Data-Driven Multi-unit Monitoring Scheme with Hierarchical Fault Detection and Diagnosis

Various areas see the trend of increasing degree of decentralization and automation. Conventional component or unit based monitoring scheme by analytical modeling could be tedious and costly in dealing with increased number of entities. A holistic data-driven scheme able to monitor multiple units and hierarchically detect and diagnose faults is proposed, in which Multi-way Principal Component Analysis (MPCA) is employed as data analysis algorithm. The proposed scheme is illustrated using data collected from an actual onshore wind farm with multiple wind turbines. The obtained fault detection and diagnosis results are validated using maintenance reports.