Adding a new function to the PowerModelsAnnex

[Rasoul]

Hi all,

I have written a function, i.e, sdp_opf, to solve the OPF problem via SDP relaxation which exports the build_sdp_opf. However, when I want to run my code via build_sdp_opf(case,model) I get the following error.

build_sdp_opf not defined

I have already written similar functions to and add them to the PowerModelsAnnex in Julia 0.6 to solve the OPF problem and I didn’t get any error. However, recently I upgraded my Julia to 1.3 and I got the above error. I will appreciate it if you let me know about possible actions that I should take to run my code in Julia 1.3. Thanks in advance!

[ccoffrin]

This seems like a repeat of a question in,

SDP relaxation in PowerModelsAneex

Hi Dr Coffrin, Thanks for the response. I is the eye matrix, defined in LinearAlgebra package, that you have directly used in your SDP code in PowerModels. You have used LinearAlgebra package in your code. But When I added that package to PowerModelsAnnex I got the above error. My problem is how I can use the LinearAlgebra library in the power model annex. I will definitely look at the Julia package development documents. All the best, Rasoul

I think the core point here is that a lot changed about how package development works between Julia v0.6 and Julia v1.0. You might start by reviewing those changes. For example,

My New Workflow with Julia 1.0

http://abelsiqueira.github.io/blog/package-development-on-julia-10/

It’s fairly similar to how it use to work, just make sure that you are working out of the “dev” directory and that your active environment is setup to use that version of the code.

[Rasoul]

Hi Dr. Coffrin,

Thanks for your response, it was useful and I really appreciate it. I fixed that error by activating the environment. As you might know I am writing the SDP relaxation for the OPF problem in rectangular coordinate in PowerModelsAnnex.
The core difference between my formulation and yours in PowerMpdels is about how the “W” matrix is defined. In my implementation I have defined a 2n*2n positive semidefinite matrix named “W” and then I extract “w”,”wr”, “wi” out of the “W” matrix to defined all other constraints based on these variables.
I have two questions. First, with the following implementation I get an error (i.e. MethodError: no method matching _get_name(::QuoteNode)) for line “@variable(model,:w)[i] = W[w_idx,w_idx]+W[w_idx+n_bus,w_idx+n_bus]”.

My second question is, even after commenting out that line I get the following error, UndefVarError: nw not defined. I don’t have any “nw” in my code except in the second line. I will appreciate it if you guide me to solve these error. Thanks in advance!

    PowerModels.standardize_cost_terms!(data, order=2)
    ref = PowerModels.build_ref(data)[:nw][0]
    bus_ids = keys(ref[:bus])
    n_bus = length(bus_ids)
    w_index = 1:length(bus_ids)
    lookup_w_index = Dict((bi,i) for (i,bi) in enumerate(bus_ids))

    WR_start_1 = zeros(length(bus_ids), length(bus_ids)) + I
    WR_start_2 = zeros(length(bus_ids), length(bus_ids))
    WR_start_3 = zeros(length(bus_ids), length(bus_ids))
    WR_start_4 = zeros(length(bus_ids), length(bus_ids))

    W_start=[WR_start_1  WR_start_2;WR_start_3  WR_start_4]
    @variable(model,
    W[i=1:2*length(bus_ids), j=1:2*length(bus_ids)], PSD, base_name="$(nw)_W", start=W_start[i,j]
    )
    @variable(model)[:w] = Dict{Int,Any}()
    for (i, bus) in ref(pm, :bus)
        w_idx = lookup_w_index[i]
        @variable(model,:w)[i] = W[w_idx,w_idx]+W[w_idx+n_bus,w_idx+n_bus]
   end

[ccoffrin]

I think the issue is with this @variable(model)[:w], I think the code should look something more like this,

    PowerModels.standardize_cost_terms!(data, order=2)
    ref = PowerModels.build_ref(data)[:nw][0]
    bus_ids = keys(ref[:bus])
    n_bus = length(bus_ids)
    w_index = 1:length(bus_ids)
    lookup_w_index = Dict((bi,i) for (i,bi) in enumerate(bus_ids))

    WR_start_1 = zeros(length(bus_ids), length(bus_ids)) + I
    WR_start_2 = zeros(length(bus_ids), length(bus_ids))
    WR_start_3 = zeros(length(bus_ids), length(bus_ids))
    WR_start_4 = zeros(length(bus_ids), length(bus_ids))

    W_start=[WR_start_1  WR_start_2;WR_start_3  WR_start_4]
    @variable(model,
    W[i=1:2*length(bus_ids), j=1:2*length(bus_ids)], PSD, base_name="$(nw)_W", start=W_start[i,j]
    )
    w_lookup = Dict{Int,Any}()
    for (i, bus) in ref(pm, :bus)
        w_idx = lookup_w_index[i]
        w_lookup[i] = W[w_idx,w_idx]+W[w_idx+n_bus,w_idx+n_bus]
   end

[Rasoul]

Thanks for your invaluable response! It fixed the error. Can you please give me some insights on how to solve the issue in the second part of my question. I still get the following error, UndefVarError: nw not defined.
I don’t have any “nw” in my code except in the second line. Thanks in advance!

[ccoffrin]

@ccoffrin

base_name="$(nw)_W"

Most likely this line should be changed to something like base_name="W"