Investigation into automating wind turbine underperformance detection using social statistics
report
This report describes the work performed as part of the PPP Toeslag project “Automated Underperformance Detection for Wind Turbines”. The project aimed to take a novel approach to wind farm condition monitoring and automated alerting, based on Bayesian statistics and the relative behaviour of wind turbines, with the following goals:
• To be able to automatically detect and quantify underperformance in a wind farm.
• To do this as much as possible using only 10 minute SCADA data, which is cheap and readily available (although often unreliable).
• To assess the additional value of a minimal set of data from additional measurements (e.g. a Lidar campaign) – in order to improve information sufficiently to take maintenance decisions. For this purpose, Fortum provided TNO with several years of commercially-sensitive SCADA data from one of their onshore wind farms in northern Europe, along with data from expensive nacelle-mounted Lidar campaigns which established yaw misalignment on three turbines.
• To be able to automatically detect and quantify underperformance in a wind farm.
• To do this as much as possible using only 10 minute SCADA data, which is cheap and readily available (although often unreliable).
• To assess the additional value of a minimal set of data from additional measurements (e.g. a Lidar campaign) – in order to improve information sufficiently to take maintenance decisions. For this purpose, Fortum provided TNO with several years of commercially-sensitive SCADA data from one of their onshore wind farms in northern Europe, along with data from expensive nacelle-mounted Lidar campaigns which established yaw misalignment on three turbines.
Topics
TNO Identifier
878591
Publisher
TNO
Collation
39 p.
Place of publication
Petten