Autonomous circuit design of a resonant converter (LLC) for on-board chargers using genetic algorithms

In the field of conductive and inductive charging systems, contrary requirements such as high power density, small installation space, low power losses and costs need to be optimized for multiple operation points taking into account customer defined power transfer profiles. In this paper the engineering experience for safe and practical operation modes (complete zero voltage switching, inductive operation region, etc.) is transferred into the mathematical domain of multiple constraints and objectives. Based on that, a new cascading penalty strategy is combined with a genetic algorithm (GA) to process the circuit design of a resonant converter (LLC) for on-board chargers autonomously. Within this self-learning design process the power losses on the primary and secondary side of the resonant converter are minimized for multiple operation points. The optimization setup reliably reaches feasible solution candidates for this highly non-linear problem and even enables the prediction of technological limits. Due to the general purpose of the method, this autonomous design process can be adapted to other circuit topologies and applications.