Advanced generator design using Pareto-optimization

Abstract

In upcoming and future more-electric airplanes more electrical loads connected to the power distribution network will behave nonlinearly and will be switching with advanced demands on the network to keep the power quality. While a lot of effort was already spent on techniques to design the network structure and filters, also the generators have to face the new demands and constraints. Trade offs between performance, weight, and latterly power quality have to be made and design choices have to be backed up by qualitative and quantitative measures. Therefore a convenient design tool chain was elaborated for an externally excited synchronous generator. Pre-existing analytic design functions containing electrical, mechanical and thermal analytical models were expanded by a new criterion for power quality. This criterion defined by industrial standards was included via virtual tests by time-domain simulation. Frequency domain criteria were gained in simulation by new signal property monitor blocks. Standards' conformance is visualized conveniently. A state of the art genetic algorithm similar to NSGA-II was applied to generate the Pareto-front of contrary design trade offs. Several convenient visualizations of the Pareto results are presented, which clearly link the parameters, tuners and output results. To ease the transition from theoretical design to the 3d mechanical layout, a CATIA model of the optimized generator candidate is generated directly from Matlab.