摘要
旋风分离器是化工行业常用气固分离装置,准确地预测旋风分离器的压降性能,并对其进行设计和放大至关重要。当前旋风分离器压降模型存在建模时间较长和预测精度较差的问题,为此采用极限学习机(Extreme learning machine,ELM)对旋风分离器压降进行了建模,并引入粒子群优化(Particle swarm optimization,PSO)算法对ELM输入层到隐含层连接权值和阈值进行了优化,以降低ELM对隐含层节点数的需求,提高模型准确度和稳定性。研究表明,优化结果较标准ELM降低了对隐含层节点数的需求,模型测试集R2和MSE分别为0.9978和2.443×10^(-4),运行时间为15.74 s,相比标准ELM模型、统计模型和人工神经网络模型,所建基于PSO-ELM的旋风分离器压降模型有更好的泛化能力和鲁棒性,极大地缩短了预测时间。PSO-ELM建模算法可以作为一种有效的方法,为旋风分离器性能分析提供指导。
The cyclone separator is a common gas-solid separation device in chemical industry.Accurate prediction of the pressure drop performance of cyclone separator is very important for its design and scale-up.At present,the pressure drop models of cyclone separator have the problems of long modeling time and poor prediction accuracy.Therefore,the extreme learning machine(ELM)was used to model the pressure drop of cyclone separator,and the particle swarm optimization(PSO)algorithm was introduced to optimize the connection weight and threshold between input layer and hidden layer of ELM,so as to reduce the demand of ELM for hidden layer nodes and improve the accuracy and stability of the model.The results show that the optimization results reduce the demand for hidden layer nodes compared with the standard ELM model.The model test sets R2 and MSE are0.9978 and 2.443×10^(-4),respectively,and the running time is 15.74 s.Compared with the ELM model,statistical model and artificial neural network model,the cyclone separator pressure drop model based on PSO-ELM has better generalization ability and robustness,and greatly shortens the prediction time.PSO-ELM modeling algorithm can be used as an effective method to guide the performance analysis of cyclone separator.
作者
王兆熙
延会波
张玮
WANG Zhao-xi;YAN Hui-bo;ZHANG Wei(College of Chemistry and Chemical Engineering,Taiyuan University of Technology,Taiyuan 030024,Shanxi,China)
出处
《天然气化工—C1化学与化工》
CAS
北大核心
2021年第4期119-125,共7页
Natural Gas Chemical Industry
基金
国家重点研发计划项目(2018YFB0604603)
山西省重点研发计划项目(201903D121027)
。
关键词
极限学习机
粒子群优化算法
旋风分离器
建模
压降
extreme learning machine
particle swarm optimization algorithm
cyclone separator
modeling
pressure drop
作者简介
第一作者:王兆熙(1993-),硕士研究生,研究方向为机器学习建模及智能优化算法,电话:17735578619,E-mail:1287861729@qq.com。;通讯作者:张玮(1973-),博士,副教授,研究方向为机器学习建模及智能优化算法,电话:13934212849,E-mail:zhangwei01@tyut.edu.cn。
