摘要
针对BP神经网络易陷入局部极小值的缺点,提出了混沌神经网络的学习算法。利用混沌的遍历性和随机性,采用混沌变量全局粗搜索与混沌变量局部细搜索相结合,得到神经网络权值的全局最优值。利用该算法对直接转矩控制(DTC)系统进行转速辨识,仿真结果表明用混沌优化BP神经网络的速度辨识器不仅具有较好的跟踪能力,还提高了运算效率,使系统具有良好的静动态性能。
To solve the disadvantage that BP neural network is liable to get into the local minimum, a novel learning algorithm that chaos neural network is proposed. By the use of the properties of ergodicity and randomness of chaos, and combining global rough search and local elaborate search of chaos variable, get the global optimization weight values of neural network. Using it in speed identification of direct torque control (DTC) system, the simulation results show that the rotor speed identification using the BP neural network optimized by chaos optimization algorithm not only has better track capability, but also enhance the operation effectiveness, it makes the system have favorable static and dynamic preperties.
出处
《电气应用》
北大核心
2008年第13期69-72,共4页
Electrotechnical Application
基金
辽宁省自然科学基金资源共享助项目(20032032)
教育部"春晖计划"合作科研项目(Z2005-2-11008)
辽宁教育厅高校科研项目(2026331)。
关键词
混沌优化算法
BP神经网络
DTC
转速辨识
Chaos optimization algorithm BP neural network DTC rotor speed identification