DGA Based Transformer Fault Analysis Using Machine Learning

In power transmission network transformer plays an important role, so its protection attains much more attention in order to have fault free supply. When the mineral oil inside the transformer is subjected to high thermal and electrical stress gases are created from the decay of mineral oil. Distinct type of fault creates diverse gases and extractions of these gases from the sample gives helpful data about the state of oil and type of fault. Early identification of fault reduces the failure rate. This work proposed Machine Learning Based Dissolved gas analysis (DGA) technique to diagnose transformer faults. In this work multiple machine learning models are developed and tested using the DGA dataset. Outputs are generated under different Validation schemes. Also the optimization algorithm with feature selection using ANOVA test and hyper parameter tuning using random optimization is proposed and acquired the Validation removes over 100% accuracy.