Predicting Linearised Wind Resource Grids using Neural Networks

Modelling the flow over terrain is a key element of wind resource assessments within the wind energy industry. Existing flow modelling methods range from fast, low fidelity analytical models to time-consuming and computationally expensive high-fidelity Computational Fluid Dynamics (CFD) software. In this work, a Grid-Kernel Neural Network approach has been developed and used to create surrogate models to emulate the WAsP wind resource software, by calculating the changes in wind speed and direction due to the orography and roughness of terrain. This data-driven approach has proven to be successful in predicting the orographic speed and direction changes at multiple heights above ground. At 100 m above ground, the mean absolute error values were 1.6% speedup and 0.4 degrees for the orographic speed and direction changes, respectively. Although the WAsP model is a linear, potential flow solver, the findings here can be counted as a first step towards creating a fully data-driven CFD wind resource model.