Zastosowanie sieci neuronowych do estymacji zmiennych stanu napędu elektrycznego z silnikiem synchronicznym z magnesami trwałymi

In this paper analysis of possibilities of neural network application for estimation of speed of permanent magnet synchronous motor is presented. In order to realize of this task Multi Layer Perceptron Neural Network are applied. Several design steps with particular emphasis on the selection of structure of neural network and organization of the input vector are described. Chosen results for prepared neural estimator are presented. High precision of speed estimation is obtained. Additionally changes of stator resistance are introduced during tests, even in such case reproduction of this state variable is very precise, tested model is robust. Calculations related to prepared model are realized in Matlab.