Boundary layer scaling of operational forecast data as a wind resource assessment tool for small and medium-scale wind turbines

Small and medium-scale wind energy is a renewable energy technology with exciting prospects as we move towards a low-carbon energy future. In order to achieve widespread deployment, however, it is vital to develop tools capable of predicting the wind resource quickly, cheaply and accurately. These tools will ensure that turbines are installed in locations which maximize the financial and environmental benefit and will allow customers to choose between competing distributed energy technologies with confidence. In this work we present the results of applying a boundary layer scaling approach to wind resource assessment driven by output from a numerical weather prediction model. The approach is capable of predicting a complete distribution of wind speeds for a prospective site using simple parameterisations of the surface roughness and without recourse to onsite measurements. The performance of the approach is evaluated using long-term wind speed measurements at 22 UK sites located in a range of terrains and its potential as a tool for small and medium-scale wind resource assessment is investigated.

Wind resource assessment is well established in the large-scale wind industry and typically involves the use of on-site anemometry to collect wind data over a period of 1-3 years [1]. These data are used to make reliable predictions of the long-term wind energy resource as required by investors. In the small-scale wind industry these timescales are often impractical and the impact of such a measurement campaign on the total investment cost may be prohibitively high. In the absence of long-term onsite measurements, indirect methods capable of low-cost and rapid deployment must be used.

Boundary layer scaling approaches that take account of the regional (>1 km) and local (< 500 m) surface roughness provide a simple and low-cost route for making preliminary wind resource assessments without the burden of long-term onsite measurements [2]. Typically, such approaches use a vertical, logarithmic wind speed profile to scale an input mean wind speed obtained from a large-scale wind atlas. The vertical wind speed profile is then used to predict the spatially and temporally averaged mean wind speed at a specific height and location, given the local and regional surface roughness. This approach was previously used to investigate the aggregated small-scale wind energy potential within the UK using a wind atlas input of gridded, 1 km2 mean wind speeds, along with local and regional land cover data [3]. A limitation of such an approach is that the stochastic variability of the wind speeds, as represented by the wind speed probability distribution, is not predicted using a single mean wind speed input. However, due to the cubic relationship between wind speed and wind power, predicting this variability is of vital importance in estimating the wind energy resource. One way of overcoming this limitation is to apply a boundary layer scaling approach to a long-term time-series of wind speeds, such as may be obtained from numerical weather prediction models. Providing the forecast data is capable of predicting the underlying variability, or shape of the wind speed probability distribution, a boundary layer scaling approach can be used to obtain a corrected time-series of wind speeds representative of the local climate.

In this work we investigate the performance of such an approach using operational forecast data from the Met Office Unified Model (UM). The UM is a world leading operational weather and climate forecast model, currently operated at resolutions of 25 km globally and at 1.5 km (previously 4 km) within the UK [4]. As a terrain-following, mesoscale model, the UM is capable of producing, local, site-specific forecasts through progressively higher resolution (12 km, 4 km and 1.5 km) models whose boundary conditions are provided by the global model. However, close to the surface, in the region where small and medium-scale wind turbines operate, such models are known to underestimate wind speeds and hence they must be used in conjunction with downscaling methods to account for the local surface characteristics [5]. For the current application, long-term (10 years) output from the 4 km resolution version of the model (UK4) is used as a reference climatology for implementation of the boundary layer scaling model. The model is used to downscale the hourly mean wind speeds using a two-stage approach: (i) regional downscaling based on estimates of the combined regional surface roughness from multiple patches of land cover and (ii) local downscaling based on estimates of the local surface roughness. Surface roughness is estimated from a database of high resolution land cover data by applying appropriate parameterisations [2, 3]. Particular attention is given to the performance of the forecast data in reproducing the form of the observed distribution of wind speeds as well as the averaged statistics of mean wind speed and mean wind power density. The impact of the forecast height is investigated by using multiple forecast heights between 10 and 500 m as input to the boundary layer scaling model. In addition, the effect of the terrain type, (urban, suburban, coastal and rural) on the accuracy of the final wind resource predictions is investigated on a site-by-site basis. This study is expected to be of interest in the field of small and medium-scale wind energy where the use of mesoscale models and downscaling techniques are receiving increasing interest as a low-cost route to wind resource assessment.

1. AWS Scientific Inc. & National Renewable Energy Laboratory (U.S.), Wind Resource Assessment Handbook - fundamentals for conducting a successful monitoring program. 1997.

2. Weekes, S.M. and A.S. Tomlin, Evaluation of a semi-empirical model for predicting the wind energy resource relevant to small-scale wind turbines. Renewable Energy, 2013. 50: p. 280-288.

3. Best, M., et al., Small-scale wind energy - technical report, UK Met Office. 2008.

4. Davies, T., et al., A new dynamical core for the Met Office's global and regional modelling of the atmosphere. Quarterly Journal of the Royal Meteorological Society, 2005. 131(608): p. 1759-1782.

5. Howard, T. and P. Clark, Correction and downscaling of NWP wind speed forecasts. Meteorological Applications, 2007. 14(2): p. 105-116.