文献类型：期刊文献

中文题名：改进的教学优化算法及其应用

英文题名：Study and application of improved teaching-learning-based optimization algorithm

作者：于坤杰[1];王昕[2];王振雷[1]

机构：[1]华东理工大学化工过程先进控制和优化技术教育部重点实验室,上海200237;[2]上海交通大学电工与电子技术中心,上海200240

年份：2014

卷号：33

期号：4

起止页码：850

中文期刊名：化工进展

外文期刊名：Chemical Industry and Engineering Progress

收录：CSTPCD;;Scopus;北大核心:【北大核心2011】；CSCD:【CSCD2013_2014】；

基金：国家自然科学基金(U1162202;61174118;61222303);中央高校基本科研业务费专项资金;国家863计划(2013AA040701);上海市科技攻关项目(12dz1125100);上海市重点学科建设项目(B504);流程工业综合自动化国家重点实验室开放课题基金项目

语种：中文

中文关键词：TLBO算法;函数优化;乙烯裂解炉

外文关键词：teaching-learning-based optimization algorithm; function optimization; ethylene crackingfurnace

摘要：针对教学优化算法(TLBO)收敛速度慢,容易陷入局部最优的问题,本文提出了一种改进的方法。算法的改进主要在两方面:一是对教学因子(TF)进行自适应调整,使TF随算法迭代减小,这样算法在搜索前期采用全局搜索,搜索空间快速收敛于最优解附近,提高搜索速度,搜索后期采用局部精细搜索以获得高精度的解。二是引入信任权重,对学生已获得的知识采取部分信任的策略,避免对已获取知识的过分信任,增加学生个体与教师及学生之间的信息共享,利于算法跳出局部最优。算法在8个标准测试函数上应用,仿真结果表明改进的算法有更快的收敛速度并且能够跳出局部最优。最后将改进的算法应用到乙烯裂解炉裂解运行效益优化中,显著提高了裂解炉的效益。
To overcome the slow convergence speed and easy trapping in local optimum of the Teaching-Learning-Based Optimization （TLBO） algorithm, an improved algorithm was proposed. First, adaptive adjustment of teaching factor TF was proposed to make the TF decrease with iterative algorithm, so the algorithm used the global search in the early search stage to make the search space rapid convergence to near optimal solution to increase search speed. In the later stage, fine local search was used to obtain the higher accuracy solution. After that, trust weight was proposed to take a partial trust strategy in the already acquired knowledge by students, avoiding too much trust in the acquired knowledge and increasing the information share between students and teachers and among the students, so that the algorithm jumped out of the local optimal. The simulation results showed that the improved algorithm had faster convergence speed and the ability to jump out of local optimal. Finally the improved algorithm was used to optimize the benefits of ethylene cracking furnace, the benefits was promoted observably.