The total energy produced by a wind farm depends on the complex interaction of many wind turbines operating in proximity with the turbulent atmosphere. Sometimes, the unsteady forces associated with wind negatively influence power production, causing damage and increasing the cost of producing energy associated with wind power. Wakes and the motion of air generated by rotating blades need to be better understood. Predicting wakes and other wind forces could lead to more effective wind turbine designs and farm layouts, thereby reducing the cost of energy, allowing the United States to increase the installed capacity of wind energy. The Wind Energy Technologies Department at Sandia has collaborated with the University of Minnesota to simulate the interaction of multiple wind turbines. By combining the validated, large-eddy simulation code with Sandia’s HPC capability, this consortium has improved its ability to predict unsteady forces and the electrical power generated by an array of wind turbines. The array of wind turbines simulated were specifically those at the Sandia Scaled Wind Farm Testbed (SWiFT) site which aided the design of new wind turbine blades being manufactured as part of the National Rotor Testbed project with the Department of Energy.
A Verification and Validation (V&V) framework is presented for the development and execution of coordinated modeling and experimental program s to assess the predictive capability of computational models of complex systems through focused, well structured, and formal processes. The elements of the framework are based on established V&V methodology developed by various organizations including the Department of Energy, National Aeronautics and Space Administration, the American Institute of Aeronautics and Astronautics, and the American Society of Mechanical Engineers. Four main topics are addressed: 1) Program planning based on expert elicitation of the modeling physics requirements, 2) experimental design for model assessment, 3) uncertainty quantification for experimental observations and computational model simulations, and 4) assessment of the model predictive capability. The audience for this document includes program planners, modelers, experimentalist, V &V specialist, and customers of the modeling results.
Rotor design and analysis work has been performed to support the conceptualization of a wind tunnel test focused on studying wake dynamics. This wind tunnel test would serve as part of a larger model validation campaign that is part of the Department of Energy Wind and Water Power Program’s Atmosphere to electrons (A2e) initiative. The first phase of this effort was directed towards designing a functionally scaled rotor based on the same design process and target full-scale turbine used for new rotors for the DOE/SNL SWiFT site. The second phase focused on assessing the capabilities of an already available rotor, the G1, designed and built by researchers at the Technical University of München.
The objective of this document is to accurately predict, assess and optimize wind plant performance utilizing High Performance Modeling (HPC) tools developed in a community-based, open-source simulation environment to understand and accurately predict the fundamental physics and complex flows of the atmospheric boundary layer, interaction with the wind plant, as well as the response of individual turbines to the complex flows within that plant
In this paper, the effect of two different turbine blade designs on the wake characteristics was investigated using large-eddy simulation with an actuator line model. For the two different designs, the total axial load is nearly the same but the spanwise (radial) distributions are different. The one with higher load near the blade tip is denoted as Design A; the other is Design B. From the computed results, we observed that the velocity deficit from Design B is higher than that from Design A. The intensity of turbulence kinetic energy in the far wake is also higher for Design B. The effect of blade load distribution on the wind turbine axial and tangential induction factors was also investigated.
The dynamic wake meandering model (DWM) is a common wake model used for fast prediction of wind farm power and loads. This model is compared to higher fidelity vortex method (VM) and actuator line large eddy simulation (AL-LES) model results. By looking independently at the steady wake deficit model of DWM, and performing a more rigorous comparison than averaged result comparisons alone can produce, the models and their physical processes can be compared. The DWM and VM results of wake deficit agree best in the mid-wake region due to the consistent recovery prior to wake breakdown predicted in the VM results. DWM and AL-LES results agree best in the far-wake due to the low recovery of the laminar flow field AL-LES simulation. The physical process of wake recovery in the DWM model differed from the higher fidelity models and resulted solely from wake expansion downstream, with no momentum recovery up to 10 diameters. Sensitivity to DWM model input boundary conditions and their effects are shown, with greatest sensitivity to the rotor loading and to the turbulence model.
Langel, Christopher M.; Chow, Raymond; Hurley, Owen F.; Van Dam, C.P.; Ehrmann, Robert S.; White, Edward B.; Maniaci, David C.
Over time it has been reported wind turbine power output can diminish below manufacturers promised levels. This is clearly undesirable from an operator standpoint, and can also put pressure on turbine companies to make up the difference. A likely explanation for the discrepancy in power output is the contamination of the leading edge due to environmental conditions creating surfaces much coarser than intended. To examine the effects of airfoil leading edge roughness, a comprehensive study has been performed both experimentally and computationally on a NACA 633 - 418 airfoil. A description of the experimental setup and test matrix are provided, along with an outline of the computational roughness amplification model used to simulate rough configurations. The experimental investigation serves to provide insight into the changes in measurable airfoil properties such as lift, drag, and boundary layer transition location. The computational effort is aimed at using the experimental results to calibrate a roughness model that has been implemented in an unsteady RANS solver. Furthermore, a blade element momentum code was used to assess the impact on the performance of a turbine as whole due to discrepancies in clean vs. soiled airfoil characteristics. The results have implications in predicting the power loss due to leading edge surface roughness, and can help to establish an upper bound on admissible surface contamination levels.
New blade designs are planned to support future research campaigns at the SWiFT facility in Lubbock, Texas. The sub-scale blades will reproduce specific aerodynamic characteristics of utility-scale rotors. Reynolds numbers for megawatt-, utility-scale rotors are generally vary from 2-8 million. The thickness of inboard airfoils for these large rotors are typically as high as 35-40%. The thickness and the proximity to three-dimensional flow of these airfoils present design and analysis challenges, even at the full scale, but more than a decade of experience with the airfoils in numerical simulation, in the wind tunnel, and in the field has generated confidence in their performance. When used on a sub-scale rotor, Reynolds number regimes are significantly lower for the inboard blade, ranging from 0.7 to 1 million. Performance of the thick airfoils in this regime is uncertain because of the lack of wind tunnel data and the inherent challenge associated with associated numerical simulations. This report documents efforts to determine the most capable analysis tools to support these simulations and to improve understanding of the aerodynamic properties of thick airfoils in this Reynolds number regime. Numerical results from various codes of four airfoils are verified against previously published wind tunnel results where data at those Reynolds numbers are available. Results are then computed for other Reynolds numbers of interest.