Hybrid Thermal Network Identification for Sensorless Temperature Monitoring of a PMSM Considering Speed-Dependent Cooling Effects

This study aims to monitor crucial temperatures, e.g., the winding and the PM temperature, inside a interior rotor PMSMwith a integrated cooling fan. Therefore, a thermal model in the form of a lumped parameter thermal network (LPTN) is identified based on a hybrid modeling approach consisting of two steps. First, the LPTN’s parameters are initialized with empiric and simplified analytic models. Special attention is given to the speed-dependent thermal transitions between motor components and their environment and between different components caused by convection, e.g., the windings’ cooling or the bearings’ heat transition. The modeling of these transitions represents a unique challenge because of a rotor-mounted cooling fan. Secondly, a genetic optimization algorithm compensates for parameter uncertainties resulting from model simplifications and necessary approximations in the used analytical models. The large amount of measurement data needed for this optimization process is generated using a fully automated test bench presented and discussed in this paper. Finally, the resulting thermal model is verified using independent measurement data.