【摘要】 针对感应电动机直接转矩控制模糊实现中模糊规则的参数难以设计等问题,本文提出了一种基于直接转矩控制的模糊神经元网络实现方法,用来自动调整模糊系统中设计参数。由于直接转矩控制模糊实现中模糊推理和解模糊化函数具有不可导性,本文提出了一种"软最大化"算子。并根据模糊实现及该算子,设计了一个六层模糊神经元网络,对该模糊神经元网络采用改进动态学习率的BP算法进行学习。仿真试验表明:利用学习过的模糊神经元网络可以改进系统的控制性能。更多还原

【Abstract】 Due to the parameters in fuzzy implementation rule based on direct torque contra for induction motor are difficult to determined,this paper proposes ’a fuzzy neural network implementation method,namely "soft maximum",a six-layer fuzzy neural network is designed. In this fuzzy neural network,an improved dynamic learning rate BP algorithm is adopted. Simulation experiment shows that this method can get good performances.更多还原