This project is intended as a Master's project for students with a background in control, sensing or mathematical modelling.
As part of the project it is encouraged for the student to spend 1-2 weeks at Vestas R&D headquarters in Aarhus.
As a wind turbine is a highly
dynamic structure, sufficient state information is key to secure an optimal
control solution. Usually, such state information is obtained by means of
selected sensors measuring e.g. rotor speed, wind speed, blade pitch angle,
nacelle yaw direction, and blade loading, combined with state estimation
Unreliable state information can
affect the structural safety of the turbine if not handled correctly and
provisions for handling faults and errors in the sensing systems must be
present as part of the turbine control system. Redundant, multi-channel sensor
system architectures is one way of securing reliable state information, but may
be constrained by cost, space restrictions, and mounting possibilities.
Therefore, it is of interest to seek solutions that provide the necessary fault
detection capabilities while avoiding adding cost.
This project is focused on
developing fault detection and sensor fusion algorithms with low detection time
and low false detection rates. Both model-based methods such as Kalman filters
and data driven methods such as Deep Neural Networks are of interest, but the
solution should handle the challenge that the turbine dynamics are highly
non-linear and that the turbine and environmental properties vary significantly.
An example could be verifying the
blade load sensors measurements by consistency with the rotor, speed, pitch
angle, produced power and measured wind speed.
It is of relevance also to
investigate how the proposed solutions comply with relevant standards for
functional safety, such as ISO 13849, IEC 62061, and IEC 61508. In this
context, architecture and avoidance of common-cause failures is of interest.
Master's student with focus on control, sensing or mathematical modelling