Hyperparameter tuning framework for calibrating analytical wake models using SCADA data of an offshore wind farm

Abstract

This work presents a robust methodology for calibrating analytical wake models, as demonstrated on the velocity deficit parameters of the Gauss–curl hybrid model using 4 years of time series supervisory control and data acquisition (SCADA) data from an offshore wind farm, with a tree-structured Parzen estimator employed as a sampler. Initially, a sensitivity analysis of wake parameters and their linear correlation is conducted. The wake model is used with a turbulence intensity of 0.06, and no blockage model is considered. Results show that the tuning parameters that are multiplied by the turbine-specific turbulence intensity pose higher sensitivity than tuning parameters not giving weight to the turbulence intensity. It is also observed that the optimization converges with a higher residual error when inflow wind conditions are affected by neighbouring wind farms. The significance of this effect becomes apparent when the energy yield of turbines situated in close proximity to nearby wind farms is compared. Sensitive parameters show strong convergence, while parameters with low sensitivity show significant variance after optimization. Additionally, coastal influences are observed to affect the calibrated results, with wind from land leading to faster wake recovery than wind from the sea. Given the assumption of constant turbulence intensity in this work, recalibration is required when more representative site-specific turbulence intensity measurements are used as input to the model. Caution is advised when using these results without considering underlying model assumptions and site-specific characteristics, as these findings may not be generalizable to other locations without further recalibration.