Optimal Feature Selection for Cleanliness Factor Prediction in USC Coal-Fired Boilers Using RF-RFECV

Fouling and slagging are major issues occurring in coal-fired power plant, leading to issues such as reduced heat transfer efficiency, metal corrosion and emergency shutdown. Therefore, predicting cleanliness factor (CF) in convective sections, which is the cleanliness metric to monitor this situation is imperative for a more systematic boiler management. Power plants typically operate with high dimensional feature datasets, yet the existing machine learning approaches particularly within ash fouling monitoring often lack integration with data mining techniques. This study aims to identify the optimal subset of features to achieve the highest CF prediction accuracy. The findings reveal that integrating random forest (RF) regressor with recursive feature elimination and cross-validation (RFECV) outperforms conventional feature selection methods. The proposed model achieved an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.91 using only 8 features, demonstrating the effectiveness of the RFRFECV method. In addition to that, this method is costeffective as it eliminates the need for hardware installation and offers an economic alternative. Ultimately, it improves the efficiency and reliability of the cleaning process while ensuring operational effectiveness of the power plant.