Supervisorer: Nikolay Dimitrov, Laura Schröder
(DTU Vindenergi)
Censor: Bjarne Skovmose Kallesøe
Titel: Design of fault detection algorithms based on virtual data sets.
Data driven methods that operate on the SCADA data from wind turbines are an attractive option to detect faults that reduce the performance and the lifetime of the turbines. For advances to happen in this area, it is important to have example datasets that help understand the nature of faults properly.
In this work, a database of virtual wind turbine operation data under faulty behavior is created through aeroelastic load simulations.
Such a database is a useful tool for training and testing the performance of various fault detection algorithms since the faults are known to the user.
The thesis presents the process of modelling the fault representation in the aeroelastic software Hawc2, as well as the performance of several algorithms for data-driven fault detection which have been applied on the simulated data.