Predictive analysis using machine learning modelling of digital twin of a Power Transformer

(ML)–driven digital twins empower utilities and asset managers to make better-informed decisions and proactively address issues before they impact transformer performance. Several research papers have been published on detecting anomalies from DGA data using machine learning algorithms. However, in addition to DGA, other real-time data and periodic test results are combined for better prediction of failures. The developed algorithm/model is tested using historical data from failed transformers. This leads to the identification of the most suitable and accurate algorithm to carry out a better and effective predictive analysis of a power transformer.

The integration of Machine Learning (ML) with Digital Twin (DT) modelling of transformers presents a transformative approach to monitoring and optimizing these critical assets in real time. The IEEE standards for DGA analysis and further historic data from a utility etc play a pivotal role in training the algorithm so that more realistic outputs can be given by the algorithms in a faster way enabling the maintenance staff to make faster decisions and to avoid irreparable damages to the transformer which is one of the costliest equipment in a substation.