Real-Time Rotor Temperature Estimation Method for Interior Permanent Magnet Synchronous Machines : A method using a Lumped Parameter Thermal Network in combination with an estimation algorithm: implementation and comparison of a Dual Kalman Filter and a D

Abstract: Accurate knowledge of the temperature of a Permanent Magnet Synchronous Machine is of high importance, not only to ensure safe operation and improve control of the operating point, but also to avoid irreversible demagnetization of rare-earth permanent magnets. However, direct measurement using thermal sensors is not suitable for high volume and low-cost applications, due to costly sensor installation and maintenance, and therefore, the magnet temperature needs instead to be estimated. The proposed real-time permanent magnet temperature estimation method is based on measurements of the stator temperature, in combination with thermal parameters of a Lumped Parameter Thermal Network, found through an offline parameter fitting using experimental data, both fed into an estimation algorithm. The aim of the algorithm is to estimate and correct the thermal parameters, which in turn enables the method to provide accurate rotor temperature estimations, even for a machine different than the one used to tune the thermal network. Two different estimation algorithms are implemented and compared; a Dual Kalman Filter and a Dual Observer. Simulation results show that the proposed method yields rotor temperature estimations within ±5 ◦C of the correct temperature, both when using the Dual Kalman Filter as well as the Dual Observer, with exceptions discussed in the thesis. The Dual Observer proved to perform better at low to medium speeds, while the Dual Kalman Filter proved better performance at medium to high speeds. The proposed temperature estimation method shows therefore to be of great potential, where the next step is to further improve the thermal network, to then test the proposed method in an experimental setting.