A Non-invasive Dual-EKF-based Rotor Temperature Estimation technique for Permanent Magnet Machines

Abstract

Demagnetization of permanent magnet machines is a crucial failure mode, which may lead to catastrophic failures. The thermal stress of the permanent magnet synchronous motor (PMSM) is the main factor leading to such malfunction. It is critical to monitor the rotor temperature by tracking the permanent flux. Based on the characteristic that the magnitude of the magnet flux linkage decreases as the temperature increases with a linear relationship, a novel non-invasive method is proposed to estimate the temperature of permanent magnet (PM) in PMSM. To solve the rank-deficient problem of the state estimation, the proposed method utilizes two Extended Kalman filters to estimate the magnet flux linkage and stator inductances information respectively. Meanwhile, the estimation results of flux and inductances message are used as known quantities in the other estimation algorithm, which forms an iterative algorithm process. When the updating of flux and inductances are executed simultaneously, the influence of magnetic saturation is compensated automatically. Simulation and experimental tests are implemented to verify the accuracy and feasibility of the proposed method. With errors less than 4 °C, within an acceptable range, the capability of providing non-invasive monitoring on the rotor temperature with high precision is proven.