Online Stator Resistance Estimation of Low-Power PMSMs Using Data Driven Methods

Accurate estimation of stator resistance is critical for the high-performance control and thermal monitoring of Permanent Magnet Synchronous Motors (PMSMs), particularly in sensorless drives where parameter mismatch leads to significant position errors.This work targets low-power PMSM drives, where the relatively large stator resistance provides favorable signal-to-noise conditions for the proposed data-driven estimation framework. Conventional online estimators often suffer from convergence lag, stability trade-offs, or high computational burdens. To address these limitations, this study presents a novel two-stage data-driven estimation framework. First, experimental data is collected offline using inverter and microcontroller at various operating points to construct a parametric model which indirectly covers the thermal dynamics of the machine. To identify this model efficiently, a Nested Non-Linear Least Squares (Nested NLS) algorithm is introduced. By structurally decoupling the linear and nonlinear parameters, this algorithm reduces the training computation time by a factor of 5.83 compared to standard optimization, enabling rapid self-commissioning. In the second stage, the identified model is utilized for online tracking via a computationally efficient recursive estimator that predicts resistance variations in real-time without signal injection. Experimental validation on a Texas Instruments TMS320F28069M MCU controlled low-power PMSM drive under aggressive load and speed cycles demonstrates that the proposed method eliminates estimation lag and achieves superior tracking accuracy, with a Root Mean Square Error (RMSE) as low as 0.0021 Ω and a maximum percentage error maintained below 0.5%.