一种改进差分进化算法的分数阶系统辨识研究  

An Improved Differential Evolution Algorithm for Fractional Order System Identification

作　　者：余伟 梁恒辉 罗映 Yu Wei;Liang Henghui;Luo Ying(School of Mechanical and Electrical Engineering and Automation,Foshan University,Foshan 528000,China;School of Mechanical Science and Engineering,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]佛山科学技术学院机电工程与自动化学院,广东佛山528000 [2]华中科技大学机械科学与工程学院,湖北武汉430074

出　　处：《系统仿真学报》2021年第5期1157-1166,共10页Journal of System Simulation

基　　金：国家自然科学基金(61803086,61733015,51975234);佛山科学技术学院研究生自由探索基金(2019ZYTS44)。

摘　　要：分数阶模型需要辨识更多的参数,为了建立高精度的分数阶模型,提出一种应用于分数阶系统辨识的改进差分进化算法。在变异策略中基向量从最优个体种群中随机选取,在搜索过程中根据成功变异个体的信息自适应调整缩放因子和交叉概率因子,提高算法的勘探和开采能力。通过求解5个测试函数,证明了改进算法具有较强的求解能力。以永磁同步电机的分数阶模型为例,辨识结果表明:改进差分进化算法在求解精度和收敛速度上具有更好的性能。In order to build a high-precision fractional-order model,which needs to identify more parameters,an improved differential evolution algorithm is proposed for the identification of fractional-order systems.In the mutation strategy,the basis vector is randomly selected from the optimal individual population,and the scaling factor and cross-probability factor are adaptively adjusted according to the information of the successfully mutated individual during the search process to improve the exploration and mining capabilities of the algorithm.By solving the five test functions,the improved algorithm is proved to have strong solving ability.Taking the fractional-order model of permanent magnet synchronous motor as an example,the identification results show that the improved differential evolution algorithm has better performance in solving accuracy and convergence speed.

关 键 词：差分进化算法 分数阶系统 系统辨识 自适应调整 永磁同步电机 

分 类 号：N945.14[自然科学总论—系统科学] TP391[自动化与计算机技术—计算机应用技术]