改进蚁群算法优化神经网络速度辨识器的研究与应用

Research and Application of Speed Identifier Based on Neural Network Optimized by Modified ACO

对基本蚁群算法进行了改进以提高其全局寻优速度,并用此改进型的蚁群算法来优化神经网络,再将该神经网络应用于直接转矩控制系统当中,实现对转速的观测,最终实现无速度传感器的直接转矩控制。仿真研究表明,改进的蚁群算法具有优良的全局优化性能,且收敛速度较快,应用其优化的神经网络在直接转矩控制系统中能较准确地辨识出系统转速,从而可以替代传统的速度传感器,实现无速度传感器直接转矩控制。

To enhance its global optimization speed,the basic ant colony optimization（ACO）was modified,then it was used to optimize the neural networks（NN）,and the optimized NN was applied to the direct torque control（DTC）system,so that the rotate speed can be observed.The DTC of sensorless was implemented at last.The research of simulation shows that the modified ACO has eminent global optimization performance,and fast convergence rate,the rotate speed of the system is able to be identified by the DTC system with the NN optimized by this method exactly,thereby,the traditional speed sensor can be replaced and the DTC with sensorless can be implement.