All of the AC OPF solvers are numerical solvers with termination tolerances. If you want more precise solutions, you have to make these tolerances smaller. Since they solve for local optima, it is always possible that different algorithms will find different local optima, but in your case I think it is a simple matter of tightening the termination tolerances. I have no idea what you intended to accomplish with your "LamQ= - LamQ in the Hessian file”.
In any case, check “help mpoption” to find the termination tolerance options for the solvers you are using. -- Ray Zimmerman Senior Research Associate B30 Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Feb 14, 2014, at 9:50 AM, amel zerigui <[email protected]> wrote: > Dear all, > > By using Matpower for OPF solution with several solvers I recognize that the > solution has a bit difference, maybe you tell me that this difference is not > important for OPF, but I found it crucial if I used the obtained point as > initial point in my research, bellow you find an example with explanation: > > The solved case is New-England power system, the obtained OPF with FMINCON > (520 in mpoption) > > Bus Voltage > Generation > # Mag(pu) Ang(deg) P (MW) Q (MVAr) > 30 1.012 -2.924 244.61 > -78.21 > 31 1.087 -0.000 570.84 > 427.49 > 32 1.030 1.981 647.05 > 194.82 > 33 1.015 3.514 636.83 > 107.29 > 34 1.022 2.268 513.32 > 137.83 > 35 1.062 5.410 656.12 > 160.27 > 36 1.090 7.896 563.70 > 116.85 > 37 1.046 2.359 539.02 > 39.18 > 38 1.045 7.701 836.45 > 19.06 > 39 1.051 -9.700 983.52 > 48.70 > The objective function (520) is: 61748.05 $/hr > > Then same case_name is solved by using MIPs solver (560) and the results are: > > Bus Voltage > Generation > # Mag(pu) Ang(deg) P (MW) Q (MVAr) > 30 1.012 -2.926 244.61 > -78.15 > 31 1.087 0.000 570.84 > 427.02 > 32 1.030 1.980 647.05 > 195.11 > 33 1.015 3.512 636.83 > 107.31 > 34 1.022 2.267 513.32 > 137.82 > 35 1.062 5.409 656.12 > 160.31 > 36 1.090 7.894 563.70 > 116.82 > 37 1.046 2.357 539.02 > 39.19 > 38 1.045 7.700 836.45 > 19.07 > 39 1.051 -9.702 983.52 > 48.74 > The objective function (of 560) is: 61748.06 $/hr > > The difference is in red. > Then what I did, because the large difference is in Q(MVAr) I took LamQ= - > LamQ in the Hessian file and I run OPF with MIPs solver, the results are > bellow: > > Bus Voltage > Generation > # Mag(pu) Ang(deg) P (MW) Q > (MVAr) > 30 1.012 -2.924 244.61 > -78.22 > 31 1.087 0.000 570.84 > 427.49 > 32 1.030 1.981 647.05 > 194.82 > 33 1.015 3.514 636.83 > 107.30 > 34 1.022 2.268 513.32 > 137.82 > 35 1.062 5.410 656.12 > 160.27 > 36 1.090 7.896 563.70 > 116.85 > 37 1.046 2.359 539.02 > 39.18 > 38 1.045 7.701 836.45 > 19.06 > 39 1.051 -9.700 983.52 > 48.71 > The objective function now (of 560) is same as above: 61748.06 $/hr > > Now the results are nearly closer to the results of FMINCON but still have > too small difference which still caused for mea problem if I take this point > as initial point of my work. > > What do you think? Is there any suggestion? And what I did is right or no? > > Thanks >
