Issue Models in FFM
Understanding how WindESCo uses physical models, machine learning, and AI to identify issues
Many of our customers have asked how WindESCo applies advanced analytics approaches appropriately - check out our general approach below to understand a little more about the step-by-step process
- First, we take physical principles and expert knowledge of wind turbines to drive analysis and results in technically sound conclusions
- This includes mechanical, aerodynamic, control system and field expertise drive specific anomaly identification
- Servo-mechanical expertise informs analysis of expected and actual behavior
- Essentially: Is the turbine controller doing what it should be doing, when it should be doing it?
- We then use innovative signal processing and machine learning applications to advance subject matter expert (SME) capabilities
- Machine learning and AI enhance models of anomaly detection and normal behavior when it is appropriate
- These approaches enable rapid automation and scaling of SME expertise
- Examples include: Clustering, Identification of derating, Changes across turbines or time
It is important that the approach is not simply a data in / data out approach, but instead uses physical understanding of behavior to drive proper input/output relationships. For example, we know that correlation does not equal causation and our team's expertise enables AI/ML to find actionable causation.