Intelligent Fault Diagnosis with Missing Data based on Optimization-Ensemble Learning

Fault diagnosis has become increasingly important in the operation and maintenance of modern power systems. As an essential part of daily electrical equipment, bearings have a crucial role to perform in the operation of smart grid. Currently, most of the bearing fault detection is based on relatively complete bearing datasets for diagnosis, and rarely considers the missing data type of situation. So based on the integrity considerations, this paper proposes a fault diagnosis model based on the TTAO (Triangulation topology aggregation optimizer) improved HistGradientBoostingClassifier while using the VMD-RCMFE (Refined Composite Multiscale Fuzzy Entropy) decomposition noise reduction and the feature extracted bearing dataset for random simulation of the missing data, which makes the whole study more practical significance of engineering. Following this, the different models are validated by comparison experiments using three different evaluation indexes, and the results show that the TTAO-HistGradientBoostingClassifier bearing fault classification model proposed in this paper outperforms the other models in terms of accuracy ($\mathbf{9 5. 8 2 \%}$), KAPPA coefficient (0.94), and MAF1 value (0.96), showing that it works better for missing data than the other models, showing more accurate and reliable performance for data missing class datasets.