摘要
针对无刷直流电机模糊控制中误差变化滞后于被控对象的状态的特点,将被控对象的神经网络模型引进模糊控制系统,根据当前时刻误差和预测误差变化值,预测下一时刻的控制输出和系统在未来时刻的误差。采用遗传算法与BP算法相结合对网络进行训练,改进网络性能,提高控制信息与实时状态的适应性。仿真和实验结果表明,这种神经网络预测模糊控制器取得了良好的控制效果。利用遗传算法和BP相结合作为神经网络的训练算法,既能避免陷入局部极小又能快速在小的解空间定位最优解。
According to BLDC fuzzy control character that change-of-error lags behind states of plant,neural network model of plant is introduced into fuzzy control system.It can forecast future output of plant,and then next sampling period change-of-error can be achieved.Based on the current error and predictive change-of-error,predictive fuzzy control output of fuzzy logic controller could be obtained by general fuzzy logic control rules.Using GA(genetic algorithm) combined with BP algorithm as the training algorithm of neural network,the network quality is improved and the control input is more fit to system states.Simulation and experience results show that predictive control is feasible.GA and BP training network avoid local nadir and quickly find optimal solution in small space.
出处
《控制工程》
CSCD
北大核心
2009年第S1期97-100,共4页
Control Engineering of China
基金
高等学校博士学科点专项科研基金资助项目(20070248010)
