[Objective] To address the issue that the control performance of model predictive current control for permanent magnet synchronous motor (PMSM) deteriorates under the working condition of motor parameter mismatch, this paper proposes a model-free control strategy for PMSM that does not require the use of motor parameters in the design process. [Method] Based on the ultra-local model of PMSM, a nonlinear extended state observer (NESO) was designed, and its stability was analyzed using Lyapunov theory. A model-free predictive current control system based on the NESO was constructed. Meanwhile, the influence of the current feedback gain in the ultra-local model on the model-free control performance was analyzed, and online identification of the current feedback gain was performed based on the sampled current iteration. The algorithm was simulated and studied using Matlab/Simulink. Firstly, under the working condition where the given current feedback gain coefficient of the system was the nominal value, the model-free control system based on the NESO proposed in this paper was verified. Then, under the working condition where the given current feedback gain parameter of the system was mismatched, a comparative simulation was conducted with the traditional methods. Finally, the online identification method of the current feedback gain parameter based on the sampled current iteration was verified. [Result] Simulation results show that when the current feedback gain parameters are mismatched in the ultra-local model, the output of the NESO is more stable than that of the traditional linear extended state observer. The current loop tracking effect of the PMSM control system based on the NESO is better, and the harmonic content in the phase current is also reduced. The online parameter identification method based on sampling current iteration can accurately identify the actual current feedback gain parameters, rapidly converge and maintain stability when the mismatch current feedback gain parameters are continuously applied in simulation. [Conclusion] Compared with the traditional control methods, the control strategy of PMSM based on the NESO reduces the sensitivity of the system’s design parameters and offers higher control performance under the working condition of motor parameter mismatch.