摘要
为了解决催化裂化装置再生器出口烟气二氧化硫质量浓度难以实时预测的问题,提出一种基于自适应粒子群优化模糊神经网络(Adaptive Particle Swarm Optimization-Fuzzy Neural Network,ADPSO FNN)算法的催化再生烟气二氧化硫质量浓度智能预测方法。首先,针对数据来源多,且来自数据采集系统(Data Collection System,DCS)与实验室信息管理系统(Laboratory Information Management System,LIMS)中多维数据时间尺度不匹配的问题,利用基于自适应回归算法实现多时间尺度的数据清洗;其次,建立基于模糊神经网络算法的二氧化硫质量浓度预测模型,提取再生烟气产排过程中的动态特性;然后,提出基于动态惯性权重和学习因子机制的自适应粒子群算法,平衡全局探索能力及局部开发能力,实现再生烟气二氧化硫质量浓度的预测;最后,利用炼厂检修前、后的数据分别建立二氧化硫预测模型并进行测试。结果显示:该预测方法实现了催化再生器出口二氧化硫的准确预测,解决了现场多时间尺度数据难以建模的问题。
To solve the real-time prediction of sulfur dioxide(SO_(2))mass concentration in flue gas at the outlet of the regenerator of catalytic cracking units,an intelligent prediction method for SO_(2) mass concentration of regenerated flue gas in fluid catalytic cracking unit based on adaptive particle swarm optimization-fuzzy neural network(ADPSO FNN)algorithm is proposed in this paper.Firstly,aiming at the multiple data sources and the time scales mismatch of multi-dimensional data from the data collection system(DCS)and laboratory information management system(LIMS),the multi-time scale data cleaning is obtained by using the adaptive regression algorithm.Secondly,the SO_(2) mass concentration prediction model based on a fuzzy neural network algorithm is established to extract the dynamic characteristics of the regeneration flue gas production and emission process.Then,an ADPSO algorithm is proposed to realize the prediction of SO_(2) mass concentration in the regenerated flue gas.A dynamic inertia weight and learning factor mechanism,using the population spacing,is proposed to obtain the distribution of particles with suitable diversity and convergence,which can balance the global exploration and local exploitation abilities of the particles.Finally,the SO_(2) mass concentration prediction model is established and tested using the data before and after the overhaul of the refinery.When the number of hidden layer neurons of the four algorithms is identical,the ADPSO FNN algorithm has a better prediction effect.Based on the comparison results of two test experiments,it can be seen that the prediction accuracy of the ADPSO FNN algorithm is the highest,reaching more than 92%.Besides,it can obtain the minimum average value of training root mean squared error(RMSE)and the minimum average value of predicted RMSE.The results show that the prediction method can accurately predict the SO_(2) mass concentration at the outlet of the catalytic regenerator,and solve the problem that it is difficult to model the multi-time scale d
作者
卢薇
杨文玉
张树才
张卫华
李焕
张子玥
LU Wei;YANG Wenyu;ZHANG Shucai;ZHANG Weihua;LI Huan;ZHANG Ziyue(State Key Laboratory of Chemical Safety Control,SINOPEC Research Institute of Study Engineering Company Limited,Qingdao 266000,Shandong,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2023年第11期4127-4135,共9页
Journal of Safety and Environment
关键词
环境学
二氧化硫质量浓度预测
多时间尺度
粒子群优化算法
模糊神经网络
数据清洗
environmentalology
sulfur dioxide mass concentration prediction
multi time-scale
particle swarm optimization algorithm
fuzzy neural network
data cleaning
