Closed cmutel closed 5 years ago
I just don't understand what the expected behaviour is here.
All the available documentation states that direct links (so-called "activity links") should be subtracted from market production volumes, and by inference this would also apply to calculating the respective shares of there inputs.
However, if we take the case of market for hard coal coke factory (chosen for its simplicity), which has only RER and GLO (changed to RoW on model application) locations, we have the following result comparison:
("model" is Ocelot, "given" is the 3.5 release).
It should be simple to get the correct market shares. Here are the values in the master data:
It is quite simple to subtract 1.322 from 4 to get the ecoinvent release version numbers:
However, the German activity "coking" has a direct link to European coke factory construction, which Ocelot calculates would take up 0.09190926275992474 of the 1,32236940655721, leading to the different ratios shown above.
@ecoinvent I desperately need some clarification as to what you want the system model to be here. Should Ocelot follow the ecoinvent release numbers, or the ecoinvent documentation? From my reading of the documentation, activity links should be subtracted from both the production volumes of both the individual activities (1.23 instead of 1.32) and the total production volume of the market (3.91 instead of 4).
Note that the above table with a zero in CZ was due to an obsolete geodata bug, and can be ignored.
There are still some issues with particular market groups, but this specific example now works correctly. Will open a new issue if another failure pathway is discovered.
This leads to incorrect allocation of inputs, e.g. for UCTE
market group for electricity, high voltage
:First guess is that this is due to direct links.