Applications of Feature Selection for Detecting Energy Discharge and Thermal Faults in Transformers Using DGA Data

Feature selection for transformers’ dissolved gas analysis (DGA) is critical for enhancing the accuracy of fault detection models. This study employs an artificial neural network fitting method for feature selection to improve model precision. Additionally, a binary genetic algorithm is utilized to further enhance the model’s accuracy. The focus is on five key dissolved gases-hydrogen $\left(\mathrm{H}_{2}\right)$, methane (CH<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</inf>), ethane (C<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> H<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</inf>), ethylene $\left(C_{2} H_{4}\right)$, and acetylene $\left(\mathrm{C}_{2} \mathrm{H}_{2}\right)$-to predict various types of energy discharge and thermal faults in transformers. The proposed model demonstrates the ability to accurately predict fault types with fewer gas samples, achieving an average accuracy of $\mathbf{9 2 \%}$. These findings have significant implications for decreasing the computational complexity of real-time monitoring systems for power and distribution transformers.