Advanced performance prediction tools for the analysis of rotating electrical machines

Abstract

This thesis seeks to advance the design tools for electrical generators. It aims to undertake an electrical generator’s design, using the Finite Element method within a defined time frame. The thesis looks at the history of generator design systems and outlines the parameters a designer must predict. These parameters are then duly calculated using various finite element methods. The thesis introduces a Pseudo Rotating Superposition system, which allows large quantities of data to be found from single static finite element simulations. Initially the system is used to predict machine saturation curves, and it is later expanded to predict the transient performance of generators. The full load performance of generators is found using a novel multivariable clustered optimisation routine. An extension using a rotating finite element solver, with pseudo rotating superposition, is then demonstrated. This creates a method which allows voltage harmonics to be quickly, accurately and validly predicted. Finally a study of iron loss is undertaken and using the above method it is shown that iron loss can be validly calculated using the quicker Radial/Tangential reference frame, rather than a slower Major/Minor frame. A collection of 48 manufactured machines are used throughout as a test group for the created methods. Results from design calculations are compared to both factory test results and to the predictions from an existing customised in house design software tool. The methods within this thesis are shown to be over 35% more accurate in the majority of cases. The whole suite of methods created can automatically calculate results for any given machine in less than 1 hour. The computer macros described in this thesis and the comparison with existing design methods and test were all made by the Author.