Sliding Mode MRAS Observer for PMSM-fed Electric Vehicle Control using Recurrent Neural Network-Based Parallel Resistance Estimator

This paper presents a sensorless speed control method for permanent magnet synchronous machines (PMSMs) driven electric vehicles (EVs). Typically, PMSMs require rotor position information for vector control, but position sensors can be unreliable, prone to failure in harsh conditions compared to other electrical components, and are expensive. Therefore, speed estimation is preferred over a speed sensor for fault-tolerant operation and cost reduction. The sliding mode observer-based model-reference adaptive control (SMO-MRAS) observer estimates the speed and stator resistance. In contrast, the SMO-MRAS observer performs well in medium and high-speed operations. However, parameter sensitivity makes the SMO-MRAS observer less effective at low speeds. Stator resistance is bound to change as the temperature varies. It is necessary to have an appropriate online identification algorithm to address this issue. A parallel stator resistance using a recurrent neural network (RNN) and rotor speed observer based on SMO-MRAS has been proposed in this article. The MATLAB/Simulink results are presented to verify the effectiveness of the overall control scheme.