Online model calibration for a simplified LES model in pursuit of real-time closed-loop wind farm control

Abstract

Abstract. Wind farm control often relies on computationally inexpensive surrogate models to predict the dynamics inside a farm. However, the reliability of these models over the spectrum of wind farm operation remains questionable due to the many uncertainties in the atmospheric conditions and tough-to-model flow dynamics at a range of spatial and temporal scales relevant for control. A closed-loop control framework is proposed in which a simplified dynamical LES model is calibrated and used for optimization in real time. This paper presents an estimation solution with an Ensemble Kalman filter (EnKF) at its core, 5 which calibrates the surrogate model to the actual atmospheric conditions. The estimator is tested in high-fidelity simulations of a nine-turbine wind farm. Using exclusively turbine SCADA measurements, the adaptability to modeling errors and changes in atmospheric conditions (TI, wind speed) is shown. Convergence is reached within 400 seconds of operation, after which the estimation error in flow fields is negligible. At a low computational cost of 1.2 s on an 8-core CPU, this algorithm shows comparable accuracy to the state of the art from the literature while being approximately two orders of magnitude faster. Using 10 the calibration solution presented, the surrogate model can be used for accurate forecasting and optimization.

Optimal closed-loop wake steering, Part 1: Conventionally neutral atmospheric boundary layer conditions

“…Here, we will employ the EnKF (Evensen, 2003) Doekemeijer et al, 2018;Mandel, 2009). The states and dimensions here represent the wake model instantiations and parameters, respectively.…”

“…The EnKF filter has been successfully used for low-order model state estimation for the purpose of receding horizon frequency regulation control (Shapiro et al, 2017) and reference power signal tracking applications (Shapiro et al, 2019). Doekemeijer et al (2018) found that the EnKF has comparable state estimation performance given either nacelle-mounted LiDAR data or Supervisory Control and Data Acquisition (SCADA) power production data alone. Since very few utility-scale wind turbines have nacelle-mounted LiDAR systems, the successful performance of the EnKF based on SCADA data alone highlights the potential for online model calibration without additional hardware installation.…”

A tutorial on the synthesis and validation of a closed-loop wind farm controller using a steady-state surrogate model

“…In some cases, a simplified state estimation algorithm has been applied for dynamic surrogate models, such as a linear Kalman filter (e.g., [18], [27]). More recently, there have been positive developments in the field of real-time model adaptation, using more sophisticated estimation algorithms that attempt to balance accuracy with computational efficiency (e.g., [28], [29]). In terms of optimization, for steady-state surrogate models, a gradient-based or nonlinear optimization algorithm is typically employed to determine the optimal steady-state control settings for the wind farm (e.g., [8], [30], [31]).…”

“…Large-eddy simulation models such as SOWFA resolve larger scale flow dynamics directly, and employ a subgridscale model for smaller eddy dynamics. SOWFA has been used on multiple occasions for surrogate model calibration (e.g., [8], [20], [44]), model validation, and wind farm controller verification (e.g., [29], [34], [35], [44]).…”

Wind Farm Modeling with Interpretable Physics-Informed Machine Learning

“…While such analytic models are lower in reliability than high fidelity simulations, they can accurately capture wind farm power production trends [4]. Low-order models are particularly relevant for the purpose of online control of wind farm power as a function of time for energy grid optimization (see, e.g., [42][43][44]) due to the computational expensive of online large-eddy simulations [45]. The physics-based, low-order model selected here is the Gaussian wake model [7] with the momentum theory prescribed by Prandtl's lifting line model [8].…”