gas demand by users is fundamental to gas utilization and scheduling activities. In this study, a least square support vector machine-based prediction model combined with the parallel strategies is proposed, in which parameter optimization is realized online by a parallel particle swarm optimization and a parallelized validation method, both being implemented with the use of a graphic processing unit. The experiments demonstrate that the online parameter optimization based model greatly improves the prediction quality compared to the version with the fixed modeling parameters. Furthermore, the parallelized strategies largely reduce the computational cost thus guaranteeing the real-time effectiveness of the practical application.

Pre One:A MKL based on-line prediction for gasholder level in steel industry

Next One:Prediction for noisy nonlinear time series by echo state network based on dual estimation