grantbuster
commented
3 years ago "Bug" is a strong word here. after much discussion, we are planning on adding a feature to re-calculate the LCOE at several steps (multi-year mean, sc-aggregation). This will be best-practice to re-calculate the LCOE from the multi-year mean AEP but not considered a "bug" if done otherwise. New feature being tracked in #277. Closing this for now.
We have a very interesting and widely systematic bug in the way reV calculates its multi-year mean LCOE.
As it turns out, taking the multi-year mean of the annual capacity factor values and the multi-year mean of the annual lcoe_fcr values does not give you consistent results. That is to say that if you re-calculated the lcoe based on the multi-year cf_mean values, you would get a different result than the reV outputs. Some of the error is due to integer storage of capacity factor values, but we’ve debunked this as the sole cause of these issues. We think this is because of the structure of the lcoe equation, basically resulting in the fact that averaging the equation outputs is not the same as inputting a multi-year mean cf_mean point value. If you don’t believe me, there’s some raw data below you can test yourself.
You can see in the plots below that the error is a strong function of the inter-year capacity factor coefficient of variation. That is, sites with high year-to-year variability have large error in the averaging of the multi-year lcoe values.
In order not to be too alarmist, I’ve also included a histogram showing that the vast majority of errors are below 2%. Errors above 10% are extremely rare. The average error is 0.37%. I believe this is for all of land-based wind in conus. Note that the errors are calculated as 100 * (reV_lcoe – recalculated_lcoe / recalculated_lcoe).
This raises the question: how do we want to calculate LCOE values? Is the correct way to calculate LCOE based on the multi-year mean capacity factor? If so, is this amount of error acceptable?
Dramatic example with cf_cov=0.375 resulting in 21% error in the multi-year mean LCOE.
LCOE = (2880398.5 0.072 + 109017.09) / (8760 2430 / 1000 * cf_mean)