摘要
空隙率是石油化工企业中非常重要的参数之一。空隙率在线测量过程中存在较大的随机性和不确定性,很难预知空隙率的变化。为了实现对空隙率的预测,提前对两相流系统进行控制和优化,提出了基于改进猫群优化(CSO)算法长短期记忆(LSTM)网络的空隙率预测算法。利用LSTM善于处理时间序列型数据的特点对空隙率进行预测,在CSO中引入模拟退火(SA)算法和平均惯性权重,改善了在预测中易陷入局部最优和全局搜索能力较弱的缺点,保证了位置的收敛性。结果表明,该算法模型具有较高的预测精度和收敛速度,可以更快更精确预测空隙率的变化,克服了数据不确定且随机的难点,对提前控制和优化两相流系统具有较高的工业应用价值。
Void fraction is one of the most important parameters in petrochemical enterprises.There is great randomness and uncertainty in the online measurement of void fraction,and it is difficult to predict the change of void fraction.In order to realize the prediction of the void fraction,the two-phase flow system is controlled and optimized in advance,a void fraction prediction algorithm based on improved cat swarm optimization(CSO)algorithm of long short-term memory(LSTM)network is proposed.Use LSTM to be good at processing time series data to predict the void fraction,introduce simulated annealing algorithm and average inertia weight in CSO,improve the shortcomings of easy to fall into local optimal and weak global search ability in prediction,and ensure the convergence of the location.The results show that the algorithm model has higher prediction accuracy and convergence speed,can predict the change of void fraction faster and more accurately,and overcomes the difficulty of uncertain and random data,It has high industrial application value for controlling and optimizing the two-phase flow system in advance.
作者
刘晓
阚哲
钱宇加
LIU Xiao;KAN Zhe;QIAN Yujia(School of Information and Control Engineering,Liaoning University of Petrochemical Technology,Fushun 113001,China)
出处
《传感器与微系统》
CSCD
北大核心
2022年第7期57-60,64,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61703191)
辽宁省自然科学基金资助项目(201602468)
辽宁省教育厅一般项目(L2020019)。
关键词
两相流
空隙率
改进猫群优化算法
模拟退火算法
平均惯性权重
长短期记忆
two-phase flow
void fraction
improve cat swarm optimization(CSO)algorithm
simulated annealing(SA)algorithm
average inertia weight
long short-term memory(LSTM)
