Enhancing self-sensing estimation accuracy via negative sequence current image registration, with evaluation on a low saliency ratio machine

Abstract

Image tracking self-sensing, which utilizes image registration to localize an arc of the negative sequence current response to a high frequency rotating voltage injection within a negative sequence template, can enhance estimation accuracy, increase tolerance to noise, and improve dynamic performance. Image registration over a full injection cycle replaces the heterodyning demodulation used in traditional point-tracking methods. Demodulation is known to produce significant harmonic content that must be filtered. Image registration mitigates dynamic degradation since there is no longer a need for low pass filtering of the harmonic content, tightly tuned filtering of the negative sequence, or decoupling of multiple saliencies from the negative sequence. This paper proposes how, by carefully considering the machine properties at an operating point, details of the current response can be used to generate a complex template image. This paper shows how image registration of the sampled image over a full injection cycle with the complex template allows for a more accurate estimate of position when compared to point-tracking methods, while removing various filters to improve dynamic performance. These methods are evaluated experimentally on a low saliency ratio SPM and compared to classical rotating HFI methods.