Comparative analysis of segmentation methods, including adaptive and HOS based algorithms

Power perturbations affect the normal operation of the grid, both from the network side and the consumer side. In order to eradicate or attenuate the harmful effects, a first step is needed: voltage or current segmentation into healthy and perturbed intervals. In this work we extend the available plethora of documented methods by implementing an adaptive version of Yule-Walker equation for autoregressive (AR) modeling based on higher-order statistics. This proposal is compared with a wide variety of existing segmentation methods: notch filters, second order AR modeling and cumulant based methods. Simulations confirm that the proposed method being much lighter computationally provides us with useful information for segmentation, besides, for some cases, this information is clearer and easier to interpret, facilitating the segmentation process.