摘要
提出一种基于改进遗传算法和递推最小二乘的非线性模糊辨识新算法。该辨识方法包含结构辨识辨出和参数辨识,结构辨识即输入空间的模糊划分,采用具有自适应性的广义高斯隶属函数;参数辨识包含前提参数和结论参数,用基于动态比例变换的改进遗传算法优化高斯函数的前提参数,用递推最小二乘辨识模糊模型的结论参数。最后通过著名的Box-Jenkins煤气炉数据仿真(仿真环境:MATLAB 6.5,计算机主频2.4 GHz,内存512 MB),并根据输入变量个数和模糊规则数,得到均方误差以证明本文方法的辨识精度,将该文辨识方法与其他方法进行比较,验证了该方法辨识精度更高。
A new method for fuzzy identification based on improved genetic algorithm and recursive least square for nonlinear system was proposed. Fuzzy identification contained structure identification and parameters identification. Structure identification means fuzzy partition of input.,space which could take adaptive extended gauss function as membership function, optimized the gauss function premise parameters by using improved genetic algorithm based on dynamic scale transform, and identified the conclusion parameters of the fuzzy model by using recursive least square. Then, the higher precision of this fuzzy identification method （simulation environment： MATLAB 6.5, CPU：2.4 GHz, Memory：512 MB） was demonstrated by the simulation results of the famous Box-Jenkins gas furnace data. The simulation got the error value according to input variables and fuzzy rules. By comparing the error result with other methods on the face of the identification and modeling precision, it validate that the proposed fuzzy identification modeling method has a higher precision.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2008年第7期881-884,共4页
Computers and Applied Chemistry
基金
燕山大学博士基金(B111)．
关键词
遗传算法
模糊划分
广义高斯函数
递推最小二乘算法
genetic algorithm, fuzzy partition, generalized gauss function, recursive least square
