Encoderless self-commissioning and identification of synchronous reluctance machines at standstill

This paper presents an identification method and modelling technique which is able to characterise the complete nonlinear cross-coupling electromagnetic flux-linkage model of a synchronous reluctance machine as a function of the direct- and quadrature axes currents within a few seconds. The presented approach is suitable for identification of the self-saturation flux-curves as well as the cross-coupling flux-maps of synchronous machines without additional testing hardware. The proposed method is performed at standstill and is suitable for encoderless and self-commissioning applications. During the identification, the reference phase voltages and measurements of the phase currents are used to estimate the flux-linkages of the machine. Afterwards, the obtained data is utilised in a neural network training routine. The trained simple neural-network represents the complete flux-maps of the machine accurately, without discontinuities and with a small amount of model parameters which has been confirmed due to comparison of the results with the measurements of a constant speed method.