Open AccessK-nearest neighbor method for power transformers condition assessment
Author(s) -
Daria Tanfilyeva,
O. V. Tanfyev,
Yury Kazantsev
Publication year - 2019
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/643/1/012016
Subject(s) - k nearest neighbors algorithm , classifier (uml) , pattern recognition (psychology) , transformer , computer science , dissolved gas analysis , artificial intelligence , data mining , naive bayes classifier , bayesian probability , reliability engineering , machine learning , engineering , electrical engineering , support vector machine , transformer oil , voltage
Reliable power supply is the most crucial task of every power utility company, making power transformers one of their chief assets. Thus, detection of power transformers abnormal condition is of high importance. Most general and approved tool of condition assessment is Dissolved Gas Analysis (DGA) of transformer insulation liquids. The identification task in fact implies conventional pattern recognition of measured parameters and their classification. The majority of prevalent classification and recognition methods require a priori knowledge of classes and symptoms, probability distribution laws and density functions, etc. This paper describes application of two fault recognition methods – Bayesian classifier and k-nearest neighbours (KNN) algorithm. Studies have shown that KNN tool allows flexibility for concentration limits adjustment with operation conditions’ altering and has a high condition classifying accuracy.