GE, Sandia Lab building better wind power turbines
Aerodynamic blade noise is the dominant noise source on modern, utility-scale wind turbines and represents a key constraint in wind turbine design
Using high-performance computing, GE engineers have overcome previous design constraints, allowing them to begin exploring ways to design reengineered wind power blades that are low-noise and more prolific power-producers.
Partnering with the Sandia National Laboratories in Albuquerque, New Mexico, GE’s work focused on advancing wind turbine blade noise prediction methods. Aerodynamic blade noise is the dominant noise source on modern, utility-scale wind turbines and represents a key constraint in wind turbine design.
Efforts to reduce blade noise can help reduce the cost of wind energy and increase power output. In fact, GE predicts a 1 decibel quieter rotor design would result in a two-percent increase in annual energy yield per turbine. With about 240 GW of new wind installations forecasted globally over the next five years, a two-percent increase would create 5 GW of additional wind power capacity.
To ensure that GE’s wind blades do not pose noise issues today, airfoil level acoustic measurements are performed in wind tunnels, field measurements are done to validate acceptable noise levels, and noise-reducing operating modes are implemented in the control system. Better modeling will help maintain the current low noise levels while boosting output.
GE’s testing involved Sandia’s Red Mesa supercomputer running a high-fidelity Large Eddy Simulation (LES) code, developed at Stanford University, to predict the detailed fluid dynamic phenomena and resulting wind blade noise. For a period of three months, this LES simulation of the turbulent airflow past a wind blade section was continuously performed on the Red Mesa HPC. The resulting flow-field predictions yielded valuable insights that were used to assess current engineering design models, the assumptions they make that most impact noise predictions, and the accuracy and reliability of model choices.