Oh, so you are not only planning to modify the formulation, but you also want to implement your own solver? I assumed you'd be using one of the existing solvers. If you are going to implement your own solver, then there is a lot more work involved than the 3 steps I outlined. Unless you have good reasons that I'm not aware of, I would recommend sticking with the existing solvers.
If you modify opf_consfcn.m and opf_hessfcn.m as suggested, there are several existing solvers than can automatically be used, namely MIPS, fmincon, IPOPT, and KNITRO. With the proper options, I believe both fmincon and IPOPT can use methods that do not require a user-supplied Hessian, so they may work just fine without updating opf_hessfcn.m, but they will likely be slower than the methods that utilize a user-supplied Hessian. -- Ray Zimmerman Senior Research Associate 419A Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Feb 17, 2012, at 5:29 AM, 델카스틸로 wrote: > > > Professor Zimmerman: > > Good day to you sir. Related to your response regarding my query in adding of > non-linear user defined constraints, you have given me three steps as > enumerated below: > > 1. add a new set of nonlinear constraints in opf_setup.m around line 243. > 2. add the implementation of the constraints and corresponding gradients to > opf_consfcn.m > 3. modify the hessian accordingly in opf_hessfcn.m > > I am planning to use Sequential Quadratic Programming for my Optimization > Method. Since this method does not require the hessian information, I may not > need to modify the hessian in opf_hessfcn.m. Is this right, sir? > > Morever, I would appreciate it sir if you could suggest an optimization > method that best fit for non-linear user-defined constraints for OPF. Thank > you very much sir. > > > Very Respectfully Yours, > > Manuelito Y. Del Castillo, Jr. > --- Original Message --- > From : "Ray Zimmerman"<[email protected]> > To : "MATPOWER discussion forum"<[email protected]> > Date : 2012/02/16 Thursday PM 10:22:14 > Subject : Re: Non-Linear User-defined constraints > > Unfortunately, as the manual states, the user-defined constraints included by > MATPOWER are only linear. If you want to include additional non-linear > constraints, you will need to get into the details of the OPF implementation > and modify the files that compute the constraints and their derivatives. It > is a non-trivial task. At a minimum, you will need to do the following ... > > 1. add a new set of nonlinear constraints in opf_setup.m around line 243. > 2. add the implementation of the constraints and corresponding gradients to > opf_consfcn.m > 3. modify the hessian accordingly in opf_hessfcn.m > > If you do end up doing this, I'd love to see the result. It would be nice to > make this process as easy as possible so your experience would be helpful. > > -- > Ray Zimmerman > Senior Research Associate > 419A Warren Hall, Cornell University, Ithaca, NY 14853 > phone: (607) 255-9645 > > > > >> >> >>
