Ahh ... in that case, I'm guessing it has to do with the starting value for your z variable. In fact, depending on the solver used, it may generate it's own initial value. If it lies in the deadband, the variable may not move at all since the derivatives of the cost there are all zero.
-- Ray Zimmerman Senior Research Associate 419A Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Jun 26, 2012, at 12:19 PM, Panagis Vovos wrote: > Dear all, > > Please allow me to correct myself. The cost function I have tested has > a negative quadratic coefficient (Cw=-1, see code below). > Therefore the 'z' variable is expected to converge either to 0 or 10. > Instead it converges to 1.9847, that is in the deadband. > Let me remind you that I am trying to add my own cost function for a > new variable 'z' in OPF,according to 5.3 of the user's manual. 'z' is > between 0 and 10, but there are no other linear or non-linear > constraints. The only thing that leads the variable to a specific > value is the cost function attached to it. It is a quadratic cost > function (d=2) with a "dead band" from 1 to 9 (r=5,k=4). Below you can > find the code I have written to implement it. You can add it to any > Matpower case and test it by running the OPF, since 'z' is an > independent variable. The problem is that the result is z=1.9847, i.e. > in the deadband! Does anyone knows why is that? I have also attached > the Matlab figure of the cost function (w,r) I believe I have created. > > %% CODE ADDED IN MATPOWER CASE FILE > > % ADDITION OF 'z' VARIABLE > NumberOfXvars=(size(mpc.bus,1)+size(mpc.gen,1))*2; > OldVarsA=zeros(size(mpc.buslink,1),NumberOfXvars); > NewVars=size(mpc.buslink,1); > NewVarsA=eye(NewVars); > mpc.A=sparse([OldVarsA NewVarsA]); > mpc.l=zeros(NewVars,1); % lower limit equals 0 > mpc.u=10*ones(NewVars,1); % upper limit equals 10 > > % ADDITION OF COST ON 'z' VARIABLES > mpc.N=mpc.A; > % fparm=[d r^ k m]; > mpc.fparm=[2 5 4 1]; > mpc.Cw=sparse([-1]); > > What am I doing wrong? > Any help would be highly appreciated. > > Best regards, > > Panagis Vovos > <NewCost.jpg>
