Non-Intrusive Induction Motor Parameter Estimation Using MRAS Algorithm and Neural Network Rotor Flux Observer

Three-phase induction motors are indispensable components of almost all industries and so it’s of crucial importance to monitor their key parameters to assure their reliable operation with minimum downtime. As such, this paper focuses on estimating the equivalent-circuit parameters of these machines. The proposed method is based on the model reference adaptive system (MRAS) and rotor flux linkage observer. A neural network state filter with feedforward multi-layer perceptron (FMLP) architecture estimates the rotor flux, and the motor speed is obtained by rotor slot harmonics or eccentricity harmonics techniques. The suggested algorithm can estimate all the motor parameters in an online, continuous, and non-intrusive way with acceptable accuracy and only needs current and voltage signals, along with motor nameplate information. A laboratory test setup is developed to validate the proposed algorithms and experimental results illustrate their performance in estimating the motor parameters with high accuracy.