摘要
两层非线性规划问题的非凸性和非可微性给全局最优解的求解带来了较大的困难,目前还缺乏成熟的全局优化策略,同时其易并行计算特性未得到重视。提出了基于模拟退火算法的两层非线性规划问题的并行全局优化策略。融合单纯形法和模拟退火算法设计了一种并行模拟退火算法,用来求解上层问题,用精确惩罚函数处理约束。下层的非线性规划问题则采用可变容差单纯形算法完成求解。所设计的组合算法有效地结合了两层非线性规划问题的易并行性,便于实现网络并行计算。计算表明算法有着可靠的全局收敛性和较高的收敛速度。
There are great difficulties in solving the general bilevel nonlinear programming problems (BNPP) with the global solutions because of their nonconvexity and indifferentiability. There isn't an effective global approach for BNPP at the present time, and furthermore, the pleasantly parallel computation property of BNPP isn't recognized yet. The parallel global approach based on simulated annealing for solving the general BNPP is proposed. A kind of Parallel Simulated Annealing (PSA) combining simulated annealing with simplex method is presented. The higher-level problem is solved by PSA and the accurate penalty function is employed to deal with its constraints. The flexible tolerance simplex method is used to solve the lower-level nonlinear programming problems. The combinational approach makes full use of the pleasantly parallel computation property of BNPP, and the network parallel computations can be easily completed. Its high reliability of global convergence and high convergence rate are testified by the numerical computations.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2005年第5期1040-1044,共5页
Journal of System Simulation
关键词
两层非线性规划
并行模拟退火算法
全局优化
并行计算
bilevel nonlinear programming
parallel simulated annealing
global optimization
parallel computation
