A Machine Learning Approach to Transformer Oil Temperature Monitoring Using Load Analysis

Abstract

Transformers are the vital components of the electrical power network, and they must be adequately monitored and examined to avoid irreversible damage. The transformers coolant, which is oil, maintain its dielectric properties for a certain temperature ranges and hence it is essential to monitor it effectively for increasing the life shell of transformer. In view of this the paper proposes a transformer monitoring system which is based on machine learning technique. For monitoring oil temperature whether low or high, various machine learning classifier like random forest, support vector machine (SVM), k-nearest neighbors (kNN), and logistic regression are evaluated in this paper. The impact of different load condition on the oil temperature is also highlighted. The performance of various classifier is validated by calculating the evaluation metrics and it can be seen from the results that kNN outperforms the random forest, SVM, and logistic regression.