Power Quality Disturbances Segmentation: An Approach Based on Gramian Angular Field

Abstract

The advanced metering infrastructure of smart grids has impacted the increase in the volume of data acquired. Therefore, the elimination of non-useful data is important to reduce the storage and processing load/cost in the utility’s control and operations center. Thus, in the management of data related to power quality, it is important to perform the segmentation of disturbances present in oscillographic records, being this a non-trivial task. In this sense, the present article proposes an adaptive approach for disturbance segmentation based on image analysis. Computational experiments were performed involving different disturbances, acquired at 512 samples/cycle, convolved with random noise. In order to demonstrate the robustness of the proposed approach, lower sampling rates were evaluated and compared. Thus, the main contribution of this article is the proposition of a novel approach for automatic and adaptive segmentation of different power quality disturbances, including single/mixed and fast events, under low sampling rates and the presence of random noise. From the results, a high precision was reached for disturbances with noise level up to 35 dB, presenting an absolute mean error range between 0.045 and 8.812 ms for signals at 512 samples/cycle and between 0.195 and 11.431 ms for signals at 64 samples/cycle.