Allow agents to decommission part of an asset

[ahawkes]

Is your feature request related to a problem? Please describe

At the time of writing, MUSE2 considers the full capacity of existing assets in investment appraisal. i.e. it takes all or nothing of an existing asset. This means that (a) a large existing asset can take most of the demand before other assets are really considered, and (b) it is possible to end up with substantial over-capacity (more than required to meet demand) if e.g. demand is reducing or the existing asset is one of the last to be adopted and is retained even though only a fraction of its capacity will actually be used.

This is not necessarily a mistake, because:

• the modeller could specify asset capacities as input to indicate the ability to decomission "chunks" of existing capacity.
• some assets are actually indivisible. For example, a big CCGT gas turnine is either decommissioned or not, you can't decommission part of it. Conversely, a fleet of gas boilers is dvisible and part of the "existing asset" could indeed be decommissioned.

Describe the solution you'd like

To cater for these situations, it would be convenient to specify whether a process type is divisible or not, and to handle partial decommissioning during investment appraisal. A related feature already in the docs is mothball_years, which is the maximum number of years an asset can remain unused before it must be decommissioned permanently.

Proposed solution is a new field in the "processes.csv" that specifies whether an asset is divisible. If it is, investment appraisal should consider the asset in tranches, similar to the way it does at present for candidate assets. Where part of the asset is not used, it should be mothballed. Where part of the asset is not used for mothball_years, it should be decommissioned (and can never come back after that).

Describe alternatives you've considered

I think this could largely be handled in input data if the modeller is careful. i.e. you could specify existing assets in chunks to enabled such behaviour.

Additional context

At the moment I think tranche size is determined via a fixed portion of required capacity (10%?). To be consistent with the suggestion above, this parameter may need to be in input data. Initial thought here is that such a parameter - capacity_tranche_size - should be associated with each commodity (commodities.csv). Then, in investment appraisal, this is the size that is used for both candidate and existing assets (where the later is divisible).

[alexdewar]

This seems interesting. Just to note that there is already an issue open for adding the mothball_years param (#649). I feel like it probably makes sense to do that part first, before we move on to making assets divisible.

I'm wondering if making this a boolean parameter (divisible vs indivisible) is the right way to do this. Presumably the thought was that we could use the existing capacity_limit_factor param to determine the size of the chunks in which assets can be decommissioned? Another way we could do this is to add a column to assets.csv called e.g. duplicates (or something) that would allow for having multiple instances of the same kind of asset, all with the capacity specified in the capacity column. This would be more flexible, but it depends whether we think it makes sense to have assets be decommissioned in chunks in any size other than given by the capacity_limiting_factor param. Users could also achieve something like this by manually adding a bunch of different assets with the same capacity to assets.csv, but it might be a nice convenience.

[alexdewar]

I've marked this issue and subissues as "on hold" pending discussion. If we do decide we want to do this for this engagement, we should work out what we want the feature to look like before anyone starts work on it.