Schema

[yasirroni]

I just noticed that @rwl has schema in csv format here. I also has similar thing in matpowercaseframes. I know that I can looks at individual idx_bus.m, idx_gen.m, etc. But, did you think that it would be better if MATPOWER has a schema for currently version 2? I think make it separate repository will also works. We can discuss introducing other data format, for example what is the best way for MOST to represent profiles, how to define DC line data, and also unbalance four lines data.

[rdzman]

Certainly. I welcome any contributions and suggestions in this regard. I'd be happy to set up another repository if that makes sense.

[yasirroni]

Thank you @rdzman. Maybe, we can start by deciding what is the best schema or structure that is the best suited as either as checker or empty template generation for MATPOWER. I don't think .mat is a good here as it is not human readable. It would be better to use csv, json, or xml. If it is in CSV, you can look at such example based on your format in caseformat/schema. Then you can decide if we start from that or make from scratch.

After that, we can start adding several data that is not properly defined like dclinecost, reserves.zones, reserves.req, reserves.cost, reserves.qty, genfuel, gentype, and time series data. All of that data, as far as I remember, exist in matpower, but scattered, unlike idx_bus.m etc.

I forget to mention that matpower also store name indexing, that is currently that I know are "bus_name", "branch_name", and "gen_name".

[yasirroni]

To put ot simply:

1. .m file require plain text parser to be read by external program.
2. MATPOWER can't save runopf results in .m.
3. .mat is binary, not human readable, thus make it hard for external program to write and human can't manually create case in that form.

[yasirroni]

I found that NREL/Sienna has a data structure that define time series and identifier that describe the time series data. See it at https://github.com/NREL-Sienna/PowerSystemsTestData/tree/master/RTS_GMLC. Maybe we can use that as a base to define what is the best way to store and define data structure for MOST and the unbalanced power flow in the future.

[yasirroni]

Hi @rdzman. I think a good start is to properly give MOST benchmark case. One format that is used by Power Grid Lib benchmark is to use json data to represent time series: https://github.com/power-grid-lib/pglib-uc.

Using .m is also discouraged by pandapower developer as they feels running MATLAB/Octave from Python is unsafe: e2nIEE/pandapower#2403 (comment).

If you need help, I would be happy to contribute a converter that run loadcase and export to other format, preferably that we could use to make benchmark for MOST. But, the format need to be decided first.

[rdzman]

I understand the downside to using M-files. My preference for any new file formats would be something like CSV. I did write a converter some time ago (this is apparently in a broken state at the moment) from M-file to CSV. It results in something like the following for case30.m. This is for the existing version 2 of the MATPOWER case format.

case30_csv_mpc2_csv.zip

Ideally, I'd like to define schema and file formats for an updated MATPOWER file format that will incorporate lots of updates, including schema for 3-phase elements, and not have to worry about backward compatibility. Unfortunately, that is a rather large undertaking and at the moment I currently have almost no time to work on it.

However, if there are others who would like to spearhead this effort, I'm happy to try to help coordinate. I think the MATPOWER Developer's List may be the best place for this discussion.