Investment fails when only remaining options fail to dispatch

[alexdewar]

When selecting the best assets for a commodity, it is possible to end up with a situation where demand is not met by already-selected assets and we are unable to calculate metrics for any of the remaining options because they fail to dispatch (appraisal produces activity of zero in every time slice). This is bad. While profitable assets should clearly be preferred to unprofitable ones, unprofitable ones should be selected in preference to terminating the simulation.

The question is what to do with these dud options.

A few ideas have been floated:

1. Keep bumping the price of the CoI until assets start dispatching
2. Just pick assets arbitrarily until demand is met in partial dispatch
3. Calculate the price increase that would be required for each asset to dispatch and use that
4. Calculate what the metric would for each asset assuming full utilisation

[alexdewar]

Am I right in thinking we decided option 4 was probably easiest @tsmbland?

[ahawkes]

There should be a warning (at least) if this happens. I like (1) the best, but also don't want to slow processing down too much. Could bump it 25% and then 50% and then give up (clear error provided). I guess such behaviour should be configurable. This might be an edge case in most well calibrated models, but I guess foolish to assume how people will use the framework.

[alexdewar]

There should be a warning (at least) if this happens. I like (1) the best, but also don't want to slow processing down too much. Could bump it 25% and then 50% and then give up (clear error provided). I guess such behaviour should be configurable. This might be an edge case in most well calibrated models, but I guess foolish to assume how people will use the framework.

Yep, we'll definitely have a warning. I'll also hide this behind a model option for now, so that it doesn't obscure other problems (e.g. example models passing tests even though they're using these non-feasible options). Once we have an implementation we can start playing with it.

[alexdewar]

I've hit an issue with this. Bumping the price like this will work in cases where the reason for the non-dispatch is economic, but only if the shadow price is non-zero! What do we think would be a principled thing to do if it's zero? Should we try with the market price first, then increasing that if it still doesn't dispatch?