Expert System for Predictive Maintenance Transformer using J48 Algorithm

Abstract

Predictive maintenance can reduce the risk of sudden transformer failure which causes the risk of plant to stop operating. One of transformer predictive maintenance technique is the Dissolved Gas Analysis (DGA) Test Oil Transformer. The gas is interpreted and analyzed to find out and get conclusions about the health condition and also possible problems in the transformer based on IEEE Standards and IEC Standards. To facilitate monitoring, a Decision Support System for Interpretation of Test Results of DGA Oil Immersed Transformer was created to form a database containing transformer data with the amount of main gas from the DGA test results. Next, decision tree was made using the J48 algorithm. The decision tree simplifying and speed up the decision-making process for recommended actions that are displayed on the system. The system also displays a trending graph of the last transformer test and quickly displays a dashboard of transformer status, i.e. normal, alarm, or danger. Engineer will get notification email if any transformer is in danger status. In addition, the system is able to provide information on possible fault types for each transformer. The benefits of this system are that the health condition of the transformer can be monitored properly and corrective action can be taken immediately on a problem based on the results of the decision support system. This will reduce the risk of shutdown and support the reliability of plant operations.