Hi Gal, I think the goal is to run an N-1 contingency constrained multiperiod dispatch problem, where the objective is to minimize the base case cost, correct? When using MOST for this, there are a number of subtle details to keep in mind.
1. All generators (and dispatchable loads) are allowed to be arbitrarily re-dispatched within ramping and reserve limits in contingency cases in order to minimize the overall expected cost. In some contingency-constrained dispatch formulations, the redispatch is controlled by participation factors rather than being done in an economically optimal way as in MOST. 2. You cannot use a dispatchable load to represent a load that must be satisfied in a zero-probability contingency, since the cost of shedding that load in that contingency will also be zero. To put it another way, adding a zero-probability contingency with shed-able load places no restrictions whatsoever on the base case dispatch and therefore has no effect on the optimization result. 3. The values in mdo.results.Pc are not physical dispatch values, so they should not be treated as such. They should not be used to compute any sort of real cost and there is no reason to expect them to sum to zero or even to be positive. The generator dispatch values are all contingency specific, so if you want a single value for each generator, it is probably mdo.results.ExpectedDispatch you are looking for. I hope this sheds some light, Ray > On Dec 5, 2016, at 1:09 PM, Gal D <[email protected]> wrote: > > Addition to my previous message: > > It does seem that the generation sums up to 240, but the dispatchable load > (generator 4) reaches -540 and the rows of mdo.results.Pc don’t sum to 0. > Does this make sense? >> On 5 Dec 2016, at 19:50, Gal D <[email protected] >> <mailto:[email protected]>> wrote: >> >> Hi Ray, >> >> Thanks for the interesting idea. However it doesn’t seem to work since I >> don’t witness the result I expect. I’ll explain: >> >> - I’m running most_ex5_mopf >> - I removed the wind completely for simplicity. There’s no storage either. >> - I included only the 3 branch contingencies in ‘contab’, all with >> probability 0 >> >> What I expect to see: generation & load always bounded by 240, since this is >> the bottleneck when line 2-3 is out. >> What I see: generation & load bounded by 540, which is the no-contingency >> case. When I inspect the verbose output, I see that per each contingency it >> is indeed bounded according to the specific case, but the overall result is >> as I described. >> >> PS. possible bug: I also tried assigning 0.25 probabilities to all three >> branch contingencies and MOST converges on negative values for generator 2 >> for some reason. >> I’m attaching my most_ex5_mopf.m and ex_contab.m for you convenience. >> >> Your help is appreciated! >> Gal >> <most_ex5_mpopf.m> >> <ex_contab.m> >> >> >>> On 5 Dec 2016, at 17:11, Ray Zimmerman <[email protected] >>> <mailto:[email protected]>> wrote: >>> >>> What about using the contab table as you mentioned, but assigning a >>> probability of 1 to the base case and 0 to the contingency cases? The >>> constraints will still need to be satisfied for the contingency cases, but >>> the objective function will only be based on the base case. You will need >>> to make sure that you only include contingencies that are feasible however, >>> so beware of isolated buses and islanding. >>> >>> Ray >>> >>> >>>> On Dec 5, 2016, at 6:38 AM, Gal D <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> >>>> Hi, >>>> >>>> I wish to run an N-1 unit commitment, considering all branch outages. How >>>> do I do that? >>>> >>>> - Notice I saw a question regarding N-1 contingency analysis by Harron >>>> Malik from July, however I’m asking a different question - I do not want >>>> to iterate on contingencies and solve per each one on its own, rather to >>>> add all N-1 constraints into a single optimization program. >>>> >>>> - I thought of doing that using the ‘contab’ table by assigning uniform >>>> probabilities on all branch failures , but I’m afraid this results in a >>>> different optimization program. >>>> >>>> >>>> Thanks, >>>> Gal >>> >> >
