Inverse Methodology for the Parameter Identification of a Lumped Parameter Thermal Network for an Induction Machine

Abstract

Thermal modelling of induction machines is becoming increasingly important with the demand for machines with ever increasing efficiency as well as compact design. The lumped parameter thermal model is a flexible and computationally cheap method for the temperature analysis inside an induction machine. However, there are a number of thermal parameter values which are difficult to determine analytically. This includes the air gap convection coefficient, the equivalent radial conductivity of the stator winding and the width of the equivalent air gap between the frame and the stator lamination. In this work, the identification of the thermal model values follows the inverse methodology: assign values to the thermal parameters by aligning temperature measurements at a specific location in the motor with the lumped-parameter model response. Simulation results show that the accuracy of the proposed thermal parameter identification scheme depends on the location of the thermal measurement and more specifically on the sensitivity of the temperature profile with respect to the unknown parameters.