abstract
- Temperature monitoring of rotating electric machines is necessary to ensure proper and safe operation. In many difficult-to-access locations, such as the rotor, direct measurements are difficult to implement due to complexity and cost limitations. Estimation methods are a valid alternative and, among the possibilities, low-order lumped-parameter thermal networks offer simplicity and computational affordability for their implementation in real time. However, a critical factor influencing the performance of these models is the accurate estimation of the initial state. This paper develops and validates a two-node thermal network that uses winding temperature feedback to estimate the rotor temperature in an outer-rotor permanent-magnet synchronous machine. The proposed method is validated using a comprehensive experimental dataset collection, including a standardized driving cycle across diverse coolant temperatures and flow rates. It is demonstrated that the proposed solution achieves a root-mean-squared temperature error of 1.09°C in average across the validation duty cycles. © 2013 IEEE.