Sistem deteksi pada transformator menggunakan Dissolved Gas Analysis (DGA) dengan metode Logistic Regression

Abstrak

This research discusses the implementation of a failure detection system in transformers using dissolved gas analysis (DGA) and the Logistic Regression method. The Logistic Regression method was chosen as a classification tool to identify gas patterns that could indicate critical conditions in the transformer. This research aims to create a system expected to increase transformer reliability, reduce the risk of electrical disturbances, and contribute to industrial practice in implementing more efficient methods for maintaining power transformers. From the training and testing DGA transformers data experiencing various levels of damage show that Logistic Regression provides excellent performance with accuracy, precision, F1-Score, and recall of 97%, 97%, 97%, and 97%. These results reflect a remarkable balance between this method's ability to classify positive and negative results. While Key gas has low accuracy, precision, and F1-Score (42%, 17%, and 24%), The recall value of 42% shows the ability of this method to detect a large number of actual positive data but with a higher error rate. Overall, these findings can potentially improve the reliability of the electric power system through early maintenance and timely detection of potential problems with transformers.