摘要
针对前馈神经网络在石油化工软测量建模中存在的不足,采用递归型Elman网络实现动态石化生产过程的建模。提出了一种新型的Elman网络——HF Elman网络,将其应用在某石化厂600 kt/a乙烯生产装置乙烯质量指标的软测量建模,并与前馈网络和Elman网络的建模性能进行了比较,仿真结果表明,HF Elman网络具有良好的动态建模能力,能够更好地实现乙烯精馏塔出口成分含量的软测量建模。
The recurrent Elman neural network is adopted to the modeling of dynamic petrochemical production aiming at the shortcomings of static feed forward neural network in soft sensor modeling, a new modified Elman neural network (HF Elman) is proposed and applied to modeling the product quality of 600 kt/a ethylene unit in one petrochemical Co.. Comparing its performance to feed forward network and other Elman including improved Elman networks, simulation results have shown that the HF Elman neural network has dynamic modeling capacity. Soft sensor can realize modeling for the product composition of the ethylene rectifying column.
出处
《石油化工自动化》
CAS
2008年第6期35-38,共4页
Automation in Petro-chemical Industry
基金
甘肃省自然科学基金资助课题(0803RJ2A026)