| In order to ensure the gains and quality of methanol production, the important technical parameter——mass fraction of crude methanol must be closely monitored in methanol synthesis process. However, due to some technical reasons and the economic conditions, the parameter can't be measured in real-time. This paper is sponsored by the methanol project designed by a chemical factory, builds a soft sensing model of methanol synthesis process based on Least Squares Support Vector Machine to calculate the mass fraction of crude methanol. and the model parameters are selected by improved Artificial Bee Colony Algorithm, so mass fraction of crude methanol can be estimated and detected on-line.First, this paper researches an intelligent optimization algorithm——Artificial Bee Colony Algorithm (ABC) which was presented in 2005, and to the deficiency of Artificial Bee Colony Algorithm, Bi-Group Artificial Bee Colony Algorithm based on Chaotic Optimization (BCABC) is brought forward. The simulation results show that the convergence rate and exact of BCABC are better than basic ABC's. Using BCABC to select LSSVM model parameters can get better combined model parameters and BCABC-LSSVM model has higher accuracy, LSSVM method based on improved Artificial Bee Colony Algorithm is feasible and effective. Second, BCABC-LSSVM is used in methanol synthesis process. Select secondary variables and build methanol synthesis process soft sensing model based on BCABC-LSSVM after studying the methanol synthesis process and analyzing the influencing factors of the methanol synthesis process by actual field data. The calculation results by the soft sensing model based on BCABC-LSSVM agree well with actual value and the soft sensing model has good generalization ability. Finally, for the methanol synthesis process, VC++6.0 and KingView6.5 are used to develop a soft sensing system the which can run on upper computer and the tested results are well. |
PDF Full Text Request