Merit order and profitability fuctionalities expose shortcomings in modeling approach of 'state-of-the-art' technologies

[AlexanderWirtz]

This is actually not a merit ticket, but a result of this new functionality.

Problem statement

The dataset for the ETM is not completely realistic (surprise!).

1. We have 'state-of-the-art plants', which are more efficient than many real plants in NL, and 'conventional plants', which are less efficient than many real plants in NL.
2. In order to still reproduce the energy balance and the electricity balance, we make a mixture of the plants mentioned above to get the correct electrical efficiency for gas plants (.i.e the right 'transformation losses' for gas power plants in NL)
3. This works fine ost of the time, but now we have a merit order and profitability module, these plants are no longer realistic. A large percentage of plants (Gas conventional) are unprofitable, but removing them results in huge 'loss of load probability'. Of course some of the 'conventional' plants are really something between 'state-of-the-art' and 'old crap'.

   Discussion of solutions

Updating the dataset to represent the actual plant efficiencies is not an easily implementable solution:

1. It would require adding many slightly different plant types and sliders
2. It would effectively not help that users can then build all these 'sub optimal' plants in the future too.

Splitting the large blocks of plant types into individual slightly different plants in merit only is also not likely to make this problem go away entirely, but it will make users less keen to eliminate all unprofitable plants.

An entire re-think of datasets and modeling is needed, I'm afraid. @wmeyers @ChaelKruip @dennisschoenmakers @JAlsem @Richard-Deuchler I want you to be aware of this issue.

[ChaelKruip]

Here's an out-of-the-box idea: perhaps the ETM should allow the user to build custom plants. We could include an extra plant (or even more than one) for the main carrier groups (coal, gas, ... ) with zero number_of_units and all other properties identical to the default plant in that category. The plant could have (in a separate slide?) several sliders that allow the user to define its most important properties:

• capacity / number of units
• efficiency
• marginal costs
• availability (?)
• ...

This would allow the expert user to alleviate the most blatant discrepancies between the ETM and reality but without changing the default dataset. Instead of a research challenge, we then have a technological challenge. But I think that making converter properties updatable with input statements is possible in principle.

[wmeyers]

Or: allow users to tweak the properties of all plants in the ETM. We could make a separate button for that next to each relevant slider, which opens a pop-up or dropdown in which you could have multiple sliders to alter the plant's properties (and even slides for different kind of properties like efficiencies, costs, etc).

[dennisquintel]

Also note that in the merit order, we want to use a distribution of plant_types, in order to make prices calculations more 'accurate', so, let's take that into account too.

I think we should put a high level project plan together in really think about how/what/when we (potentially) want to do with this and how this fits with any of our clients/partners.

[ChaelKruip]

I think we should put a high level project plan together in really think about how/what/when we (potentially) want to do with this and how this fits with any of our clients/partners.

+1

[Richard-Deuchler]

This issue may also be relevant to the discussion. https://github.com/quintel/merit/issues/65