Partial Discharge Separation by Using Pulses Cross-Correlation

Abstract

To date, continuous monitoring of power grid elements susceptible to aging processes is essential to increase the reliability of these systems and avoid failures. In the field of energy transmission, an index of the state of a cable insulation system is partial discharge activity, which is influenced by the presence of structural defects (e.g. voids within the material). In High-Voltage-Direct-Current applications, given the dependence of dielectric conductivity on temperature, this phenomenon is also influenced by the load. Thus, partial discharge activity is strongly related to the operating condition of the line and not only by the aging of materials. Given the importance of partial discharges monitoring, it is necessary to develop techniques for analyzing them that allow the identification of different types of discharges. The lack of a reference standard prompts researchers to propose different approaches to achieve these goals. This paper describes a comparison of two separation algorithms both based on cross-correlation. Data used for comparison have been obtained through a partial discharges measurement, under DC voltage stress, on XLPE model cable. The first algorithm evaluates the similarity among the pulses and separates them into clusters following the acquisition order. The second one, on the other hand, uses the correlation matrix as input data, which must be calculated before running the algorithm. The results show that both algorithms succeed in identifying the same phenomena. The former with less accuracy but employing less computation time than the latter. The latter, on the other hand, provides higher accuracy but requires longer computation time. The choice in using either algorithm can thus be traced back to the desired accuracy/time ratio.