A quick-and-easy way to generate datasets with some time series structure would be to use some publicly-available total load data (e.g. form an ISO) to define the global scaling factor in our current data-generation code.
This would enable us to have time-stamped data points, which can then be solved as usual.
Data volume-wise:
- 1 year, hourly granularity is ~8700 data points
- 1 year, 5-min granularity is ~100k data points
--> that's ~20GB ish for a 6k bus system (just for the primal AC solutions), so we'll need to be cautious with how much data we generate
Another hurdle would be that we don't have ramping rates in PGLib, so these would need to be generated somehow.
cc @allensctong who might benefit from that.
A quick-and-easy way to generate datasets with some time series structure would be to use some publicly-available total load data (e.g. form an ISO) to define the global scaling factor in our current data-generation code.
This would enable us to have time-stamped data points, which can then be solved as usual.
Data volume-wise:
--> that's ~20GB ish for a 6k bus system (just for the primal AC solutions), so we'll need to be cautious with how much data we generate
Another hurdle would be that we don't have ramping rates in PGLib, so these would need to be generated somehow.
cc @allensctong who might benefit from that.