Re: Solved case does not converge in zero iterations

[Ray Zimmerman]

Hi Dominic, I don’t think there is any bug here. First of all, the issue of the power flow converging in 1 iteration as opposed to zero is simply an issue of the termination tolerances. The issue of non-matching reactive power outputs is related to the fact that there are two generators at the same bus, either of which could generate the reactive power. The total reactive power injected at the bus is the same in both solutions, so the power flow equations are satisfied in both cases. The OPF solution has all of the reactive supply at the bus coming from one of the two generators. The power flow code splits the reactive injection between multiple units proportional to the reactive range of each unit. The following code (also in attached script file) results in the output below and will hopefully qualm any doubts ... define_constants; load mpc123b mpopt = mpoption('opf.ac.solver', 'MIPS', 'mips.comptol', 1e-10, 'mips.step_control', 1); mpopt = mpoption(mpopt, 'out.all', 0); r1 = runopf(mpc123, mpopt); mpopt = mpoption(mpopt, 'verbose', 2); r2 = runpf(r1, mpopt); compare_case(r1, r2) k = find(r1.gen(:, GEN_BUS) == r1.gen(220, GEN_BUS)) [r1.gen(k, QG) r2.gen(k, QG)] sum([r1.gen(k, QG) r2.gen(k, QG)]) >> test123 MATPOWER Version 5.0, 17-Dec-2014 -- AC Optimal Power Flow MATLAB Interior Point Solver -- MIPS-sc, Version 1.1, 17-Dec-2014 Converged! MATPOWER Version 5.0, 17-Dec-2014 -- AC Power Flow (Newton) ----------------  --------------  --------------  --------------  -----  matrix / col         case 1          case 2        difference     row  ----------------  --------------  --------------  --------------  ----- bus   VM                       1.1             1.1     2.22045e-16     44   VA                   35.2787         35.2787     1.42109e-14    180 * gen   PG                6.4271e-10     6.38956e-10     3.75491e-12   3414   QG               1.02725e-10         494.118         494.118    220 * branch   PF                   75.2832         75.2832     2.81496e-09   2148 *   QF                     8.577           8.577     1.48452e-09   2148   PT                  -75.2831        -75.2831     2.81496e-09   2148   QT                  -8.65882        -8.65882     1.48449e-09   2148 k =    219    220 ans =   867.6421  373.5236     0.0000  494.1184 ans =   867.6421  867.6421 --  Ray Zimmerman Senior Research Associate B30 Warren Hall, Cornell University, Ithaca, NY 14853  USA phone: (607) 255-9645

test123.m
Description: Binary data

On Jan 28, 2015, at 1:42 PM, Hewes, Dominic <dominic.he...@tum.de> wrote: Dear Matpower Community,   I am observing a strange problem whereby the results from a successful 'runopf()' do not seem to present a solved power flow case. I want to verify the power flow solution from an OPF by running a PF with the OPF results as the mpc struct. Firstly, the 'runpf()' converges in 1 iteration, whereas i would expect a solved case to converge in 0 iterations- am i mistaken here? Secondly, when i use the 'compare_case()' command to compare the OPF results with the resulting PF results, I see that there are large differences between the solutions. My code is as follows:   resultsOPF=runopf(mpc123); resultsPF=runpf(resultsOPF);   compare_case(resultsOPF, resultsPF)   The comparison shows a maximum reactive power difference of 494 MVAR between the generator results:   gen   PG               5.00578e-10     9.67145e-09     9.17088e-09    614   QG               7.99959e-11         494.118         494.118    218 *   The OPF command converges successfully with no error warnings, and so I assumed that running the results struct through a PF command should give the same power flow solution. Am i mistaken here?  If my thinking is correct, this would indicate that the optimiser has provided results that do not represent a feasible power flow solution - is this potentially a bug? I have tested the same method on the 'case14.m' file and observe that this also requires 1 iteration to converge and gives very small differences (~2e-7) between the OPF and PF results.   I am using the latest matpower release (5.0), the TSPOPF 5.0 solver and a windows 7 machine (see output of mpver at end of email). I have tried with other solvers and observe the same problem. I am working with a large model (>5000 bus) that i have attached below. Could the size of my model be the cause of this problem?   Presently I am not sure whether I am making a simple mistake or if there is a bug in the solver that is causing the output of inaccurate results.   Has anyone experienced similar problems? I would very much appreciate any advice on the cause of the problem.   Kind Regards,   Dominic     mpver   MATPOWER               Version 5.0        17-Dec-2014 MATLAB                 Version 8.4        08-Sep-2014   Release: (R2014b) Optimization Toolbox   Version 7.1        08-Sep-2014   Release: (R2014b) MIPS                   Version 1.1        17-Dec-2014 SDP_PF                 -- not installed -- YALMIP                 -- not installed -- BPMPD_MEX              -- not installed -- CPLEX                  Version 12.6.0.0                 PCWIN64 Gurobi                 -- not installed -- GLPK                   -- not installed -- IPOPT                  -- not installed -- KNITRO                 -- not installed -- MINOPF                 -- not installed -- MOSEK                  -- not installed -- PDIPMOPF               Version 5.0        17-Dec-2014   PCWIN64 SCPDIPMOPF             Version 5.0        17-Dec-2014   PCWIN64 SDPT3                  -- not installed -- SeDuMi                 -- not installed -- TRALMOPF               Version 5.0        17-Dec-2014   PCWIN64 Architecture:          PCWIN64 <mpc123.mat>