A RBF Neural Network Based Sensorless Control Scheme for Switched Reluctance Motor

摘要

A new sensorless control method for Switched Reluctance Motor (SRM) is proposed in this paper. To simplify the measuring procedure and enhance the accuracy of the flux linkage, a new flux linkage calculation method is developed by combing the three-dimensional Finite element method and the measurement of the aligned flux-linkage. The Radial Basis Function (RBF) neural network is utilized for modeling the nonlinear flux linkage characteristics, and a 5-order polynomial fitting model is developed for modeling the flux/current characteristics at the turn-OFF angle. Thus, using the Turn-on angle, Turn-OFF angle and the phase current, the flux-linkage thresholds at the Turn-ONangle of the next driving phase and the Turn-OFF angle of the current driving phase can be estimated. By comparing the estimated flux linkage with the Turn-ON and Turn-OFF flux linkage thresholds, the driving signals of each phase can be estimated, which can be used for rotor speed estimation and sensorless control. Experiments verify the validity of the sensorless scheme.