Wind Turbine Tower Structural Health Monitoring

Strong winds bend and thus strain the wind turbine tower (see figure to the left). Both the instantaneous stress and the cumulative effect over time can in principle be assessed by continuously measuring the 3D displacement of the nacelle, e.g. by using a real-time kinematic positioning (RTK) module based on satellite navigation. However, high-quality RTK data can in practice only be gathered 10-20 % of the time, rendering the entire calculation unreliable.

To monitor the operation of the wind turbine, a large amount of operating data, such as wind speed and power output, is continuously recorded. These data are somehow corre-lated to the displacement, but the relation cannot be analytically described. The same is true for acceleration data which can easily be gathered by adding a measurement module. The invention makes use of this connection by training a neural network with RTK data as labels and other data as features when both are available with high quality and subseq-uently using the network to calculate nacelle displacement when RTK data are unavailable.