神经网络用于异步电机转速估计的仿真研究

Research on the Feasibility of Asynchronous Motor Speed Estimation Using Neural Network

在异步电机的无传感器矢量控制中,转速估计是至关重要的一环。理论上有多种估计方法,并且已有相关的产品面市。而神经网络控制作为一种新兴的控制模式,具有超强的自学习、自适应和泛化能力,理论上能逼近任意非线性函数,特别适用于异步电机这种多变量、强耦合的系统。文章将尝试建立神经网络模型来对异步电机转速进行估计,并分析不同的网络结构及训练方法对估计精度的影响。

In the sensorless vector control of asynchronous motor, the estimation of speed is a crucial part. There are many speed estimation methods, and related products have been available. Neural network control is emerging as a control model with superior self study and generalization. It can approximate any nonlinear function in theory, particularly the strong coupling system with multi- variables as induction motor. This paper attempts to establish neural network models to estimate the speed of the asynchronous motor, and analyses the influence of different network structure and training methods on the precision of estimation.