To enhance the robustness and minimize the discrepancies between the estimated and actual values of speed and position in conventional Model Reference Adaptive System (MRAS) observers for Permanent Magnet Synchronous Motor (PMSM) control, this study introduces an advanced MRAS algorithm. To address the inherent limitations of the Proportional-Integral (PI) structure in the speed adaptive rate of conventional Model Reference Adaptive Systems (MRAS), which often lead to reduced accuracy in speed identification and compromised robustness, this study proposes an innovative approach. Specifically, the PI structure of the speed adaptive law is replaced with a sliding mode variable structure. Recognizing that the sliding mode variable structure may induce chattering, this paper introduces a super-twisting sliding mode algorithm derived from higher-order sliding mode algorithms to replace the PI structure in the MRAS speed adaptive law. Furthermore, an adaptive coefficient is incorporated into the integral term to enhance the precision of motor speed identification and mitigate chattering effects. The efficacy of this approach is demonstrated through experimental simulations conducted in MATLAB/SIMULINK, which reveal a significant reduction in estimation errors during the observer’s identification of motor parameters, thereby improving the response speed of the identification process.