Partial-discharge detection in power systems using transformers, an emerging machine learning technique

• Main Author: Tan, Ryan Zhiyang
• Other Authors: Xu Yan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/157731
• Institution: Nanyang Technological University

Artificial Intelligence is a fast-growing industry and has found a foothold in almost every industry. Some known works of Artificial Intelligence are namely, Natural Language Processing (NLP) for language translation and classification algorithms like Support Vector Machines (SVM). This Final Year Project Report examines how a model that is commonly used for NLP will perform when performing binary classification of partial discharge data. The model used will be Transformers which has been introduced in 2017 and published with the paper “Attention is All You Need” [4]. Partial Discharge is a key indicator on the “health” of electrical machinery and equipment. Early detection of Partial Discharge is important to save costs and time. There have already been many models used for Partial Discharge detection. However, little has been done with the Transformer model in signal classification. In this paper, personal research on Artificial Intelligence and Transformer models will be presented. In addition, the self-sourced data will be explained. The model used, alterations to the Transformer model and the process of training and testing will be shown. Conclusions will also be made of the current state of the project and what can be done in the future to achieve better results. Works that were still in progress but could not achieve meaningful results will also be stated in this paper.