Stability Analysis of Sensorless Speed Control for PMSM Considered Current Control System

Automatic Tuning Method for PMSM using Big Data based Artificial Neural Network

“…These training data are the control gains and other data of the PMSM that have been adjusted in the past. During training, the weighting of each neuron is performed by back-propagation to minimize the error evaluation function E shown in equation (1). y(NM, Np1, Np2) is the output of the ANN for NM PMSMs with Np1 and Np2 parameters set, and t(NM, Np1, Np2) is the stable discrimination result, which is the teacher signal and takes the value of 0 or 1.…”

“…In this paper, the cross-angle frequency of the current controller ACR is set to 628[rad/s], and the other two controllers are configured to adjust to the response of the current controller. In order to adjust all three controllers simultaneously, ACR should be added as Np3 in equation (1). Such a configuration enables even more optimal adjustment.…”

“…A method based on linear analysis with equilibrium point approximation has been studied as a general gain design method for stable control of permanent magnet synchronous motors (PMSMs), however it is difficult to apply to operating regions and control conditions that cannot be linearly approximated (1) . For example, it is necessary to sufficiently reduce torque and speed ripple during controller switching, which is performed in many products.…”