The Impact of PMU Data Precision and Accuracy on Event Classification in Distribution Systems

Abstract

The use of synchrophasor data is recurrent in electrical power systems and has an increasing demand in distribution networks. The quality of this data can directly influence later stages related to decision-making. In this sense, this work presents the impact analysis of synchrophasor data quality on the event classification in distribution networks. We propose a controlled environment, simulating 37 200 events and including errors in the data with different levels, distributions, and biases, enabling evaluations according to the data’s precision and accuracy. Additionally, we present 2 160 classification results involving six machine learning methods, three different groups with 4, 8, and 62 events (classes) per group, and several characteristics of phasor errors. In addition to comparative analyzes, the research scope makes it possible to map the influence of multiple sources of errors, disclosing the data quality relevance and how it affects the event recognition in the distribution systems – a subject underexplored in the recent literature.