AFSA-BP速度辨识器在DTC系统中的应用研究

AFSA-BP Network Speed Identifier and Application in Direct Torque Control System

为实现无速度传感器直接转矩控制,有时采用神经网络转速辨识器,但前馈神经网络结构难以确定,运用BP算法时又极易陷入局部解。将人工鱼群算法与BP算法相结合的AFSA+BP算法,实现了人工鱼群算法的全局搜索能力与BP算法的局部寻优性能的互补结合。将所设计的神经网络转速辨识器运用到直接转矩控制系统当中,利用Matlab/Simulink实现无速度传感器控制系统的仿真实验结果表明,该算法具有良好辨识效果。

To achieve speed-sensorless direct torque control （DTC） system, neural network （NN） was ap- plied to design speed identifier. But, feed forward network is difficult to confirm the structure and BP algo- rithm is easy to plunge into local solution. To combine AFSA with BP, which is called AFSA ＋BP algorithm and it realized the combination of AFSA＇s global search capability and BP algorithm＇s local optimize perform- ance. The performance of the proposed scheme was carried out by simulation experiment using Matlab/Simu- link.