Data Analysis and Preprocessing Method of Medium Voltage Distribution Network Feeders

Abstract

The investment in the energy sector aims to ensure a continuous, reliable, and quality supply of electrical energy imposed by the electricity regulatory agency with maximum economic-financial balance. This paper discusses the challenges of processing data from medium voltage distribution feeders to use on the distribution network planning. The analysis of missing data and outliers is made on the three-phase voltage, current, and power factor of 459 time series of real feeders. Furthermore, it is proposed a method of preprocessing, and missing data imputation using the unbalanced characteristic between phases, interpolation, and the normalized scaled standard weekday curve.  The results show that most missing data are three-phase, however, with a significant amount of single and dual-phase loss that can be filled by the proportion between phases. Hence, the challenge is to fill multiple weeks of missing three-phase data, and for that, it is proposed the use of the standard curve for each day of the week. The method proposed is a promising alternative for data imputation in medium-voltage feeders. The technique is tested using real feeder data degraded by its missing data probability function, and compared with the Naïve approach.