Multi-target normal behaviour models for wind farm condition monitoring

The trend towards larger wind turbines and remote locations of farms fuels the demand for automated condition monitoring strategies that can reduce operating cost avoid unplanned downtime. Normal behaviour modelling has been introduced to detect anomalous deviations from normal operation based on turbine’s SCADA data. A growing number machine learning models turbine subsystems are being developed by farm managers this end. However, these need be kept track of, maintained require frequent updates. This research explores multi-target as a new approach capturing behaviour. We present an overview regression methods, motivate their application benefits in SCADA-based monitoring, assess performance case study. find advantageous comparison single-target they effort practical without compromising accuracy. also outline some areas future research.