Title: FCM-LSSVM modelling for ethylene loss rate of distillation column with respect to operation conditions

Authors: Cheng Shao; Xiaoyun Dong; Li Zhu

Addresses: Institute of Advanced Control Technology, Dalian University of Technology, Dalian 116024, China ' Institute of Advanced Control Technology, Dalian University of Technology, Dalian 116024, China ' Institute of Advanced Control Technology, Dalian University of Technology, Dalian 116024, China

Abstract: Ethylene loss rate is an important evaluation index in ethylene distillation operation, which directly relates to the comprehensive energy efficiency level of ethylene production. Therefore, the accurate prediction of ethylene loss rate is helpful in optimising energy consumption of ethylene production in distillation columns. It is observed that some working conditions such as the composition of cracking gas, unit load and the refrigerant temperature have a great influence on the ethylene loss rate. In this paper, the prediction issue of ethylene loss rate is thus concerned and a new method with FCM-LSSVM is proposed for modelling the ethylene loss rate of distillation column with respect to the operating conditions. The clustering method is employed to classify optimally the ethylene distillation operation database, and then the LSSVM is suggested to establish the ethylene loss rate prediction model with different working conditions. Finally, simulations and comparative analyses for the proposed method are carried out by using real distillation data, which demonstrates that the model set up under multi-conditions would be more effective due to better predictive precision and generalisation ability.

Keywords: ethylene distillation column; operating condition; fuzzy C-means clustering; loss rate; LSSVM.

DOI: 10.1504/IJCAT.2018.093535

International Journal of Computer Applications in Technology, 2018 Vol.57 No.4, pp.302 - 311

Received: 05 Jan 2017
Accepted: 09 May 2017
Published online: 27 Jul 2018 *

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