High Disturbance-Resistant Speed Control for Permanent Magnet Synchronous Motors: A BPNN Self-Tuning Improved Sliding Mode Strategy Without Load Observer

Sliding mode control (SMC) provides robustness and disturbance rejection in permanent magnet synchronous motor (PMSM) control but faces the challenge of speed degradation during sudden load disturbance changes without a load observer. This paper proposes a backpropagation neural network-adjusted improved SMC (BPNN-ISMC). A simplified PMSM model is established by ignoring the disturbance term. An improved arrival law is developed by optimizing the constant-speed approach term of the traditional exponential arrival law and embedding an adaptive term. A BPNN is designed with performance metrics including speed error, its derivative, and maximum error to improve training efficiency. Speed/position estimation combines a sliding mode observer with an extended Kalman filter to suppress jitter. Simulation results demonstrate the significant advantages of the BPNN-ISMC method: in set-point control, overshoot suppression is evident, and the relative error during the stable phase after sudden load disturbance increases is reduced by 93.62% compared to the ISMC method and by 99.80% compared to the SMC method. Compared to the ADRC method, although the steady-state errors are the same, the BPNN-ISMC method exhibits smaller speed fluctuations during sudden changes. In servo control, the root mean square error of speed tracking is reduced by 18.83% compared to the ISMC method, by 89.70% compared to the SMC method, and by 37.14% compared to the ADRC method. This confirms the dynamic performance improvement achieved through adaptive adjustment of neural network parameters.