An improved approach for transient thermal modeling using lumped parameter networks

For thermal modelling of electrical machines often lumped parameter thermal models are used. This simple lumped parameter approach leads to overestimation of temperatures when inner losses are present. A new approach is derived using a node shifting technique, compensating the temperature overprediction and delivering the exact mean temperature in one dimensional cases. The application to a two dimensional lumped parameter model for electrical machines stator cross section is presented. The results for the one and two dimensional case are evaluated together with the basic lumped parameter approach and the common T-equivalent network compensation approach. It is shown that the desired compensation effect is achieved by the new model and that some disadvantages of the common compensation approach are avoided.