Accurate flux linkage information is crucial for achieving high sensorless control performance in interior permanent magnet synchronous motors. The integration of back electromotive force (EMF) can yield flux linkage; nevertheless

direct integration suffers from issues such as DC bias and incorrect initial integral value. Additionally

the approximation error associated with low-pass filters is significant

whereas the second-order generalized integrator- frequency-locked loop (SOGI-FLL) exhibits superior integration performance. However

DC bias and harmonics in EMF and discretization errors during digitization can lead to biased estimates of synchronous angular frequency in FLLs. This deviation problem becomes more pronounced during speed changes

adversely affecting the accuracy of flux and rotor position estimation. This paper proposes a flux observation algorithm utilizing a third-order generalized integrator (TOGI) combined with a moving average filter (MAF). TOGI effectively mitigates DC bias and high-order harmonics in EMF

and MAF further eliminates the odd harmonics and frequency coupling components in the output flux

thereby improving the estimation accuracy of the flux and rotor position. Simultaneously

the compensation algorithm for phase angle deviation is presented. Theoretical analysis alongside experimental validation demonstrates that the proposed flux observation scheme accurately estimates flux and rotor position over the range of 3% to 100% of the rated speed.