Exploring the application of reinforcement learning to wind farm control

Henry Korb,Stefan Ivanell,Henrik Asmuth,Merten Stender

doi:10.1088/1742-6596/1934/1/012022

Abstract

Optimal control of wind farms to maximize power is a challenging task since the wake interaction between the turbines is a highly nonlinear phenomenon. In recent years the field of Reinforcement Learning has made great contributions to nonlinear control problems and has been successfully applied to control and optimization in 2D laminar flows. In this work, Reinforcement Learning is applied to wind farm control for the first time to the authors’ best knowledge. To demonstrate the optimization abilities of the newly developed framework, parameters of an already existing control strategy, the helix approach, are tuned to optimize the total power production of a small wind farm. This also includes an extension of the helix approach to multiple turbines. Furthermore, it is attempted to develop novel control strategies based on the control of the generator torque. The results are analysed and difficulties in the setup in regards to Reinforcement Learning are discussed. The tuned helix approach yields a total power increase of 6.8% on average for the investigated case, while the generator torque controller does not yield an increase in total power. Finally, an alternative setup is proposed to improve the design of the problem.

Highlights

The current control paradigm most often employed for wind turbines is standard greedy control, taking into account only a single turbine
Wind farm control The power P generated by a turbine depends on the rotor speed ω and the torque exerted by the generator, Mgen
When applied to the optimization of the helix approach it successfully improved some of its parameters the high computational cost forbid a full optimization

Summary

Introduction

The current control paradigm most often employed for wind turbines is standard greedy control, taking into account only a single turbine. In [4] they mimicked the optimal behaviour with a sinusoidal variation of the thrust force. A similar approach that is based on the sinusoidal variation of control parameters, is the helix approach developed in [6]. In this approach, the blades are pitched individually in such a way that a sinusoidal tilt and yaw moment acts on the wake. The blades are pitched individually in such a way that a sinusoidal tilt and yaw moment acts on the wake This results in a helical deflection of the wake. A new optimization approach, based on Reinforcement Learning (RL), will be applied to wind farm control.

Methods

Findings

Discussion

Conclusion