This python script takes two csv files containing weather data and compares the wind speeds and direction.
The mast_data.csv file contains one years wind speed and direction data from a weather mast placed at a potential wind farm site.
The reference_site_Wind_speed_dir_1980_2013.csv file contains weather data from a nearby met station from 1980 - 2013.
The aim of the exercise is to compare the data and make a long term (20 year) prediction for wind speed and direction at the proposed wind farm site.
The python script uses a number of libraries.
- scipy - for statistical functions
- numpy - for mathematical functions
- pandas - to help manage tables of data
The two csv files are loaded into pandas dataframes.
There a number of instruments on the mast including two anemometers at 60m. The maximum value at this height is used. The direction at 58m is used. Other data is ignored. The timestamps are parsed correctly.
The mast data is cleaned to remove any invalid data. For a number of days, the instruments recorded 0's across all instruments. This indicates the instruments are not working. This data is ignored.
Hourly data is available from the reference site. The mast takes measurements every 10mins. The mast data is averaged out across the hour.
The reference site data gives wind speed in knots. This is converted to m/s. The reference data is taken at 10m. This is extrapolated to 60m using a shear coefficient of 0.2 to give the equivalent wind speed at 60m.
The hourly mast data and extrapolated reference data are joined on the timestamp.
The linregress function from scipy is used to get the slope, intercept and pearson correlation coefficient.
The slope and intercept are used to predict 20 years of data at the mast site. The predicted wind speed annual averages are taken. The RMSE between the predicted and measured wind speed is calculated.
The wind direction bias can be calcuated by comparing the wind direction at the mast site with the wind direction at the reference site. Unfortunately there is an issue calculating the average wind direction per hour at the mast site. The average of the 6 values per hour may not be valid due to wrap around issues. This is
Based on these data files, the following values are calculated:
Pearson correlation coefficient: 0.77827227059 Slope: 0.721048002783 Intercept: 0.770927897863 Predicted mean wind speeds per year: ref_speed_ms extr_speed_ms site_speed 1980 5.188626 7.424764 9.228007 1981 5.405039 7.734443 9.657492 1982 5.594385 8.005393 10.033264 1983 6.104924 8.735957 11.046461 1984 6.296207 9.009678 11.426077 1985 6.571455 9.403549 11.972325 1986 6.848463 9.799939 12.522067 1987 6.094831 8.721515 11.026433 1988 6.135231 8.779326 11.106609 1989 6.311345 9.031340 11.456119 1990 6.638697 9.499771 12.105772 1991 5.166228 7.392713 9.183556 1992 5.074989 7.262153 9.002486 1993 5.022296 7.186751 8.897914 1994 5.184418 7.418741 9.219654 1995 4.998709 7.152998 8.851102 1996 5.161768 7.386330 9.174704 1997 5.090302 7.284064 9.032875 1998 5.235876 7.492377 9.321778 1999 5.213286 7.460051 9.276946 2000 5.042630 7.215848 8.938268 2001 4.773570 6.830831 8.404299 2002 5.135365 7.348548 9.122305 2003 4.965029 7.104803 8.784262 2004 4.915301 7.033643 8.685573 2005 4.984216 7.132259 8.822340 2006 4.915800 7.034357 8.686564 2007 4.734902 6.775499 8.327561 2008 4.950878 7.084553 8.756179 2009 4.902422 7.015214 8.660014 2010 4.614441 6.603122 8.088497 2011 5.035440 7.205559 8.923998 2012 4.850758 6.941285 8.557484 2013 5.154962 7.376592 9.161198 2014 5.140000 7.355181 9.131505 Measured mean wind speed from mast site: 8.47490142034 The Root Mean Square Error between predicted and measured windspeed: 2.9502990692 The Bias error in degrees: -8.88163985661 The long term uncertainty (standard deviation of predicted speeds): 1.18079234741
If temperature data was available, it would be good to remove mast data when the temperature is less than 0 as icing on instruments may be an issue.