WIP: sketch of 2 variants to add commodity flow of capacity variables

[volker-krey]

The PR sketches two alternative variants that allow connecting a commodity flow to capacity addition and retirement.

• variant 1: commodity input and output strictly associated with construction and retirement of capacity (commodity flow is strictly triggered in periods when new capacity comes online or is retired, in case of new periods being added, no user action is required)
• variant 2: commodity input and output associated with operation of capacity at any period (more flexible as commodity flows can also be triggered before construction or after retirement, but requires adjustments by user in case new periods are being added to a model)

Comments on the two variants of adding commodity flow of capacity variables are appreciated.

[GamzeUnlu95]

For variant 2, for the case of retirement, the output value in the year (vintage + lifetime) will specify the material that is released upon retirement. I was wondering with the current formulation, isn't it the case that if the capacity is retired earlier than its lifetime, the material is still going to be released at the end of its lifetime even though it is retired earlier?

[volker-krey]

@GamzeUnlu95 good point. The implications of premature retirement for continuous commodity flow could be addressed in variant 2 by switching from the CAP_NEW to the CAP variable (equivalent to fix_cost), but to ensure that end-of-life release of commodities is treated correctly, we either have treat commodity flow beyond the lifetime in a special way or switch to variant 1.

[volker-krey]

@GamzeUnlu95 please have a look at the revised formulation of variant 2. This now consistently deals with commodity flows during lifetime, but the issue of commodity flows after end-of-life remains unsolved. The question is whether this is an acceptable error.

[volker-krey]

One issue that still needs attention are commodity flows upon capacity retirement if the vintage period (when the capacity was built) has a different length than the retirement period (period after last available of capacity). Since the CAP variables are normalized to annual construction, in such a case, the flow triggered upon retirement can be larger or smaller than intended. For the global model this is relevant in the transition from 5- to 10-year steps after 2060. A rescaling will need to be introduced.

[khaeru]

In today's MESSAGE meeting, regarding the new items (sets, variables, equations, and parameters) please look at #190, in particular the additions to message_ix.models.MESSAGE_ITEMS:

message_ix/message_ix/models.py

Lines 20 to 44 in 3203190

	# Common indices for some parameters in MESSAGE_ITEMS
	_idx_common = ['node', 'technology', 'level', 'commodity', 'year', 'time']
	# NB only a partial list; see https://github.com/iiasa/message_ix/issues/254
	#: List of ixmp items for MESSAGE.
	MESSAGE_ITEMS = {
	# Index sets
	# Storage level
	'level_storage': dict(ix_type='set'),
	# Storage reservoir technology
	'storage_tec': dict(ix_type='set'),
	# Mapping set: mapping of storage reservoir to charger/discharger
	'map_tec_storage': dict(ix_type='set',
	idx_sets=['node', 'technology', 'storage_tec',
	'level', 'commodity']),
	# Parameters
	# Order of sub-annual time steps
	'time_order': dict(ix_type='par', idx_sets=['lvl_temporal', 'time']),
	# Initial amount of storage
	'storage_initial': dict(ix_type='par', idx_sets=_idx_common),
	# Storage losses as a percentage of installed capacity
	'storage_self_discharge': dict(ix_type='par', idx_sets=_idx_common),
	}

New additions due to this PR should be handled in the same way.

[Jihoon]

I've tested the functionality of this branch with the main MESSAGEix-Materials model, and it is working as it is expected without generating additional issues.

[Jihoon]

This one is now superseded by #451.