Heterogenous time resolutions

[hellemo]

We discussed supporting different time resolution (e.g. per technology) briefly at JuMP-dev last year with the Tulipa team. I was reminded about this topic by their recent paper: https://doi.org/10.1016/j.apenergy.2025.126267

I'm creating this topic to gather discussion around what could be a good way to support this in TimeStruct. Could we for example provide some kind of view/lens for interpreting a certain timestruct on an aggregated/disaggregated level? I remember noting earlier that we would need some support for calculating intersections and identify overlapping and non-overlapping time periods.

Ideas?

[JulStraus]

I have not really looked into it too much, but I know that SpineOpt allows it as well. I am personally not the biggest fan of differing temporal resolutions although it seems that the Tulipa team solved the issues associated with the common pitfalls regarding implicit storage, capacity constraints, and energy balance as outlined in Section 2.

One question I have is the key gain compared to potential flexibility imposed through a varying temporal resolution which is the same for all variables. We can achieve this through, e.g., SimpleTimes(7, [1, 2, 3, 4, 3, 2, 1]). This case is not compared in the publication (only for solar PV as single technology with a differentiated temporal resolution), but can also provide us with a significant benefit.

A more fundamental question is: Is that something we can include in TimeStruct or is something which must be supported on the model level?