Measured and Simulated Turbulence - Compared at a Section of a Rotating Wind Turbine Blade

Computer models for prediction of wind turbine response are often based on time simulation. In such models it is common practice to use a stochastically generated, synthetic, turbulent wind field as input to the response calculation.

The present work investigates the quality of such generated synthetic turbulence by comparing it with measured turbulence, represented as the local wind vector at a cross section of a rotating wind turbine blade in natural wind. This procedure has the great advantage that the important spatial structure of the turbulence is tested for through the comparison of the rotationally sampled wind vector. The comparison is extended to cover the normal force on a blade segment, which is derived in a simplified way from the wind vector. In particular, the importance of including the transversal turbulence components is addressed.

A recently developed simulation method, which includes all three turbulence components, shows excellent agreement with the measurements and is superior to a model, which includes only the longitudinal component. It is demonstrated that inclusion of all three components is important for the simulation of the local wind vector, represented in the present context as the relative velocity and the angle of attack. The model, which includes only the longitudinal component, gives rise to underestimation of the relative velocity variations and overestimation of the angle of attack variations originating from turbulence.

In order to demonstrate the effect of these discrepancies on the resulting force on the blade segment, the force is derived in a simplified way by use of linearized normal force coefficient curves. The actual influence of the discrepancies depends on the slope of the coefficient curve. Generally, high slopes (below stall onset) result in overestimation of the force variation, and low slopes (above stall onset) result in underestimation of the force variation. Rainflow counting and estimation of fatigue damage show that the transversal components in general are of great importance for the prediction of fatigue.

The investigation has not directly dealt with the influence of turbulence on the load calculation models, as for example dynamic stall models, but the obtained results suggest
that the quality of the turbulence field should be of great concern when applying and evaluating these models.