The latest MATPOWER (v7.1) uses
MP-Opt-Model<https://github.com/MATPOWER/mp-opt-model> to build and solve the
mathematical model, so it would probably be helpful to familiarize yourself
with that by having a look at the MP-Opt-Model User’s
Manual<https://matpower.org/docs/MP-Opt-Model-manual-3.0.pdf>. However, I
realize that as we have abstracted and generalized the OPF design over time to
more cleanly separate problem formulation from solvers, this structure is not
as transparent as it once was, which calls for a bit more documentation to make
this clear. I expect this will be addressed when we integrate the new
generalized network and element modeling layer,
MP-Element<https://github.com/MATPOWER/mp-element/> into the documentation.
But in the mean time, here is the basic calling structure of the main functions
involved in setting up and solving the OPF, in the case of an AC OPF solved by
fmincon.
runopf()
opf()
opf_setup()<https://matpower.org/docs/ref/matpower7.1/lib/opf_setup.html> -
builds the mathematical model (om)
om.add_var()
om.add_lin_constraint()
om.add_nln_constraint()
om.add_quad_cost()
...
opf_execute()<https://matpower.org/docs/ref/matpower7.1/lib/opf_execute.html> -
solves the mathematical model (om)
om.solve()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/@opt_model/solve.html>
@nlp_costfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_costfcn.html>
om.eval_quad_cost()
om.eval_nln_cost()
@nlp_consfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_consfcn.html>
om.eval_nln_constraint()
@nlp_hessfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_hessfcn.html>
om.eval_nln_constraint_hess()
nlps_master()
nlps_fmincon()
fmincon()
So the building of the model for the OPF occurs in
opf_setup()<https://matpower.org/docs/ref/matpower7.1/lib/opf_setup.html>, Then
opf_execute()<https://matpower.org/docs/ref/matpower7.1/lib/opf_execute.html>
calls the
solve()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/@opt_model/solve.html>
method for the model, where you can see the definitions of the function
handles used to evaluate the costs, constraints and hessians, namely
nlp_costfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_costfcn.html>,
nlp_consfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_consfcn.html>,
and
nlp_hessfcn()<https://matpower.org/docs/ref/matpower7.1/mp-opt-model/lib/nlp_hessfcn.html>,
respectively. Each of these uses the appropriate evaluation functions on the
mathematical model object to compute the cost, constraint or hessian function
values.
I hope this helps,
Ray
On Nov 4, 2020, at 1:53 AM, yangyang
<[email protected]<mailto:[email protected]>> wrote:
Dear all,
I am learning how matpower implements OPF using fmincon(). This matlab
optimization function requires a bunch of numerical input arguments to start
the optimization, like start point of the variable vector, equality,
constraints, etc. The examples on matlab website also shows how to implement
optimization using fmincon(). However, I really cant find which is the m file
to convert a matpower case to a bunch of numerical things to start fmincon(). I
have searched the matpower folder and the online function reference and I still
cannot find how it is implemented. It seems that matpower uses a lot of
function handles for implementation; and even setting breakpoints in fmincon.m
will not show the intermediate variables during the optimization! Can some one
show me the key m files to convert a matpower case into formats compatible with
fmincon? Thank you for your patience!
Best Regards,
Yang Yang
