Parallel stator resistance estimator using neural networks for rotor flux based model reference adaptive system speed observer

Abstract

Model reference adaptive system schemes offer simpler implementation and require less computational effort compared to other speed sensorless methods. The performance of rotor flux based model reference adaptive system schemes at low-speed operation is poor because of parameter sensitivity and presence of the integrator in the reference model. As stator resistance inevitably varies with temperature, for accurate operation at low speeds, an appropriate online identification algorithm for the stator resistance is required. In this article, a neural network based parallel stator resistance and rotor speed estimator has been proposed to simultaneously rectify the limitation of model reference adaptive system schemes, i.e., stator resistance variation and DC offset due to integrator, employing a neutral network in stator resistance estimator and modifying the reference model by adding a compensating voltage term. An indirect sensorless vector control scheme has been simulated and experimentally validated using the dSPACE DS-1104 R&D controller board (dSPACE GmbH, Paderborn, Germany) to verify the performance of drives at different operating conditions. Copyright © Taylor & Francis Group, LLC.