Power plant induced-draft fan fault prediction using machine learning stacking ensemble
Tlamelo Emmanuel, Dimane Mpoeleng, Thabiso Maupong
Abstract
The improvement of fault prediction and diagnosis in industrial systems is crucial to minimize unscheduled shutdowns. However, the predictive performance of current models for thermal power plants is limited due to their reliance on single algorithm approaches. Furthermore, there is a shortage of experiments on thermal fired power plant equipment, as most research focuses on nuclear power plants. In this study, we propose a fault predictive stacking approach for a thermal power plant induced draft fan and evaluate the performance of base learners, including Support Vector Machines (SVM), K Nearest Neighbors (KNN), and Random Forests (RF). Our proposed stacking ensemble approach achieved a higher prediction accuracy of 99.89% compared to the base algorithms. Additionally, the stacking ensemble method showed superior prediction performance compared to the base methods.