<![CDATA[Thermal sensitivity incidents on generator rotors at Muara Tawar Power Plant have increased over the past five years, which will have an impact on the overall performance of the power plant. The general method of conducting thermal sensitivity testing requires the generating unit to be in a certain operating pattern, thus limiting the analysis of anticipating events in real time. Correlation analysis between excitation current variables, reactive power, vibration, and temperature needs to be carried out periodically. The acquisition of these operating parameters was carried out on three generator rotors for 14 days per minute and will be implemented into a machine learning model. This study uses the Random Forest model to predict vibrations on the rotor and determine featureimportance values, with the addition of Long Short-Term Memory (LSTM) modeling to predict future trends based on feature importance. The results show that the Random Forest model can predict vibrations in the rotor and determining the importance of the features used, with an average evaluation metric RMSE of 0.92% and R2 of 81.62% on the exciter side, and RMSE of 2.75% and R2 of 61.42% on the turbine side. The LSTM model also demonstrates good capability in predicting future trends in thermal sensitivity identification based on exciter current features with an RMSE of 7.29% and for reactive power features of 6.52%, indicating that the proposed modeling implementation allows a better understanding of the variables relevant to thermal sensitivity, thus predicting them in the future can produce comprehensive operation and maintenance strategies.]]>
Copyrights © 2024
