Pseudo-2D RANS: A LiDAR-driven mid-fidelity model for simulations of wind farm flows

TL;DR

The paper introduces Pseudo-2D RANS, a computationally efficient wind farm flow model calibrated with LiDAR data, improving accuracy over existing engineering wake models for wind farm simulations.

Contribution

It presents a novel pseudo-2D RANS model that combines theoretical foundations with calibration to LiDAR measurements, enabling accurate and low-cost wind farm flow simulations.

Findings

Achieved a 5.6% error in power prediction for an onshore wind farm.

Improved efficiency predictions by 24% over existing engineering wake models.

Validated the model against extensive real-world wind farm data.

Abstract

Next-generation models of wind farm flows are increasingly needed to assist the design, operation, and performance diagnostic of modern wind power plants. Accuracy in the descriptions of the wind farm aerodynamics, including the effects of atmospheric stability, coalescing wakes, and the pressure field induced by the turbine rotors, and low computational costs are necessary attributes for such tools. The Pseudo-2D RANS model is formulated to provide an efficient solution of the Navier-Stokes equations governing wind-farm flows installed in flat terrain and offshore. The turbulence closure and actuator disk model are calibrated based on wind LiDAR measurements of wind turbine wakes collected under different operative and atmospheric conditions. A shallow-water formulation is implemented to achieve a converged solution for the velocity and pressure fields across a farm with computational costs comparable to those of mid-fidelity engineering wake models. The theoretical foundations and numerical scheme of the Pseudo-2D RANS model are provided, together with a detailed description of the verification and validation processes. The model is assessed against a large dataset of power production for an onshore wind farm located in North Texas showing a normalized mean absolute error of 5.6% on the 10-minute-averaged active power and 3% on the clustered wind farm efficiency, which represent 8% and 24%, respectively, improvements with respect to the best-performing engineering wake model tested in this work.