Abstract: Individual horizontal axis wind turbines (HAWTs) operate by using the aerodynamic force of wind to spin the rotor blades and generate power. Currently wind farm control methods in practice use look-up tables based on offline optimization solutions. However, once multiple turbines are grouped together to form a wind farm, wake dynamics cause the system to behave differently.
One novel insight guides this research talk: treating wind farms as a collective entity and noting the wind acts as a shared resource. Through that lens and acknowledging upstream effects as the dominant factor leading to diminished power extraction guides the construction of distributed control schemes.
Motivated by Model Predictive Control (MPC) methods, we present recent work under a multi-objective optimization framework. That is, we design a centralized controller to maximize the power of a wind farm under a set of real-world derived constraints using data-driven estimation of wind dynamics. Furthermore, we prove such a controller through Lyapunov theory and validate through simulation results.
Bio: Lucas Buccafusca is a PhD candidate in Industrial and Systems Engineering after completing his Master’s in Electrical and Computer Engineering, both at the University of Illinois at Urbana-Champaign. His research focuses on distributed control, learning, and optimization; primarily applying these techniques on the study on maximizing power extraction of wind farms. His work has been recognized nationally, winning the INFORMS Poster Presentation competition and an Editor’s Choice award in the Journal of Renewable and Sustainable Energy.