SILSOE RESEARCH INSTITUTE Spatial variation of wind speeds between sites Andrew Quinn
SILSOE RESEARCH INSTITUTE 2 Why consider wind? Many problems require a knowledge of the natural environment Design and construction, Transport risk analysis Pollution control, Pollen and aphid movement Wind Power, Forestry management, etc etc Some require single design values Others require distributions
SILSOE RESEARCH INSTITUTE 3 Wind speed distribution How does this change with site? Data from Shap Meteorological site, hourly maximum gust 4/1994 – 3/2003
SILSOE RESEARCH INSTITUTE 4 General form of the CDF Weibull distribution Where c and k are site specific constants Fits both mean hourly data and gust data
SILSOE RESEARCH INSTITUTE 5 Extreme value analysis Majority of previous studies Typical approach is to take linear approximation to the tail of the distribution Where LHS is known as reduced variate
SILSOE RESEARCH INSTITUTE 6 Gumbel plot Data from Shap meteorological site - hourly mean and gust up to 1 year return period
SILSOE RESEARCH INSTITUTE 7 Extreme value analysis Shap data hourly mean and gust up to 1 year return period Gumbel plot of Shap gust extrapolated to 50-year return period Peak values Such approaches require long data sets Therefore not local to sites of interest
SILSOE RESEARCH INSTITUTE 8 Wind Speed map methods BS6399:Part2, Eurocode 1, ESDU Wind Atlas method Miller et al (1998) Fig. 5. Estimated 50-year return hourly-mean wind speed for the United Kingdom. Values given in m/s. After Abild et al (1992) Design wind speed methods for dealing with spatial effects
SILSOE RESEARCH INSTITUTE 9 Objective Obtain a wind speed distribution for a site: Objective (no subjective estimates of parameters) Realistic (rather than a conservative design value) Based on short-term data records Consistent with other methods (EVA)
SILSOE RESEARCH INSTITUTE 10 Approach Consider the wind distribution at two sites Where c A and k A known (from long records) Define U B + such that
SILSOE RESEARCH INSTITUTE 11 Approach For consistency with standard EVA where assume k A = k B (i.e. distributions same shape) General form E(U B + ) E(γU A )
SILSOE RESEARCH INSTITUTE 12 Evaluating γ from short term records using ranked simultaneous data 9 years data 3 months (winter) data
SILSOE RESEARCH INSTITUTE 13 Reliability of short term data
SILSOE RESEARCH INSTITUTE 14 Solution Thus we know Long term wind speed probability distribution from a reference site Can calculate expected wind speed & return period Relationship between two sites Objective Realistic Small data set
SILSOE RESEARCH INSTITUTE 15 Example anomalous results
SILSOE RESEARCH INSTITUTE 16 including wind direction
SILSOE RESEARCH INSTITUTE 17 Conclusions Method for objective, realistic estimates based on short term site data Data from 8 sites used and estimates for hourly mean and gust wind speeds Accuracy level similar to direct MO records Wind direction can be a significant factor
SILSOE RESEARCH INSTITUTE Spatial variation of wind speeds between sites Acknowledgements Roger Hoxey, Chris Hampson, Nick Teer and the other members of the project team at SRI Russell Pottrill, William Bradbury and David Deaves (Atkins)