Final Report Summary - ACTIVEWINDFARMS (Active Wind Farms: Optimization and Control of Atmospheric Energy Extraction in Gigawatt Wind Farms)
With the recognition that wind energy is becoming an important contributor to the world’s energy portfolio, large wind farms are expected to exceed capacities of 1 gigawatt in the future. At this scale, turbine wake interactions play an important role in wind-farm performance. In addition, farms of gigawatt size slow down the Atmospheric Boundary Layer (ABL) as a whole, reducing the availability of wind at turbine hub height. This leads to underperformance of turbines which can reach levels of 40%–50% compared to the same turbine in a lone-standing case. At the same time, turbulence levels increase, leading, e.g. to increased loading of turbine components. For large wind farms, the complex turbulent flow physics of the atmospheric boundary layer and the wake interactions are crucial ingredients in their design and operation. The major ambition of Active Wind Farms was the development of new control strategies that improve energy extraction in large wind farms, and reduce operating costs by better interacting with the turbulent flow in the boundary layer. To this end, the project envisaged the use of supercomputing, and the use of optimal control techniques that optimize the interaction between large wind farms and the ABL, taking into account the three-dimensional dynamics of the turbulent flow in the farm.
During ActiveWindFarms, the simulation and optimization platform SP-Wind of KU Leuven was extended and further developed. It consists of a highly efficient large-eddy simulations (LES) code of the atmospheric boundary layer, enriched with a flexible multi-body representation of the wind turbines, and allowing coupling to meso-scale atmospheric simulations. Moreover, an adjoint-based optimization framework was developed in the LES code, that is unique in the field, and allows for efficient optimization of wind farm turbulence interaction. Next to that, a wind-tunnel experiment of a large miniaturized wind farm was set-up and carried out for validation of simulation and optimization results.
Using the tools developed during ActiveWindFarms, it was shown that wind-farm energy extraction can be significantly increased by dynamic coordinated control of the turbines, leading to increased wake recovery, and power gains of up to 20% for some cases. Such dynamic approaches are much more effective than static set-point optimization, which is currently the main research paradigm for improving wind-farm performance. Moreover, novel control physics were discovered that increase wake mixing, and can be synthesized in simple open loop control strategies, i.e. based on vortex shedding using dynamic thrust control, or exciting wake meandering using dynamic turbine yawing.
Next to maximizing energy extraction, ActiveWindFarms also investigated the provision of ancillary services to the power grid by large windfarms, in particular secondary reserves were considered. To this end, new dynamic wake models were developed that can be used as control model for optimal control of power tracking. It was demonstrated that wind farms are very good at delivering such services when using these models, and may require less prior downrating than conventional power plants for same levels of grid support
Finally, during ActiveWindFarms, the simulation tool SP-Wind and its boundary conditions were continuously improved. As a result of this process, it was found that large wind farms can excite atmospheric gravity waves that may have a profound impact on power extraction. These are meso-scale phenomena at a scale that was previously considered too large to be influenced by wind energy installations. These findings will require the rethinking of siting, forecasting and design tools for large wind farms, and first steps in this direction were already accomplished at the end of the project.
During ActiveWindFarms, the simulation and optimization platform SP-Wind of KU Leuven was extended and further developed. It consists of a highly efficient large-eddy simulations (LES) code of the atmospheric boundary layer, enriched with a flexible multi-body representation of the wind turbines, and allowing coupling to meso-scale atmospheric simulations. Moreover, an adjoint-based optimization framework was developed in the LES code, that is unique in the field, and allows for efficient optimization of wind farm turbulence interaction. Next to that, a wind-tunnel experiment of a large miniaturized wind farm was set-up and carried out for validation of simulation and optimization results.
Using the tools developed during ActiveWindFarms, it was shown that wind-farm energy extraction can be significantly increased by dynamic coordinated control of the turbines, leading to increased wake recovery, and power gains of up to 20% for some cases. Such dynamic approaches are much more effective than static set-point optimization, which is currently the main research paradigm for improving wind-farm performance. Moreover, novel control physics were discovered that increase wake mixing, and can be synthesized in simple open loop control strategies, i.e. based on vortex shedding using dynamic thrust control, or exciting wake meandering using dynamic turbine yawing.
Next to maximizing energy extraction, ActiveWindFarms also investigated the provision of ancillary services to the power grid by large windfarms, in particular secondary reserves were considered. To this end, new dynamic wake models were developed that can be used as control model for optimal control of power tracking. It was demonstrated that wind farms are very good at delivering such services when using these models, and may require less prior downrating than conventional power plants for same levels of grid support
Finally, during ActiveWindFarms, the simulation tool SP-Wind and its boundary conditions were continuously improved. As a result of this process, it was found that large wind farms can excite atmospheric gravity waves that may have a profound impact on power extraction. These are meso-scale phenomena at a scale that was previously considered too large to be influenced by wind energy installations. These findings will require the rethinking of siting, forecasting and design tools for large wind farms, and first steps in this direction were already accomplished at the end of the project.