Technology-Oriented Wind Farm Configuration Selection Using an RFR-Based Turbine Power-Output Surrogate Model: A National-Scale Assessment for Brazil

Abstract

This work presents a framework for national-scale onshore wind farm screening in Brazil based on turbine technology. The approach integrates ERA5 hourly wind fields, turbine technical specifications, a Random Forest Regression (RFR)-based turbine power-output surrogate model, and GIS-based spatial exclusion analysis. The RFR model was trained using 821 empirical power curves, enabling different real turbine models to be evaluated within a unified framework. Seven representative turbines were evaluated across five wind farm configurations under capacity-factor thresholds of 0.20, 0.25, and 0.44. Results show that turbine technology substantially changes site suitability: for CF≥0.20, viable locations ranged from 116 to 1846, and for CF≥0.25, from 19 to 1175. Only 298 locations nationally reached CF≥0.44, with the two best-performing configurations accounting for 297 of them. Even between these similarly rated turbines, neither dominated nationally: one outperformed the other at 58.5% of viable locations, with the reverse holding for the remaining 41.5%. Under the cost assumptions adopted, only the best-performing configurations achieved levelized costs of energy competitive with recently contracted wind energy prices in Brazil. Validation against the Icaraizinho wind farm yielded an MAPE of 6.40%. Overall, the results demonstrate that wind potential is technology-dependent and should be assessed using turbine-specific performance models.