Optimization of Parameters for FTS Kinetic Model by Genetic Algorithm

Han Rui feng,  Zhang Yong kui^＃
(Xinzhou Teacher’s University,  Xinzhou  034000  ^＃Computer Science Department, Shanxi University,  Taiyuan  030006)

Abstract  Detailed kinetic model is one of the most important basic research items for Fischer-Tropsch Synthesis(FTS). LM(Levenberg-Marquardt) algorithm is used a lot in estimating parameters of the kinetic model. However, as an unlimited algorithm, LM often makes an inaccurate conclusion because of parameters exceeding the limit. Its computation greatly depends on the initial point, and easily falls into non-global optima. It is a new attempt to apply Genetic Algorithm(GA) to the solutions of optimization problems of FTS parameters. The conclusion comes that GA can be applied to such problems, which having obtained three-comparatively-satisfying-group results of parameter-estimating through a number of systemic tests.
Key words  Fischer-Tropsch synthesis,  Parameter estimation,  Genetic algorithm

遗传算法用于费托合成反应动力学参数优化

韩瑞峰  张永奎^＃
(忻州师范学院信息网络中心  忻州  034000  ^＃山西大学计算机科学系  太原 030006)

摘要  详细动力学模型是费托合成(Fischer-Tropsch Synthesis，FTS)反应技术从实验室走向工业化过程中最关键的基础研究项目之一。对动力学模型的参数估算一直采用传统的无约束优化算法LM(Levenberg-Marquardt)算法，在计算中容易因参数越界而使计算失败，计算结果强烈依赖于初值，且容易限于局部最优。运用遗传算法来解决费托合成反应详细动力学模型的参数优化问题是一种全新的尝试，通过系统的实验获得了三组比较满意的参数估算结果，证明遗传算法用于解决该参数优化问题是适宜的。
关键词  费托合成  参数优化  遗传算法

韩瑞峰  硕士，讲师，主要从事人工智能和遗传算法的研究。
山西省高校科技研究开发项目(200358)
2003-04-02收稿，2003-11-28接受 

(PDF全文下载c04041)