Inconsistency between active power lower bounds and time-series of area loads

[youngdae]

Hello,

Are the loads in REAL_TIME_regional_Load.csv supposed to work with the grid network specified in RTS_GMLC.m?

I was trying to solve an ACOPF with a grid network from RTS_GMLC.m and load from REAL_TIME_regional_Load.csv. However, I found that the sum of the lower bounds of active power of generators in RTS_GMLC.m could be greater than the sum of area loads, resulting in infeasibility. For example, the sum of the lower bounds of active power (i.e., the sum of Pmin of generators with status being 1) in RTS_GMLC.m is 3744.99 MW, but the sum of area loads of the first time period is 3289.53 MW. In this case, no feasible active powers can be found.

I wonder if this is correct.

[claytonpbarrows]

It's known that the real-time load is quite bad. There are a number of discrepancies between the day-ahead and real-time load profiles that represent unrealistically large forecast errors. However, I think the issue that you present is one that requires a commitment adjustment to maintain feasibility. If I understand your issue, you are trying to run an ACOPF with all units committed. I have no expectation that the "commit all" solution is feasible for all setpoints. However, I'm also aware that the current day-ahead, and real-time load, wind, and solar profiles along with the defined reserve requirements also do not always result in a feasible commitment pattern (see #129 , #114). Finally, I'm also aware that the the (lack of) reactive power time series, and the network design itself is not sufficient to support the wind and solar penetration that has been added to the system.
PRs are welcome to address some of these issues