一种改进Wolef-BFGS-SQP法对随机规划的研究  

Study on Stochastic Programming with an Improved Wolef-BFGS-SQP Method

作　　者：崔迪[1] 孙祥斌[2] 张伟[1] 

机构地区：[1]山东科技大学信息科学与工程学院,山东青岛266510 [2]山东工商学院人事处,山东烟台264005

出　　处：《山东科技大学学报（自然科学版）》2005年第2期94-96,共3页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金资助项目(10171055)

摘　　要：基于拟蒙特卡洛随机模拟的Wolef-BFGS-SQP法对随机规划的再研究。对于求解LC1函数优化的BFGS-SQP算法,改用Wolef线性搜索原则,同样得到了该算法的全局收敛性,并把它推广到解决随机规划的问题上去。在随机模拟过程的处理上采用拟蒙特卡洛随机模拟技术,鉴于计算量比较大的问题,采用仅多加一个线性约束的信赖域技术计算,即可克服收敛慢等的缺点。This paper studies two-stage stochastic programming problem with a new Wolef-BFGS-SQP method based on the Quasi-Monte-Carlo stochastic simulation techniques. In particular, for the BFGS-SQP algorithm used to solve the LC function optimization, its global convergence of the algorithm can also be obtained by means of the Wolef linear search principle, which can be popularized to solve the stochastic programming problem. The Quasi-Monte-Carlo technology has been applied to handle the stochastic simulation process. Since the calculating amount is quite large, trust-region technology in which only one linear constraint is added can be used to overcome the slow convergence problem.

关 键 词：随机规划 拟蒙特卡洛 随机模拟 Wolef-BFGS-SQP法 

分 类 号：O221.5[理学—运筹学与控制论]