Clasificación de zonas en fallas en sistemas de distribución de energía utilizando algoritmos de agrupamiento

Abstract

Introduction: The main cause of disruptions in the energy service is generated in the distribution grids. The faults need repairs before restoring the service, which requires their rapid location, since this process influences the duration and frequency of the disruptions. Many companies assess power service quality solely by measuring the equivalent interruption frequency and duration (DES and FES), as they lack the resources to invest time and money in strategies to improve system reliability.

Objective: Determine the location of the fault zone in medium voltage power distribution systems.

Method: The clustering methods of k-means and the Gaussian Mixture Model (GMM) based on the expectation-maximization algorithm were implemented, to create groups based on their characteristics, thus defining the probability of belonging to each one.

Results: A methodology for the detection of failures efficiently was obtained, which serves as support for planning processes and execution of action plans, facilitating the taking of corrective measures related to the continuity of the service and decreasing the system restoration time.

Conclusions: The application of the k-means and GMM algorithms allows identifying possible fault zones according to the test data. Although it does not show a single fault zone, since it makes estimates according to the data, it is a tool to make decisions based on the estimates obtained.