Risk-averse wake steering optimization for energy and power maximization under uncertain wind direction changes

Abstract

Wind farm flow control strategies aim to manipulate the flow between the turbines to achieve a farm-wide goal. Wake steering is one such strategy, typically employed to increase the yield of a wind farm by redirecting upstream turbines’ wakes away from downstream ones. This control approach is sensitive to the wind direction, and frequently varying wind directions create the need for robust yaw steering control setpoints. In the past, this has been achieved in the steady-state domain, as dynamic simulations are typically deemed too expensive to perform a large grid-search for optimal setpoints. This paper utilizes a computationally cheap dynamic wake model and explores how robust control setpoints can be derived in the time domain. To this end, the paper presents a methodology for generating synthetic wind direction changes with a prescribed variation in wind direction. These wind direction time series are then used to create a database of a two-turbine wind farm, which allows the exploration of different cost functions in time. The database provides both the expected value and the uncertainty of both power and energy. The obtained data is then used to explore four cost functions to derive robust setpoints. Comparing energy and power performance, we define useful quantities of interest to connect the two and to highlight necessary assumptions made when using steady-state setpoints. The paper concludes by applying the resulting look-up table controllers in a ten-turbine wind farm. The performance shows that maximizing for the expected power is the best approach to increase the farm efficiency. The results also show that an alternative cost function, which avoids losses, does lead to similar but smaller gains at a much lower yaw angle investment.