A Data-Driven Model Predictive Control for Wind Farm Power Maximization

Abstract

This paper presents a data-driven approach to maximize the power of a wind farm by developing a dynamic mode decomposition with input and output for reduced order model (DMDior)-based reduced order model (ROM) for model predictive control (MPC). The main goal of this research is to efficiently model and manage the complex flow field within a wind farm to enhance power production. We leveraged DMDior to transform extensive high-dimensional flow data into an accurate yet simplified ROM, which successfully represents the essential dynamic features of wind flow, including the critical interactions between turbines and their adaptive response to environmental changes. Based on this ROM, the MPC framework was carefully designed. MPC uses this model to dynamically adjust the yaw angle of a wind turbine to optimally match changing wind patterns to maximize power output. The system also incorporates an adaptive Kalman filter designed for the state estimation in MPC applications. This estimation is critical to the effective execution of the MPC in each iteration. This ensures that the MPC operates based on the most up-to-date and accurate representation of the wind farm’s state, improving the overall reliability and efficiency of the control strategy. This approach demonstrates a practical and effective way to increase the power output of a wind farm, with experimental results indicating a power increase of about 4.72%.

Keywords

• Wind farm control
• data-driven approach
• dynamic mode decomposition with input and output
• reduced order model
• model predictive control
• adaptive Kalman filter