Improving wind farm flow models by learning from operational data

Abstract

This paper describes a method to improve and correct an engineering wind farm flow model by using operational data. Wind farm models represent an approximation of reality and therefore often lack accuracy and suffer from unmodeled physical effects. It is shown here that, by surgically inserting error terms in the model equations and learning the associated parameters from operational data, the performance of a baseline model can be improved significantly. Compared to a purely data-driven approach, the resulting model encapsulates prior knowledge beyond the one contained in the training data set, 5 which has a number of advantages. To assure a wide applicability of the method -including also to existing assets-learning is here purely driven by standard operational (SCADA) data. The proposed method is demonstrated first using a cluster of three scaled wind turbines operated in a boundary layer wind tunnel. Given that inflow, wakes and operational conditions can be precisely measured in the repeatable and controllable environment of the wind tunnel, this first application serves the purpose of showing that the correct error terms can indeed be identified. Next, the method is applied to a real wind farm situated 10 in a complex terrain environment. Here again learning from operational data is shown to improve the prediction capabilities of the baseline model. IntroductionKnowledge of the flow at the rotor disk of each wind turbine in a wind power plant enables several applications, including wind farm control, the provision of grid services, predictive maintenance, the estimation of life consumption, the feed-in to digital 15 twins and power forecasting, among others. This paper describes a new method to estimate turbine inflow within a wind farm. The main idea is to use an existing wind farm flow model to provide a baseline predictive capability; however, as all models contain approximations and may lack some physical phenomena, the baseline model is improved (or "augmented", which is the term used in this work) by adding parametric correction terms. In turn, these extra elements of the model are learnt by using operational data. The correction 20 terms capture effects that are typically not present in standard flow models (as, for example, secondary steering or wind farm blockage (Bleeg et al., 2018)), or that are highly dependent on a specific site or difficult to model upfront (as, for example, non-uniform inflow caused by local orography and vegetation).Various wind farm flow models have been developed and are described in the literature. While Direct Numerical Simulation (DNS) is still out of reach for practical applications due to its overwhelming computational cost, Large Eddy Simulation (LES) 25 methods are now routinely used for the modeling of wind farm flows Breton et al., 2017). Although 1 https://doi.org/10.5194/wes-2019-91 Preprint. Discussion started: 2 December 2019 c Author(s) 2019. CC BY 4.0 License.invaluable for the understanding of the behavior of the atmospheric boundary layer and of wakes, LE...

Field experiment for open-loop yaw-based wake steering at a commercial onshore wind farm in Italy

“…This leads to an underestimation of the potential benefits of wake steering and consequently to suboptimal yaw misalignment setpoints. Historical operational data may also be used to reduce the model-plant mismatch (Schreiber et al, 2019).…”

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

“…As shown in Figure A1(b,c), the EnKF state estimation combined with the lifting line model are able to reproduce the power production for the artificial data to high accuracy. The ability for a one or two parameter analytic wake model to capture arbitrarily generated power production profiles should be investigated in future studies as the model may enforce unrealistic model parameters to represent neglected physics (Schreiber et al, 2019). The validity of this data-driven framework is validated in the LES test cases in a comparison between model power predictions and LES power measurements (Section 5).…”

“…Shapiro et al (2019) proposed the use of canonical turbulent wake mixing and a prescribed mixing length model to estimate k w . Schreiber et al (2019) utilized a data-driven approach where error terms are added to the engineering model and SCADA is used for data assimilation to correct the wake model inaccuracies. Gradient optimization-based SCADA data assimilation was used by to select the model parameters which minimize the model error in producing the site-specific wind farm greedy baseline power production.…”

Field testing of a local wind inflow estimator and wake detector

“…Even though these measurements might be accurate enough for these simple tasks, the actual complexity of the turbine inflow remains completely beyond the reach of such sensors. In addition, wind vanes and anemometers provide pointwise information, while wind conditions exhibit significant spatial variability not only at the large scale of the farm, as in off-shore plants (Peña et al, 2018) and at complex terrain sites (Lange et al, 2017;Schreiber et al, 2019), but also at the smaller scale of the individual turbine rotor disk (Murphy et al, 2019). More sophisticated measurements can be provided by lidars (Held and Mann, 2019) and other remote sensing technologies, which are however still costly and -being mostly used for assessment, validation and research-are not yet commonly used for production installations.…”