I think the facility count curves for edu, health, maybe street lighting? should default to zero to avoid residual connection costs working into calcs even if edu and health demand is zeroed out #105
I wanted to note here an NP issue that I just noticed, and I think we should be careful about. The problem relates to grid initial internal cost calculations. Specifically, the results you get from a seemingly correct estimate of initial internal costs disagree with NP results by a few percent. It's not a huge deal, but it's big enough to notice when costs don't add up -- maybe 3% of initial internal costs, so probably less than 1% of total initial. We've kind of waived it away in the past since we're almost always more concerned about MV costs than local/LV costs. But now that I know what it is, I think it's avoidable. But a note to Chris: I think it may justify a change in the default settings.
The essence is: even when you zero-out the social infrastructure DEMAND settings, there is still a few logistic "facility count" curves running in the background, estimating a number of schools, clinics and such. So even if the demand per facility is set to zero, NP will still estimate that these facilities exist, and will include them for initial costs, like connections. So these facility count numbers work their way into initial internal costs through things like connection costs. It's hard to see where this happens, since facility connection costs are buried within other calcs, and so you don't see them until you really track these variables 2-3 layers into the NP code. [I give a bit more detail below]
So... a couple options:
1) each time you do a network planner run with "zero" social infrastructure, you need to manually zero out the logistic facility count curves. the defaults for social infrastructure are now set to zero, but that's not enough.
2) another option would be to have Chris re-set the defaults for these to make all the second numbers in the facility count curve number pairs zero, flattening out the curve.
3) just ignore the problem. as I say, results are only off by about 1-2%. so no one (including us) really notices. but in the rare case that you're doing a breakdown of all the grid initial costs, if you compare what seems to be the sensible result with what NP gives you, the difference is noticeable, and can be distracting. so if we're in a meeting with Vijay, and he notices, now we know....
Edwin
detail:
The grid initial cost has 4 consituents:
1) transformer cost (varies a bit by settlement)
2) cost of LV line equipment per connection (constant--should equal the input parameter)
3) costs of LV line per connection (constant -- should equal the input parameter X mean inter-HH distance)
4) installation costs per connection (SHOULD be constant, and SHOULD track with the number of target HHs, but in practice doesn't quite agree -- this is what I was working to figure out.)
for this last point: the number of connections turned out to be higher than expected, sometimes by 5-10%. I'm pretty sure this is due to the facility count curves for schools, clinics, etc.
I wanted to note here an NP issue that I just noticed, and I think we should be careful about. The problem relates to grid initial internal cost calculations. Specifically, the results you get from a seemingly correct estimate of initial internal costs disagree with NP results by a few percent. It's not a huge deal, but it's big enough to notice when costs don't add up -- maybe 3% of initial internal costs, so probably less than 1% of total initial. We've kind of waived it away in the past since we're almost always more concerned about MV costs than local/LV costs. But now that I know what it is, I think it's avoidable. But a note to Chris: I think it may justify a change in the default settings.
The essence is: even when you zero-out the social infrastructure DEMAND settings, there is still a few logistic "facility count" curves running in the background, estimating a number of schools, clinics and such. So even if the demand per facility is set to zero, NP will still estimate that these facilities exist, and will include them for initial costs, like connections. So these facility count numbers work their way into initial internal costs through things like connection costs. It's hard to see where this happens, since facility connection costs are buried within other calcs, and so you don't see them until you really track these variables 2-3 layers into the NP code. [I give a bit more detail below]
So... a couple options:
1) each time you do a network planner run with "zero" social infrastructure, you need to manually zero out the logistic facility count curves. the defaults for social infrastructure are now set to zero, but that's not enough.
2) another option would be to have Chris re-set the defaults for these to make all the second numbers in the facility count curve number pairs zero, flattening out the curve.
3) just ignore the problem. as I say, results are only off by about 1-2%. so no one (including us) really notices. but in the rare case that you're doing a breakdown of all the grid initial costs, if you compare what seems to be the sensible result with what NP gives you, the difference is noticeable, and can be distracting. so if we're in a meeting with Vijay, and he notices, now we know....
Edwin
detail:
The grid initial cost has 4 consituents: 1) transformer cost (varies a bit by settlement) 2) cost of LV line equipment per connection (constant--should equal the input parameter) 3) costs of LV line per connection (constant -- should equal the input parameter X mean inter-HH distance) 4) installation costs per connection (SHOULD be constant, and SHOULD track with the number of target HHs, but in practice doesn't quite agree -- this is what I was working to figure out.)
for this last point: the number of connections turned out to be higher than expected, sometimes by 5-10%. I'm pretty sure this is due to the facility count curves for schools, clinics, etc.