<![CDATA[Mechanical speed sensors increase cost, wiring, maintenance requirements, and potential failure points in induction-motor drives, while sensorless estimation remains difficult during low-speed operation, load transients, resistance drift, and speed reversal. This study proposes an excitation-aware innovation-adaptive dual extended Kalman filter (EAIA-DEKF) for simultaneous estimation of mechanical speed, load torque, rotor flux, and stator resistance. The observer combines a six-state extended Kalman filter, a scalar resistance estimator, an amplitude-based excitation gate, bounded innovation-energy covariance adaptation, parameter projection, and the Joseph covariance update. The resistance-dependent current model is retained in both prediction and linearization. A 4.2 s numerical study includes low-speed startup, frequency transitions, load changes, a 52% thermal rise in stator resistance, and speed reversal. Three EAIA-DEKF parameter sets are evaluated under identical data and compared with a fixed-covariance EKF and a current-model MRAS benchmark. The selected set achieves a speed root-mean-square error of 0.2213 rad/s, compared with 2.5112 rad/s for the fixed EKF and 12.3041 rad/s for MRAS. Relative to the fixed EKF, the load-torque and rotor-flux errors decrease by 54.82% and 90.11%, respectively. The coordinated state and parameter branches therefore improve transient tracking without allowing the estimated quantities to diverge. The results show that the proposed observer maintains bounded, accurate estimation under the investigated electrical, mechanical, and parameter variations.]]>
Copyrights © 2026
