Development Of Random Forest Model For Identification Of Transformer Faults Based On The Duval Triangle Method
Abstrak
A transformer is a type of electrical device that uses to change the electrical energy in electrical circuits into another electrical circuit. The resistance that occurs in the flow of energy distribution which needs consideration in this situation is the liquid insulation of transformer oil. Aging of the oil occurs when transformer oil is used to carry heat at high temperatures for an extended length of time. One of the preventative maintenance procedures that must be performed is DGA testing. One of the methods of DGA is the Duval Triangle method. The levels of acetylene, ethylene, and methane (C2H2, CH4, and C2H4) are special circumstances that are taken into account using this procedure. However, Duval's Triangle method still has flaws, especially since utilizing and comprehending graphs is challenging, which prolongs the analytical process. Therefore, to make it easier to understand using the Random Forest method where it’s one of the most widely used creates several distinct decision trees, each of which serves as a classifier in and of itself. The important aspect when developing Random Forest networks is picking the right tuning parameters. Random forest parameters can help the model be more predictive or make the model's training process simpler. As a result, the level accuracy of combining between Duval Triangle and Random Forest methods has over 90%. The application of the three transformers at UPT Malang proves the accuracy of the proposed model. Where the result between calculation and program has a same result. So, the use of Random Forest model could make the utilizing and comprehending graphs of Duval Triangle easier.