`is_time_variant` vs `is_time_variant_term`

[luke-kiernan]

Just realized I've introduced a bug in the lk/mbc_iec_refactor branch. I failed to recognize the distinction between is_time_variant and is_time_variant_term. Example: if x::TimeSeriesIncrementalCurve has x.initial_input = nothing, then:

• is_time_variant(x) = true: it's a time-varying curve object.
• is_time_variant_term(x, OnVariable) = false: the field of x that becomes the coefficient of OnVariable (the initial input field) is not time varying.

edit: they're both for deciding whether to add terms to the variant or invariant objective function terms. is_time_variant_term is used in proportional_cost alone. Other places (like add_variable_cost_to_objective!) use is_time_variant.

[luke-kiernan]

It turns out the current behavior (in that branch) is overly cautious: if the curve is time varying, we're right now treating ALL its terms as time-varying. If I'm understanding things correctly, even though we've separated static vs time varying more cleanly, that's still not quite right. E.g. the case mentioned above.

edit: oh wait. If we're using a formulation that has an OnVariable, though, does it even make sense to have initial_input = nothing?

[jd-lara]

I don't understand what are you trying to answer here... The multiplier to the on variable in the cost function could be time variant

[luke-kiernan]

Following up in case someone else comes across this:

edit: oh wait. If we're using a formulation that has an OnVariable, though, does it even make sense to have initial_input = nothing?

No, it does not. So the above example isn't actually a concern.

There is still one case in where the "time varying field" versus "time varying struct" comes up: FuelCurve{T <: ValueCurve}'s fuel_cost::Union{Float64, TimeSeriesKey} field. So you can't tell whether a FuelCurve is time-varying or static simply from looking at its type: it could have a static curve as its value_curve::T, yet still have time-varying fuel_cost.

I believe I've resolved the issue. Here's the fix:

• IS.is_time_series_backed on cost curves answers "is any field time-varying?" [So for FuelCurves, it checks fuel_cost too.]
• IOM.is_time_variant forwards to IS.is_time_series_backed
• Rename is_time_variant_term to is_time_variant_proportional: it answers the question "is the coefficient of OnVariable time-varying?". [If the value curve is time-varying, then the answer is certainly yes; but it's also "yes" for a FuelCurve with a static value curve and time-varying fuel_cost.]