Meaning of milestone years?

[alexdewar]

I've got some questions about milestone years and I thought I'd put them here so we have a record of any responses for posterity. It hasn't been an issue up until now, but I'm currently stuck working on demand-related issues because I don't think I entirely understand their meaning.

In the glossary, milestone years are described as:

Milestone Years: A set of years in the model time horizon where outputs are recorded. For
example, with a 2025 Base Year and Time Horizon to 2100, a user might choose to record outputs in
5-year steps.

I've got a few questions:

1. Even if output is only recorded for milestone years, does that mean that the simulation runs for other years? I.e. if your milestone years are 2020 and 2030, will the simulation only run for those two years? Or will it run for all the intervening years and only record data for those two? If the latter, is the optimisation performed for each year individually or for all years in between milestones simultaneously?
2. There are a number of places in the code where we validate "year" fields by checking whether it is after the first milestone year and before the last (e.g. commodity costs, demand etc.), but I'm now not sure if this is correct -- at least not in every case. Instead, should we be checking that the year specified belongs to the set of milestone years? E.g. does it make sense for a user to specify demand for the year 2021 if the milestone years are 2020 and 2030?

[ahawkes]

Hi Alex, My thoughts:

1. The simulation will only run for the milestone years. This could be revisited later on as I would expect simulation to be less realistic with big gaps between milestone years, but for now it's OK (and is the approach in all other similar modelling frameworks). Also there are some issues such as what it technology lifetime is less than the gap between milestone years.
2. Yeah I think we should not do this validation. Year field can indeed be before or after the simulated period. In some cases such data would be important (e.g. the commissioning year of an asset being before the base year would be almost universally true - and determines when that asset is decommissioned). In other cases the model would ignore it (e.g. a demand past the last milestone year).
3. Related to point (2), we need rules for interpolation and extrapolation where specific inputs are entered for non milestone years, or a milestone year is missed out etc. For now, this can be only for the parameter listed below, and can we (a) interpolate linearly between input data points (e.g. inputs are 2020 = 2, and 2050 = 8. Milestone years are 2020, 2030, 2040 and 2050. Interpolated values therefore are 2, 4, 6, 8 for respective milestone years). And where we need to extrapolate we (b) take the last value input and extrapolate flat-forward, and (c) assume zero in milestone years up to the first value provided. There can be a lot of different interpolation and extrapolation rules, so a placeholder issue to revisit this would be useful! Also user needs to be able to configure how these are applied, in some later version of MUSE2.

Inputs that need interpolation and extrapolation

• Demand
• Commodity cost
• Process investment constraint values
• commodity constraint values
• Some agent-related parameters will need this, but I'm restructuring input tables here so let's leave that for now.

[alexdewar]

Hi Adam. Thanks for the reply! There's lots of useful information in there.

I think what I'll probably do for demand and commodity cost -- which are the only inputs we have code for currently -- is to constrain it (for now) so that users must supply values for each milestone year and forbid them from supplying values from other years, because that's what'll be easiest to do on the software side. I'll open issues re interpolation and allowing out-of-range years and then we can lift the restriction when we write the code to add these features.

[alexdewar]

I think this discussion is resolved.