Novel High-Order Lumped-Parameter Thermal Models for Predicting Spatial Temperature Distribution in Electrical Machines

High-order lumped-parameter thermal models (LPTMs) can be used to predict spatial temperature distribution and hot spot temperatures for critical components and account for spatial temperature variations in medium-to-large-sized electrical machines. However, all existing high-order LPTMs require overdiscretization levels to reduce estimation errors due to the assumption of concentrated loss generation and/or the inappropriate definition of compensation thermal resistances. Such overdiscretization, however, leads to excessive burdens in thermal modeling and computation. In this article, a novel high-order thermal modeling method is proposed for both cuboidal and hollow cylindrical components, addressing 2-D and 3-D heat transfer problems in electrical machines. The proposed method introduces an individual compensation thermal resistance for each discrete element, which is solely dependent on geometric and thermal parameters and obtained based on the multidimensional heat transfer theorem. As a result, the proposed models can fully consider the distributed loss generation within the component and the thermal continuities between adjacent discrete elements. The proposed models can maintain high estimation accuracy under different discretization levels without increasing computation time. They are further applied to winding regions in a wound lamination core and a permanent magnet synchronous machine (PMSM) for experimental validation.