Large scale optimization of electronically controlled synchronous reluctance machines using CE-FEA and differential evolution

Abstract

An automated design procedure for current regulated synchronous reluctance machines supplied from power electronic converters is proposed. An ultra-fast computationally efficient electromagnetic FEA, which uses only a minimum number of magnetostatic solutions in order to comprehensively evaluate performance is employed. The optimization algorithm is based on differential evolution and uses as independent variables the torque angle and ratios for a generic rotor topology with four flux barriers. Two problems, one with two and another one with three objectives, are studied and results compared. Global performance indices and objectives incorporate the effect of average torque output, losses, torque ripple, and power factor at fixed cost. It is shown that through optimal studies with more than 5,000 candidate designs, high output power, high efficiency, and low torque ripple can be achieved, while the relatively low power factor remains an inherent limitation of synchronous reluctance technology. Simulations are validated versus tests from a 10hp 1,800rpm prototype.