Engineering approach to turbine interactions

As the wind industry evolved over the past decades, so did the understanding of aerodynamic phenomena that occur across and around wind farms. Hence, it is a common practice to combine multiple models to properly account for the different components of aerodynamic losses in a large, contemporary wind farm. This practice is a consequence of the industry's realisation that certain phenomena became non-negligible with the increase in wind farm size - which prompted corrections to the pre-existing modelling approaches rather than replacement of those. Such was the case when DNV introduced the LWF correction model in an attempt to capture the effect large wind farms have on the atmospheric boundary layer. Similarly, when the industry realised the magnitude of the wind farm blockage effect, DNV created the BEET model to correct for losses unaccounted for in the pre-existing models.

An important requirement within that approach, given the simplification in the engineering models and related corrections, is to validate and tune the models to wind farm SCADA data so that overall accuracy of the model chain is satisfactory.

Wind farm modelling within the above described traditional modeling paradigm is possible in WindFarmer:Analyst using engineering wake & blockage models.

Alternatives

An alternative to the traditional approach is to use computational fluid dynamics (CFD) methods to resolve wakes & blockage (for example DNV's RANS CFD). CFD is a much higher-fidelity model, one based on first principles and hence not requiring any tuning to operational data. DNV recommends RANS CFD as the lowest-uncertainty method for modeling wind farm flows in the context of energy production assessment.

However, RANS CFD comes at a much higher computational cost - making it feasible primarily in the context energy production assessments at the final stages of project developement. During initial (frequent and numerous) design iterations, a much faster surrogate model to DNV's RANS CFD, the CFD.ML is a worthwhile option.