On-Line Monitoring for the Operation Conditions of Hydro Power Generators Based on Multi-Source Data Fusion and Dynamic Principal Component Analysis

This study presents a dynamic principal component analysis method for real-time monitoring and anomaly detection in hydro power generators. By using dynamic principal component analysis (PCA) as the core algorithm integrated with sliding windows and incremental updates, the approach efficiently captures short-term variations in high-dimensional time-series data. It identifies anomalies through various methods, and then identify the sources of the anomalies. The method is computationally efficient and suitable for large-scale monitoring, enabling precise localization of anomalies and ensuring robust operational assessments.