Sensorless Speed Control of SRM Drive Using Optimized Neural Network Model for Rotor Position Estimation
Amarnath Yalavarthi, Bhim Singh
Abstract
This paper presents an artificial neural network (ANN) approach for the development of position/speed sensorless switched reluctance motor (SRM) drive. A precise estimation of rotor position is essential for reliable operation of the drive. Due to its non-linear saturation characteristics, SRM poses a challenging task in modelling mathematically for developing control algorithm. In this work, a neural network uses a real-time magnetic data, which is mapped through supervised training by using Levenberg–Marquardt (LM) based back propagation (BP) learning algorithms in a Simulink environment. Neuron number play a dominant role while training the network and impacts fitting level. Therefore, network model is optimized for number of neurons in the internal layers. Besides position sensorless scheme, the system uses an advance angle control for controlling the speed of SRM drive, eliminating the switching losses incurred by conventional hysteresis controller. Response of the drive during steady state and transient conditions is validated experimentally, accounting its suitableness for the control of SRM.