Sensorless control of permanent magnet synchronous motor using ANFIS based MRAS

Many high performance industrial and traction permanent magnet synchronous motor (PMSM) drives require position sensorless operation. Most of the sensorless control techniques are dependent on motor parameters, such as stator resistance, inductance, and torque constant. Thus, the performance suffers greatly in harsh and highly dynamic operating conditions, where the motor parameters are changing. To overcome this problem, Model Reference Adaptive System (MRAS) based algorithms have been developed, which offer adaptive and simple solution. In this paper, an Adaptive Network-based Fuzzy Inference System (ANFIS) based MRAS observer is proposed, where the adaptive model and adaptation mechanism of the conventional MRAS is replaced by ANFIS. The combined capability of neuro-fuzzy controller in handling uncertainties and learning from processes is proven to be advantageous in modeling highly nonlinear systems. Thus, to neutralize the effect of parameter variations, a novel online tuned ANFIS architecture is developed, which is optimized for Surface PMSM MRAS. This architecture tracks the rotor position and speed accurately in the entire speed range. Furthermore, a detailed comparative simulation and experimental study is carried out for ANFIS and sliding mode observers. The proposed ANFIS based estimation technique shows better performance and immunity to parameter variation compared to sliding mode observer.