It looks like the problem is infeasible due to the voltage angle limits you put on the branches. You set the ANGMIN to 360 and the ANGMAX to -360 (backwards). Either correct the sign on those limits or set the OPF_IGNORE_ANG_LIM option to zero and it should solve just fine.
-- Ray Zimmerman Senior Research Associate B30 Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Jan 30, 2014, at 10:43 AM, Angelina <[email protected]> wrote: > Dear Dr Zimmerman, Shri; > (Eventually, my previous email wasn’t sent ,sorry for any confusion. I attach > it again to you:) > > After struggling with the problem I found out that the ac opf could converge > in matpower version 3.2, but not in later versions. I can’t really understand > why. The solver that I used is fmincon, although it doesn’t seem to make any > difference in case I use another solver. However, could I use MIPS in 3.2 > version? > Additionally, in version Matpower 4.1 I used to write : > [results,success]=runopf(mpc,mpopt) > in order to save and analyse the results after the opf. I can see that this > command doesn’t work for version 3.2. Does anybody know if there is an > equivalent one?; Please provide me any assistance,; Angelina > > From: Angelina [mailto:[email protected]] > Sent: 30 January 2014 15:38 > To: 'MATPOWER discussion forum' > Subject: RE: numerically failed opf > > > As far as your reply to my yesterday’s questions I tried yesterday to use the > pf solution as an initial guess for opf with that command : > > [pf_soln,successpf] = runpf(mpc); > opt = mpoption('OPF_ALG', 500); > [r,successopf] = runopf(pf_soln, opt); > > But it didn’t help. > Just now I tried with the mipsopf_solver as you proposed like the following : > [pf_soln,successpf] = runpf(mpc); > opt = mpoption('OPF_ALG', 560); > [r,successopf] = mipsopf_solver(pf_soln, opt); > > but after the convergence of the power flow, it stops with the error > undefined function 'get_mpc'. Where could I find that function? (I found one > get_mpc online , but it is something related to clusters and it seems > irrelevant) > I attach to you the case file, so If you can have a look at it. However, as > you will may see by the email that I sent you a while ago, I can solve the > acopf with the matpower version 3.2. Although I have some questions , that I > would be grateful if you could eventually address them. > Kind regards, > Angelina > > > > From: [email protected] > [mailto:[email protected]] On Behalf Of Abhyankar, > Shrirang G. > Sent: 30 January 2014 14:54 > To: MATPOWER discussion forum > Subject: Re: numerically failed opf > > Try using the solution of the power flow as the initial guess for OPF. The > initial guess, with the mips solver, is computed in mipsopf_solver. Please > send your case file if you still have problems. > > Shri > > From: Angelina <[email protected]> > Reply-To: MATPOWER discussion forum <[email protected]> > Date: Wed, 29 Jan 2014 22:05:48 +0000 > To: 'MATPOWER discussion forum' <[email protected]> > Subject: RE: numerically failed opf > > Dear Dr Zimmerman, Shri > I tried also with the other solvers (like PDIM,SCDIPM,TRALM) but the problem > cannot have a solution. Even when I relax the voltage constraints ( from > [0.94-1.06] to [0.9-1.1]) nothing changes. Additionally, even when I have the > problem in its simplest form, that is only one slack bus with the grid infeed > and no other generators in the rest of the nodes the problem does not > converge. When I run the same problem with power flow it converged normally, > without any violations. This confuses me a lot. > Could I send you the case file and have a look at it? > I would be really grateful if you could help me. > Regards, > Angelina > > From: [email protected] > [mailto:[email protected]] On Behalf Of Abhyankar, > Shrirang G. > Sent: 29 January 2014 15:53 > To: MATPOWER discussion forum > Subject: Re: numerically failed opf > > The numerical divergence could also possibly be due to the infeasibility of > the problem. Can you try relaxing some of the constraints (especially the > voltage magnitude constraints) and see if the OPF converges? > > Switching to another solver is also an option. See the list of available > optimization packages that MATPOWER supports here > http://www.pserc.cornell.edu//matpower/#optionalpackages > > These archive threads may be relevant for your problem. > http://www.mail-archive.com/[email protected]/msg02915.html > http://www.mail-archive.com/[email protected]/msg01298.html > > Shri > > From: Angelina <[email protected]> > Reply-To: MATPOWER discussion forum <[email protected]> > Date: Wed, 29 Jan 2014 15:35:23 +0000 > To: MATPOWER discussion forum <[email protected]> > Subject: numerically failed opf > > Dear Dr Zimmerman, > I am trying to solve an ac opf with the mips solver (560) in a lightly > meshed 6.6 kV distribution system of 32 branches. The range for the > reactances is [0.05-1.9 pu], whereas some of them have almost negligible > values. When I am solving the simple ac power flow the problem converges with > no problems to be appeared. However, when I am trying to solve the same > problem with optimal power flow, assigning some costs to the generators, the > ac opf fails numerically or the message ‘ the matrix is singular to working > precision, the results may be inaccurate’ appears. > I suspect that the problem is the solver. Could you recommend me another > solver to use? > Any guidance would be appreciated. > Kind Regards, > Angelina
