Among　the　existing　sensorless　control　algorithms　for　permanent　magnet　synchronous　motors　(PMSMs),　the　model　reference　adaptive　system　(MRAS)　is　widely　utilized　due　to　its　merits　of　good　robustness,　high　stability,　and　rapid　response.　Nevertheless,　the　performance　of　the　sensorless　control　is　largely　determined　by　the　identified　inductance　parameters.　The　virtual-rotary-axis　high-frequency　injection　(VHFSI)　method　is　an　online　identification　strategy　for　PMSM　inductance　parameters　which　is　not　affected　by　the　observed　rotor　position　errors　among　inductance　identification　algorithms.　However,　the　existing　PMSM　inductance　online　identification　algorithm　based　on　VHFSI　has　ignored　the　influence　of　the　error　components,　resulting　in　low　accuracy　and　the　weak　dynamic　performance　of　the　inductance　identification　algorithm.　In　response　to　the　problems　existing　in　the　original　method,　this　paper　improves　VHFSI　by　designing　a　feedforward　decoupling　algorithm　to　compensate　for　the　error　components.　Theoretical　verification　and　simulation　results　indicate　that　the　improved　identification　algorithm　enhances　the　accuracy　of　inductance　parameter　identification,　significantly　improves　the　dynamic　tracking　performance　of　inductance　identification,　and　combines　it　with　the　MRAS　sensorless　control　method　to　constitute　a　sensorless　control　system　for　the　motor,　thereby　enhancing　control　performance　and　system　stability.