[FEATURE] 2-D or cross-join of batch updates

[pieterminnaar]

We need to do an hourly calculation for a full year of multiple N-1 scenarios.

You now have to create a new model for each N-1 situation and then do a batch update for the timeseries.

It would be convenient if this can be done in one step which will in effect be a cross-join between the two update sets.

n_min_1_and_timeseries = {
ComponentType.sym_load: load_profile, # Contains an hourly value for each load
ComponentType.line: line_status, # Contains a network mutation for a cable
}
output_data = model.calculate_power_flow(update_data=n_min_1_and_timeseries)

[mgovers]

Hi @pieterminnaar,

Thank you for requesting this issue. Looks like an interesting feature.

Unfortunately, this would be a significant change in both the Python library, the C API and the core (if we want to officially support this on all levels). EDIT: we will brainstorm on this to see how much effort this would be.

In the meantime, it should be relatively easy to do this on your side using np.reshape. See also this stackoverflow thread: https://stackoverflow.com/questions/18757742/how-to-flatten-only-some-dimensions-of-a-numpy-array

We will bring this to the team to see how we can approach this on our end as well.

Hope this helps.

[TonyXiang8787]

@mgovers, I think implementation in the core/C-API can still be feasible with relatively easy effort. Maybe we need a brainstorm.

[TonyXiang8787]

Hi @pieterminnaar,

I have some experimental implementation in #1186, you can give it a try by manually downloading the wheel file from the CI:

https://github.com/PowerGridModel/power-grid-model/actions/runs/19290638431?pr=1186

Just scroll down into the bottom and find the wheel file for your OS for manual install. It works in a way that you can support a list of update batch dataset to the calculate_* function. The model will then make a cross-join of the update to produce all the combinations in the results. The results are still flattend in 1D-batch (2D with object). You can of course reshape it to the multi-dimension. Let me know if you would prefer a direct return lf multi-dimension array. But that would mean the returned dataset will not work into the serializer at this moment.

n_min_1 = {
ComponentType.line: line_status, # Contains a network mutation for a cable
}

timeseries = {
ComponentType.sym_load: load_profile, # Contains an hourly value for each load
}

# returns 1D batch results of size_n_min_1 * size_n_time_series
output_data = model.calculate_power_flow(update_data=[n_min_1, timeseries])