Online Non-Invasive Estimation of Permanent Magnets Temperature for PM Synchronous Machines

This thesis deals with the design and testing, both in simulation and real testbench environments, of two non-invasive methods for the online estimation of the Permanent Magnets (PMs) temperature in Synchronous machines. The Interior Permanent Magnet (IPM) motors present the best cost versus performance tradeoff for traction applications; this translates into very high efficiency and torque density, associated to Constant Power Speed Range capability. This thesis is in collaboration with a company, which provided the test case and related Matlab - Simulink model and one of the two PMs temperature estimation methods under investigation; due to confidentiality issues, the description and implementation of this method are not presented, while simulation results will be. At first, the effect of the PM temperature on the machine performance is studied, along with an overview of the various PMs temperature estimation methods that can be found in literature, focusing on the complexity, feasibility and invasiveness of each of them. The Flux Observer based methods emerge from the literature review as the most promising candidates, also in consideration that the considered torque control algorithm is the stator-flux-oriented scheme called Direct Flux Vector Control (DFVC), which is already flux-observer dependent. A new second method of temperature estimation, based on the Vector Projection technique mutated from sensorless control, has been developed, simulated and tested during the thesis. The related results are openly presented. This also includes the possibility of stator resistance adaptation. The core of the thesis is the implementation and comparison of the two online non-invasive methods for the estimation of the PMs temperature, the second one with or without stator phase resistance adaptation "add-on". After comparison of the steady-state simulation results obtained with the two techniques for a large number of reference working conditions of the EM, a hybrid approach is proposed, where the two methods are activated and deactivated depending on the operating condition, so to maximize the estimation accuracy. Finally, the proposed methods were experimentally tested on two testbenches, the first one referring to an off-the-shelf traction motor by BRUSA GmbH, and the second one mounting the IPM motor of the partner company.