WindPredictor

Accurate wind power forecasts are crucial in order for wind farm operators to manage operations, meet regulatory requirements, reduce penalties and losses, and maximize revenue. Yet wind power production is difficult to forecast accurately. Wind power itself scales with the cube of the wind speed, and the response of a wind turbine is a complex power output curve. Depending on wind speed and the characteristics of the wind turbine, abrupt changes in wind speed (“ramps”) can cause large changes in power output. Complex terrain and meteorological phenomena such as low-level jets, land-sea breezes, and thunderstorm winds increase the forecasting challenge.

In 2008, Weather Decision Technologies, Inc. (WDT), and NanoWeather, Inc., became part of a consortium of academic, public, and private enterprises that was awarded a significant two-year grant from the State of Oklahoma’s Economic Development Generating Excellence (EDGE) program. The consortium is led by the University of Oklahoma’s Oklahoma Wind Power Initiative (OWPI). The grant was made to WDT and NanoWeather to conduct research into and to develop a wind energy forecast and assessment system in support of the state’s rapidly growing wind energy industry.

Specifically, WDT and NanoWeather are combining their areas of core expertise to produce WindPredictor™, a unified forecasting system ideally suited for wind power forecasting. WDT brings its knowledge and expertise in years of operational mesoscale forecasting, notably the Weather Research and Forecasting (WRF) Model. WDT was an early adopter of WRF and was the first company to deploy WRF internationally.

NanoWeather brings the Uncoupled Surface Layer (USL) Model, a microscale numerical weather prediction (NWP) model that focuses entirely on conditions in the lowest levels of the atmosphere. This forecast model regularly beats all other models, and the vast majority of human forecasters, at WxChallenge, the North American collegiate weather forecasting contest. It was awarded the Grand Prize at the 2006 Collegiate Inventors Competition.

The combination of mesoscale (WRF) and microscale (USL) models forms the core of WindPredictor™. Other components include: a forecast ensemble system; a web-based user interface; skill scores for recent forecasts; and a method for detecting and predicting wind ramps. WindPredictor™ can utilize observational data from many sources, including instrumented meteorological towers, sodars, lidars, thermodynamic profiling radars, and wind power measurements.

The Uncoupled Surface Layer (USL) Model

The USL model focuses entirely on conditions in the lowest portion of the boundary layer. It utilizes a mesoscale model for its upper and lateral boundaries. Its efficient design enables it to run at very high resolution (tens of meters). The USL model has the following desirable qualities:

• Capable of running efficiently at very high resolution (tens of meters)
• Accurately deduces land surface properties
• Employs realistic surface- and boundary-layer physics

Surface layer parameterizations in standard NWP models are typically based on Monin-Obukhov similarity theory. Such schemes require turbulence for numerical stability, but this is often unrealistic, particularly in stable nighttime conditions. The calm layer presents a difficult challenge for wind energy forecasting. The wind shear above this layer can generate intermittent turbulence, resulting in wind “spikes” throughout the night as well as extreme horizontal wind speed gradients. Knowing the presence and depth of this layer is essential for predicting the variability of nighttime winds on small spatial and temporal scales. Other models, which assume relatively steady turbulence throughout the night, are simply not capable of predicting this type of variability.

The USL model uses a continuously running, observation-based tuning process that directly calculates various soil and vegetation parameters. It thereby avoids lookup tables, which assume uniform soil and vegetation types across large areas. NDVI is utilized in the process. Derived tuning parameters include:

• Radiative cooling efficiency
• Soil conductivity, density, and specific heat
• Vegetation density
• Efficiency of vertical mixing based on surrounding terrain and vegetation
• Terrain blocking of wind by direction.

WDT and NanoWeather were invited to contribute USL forecasts to the WMO World Weather Research Programme project entitled the Science of Nowcasting Winter Weather for Vancouver 2010.

USL forecast of surface wind in the vicinity of Whistler, BC, Canada. Forecasts are produced on a 100 m grid.

For more information, contact

Brandon Wilkes (405) 801.3910 or
Dr. Richard Carpenter (405) 579. 7675 x229