Multi-Sensor Feature Fusion and Grey Wolf Optimizer-Based Support Vector Machine for Transient Fault Detection in a Once-Through Power Plant

In this study, a multi-sensor feature fusion based fault detection approach was presented to deal with abnormal operating conditions in a sub-section of a Benson® type boiler. Four different filter-type feature selection were employed for ranking features. Dempster-Shafer evidence theory was employed to fuse the selected features and to increase the confidence in selecting the appropriate features. The selected features were used for training support vector machine classifiers based on experimental data. The parameters of the classifiers were adjusted using an exploration-enhanced grey wolf optimizer. To achieve the best possible performances over training dataset. The performance of the designed fault detection system was evaluated using the testing dataset. The obtained results indicate superior performances of the proposed approach in early recognition of faults during transient conditions.