Wind turbine load validation under wake conditions using Doppler lidar

A significant source of uncertainty in energy productions and lifetime predictions of wind farms is the wake-induced effects by neighbouring turbines. In the current state-of the-art procedure, engineering wake models, i.a., the Dynamic Wake Meandering (DWM) model, are typically used to simulate flow fields within wind farms.

These models are suitable for carrying out hundreds to thousands of aeroelastic simulations required in an iterative design and optimization study. However, engineering wake models are subjected to a significant uncertainty level due to the simplistic flow modelling assumptions, incorrect calibration of the model parameters, and unknown confidence in the overall model prediction ability due to lack of adequate experimental data for model validation.

As wake-induced power and loads are an essential design factor, there is a strong need to accurately simulate wake fields to ensure reliable wind turbine designs and optimized wind farm layouts and operational strategies. The present thesis focuses on wind turbine load validation procedures under wake conditions using nacelle-mounted lidars (LIght Detection And Ranging).

These lidars can measure the inflow wind at a high spatial and temporal resolution, yielding much more insight into the actual inflow approaching the turbine rotor. This thesis’s primary purpose is to improve the accuracy and reduce the uncertainty in load assessments under wake conditions by demonstrating load validation procedures using measurements from nacelle lidars.

Two hypotheses are formulated: (I) incorporating nacelle-lidar measurements in the wake field reconstruction methods improves the accuracy of power and load predictions compared to engineering wake models, (II) calibrating engineering wake models using high-resolution nacelle-lidar measurements improves the accuracy in both wake simulations and power and load assessments.

Three lidar-based wake field reconstruction procedures are defined and evaluated numerically and experimentally to verify the first hypothesis. First, the wake is modeled by means of time-averaged wind field characteristics estimated by a model-fitting technique combined with multiple measurements performed by the Avent 5-beam Demonstrator and the ZephIR Dual-Mode nacelle lidars, which were installed at the Nørrekær Enge (NKE) wind farm in Denmark.

Second, the wake field is modeled as a time series of wake deficits, whose properties such as widths, depths, and center locations are estimated by fitting nacelle-lidar measurements to a bivariate Gaussian shape function. Third, the nacelle-lidar measurements of the wake field are incorporated as constraints into turbulence fields serving as inputs to aeroelastic simulations.

The latter two methods are evaluated through numerical studies. The lidar-based load validation procedures indicate that reconstructing the wake-induced velocity deficit and its meandering (displacement in the lateral and vertical directions) is fundamental to ensure accurate power and load predictions.

Further, incorporating a sufficient number of nacelle-lidar measurements as constraints into turbulence fields is the most accurate and robust method. This approach reconstructs wake fields with strong similarities to the observed inflow, and as a result, reduces the statistical uncertainty in power and load predictions compared to conventional engineering wake models.

To verify the second hypothesis, high spatial and temporal resolution SpinnerLidar measurements of the wake field, which were collected at the Scaled Wind Technology Facility (SWiFT) in Texas, are analyzed. The lidar-derived wake characteristics include the velocity deficit, wake-added turbulence, and wake meandering in both lateral and vertical directions under varying inflow wind and atmospheric stability conditions.

Based on the lidar observations, a probabilistic calibration of the DWM model is carried out using Bayesian inference, where the resulting joint distribution of parameters allows both for model implementation and uncertainty assessment. The performance of the DWM model is then evaluated using power and load measurements collected at the SWiFT and NKE sites.

The results show that the DWM model can accurately predict power and fatigue loads statistics, as long as properly-calibrated parameters are used and wake meandering time series are precisely replicated. In summary, this work demonstrates the applicability of nacelle-mounted lidars to improve wind turbine power and load assessments under wake conditions.

Further, it verifies that incorporating nacelle-lidar measurements in the wake field reconstruction procedure and the calibration procedure for engineering wake models improves the accuracy in energy productions and lifetime predictions of wind farms.