<< Click to Display Table of Contents >> Navigation:  Meteo Database > TMY data generation > TMY generation algorithms

There exists various norms and methods to generate TMY data, but the rule of having a minimum of 10 years and the statistical approach on which the generation is based is common between all methods.

Non-exhaustive list of TMY methods: ISO norm 15927-4_2005 (source), Sandia, NREL NSRDB TMY2/3 (source NREL/TP-581-43156), NREL NSRDB gridded TMY.

 

Common algorithm to generate TMY

•Take 10 or more years of hourly data for the same place and from the same source (e.g. data from the same source for 2011 to 2020).

•Compute daily maximum, minimum and mean of selected variables (cf. weight list below).

•Compute the cumulative distribution function (CDF) of each variable for each month:

oone cumulative distribution function for each variable, each month and each year of data (e.g. for the GHI: one for Jan. 2011, one for Jan. 2012, one for Jan 2013, ... and for each month, the same for TAmb, or other variables).

oone long-term cumulative distribution function for each variable and each month (e.g. one for GHI for January containing all daily values for 2011 to 2020).

•Compute the weighted sum (WS) of the Finkelstein-Schafer statistic (FS) for each variable:

oCompute FS, the sum over n days of a month the absolute difference between the long-term CDF and the candidate month CDF at value xi

oCompute WS, the weighted sum of FS for each month of each year

•Rank each months by lowest weighted sum WS (rank every January, every February, ...)

•Select each month based on various criteria defined in the different norms/methods

oThe final step for choosing months in the ISO norm is to compare the wind speed of the best 3 months from the ranked WS to the long-term average and choose the one with the lowest difference.

oFor the Sandia and NREL methods, the best 5 months from the ranked WS are re-ranked by their closeness to the long-term average and median. The 5 months are then filtered by analyzing the frequency and length of extrema in ambient temperature and global horizontal irradiance.

•Concatenate the selected months into a single continuous year (e.g. Jan 2015, Feb 2011, Mar 2017, etc...), interpolate the values of different variables at the month boundaries to smooth out discontinuities.

 

Variables weight for the FS statistics

* Although the wind velocity is not used in the CDF for the ISO norm, it is used as the final criteria