摘要
DE算法是一类基于种群的启发式全局搜索技术,该算法原理简单,控制参数少,鲁棒性强,具有良好的优化性能。首先利用DE算法对Wiener模型参数进行辨识,分析了算法中变异率F对辨识过程中的全局并行搜索能力和收敛速度的影响;其次运用一种自适应变异差分进化算法(ADE)进行Wiener模型参数辨识,该算法在初期变异率较高,种群具有多样性,避免过早收敛于局部最优解;在进化过程中,变异率逐渐变小,优良个体得以保留,避免最优解遭到破坏。运用ADE算法对Wiener模型的数值仿真结果表明了ADE算法在参数辨识问题中的有效性,以及较PSO算法更强的非线性系统辨识能力。与一般的DE算法相比较,ADE算法辨识到全局最优解的精度和概率有较大提高,对算法参数的敏感性降低。
DE algorithm is a population-based heuristic global search technology. The algorithm has simple principle, fewer control parameters, strong robustness, and good optimization performance. Firstly, differential evolution algorithm for parameters identification of Wiener model was used. The influence of mutation rate F on global parallel search ability and convergence in the process of identification were analyzed. Secondly, an adaptive differential evolution algorithm (ADE) was used to identify parameters of Wiener model. The algorithm keeps individual diversity to avoid premature convergence during the early stage and reduces the mutation rate gradually so as not to damage the optimal solution obtained during the later stage of the search process. Finally, numerical simulation was performed on Wiener model. The results show that ADE algorithm has more effectiveness in parameter identification problem than PSO. On the other hand, compared with the general DE algorithm, ADE algorithm identifies the parameters of Wiener model with higher precision as well as shows lower sensitivity to the al^zorithmic parameters.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第5期969-974,982,共7页
Journal of System Simulation
基金
国家自然科学基金项目(21206053
21276111)
博士后基金项目(1101021B
2012M511678)