Lognormal distribution for Monte Carlo Simulation

Question

Hello everyone, 

I wanted to use the Monte Carlo Simulation to verify the results for producing 1 kg of NCA cathode in China ("cathode production, NCA, for Li-ion battery | cathode, NCA, for Li-ion battery | APOS, U"). For that I used the EcoInvent v3.10 APOS Unit-Process and openLCA 2.2.0. I haven't changed any of the upstream processes for the simulation and also used the default TRACI LCIA Method. This leads to the shown diagramm for the GWP where I more or less only have a mean value but also negative values. This does not seems to be right but I also tried it for the whole NCA battery which leads to a similar result. Is this a bug, because I saw in the forum other monte carlo simultions which looked alright.

Answer 1

Unfortunately, these large outliers come from unrealistic high uncertainties inside the ecoinvent exchanges. In upcoming ecoinvent versions we will correct this by setting a threshold to remove unrealistic high uncertainties, as it is done by others, too. 

It is not a perfect solutions, since these values are inherent / original to the ecoinvent database and it would be better that these wrong values get corrected, but it will be a good intermediate solution until these too high uncertainties are updated by ecoinvent.

A threshold could be for example to remove the gsigma (geometric sigma in log-normal) values above 2.5. If you want to remove uncertainties above 2.5 in the mean time, you can run try the following SQL command and check your MCS results:

UPDATE TBL_EXCHANGES
SET PARAMETER2_VALUE = 1.0
WHERE PARAMETER2_VALUE > 2.5
AND DISTRIBUTION_TYPE = 1;