Health Index Stator Generator Evaluation with Random Forest Algorithm: A Case Study of PLTU Tanjung Awar-Awar

The stator generator is the most critical component in equipment within a 300 MW coal-fired power plant system. To ensure the reliability of the generator stator, a health index assessment is required as a requirement for operational feasibility using the Analytical Hierarchy Process (AHP) method. Using the AHP method, the health index assessment is conducted by assigning weights to maintenance and testing parameters, while operational parameters are used for validation. However, this approach only represents a corrective health index. This study aims to develop a preventive health index assessment for the generator stator by assigning weights through the integration of three parameters, namely testing, maintenance, and operational parameters. All data are initially assessed manually and subsequently used as input for machine learning. Machine learning is capable of capturing patterns of parameter changes that cause stator damage. This study employed three machine learning models, K-Nearest Neighbors (K-NN), Decision Tree and Random Forest (RF) algorithm. The algorithms were evaluated using several parameter to determine the best-performing model. Random forest has lowest error 4,55%. It can explain most variability in the observed data. Furthermore, the RF will be used to evaluate the stator health index at Coal-Fired Power Plant Tanjung Awar-awar.