How is storage_state_fix (datastructure_update) supposed to work with time series?

[Karlis-Baltputnis]

Describe your problem

I have a simple model in which the initial and final hydro-reservoir state is set not by storage_state_initial and the cyclic_condition (with == sense), but instead by storage_state_fix with two timesteps like this:

So the first value is for time step before model start, and the final is for model end.

In results, I would expect the inbetween node_state to vary. However, instead I get the same node_state trajectory as in the input visualization.

So it seems that the constraint has been created by forward-filling for the inbetween time steps, instead of just setting it for the specific, defined time steps.

If I delete the first storage_state_fix row and instead define the beginning level via storage_state_initial, I do get varying node_state inbetween in the results, as expected.

I would have expected both approaches to lead to the same results, but perhaps I misunderstand the intention of storage_state_fix

Attach your model as a JSON file (encouraged)

simple_HPP.json

What versions are you using?

SpineOpt: 1.0.0 (git hash: 03a9317) / datastructure_update
SpineInterface: 0.18.0
SpineToolbox:0 .10.9.dev13
Julia: 1.12.6

[tarskul]

My suspicion is that the variable resolution of the fix state glues the values to each other instead of treating them as separate fixed values. I was trying to look for this in the code, but in the half hour that I searched, I was not able to pinpoint exactly how the fixed value is treated during the variable creation. Maybe I'll find another random half hour some time later.

[Karlis-Baltputnis]

It makes sense though. I should have just put a nan value for storage_state_fix at the model start timestamp (2030-01-01T00:00:00, i.e., as a new second row in storage_state_fix) and I wouldn't have this issue....