Parameter-Free Cascade Model Predictive Control of PMSM Drives Considering Stagnation Propagation

Abstract

Model predictive control (MPC) is a reliable strategy for motor drives, and cascade MPC (CMPC) integrating speed and current loops offers superior dynamic performance. However, CMPC is highly sensitive to parameter mismatches and load disturbances. To eliminate the dependence on motor parameters and avoid observers, this paper presents a parameter-free CMPC scheme for permanent magnet synchronous motors based on signal differences. An ultra-local model is first established, and its lumped parameters are directly computed from speed, voltage, and current errors. A convergence analysis of the algebraic parameter estimator reveals that vanishing voltage and current differences cause the estimates to diverge, which would lead to current spikes and speed oscillations. The cross-loop stagnation propagation mechanism inherent in cascade parameter-free identification is then revealed. When the speed-loop output-difference stagnates, the $q$ -axis current reference increment freezes, forcing the current loop to repeatedly select the same voltage vector and inducing input-difference stagnation, which forms a positive-feedback loop that can lock the entire system. To actively break this propagation chain, a cascade-aware, current-error-driven voltage correction strategy is developed, which directly employs instantaneous current tracking errors to guide the desired voltage direction and constructs a reduced candidate set that explicitly excludes the stagnated vector. Comparative experiments validate that the proposed method effectively suppresses current harmonics and maintains robust performance without requiring any motor parameters or additional observers.