Model Order Reduction Suitable for Online Linear Parameter-varying Thermal Models of Electric Motors

For the control of high performance electrical machines, it is desirable to estimate the hot-spot temperatures in order to exploit the machine up to its thermal limits. An accurate temperature estimation can be achieved by means of high-order lumped parameter thermal networks. In variable-speed drives the thermal networks are parameter-varying due to the speed dependency of the convective heat transfer. These high-order parameter-varying models require a high calculation effort. For online temperature estimation it is desirable to reduce the order of the model. This paper presents an easy-to-implement model order reduction method for linear parameter-varying thermal model, which makes the model suitable for real-time online temperature estimation. The concept is exemplarily shown on an air-cooled induction motor for automotive applications. A high-order thermal model is built up and reduced using the proposed algorithm. The accuracy of models is validated by simulations and measurements.