SMO-Residual-Driven Event-Triggered Sensorless FCS-MPC for PMSM Drives

Periodic sensorless finite-control-set model predictive control (FCS-MPC) evaluates all eight voltage vectors every $10\mathsf{\mu }\mathrm{s}$, even when the applied vector is still adequate. This study develops a controller that reuses the applied vector when one held-vector prediction satisfies the descent and current constraints. A double-power saturated sliding-mode observer supplies position and speed, and its sliding residual tightens or relaxes the descent threshold. Under bounded lumped disturbances, the sliding error reaches a prescribed boundary layer in bounded time, while the compensated-current error remains locally uniformly ultimately bounded. At the largest simulated initial error, the double-power law shortens boundary-layer entrance from 2.11 to 1.08 ms and post-load-step speed-estimation recovery from 55.23 to 24.35 ms. Across the composite profile, 39.42% of periods require full enumeration, reducing candidate-state checks by 60.6%. With a similar update budget, position RMSE falls from $1.{035}^{\circ }$ to $0.{949}^{\circ }$. On the hardware bench, full-enumeration checks fall by 21.6% and equivalent switching frequency by 14.9%; speed-tracking and electrical-angle RMSE increase by $1.22\mathrm{r}/\min$ and $0.{400}^{\circ }$, respectively. These measurements show the computational saving and its accuracy cost.