Enhanced Multi-objective Model Predictive Control for Permanent Magnet Assisted Synchronous Reluctance Machines: A Robust Implementation with Sparse Parameter Identification

Model predictive control of permanent magnet machines is still attracting attention due to the need of simultaneously addressing computational burden, robustness to parameter variance and mismatch, variable switching frequency, optimal switching sequence, current harmonic distortion, and overmodulation. This paper proposes a novel unified parameter estimator and an enhanced predictive controller to deal with such problems. A very fast multi-objective, finite control set - long-horizon model predictive controller is constructed on a field programmable gate array that allows very accurate current and switching frequency control with optimal switching over 3 μs period through a 10-step horizon, improving robustness, dynamic response, reference tracking, current distortion, efficiency and maximum switching frequency capability. The rise times of the d&q-axes currents are reduced by 16% and 32.8% respectively, while the base speed is extended by 9.27% due to improved overmodulation. The proposed non-invasive, on-line, data-based sparse estimator is capable of estimating all motor parameters with high accuracy under two to four electrical periods. The novel estimator has errors of 1.22%, 2.84%, 4.15% and 3.09% respectively for stator resistance, d&q-axis inductances and magnet flux, improving current trajectory generation and prediction accuracy over the whole operation region.