Dear Enrico, I’m happy to help.
The resistance and reactance values don’t necessarily look too high or low. Nothing stood out as obviously wrong to me at least. I reduced the values just to see if the convergence of the ACOPF was a topology issue. I find that simplifying a system until it works is useful when dealing with problems like this. Another thing you could try is to solve parts of the system (e.g. simplify your network to only 380kV nodes, or a small area) and gradually add complexity back into your model until it fails. In response to your last question, you’re almost correct. In an OPF the values you put in to mpc.gen(:,PG) will change, as PG is an optimisation variable, but the initial guess will affect the convergence of the algorithm. In some cases (possibly yours!) a bad initial guess for PG and QG can prevent the ACOPF from converging to a solution. Kind regards, Samuel From: bounce-121259537-76238...@list.cornell.edu [mailto:bounce-121259537-76238...@list.cornell.edu] On Behalf Of Enrico Vaccariello Sent: 21. febrúar 2017 00:48 To: MATPOWER discussion forum <matpowe...@list.cornell.edu> Subject: Re: Sequential OPF: Non-convergence of both OPF and PF Dear Samuel, I really appreciate your help, thank you. After reading your reply I have made some further tests, and I believe that the very issue is the branch matrix too. The ACOPF actually converges with your corrections on the branch matrix, and I have noticed that it still works even keeping the original values for the lines' resistances. Keeping the original values of mpc.branch(:,BR_R) results in active power losses accounting for around 3% of the total generated active power, which seems reasonable to me, suggesting that the p.u. values of the resistances are acceptable. So the problem should lie in the values of X and of the MVA ratings. Did u reduce the values of X because they looked too high to you or just to get the ACOPF to converge? As for the MVA ratings given in mpc.branch(:,RATE_A), setting them all equal to zero, as you suggested, and running the ACOPF clearly shows that in the optimal solution some branches have power flows much higher than their limit in the original case. Some of these branches are the same that you indicated (150,369,...). What I think is weird is that all these branches link PQ buses, with no power generation. So, it's difficult to understand what is the logic behind it. Anyway, I have checked all the branches and their From and To indexes are all correct. I hope I will find a way to fix this. Until then, thank you again. One last thing I wanted to ask about your reply (first paragraph of your mail): I have understood that the values that you use to initialize mpc.bus(:,PG) do affect the solution of the PF problem, while they are not considered by the OPF solver. Is that right? Best regards, Enrico 2017-02-20 12:31 GMT+01:00 Samuel Perkin <samu...@landsnet.is<mailto:samu...@landsnet.is>>: Hi Enrico, I had a quick look at your case. Firstly, the results you get when running an DC PF on your case are normal. The active power is fixed at all Load and Generator buses, and therefore only the slack bus can add/remove active power to the system in order to find a solution. Given that all your generators begin with 0 MW of active power production, you find that all power must be produced at your slack bus. I tried running an ACOPF on your model, using: mpopt = mpoption('verbose', 3, 'out.force', 1,'pf.nr.max_it',10); runopf(mpc,mpopt); This doesn’t converge, but the output suggests a few issues with your model. If you look at the branch constraints, you’ll find that some of your branches with the greatest capacity are only being utilized <5%. Whilst a few low capacity branches (150, 290, 369, 370) have absurdly large flows. This implies to me that you either have some branches connected to the wrong buses, or have some errors in the electrical parameters of your lines. Note that it is possible to get an ACOPF to converge on your case if you apply the following modifications/simplifications, which supports my argument that there is probably a few errors in your branch data: mpc.branch(:,RATE_A) = 0; mpc.branch(:,BR_R) = 0.1*mpc.branch(:,BR_R); mpc.branch(:,BR_X) = 0.001*mpc.branch(:,BR_X); I hope this helps you debug your case, and good luck with your thesis. Kind regards, Samuel Perkin From: bounce-121256175-76238...@list.cornell.edu<mailto:bounce-121256175-76238...@list.cornell.edu> [mailto:bounce-121256175-76238...@list.cornell.edu<mailto:bounce-121256175-76238...@list.cornell.edu>] On Behalf Of Enrico Vaccariello Sent: 20. febrúar 2017 02:25 To: matpower-l@cornell.edu<mailto:matpower-l@cornell.edu> Subject: Sequential OPF: Non-convergence of both OPF and PF Dear All MATPOWER developers and users, I really hope some of you can give me some advice. I am working on sequential OPF simulations (not using MOST, but simply running MATPOWER's runopf in a loop and manually updating some values the MATPOWER case at each time-step t) to be performed on a case(s) that I have been working a lot to prepare. Anyway, unfortunately, now that the time of running the simulation has finally come, I have some serious convergence issues. With the 285-bus case structs that I have built, the OPF simulation running with the default MIPS does not converge. The case structs differ one from another for some parameters as a function of t (time-step of the loop). Anyway, whatever the t I set, the OPF won't converge in any case, so there must be some structural issue of my cases I guess. I'll attach the case of t=1 to this mail for who wishes to have a look. Making some tests, I have found out that the PF simulation does not converge either. Among the (non-converged) results of the PF it seems that the slack bus is the only trying to generate active power (a lot of), whereas all the others do not dispatch, i.e. results.gen(:,PG) is zero for all the generators of the PV buses. What is more, the same generators dispatch very high quantities of reactive power, much beyond their lower and upper limits. If someone could help I would be very grateful... Last days for my thesis, very hard times! Thank you and best regards. Enrico
