Optimizing wind farm control through wake steering using surrogate models based on high-fidelity simulations

Abstract

Abstract. This paper aims to develop fast and reliable surrogate models for yaw-based wind farm control. The surrogates, based on polynomial chaos expansion (PCE), are built using high-fidelity flow simulations coupled with aeroelastic simulations of the turbine performance and loads. Developing a model for wind farm control is a challenging control problem due to the time-varying dynamics of the wake. The wind farm control strategy is optimized for both the power output and the loading of the turbines. The optimization performed using two Vestas V27 turbines in a row for a specific atmospheric condition suggests that a power gain of almost 3%±1% can be achieved at close spacing by yawing the upstream turbine more than 15∘. At larger spacing the optimization shows that yawing is not beneficial as the optimization reverts to normal operation. Furthermore, it was also identified that a reduction in the equivalent loads was obtained at the cost of power production. The total power gains are discussed in relation to the associated model errors and the uncertainty of the surrogate models used in the optimization, as well as the implications for wind farm control.

Wind-Farm Power Tracking Via Preview-Based Robust Reinforcement Learning

“…Due to the strong aerodynamic couplings, de-rating upstream turbines can mitigate wake effects and lead to the power increase of the downstream turbines or even the whole wind farm. Many studies [1]- [7] have shown that the greedy mode (i.e. all turbines in the farm working in the MPPT mode) cannot maximize the whole farm's power generation.…”

“…A commonly-used wind farm control strategy is to establish analytical or parametric wind-farm models firstly and then design controllers based on them. Following this pattern, many model-based methods have been proposed to optimize the power generation of wind farms [1], [2]. However, due to high system complexities, modelbased wind farm control methods suffer from uncertainties and unmodelled dynamics, and thus in practice they could have quite different performance compared with theoretical results.…”

“…The proposed PR-DRL method addresses this challenging problem by employing only the system's input & output data without requiring any analytical model. It overcomes the drawbacks of model-based wind farm control methods [1], [2], [9], [12], [13], [14], [15], [16] that are sensitive to modelling errors and uncertainties.…”

Wind farm control ‐ Part I: A review on control system concepts and structures

“…In [133], a surrogate model using polynomial chaos expansion is built. The data to train the model are generated with the high-fidelity flow solver Ellipsys3D LES and the aeroelastic tool FLEX5.…”

Real-time optimization of wind farms using modifier adaptation and machine learning

“…The development of surrogate models is an active research field. One of the most popular wake models is the Jensen Park model (Jensen, 1983;Katic et al, 1987). Jiménez et al (2010) developed one of the first steady-state wake models that described wake deflection due to yaw.…”