Application of machine learning in the transformer health index analysis
Abstract
The paper will present the application of several different machine learning algorithms in the analysis of the health index of power transformers. The health index defines the general state of health of the transformer at a certain point in time in its lifetime and is obtained by aggregating a number of indicators, results of various tests, expert assessments of the condition of the vital parts of the transformer, as well as historical data on loading. These are the following algorithms (English): multinomial logistic regression, extremely randomized trees, k-neighbors classifier, random forests, gradient boosting classifier, support vector machines, artificial neural network. The application of machine learning in health index analysis will be demonstrated on a specific set of thirty energy transformers of different power and transmission ratio.
