[Objective] In order to reduce the dependence of maximum torque per ampere (MTPA) control of permanent magnet synchronous motor (PMSM) on motor parameters, the incremental model predictive current control (MPCC) MTPA for PMSM based on virtual signal injection is established in this paper. And model reference adaptive system (MRAS) is used to identify parameters to improve parameters robustness. [Methods] The traditional formula method was adopted to achieve MTPA, which required three parameters—stator dq-axis inductance and rotor flux. However, mismatches in these three parameters all had a large impact on the MTPA calculation results. An incremental MPCC for PMSM was established, which required three parameters—stator dq-axis inductance and stator resistance. Among them, the mismatch in stator resistance had a relatively small impact. The virtual signal injection method was adopted to achieve MTPA, which required two parameters—stator d-axis inductance and stator resistance. Among them, the mismatch in stator d-axis inductance had a large impact, while the mismatch in stator resistance had a small impact. In summary, an MTPA based on virtual signal injection combined with incremental MPCC for PMSM was established, and the MRAS was adopted to identify the stator dq-axis inductances. The identified parameters were used in both the incremental model predictive current control and the virtual signal injection MTPA to improve parameter robustness. A simulation model was built based on Matlab/Simulink, and the above methods were simulated and analyzed. [Results] The incremental MPCC for PMSM based on the virtual signal injection MTPA adopted MRAS to identify dq-axis stator inductance. Under both matched and mismatched parameter conditions, the error of the current vector angle βMTPA calculated by the virtual signal injection MTPA was small, and the impact of stator resistance mismatch was small. [Conclusion] The virtual signal injection MTPA only requires d-axis stator inductance and stator resistance, has low parameter dependency, and the impact of stator resistance mismatch is small. The incremental MPCC MTPA for PMSM based on virtual signal injection requires only the identification of the dq-axis inductances. The virtual signal injection MTPA using MRAS for parameter identification has a small error, which improves parameter robustness.