基于遗传算法的小波神经网络DTC转速辨识  

Rotor speed identifying on DTC system based on wavelet neural network of genetic algorithms

作　　者：周杰[1] 王蕴恒[1] 潘洪亮[1] 

机构地区：[1]黑龙江科技学院电气与信息工程学院

出　　处：《黑龙江科技学院学报》2009年第3期240-243,共4页Journal of Heilongjiang Institute of Science and Technology

基　　金：黑龙江省自然科学基金资助项目(E2007-34)

摘　　要：由于直接转矩控制(DTC)系统的参数难以建立精确的数学模型,提出了基于遗传算法(GA)的小波神经网络(WNN)DTC系统参数辨识方法。利用GA能够搜索全局最优解且不受搜索空间的限制,再加上小波神经网络表现的良好的时频局部化特性,以及多尺度的功能,通过遗传算法对WNN的权值、伸缩因子和位移因子进行优化,实现低速运行时对转速变化的精确控制,改善了DTC系统的转速动态特性。仿真实验结果表明:该方法能提高DTC系统参数的辨识精度,基于GA的WNN具有良好的辨识效果。The difficulties arising from building the exact math model using the parameter of direct torque control system result in the parameters identification on DTC system of wavelet neural networks （ WNN）, based on genetic algorithm （GA）. The use of genetic algorithm for global optimal solution without being limited by search space, combined with better time-frequency localization characteristic of the WNN and better multi-scale function, renders possible the accurate control of speed variation in the case of the low speed, due to optimizing WNN weight, dilation and displacement factors by using GA, thus improving the speed dynamic characteristics of DTC system. The simulation on DTC system indicates that the method, capable of increasing identification accuracy on DTC system parameter proves that the system exhibits a better performance.

关 键 词：DTC 小波神经网络 遗传算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]