Modeling semiconductor losses for population based electric machinery design

This work sets forth a semiconductor loss modeling approach suitable for use in population based design of electric machinery. Population based machine design typically employs a genetic or particle swarm algorithm which maximizes a fitness function to optimize the design. The fitness function often includes multiple objectives such as machine mass and machine loss. This work incorporates the consideration of semiconductor loss into the machine design using an embedded simulation in which a time-domain simulation is conducted as part of a fitness function evaluation. The embedded simulation predictions are compared to commercial software solutions over a wide variety of operating conditions and consistent results are obtained. A design study using the embedded simulation within a population based optimization of a 3-phase current controlled permanent magnet synchronous machine (PMSM) drive is demonstrated.