Robust Rotor Temperature Estimation of Permanent Magnet Motors for Electric Vehicles

As the mainstream powertrain of electric vehicles, permanent magnet motors are facing the challenge of durability and thermal failure. Therefore, the real-time rotor temperature monitoring plays a critical role, however, it is hard to online measure with sensors. To this end, a rotor temperature estimation method based on the lumped-parameter thermal networks and dual H infinity filters is proposed. Firstly, the lumped-parameter thermal network of three nodes, such as the stator, rotor and bearing, is numerically formulated to determine the power loss. Accordingly, the discretized state-space expressions are specified for the time-step iterative solution. Then, to address the uncertainty of model parameters, the dual H infinity filters are used in the rotor temperature estimation process. Finally, the simulation and experimental tests are performed to validate the effectiveness and the real-time executability of the proposed method. The test results show that the proposed method can well track the actual temperature tendency with estimation errors of less than 7.5 °C. Compared with the existing methods, the worst-case estimation accuracy has been improved by at least 25%; besides, the proposed method presents good robustness against the parameter uncertainty; meanwhile, the higher estimation convergence is made in the face of huge model deviations.