利用人工神经网络进行感应电机转子磁链估计

为了准确估计感应电机转子磁链，实现感应电机高性能矢量控制，提出了两种基于人工神经网络的感应电机转子磁链的估计模型：采用串联结构的神经网络模型和采用复合结构的神经网络模型。前者是由递归网络和多层前馈网络串联组成；后者则是由这两种网络耦合而成。复合神经网络包含一个递归单元层和多个前馈单元层，类似于一个状态反馈系统。这两种网络模型主要是针对不同训练方式而建立起来的，针对复合结构神经网络还提出一种新的有效学习算法。仿真结果表明，这两种模型能够快速准确地估计在负载变化条件下感应电机的转子磁链。

Rotor flux estimation of an induction motor using artificial neural networks

In order to estimate the rotor flux of an induction motor accurately and to achieve a high vector control performance, two estimation models based on artificial neural networks are presented. Two networks are employed in series in one model, and coupled in the other. Both are comprised of a recurrent neural network and an MLP network. The coupled networks, which are similar to a state feedback system, include one recurrent layer and several feedforward layers. These two network models are designed in accordance with two different training methods. And a new efficient algorithm suitable for the coupled networks is developed. The simulation results show that the models can instantaneously and accurately estimate the rotor flux magnitude of an induction motor under a variety of load conditions.