|
标题：改进重置粒子群算法在MPC调速系统中的应用 |
|
作者：王永宾，许军，周奇勋 |
|
| 2015年第5期 访问次数：312次 |
|
摘要：提出一种改进重置粒子群优化算法，用于模型预测控制器多参数的整定优化。改进算法分为三个搜索阶段："全局搜索"阶段采用Neumann邻域粒子群优化算法，实现全局搜索能力和收敛速度间的均衡；"局部搜索"阶段采用无导数拟牛顿BFGS算法，提高收敛速度和收敛精度；"动态重置"阶段对满足重置条件的子群进行重置，解决收敛早熟问题。利用标准测试函数对4种粒子群优化算法进行比较分析，结果表明改进重置粒子群优化算法在收敛效率、收敛精度与通用性方面占据优势。将改进重置粒子群优化算法应用于模型预测控制永磁同步电动机调速系统，仿真结果表明最优参数能够保证计算得到优化控制律，从而实现系统性能的改善。
关键词：型预测控制；粒子群；全局最优；重置；拟牛顿算法
Abstract:This paper proposed a improved reinitialization particle swarm optimization (PSO) algorithm to tune parameters of MPC. The optimization process was divided into three searching stages. In the global searching stage, Neumann PSO method was adopted to give a compromise between global search ability and convergence speed. In the local searching stage, derivativefree quasiNewton method, known as BFGS algorithm, was adopted to improve the convergence speed and accuracy of parameters. In the dynamic reinitialize stage, the particles in subgroups which satisfied the reinitialization conditions were reinitialized within the searching space. This is very useful to solve premature convergence. Four improved PSOes were tested with the standard test functions together, and simulation results show that the BFGSR proposed has better performance in convergence efficiency, convergence precision and versatility. The BFGSR was adopted to optimize the parameters of MPC in the PMSM speed control systems, and simulation results showed that the static and dynamic performance of the system achieved significant improvement with the optimal adjustable parameters.
Key words: MPC; PSO; global optimum; reinitialization; quasiNewton method |
|
| |
|
