Advanced islanding Detection for Power Grids using Transformer Neural Networks

Abstract

The increasing use of solar PV systems is causing distribution companies to worry about accidental islanding more and more. Due to unintended energization and low power quality, these accidental islands endanger workers and sensitive equipment. There are non-detection zones in conventional detection technologies. Due to the need for human feature extraction from data, new machine learning-based approaches to accidental island detection are difficult to predict. Our proposed end-to-end solution automates the feature extraction portion of the modeling process using Time Series Transformers, which simplifies the process overall. Time Series Transformer provides a simple but trustworthy detection model for islanding detection, as shown by our findings, which surpass those of existing machine learning approaches. Jhoanna Rhodette Pedrasa is a professor at the University of the Philippines' Diliman campus in Quezon City, Philippines, who specializes in electrical and electronics engineering. [email protected] a straightforward and precise model for islanding detection based on the features extracted from the data. The structure of this article is as follows. Part II provides a summary of the most up-to-date approaches to islanding identification using machine learning and deep learning. In Section III, the dataset, TST architecture, additional models, and assessment measures are detailed. Discussion and outcomes of the experiments are presented in Section IV. The article is concluded in section V.