I am attempting to look at the value of reduced forecasting error by
running a deterministic 24-hour UC using MOST, followed by a re-dispatch
and each hour (where it runs the same UCED but with the correct demand
value at the first time point in the redispatch, and then the forecasted
value for the rest fo the timepoints). I am currently running into a
problem where the re-dispatch doesn't have the correct initial conditions.

For each redispatch, I was able to assign the correct ramping constraints
at t=0 using profiles that adjust the PMAX and PMIN values of each
generator. However, I then ran into a problem where the re-dispatch is
choosing to de-commit generators at t=0 that were previously committed. I
used the profiles again to enforce the units that are already committed by
changing the CommitKey in the xgd table. I believe there may have been a
bug in apply_profile.m that I found that prevents people from updating the
xgd table with a profile, but I think I found the solution and submitted it
as an issue on github.

My end goal is to look at the added system cost of forecasting error, much
like was done in this paper <http://ieeexplore.ieee.org/document/5428351/>,
where they conduct hourly dispatch using MATPOWER opf. My current issue is
in regards to adding reserves. I don't want to add fixed zonal reserves, as
that won't help with the forecasting error. Instead, I'd like to add some
form of ramping reserves. I've played with the
PositiveLoadFollowingQuantity in the xgd table, but that seems to provide
an additional ramping constraint on each generator rather than guarantee a
level of ramping reserve.

I seem to have functioning code for the small tutorial 3 bus test grid, but
when I move up to the iEEE 30 bus network, I am still running into several
issues where I either see no difference in cost or there is significant
load shedding.

Does anyone have any advice in regards to how to compare an imperfect UCED
with hourly dispatch vs. a perfect UCED in Matpower MOST? Perhaps I am
going about this all wrong. Any advice in terms of generally how to conduct
that comparison would be greatly appreciated.

Thanks!
Stephen Suffian

Reply via email to