优化变结构径向基函数网络研究及其在MDEA水溶液CO_2吸收能力预测中的应用  

Optimized RBF neural network and it's application in predicting solubility of carbon dioxide in methyldiethanolamine solutions

作　　者：张健[1] 郭庆华[1] 廖敏[1] 梁钦锋[1] 周志杰[1] 于广锁[1] 

机构地区：[1]煤气化教育部重点实验室,华东理工大学,上海200237

出　　处：《计算机与应用化学》2010年第4期507-511,共5页Computers and Applied Chemistry

基　　金：上海市曙光计划(06SG34);教育部新世纪优秀人才支持计划(NCET-06-0416)

摘　　要：CO_2排放导致的全球气候变化日益受到关注,通过技术手段解决温室效益被广泛讨论。MDEA溶液吸收烟气中CO_2是1种传统有效的减排技术。因此,在工程设计、操作和优化中,MDEA溶液在不同条件下CO_2吸收能力的预测至关重要。本文提出新型可逆跳跃马尔可夫-蒙特卡洛法优化的变结构径向基函数网络模型。并用该模型关联文献中的CO_2-MDEA-H_2O体系在压力0.1～4559.5 kPa和温度298～393 K之间的气液平衡数据。经随机抽取的8组数据验证,模型推算结果与实验数据十分接近.该模型精度可满足工程设计要求。There is growing concern that anthropogenic carbon dioxide（CO_2） emissions are contributing to global climate change. Therefore,it is critical to develop technologies to mitigate this problem.One very traditional and efficient approach to reducing CO_2 emissions is using MDEA to absorb flue gas containing CO_2.It is very important to find an efficient way to predict solubility of carbon dioxide in MDEA in the project design,operation and optimization.In this paper,we explored a new RBF neural network optimized by reversible jump Markov-Chain Monte Carlo steps.The model trained by reported data in literature which between（0.1～4 559.5） kPa pressure and（298～393） K temperature.Eight random data were used to test the model.The result shows that the algorithm is effective and fit the design requirement.

关 键 词：CO_2溶解度 MDEA水溶液 人工神经网络技术 可逆跳跃马尔可夫-蒙特卡洛法 

分 类 号：TQ015.9[化学工程] O6-39[理学—化学]