基于Simplex-annealing混合方法的模型参数估计

针对现有算法的导数依赖性及其局部优化性能 ,为控制工程中的模型参数估计课题提供一种新思路。把具有概率突跳特性模拟退火 (SA)和基于高维 Euclid空间中凸多面体结构的单纯形搜索法 (SM)有机地结合 ,通过对搜索操作和参数的有效设计 ,提出了一种基于 Sim plex- annealing混合算法 (SMSA)的模型参数估计方法。对以传递函数、状态空间和自回归滑动平均 (ARMA)模型形式表达的不同典型对象进行仿真 ,结果表明 :SMSA方法在模型结构已知的情况下可准确地估计参数 ,其性能明显优于单一遗传算法(GA)

Simplex-annealing hybrid approach for model parameter estimation

A hybrid method is presented for model -parameter estimation, one of the important issues in control engineering, which avoids the derivative -dependence and local optimal performance of current methods. The probabilistic jumping property of simulated annealing (SA) was combined with the simplex method (SM) based on convex polyhedrons in N -dimensional Euclidean space to form a simplex -annealing hybrid approach (SMSA) for model -parameter estimation with well -designed search operators and parameters. Simulation results using different plants with transfer function, state space and ARMA representation respectively show that SMSA can accurately estimate the parameters when the model structure is known and that the performance is better than that of the simple genetic algorithm (GA).