Machine Learning Algorithms Fusion Based on DGA Data for Improving Fault Diagnosis of Electrical Power Transformer

Abstract

Abstract Dissolved Gas Analysis (DGA) continues to be widely recognized as a valuable method in recent times for the early identification of issues in oil-filled power transformers. It has gained extensive adoption as a primary approach for the early discovery of these issues, relying on the analysis of dissolved gases. This contributes to enhancing the dependability of electrical systems. This paper proposes an efficient fusion method based on DGA data using the two best Machine Learning algorithms, the neural network (MLP), the naïve Bayes (NB) throughdata input vector ppm, a percentage input vector, and an Logarithmic input vector. The fusion method predictively combined the two classifiers and obtained a statistical evaluation: accuracy, recall, precision, and F-measure higher than both classifiers separately. The proposed fusion method was evaluated for performance using a test database and compared with conventional and smart methods. Results showed that the proposed model outperformed both traditional and intelligent methods in terms of diagnostic accuracy when using percentage and logarithmic input vectors. The Prediction Based Fusion (PBF) vector Percentages achieved an accuracy rate of 97.22%, while PBF vector Logarithmic achieved an accuracy rate of 95.83%. These rates were higher than those achieved by traditional methods, such as the Modified RRM/CEGB method 91.67% and Modified RRM/IEC method 90.28%. Additionally, the proposed model surpassed the accuracy rates of intelligent methods, such as CSUS ANN 88.89% and Conditional Probability 93.06%.