Magnetic model identification for synchronous reluctance motors including transients

The paper deals with the flux linkages estimation of a synchronous motor during speed or load transients. The proposed method is based on the radial basis function neural network, which returns continuous functions between currents and magnetic flux linkages, considering all the nonlinearities. The training of the neural network weights is realised by an innovative training algorithm able to cope with both steady state and variable working conditions, conversely to most of the identification techniques. The method was validated throughout several tests performed on a synchronous reluctance motor electric drive.