POWER TRANSFORMER INSULATION SYSTEM HEALTH INDEX WITH MISSING DATA PREDICTION USING RANDOM FOREST

• Main Author: Chintia, Geby
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77947
• Institution: Institut Teknologi Bandung

Health Index Transformer is currently the most common way to predict the health condition of transformers, to avoid the failure of the transformer, anticipation is needed by monitoring and maintaining the condition of the transformer. This study presents information regarding the diagnosis of transformer conditions based on the Transformer Health Index method using Random Forest. The Transformer Health Index method provides a comprehensive assessment of the condition of the transformer. Then generally so many development methods are also used. To avoid failure of the power transformer anticipation, monitoring and maintaining the condition of the transformer is needed. This study presents information regarding the diagnosis of transformer conditions based on the Transformer Health Index method using Random Forest. The Transformer Health Index method provides a comprehensive assessment of the condition of the transformer. This paper discusses the prediction of transformer health conditions using the Missing Data Replacement Method in five ways, which are Removed Parameter, Average value, Assume Good, SLR, and Random Forest Prediction, with seven proposed combinations based on three parameters, namely 2FAL, IFT and Water Content while considering the level of accuracy. The results from 504 Transformer analysis state that the Health Index calculation using the Random Forest method has the highest accuracy rate among other methods with a value of 92%. As we know that Random Forest Metode is an ensemble classification technique that often produces superior performances over individual classifiers, that fits some decision tree classifiers on various subsamples of the dataset and uses averaging to improve the predictive accuracy. From each combination applied, explaining the availability of several parameters could be an influence on the Health Index Transformer assessment, but it can still be done with the proposed method.