A Survey on Machine Learning Approaches for Transformer Fault Estimation

Abstract

With the advent of contemporary technology, the need for a steady flow of electrical power has skyrocketed across all industries, necessitating virtually faultless operation of power grids. The protection relays for power transformers must function flawlessly and erratically to reduce the occurrence and length of unwelcome power outages. Operating speed with short fault detection and clearance time and reliability with no false tripping are two parameters that contribute to the high-pointed demand. Absolute dependability cannot be achieved alone by performing timely maintenance on the distribution components, especially transformers. Compared to traditional approaches, heuristic prediction offers proactive fault management, potentially reducing downtime and operational disruptions. In addition, power outages are common during maintenance and replacement due to faults. To sidestep this problem, heuristic prediction of fault probabilities over time is the way to go. This paper explores many modern methods for estimating transformer faults.