MATPOWER Network Data in Julia

Liberating MATPOWER caseformat data.

The MATPOWER caseformat is frequently used to store power system data, but Julia (with its GraphViz package) is better suited for network visualization than MATLAB. I decided to write a Julia package to import MATPOWER’s built-in network data into a Julia environment.

It is possible to run MATLAB code using Julia, but not to translate arbitrary MATLAB code. Rather than working with MATPOWER directly, I used PYPOWER, a Python port under the GPL v3 license. I cloned the PYPOWER repository:

git clone --recursive https://github.com/rwl/PYPOWER
cd PYPOWER

Next, I copied all .py files containing network data into a new directory. These files begin with “case”, so it’s easy to glob them together:

mkdir case-only
cp pypower/case*.py case-only

Here are the contents of case-only:

case118.py
case14.py
case24_ieee_rts.py
case300.py
case30pwl.py
case30.py
case30Q.py
case39.py
case4gs.py
case57.py
case6ww.py
case9.py
case9Q.py
caseformat.py

Each of these Python files contains a single function with a predictable name – piece of cake. After a bit of experimentation, I came up with the following Julia module, which uses PyCall.jl to handle the Python:

module Power
using PyCall

export loadcase

function loadcase(caseName::ASCIIString)
    """
    Return a Dict containing power system data
    in MATPOWER's format. Valid cases:
    case118
    case118
    case14
    case14
    case24_ieee_rts
    case300
    case30pwl
    case30
    case30Q
    case39
    case4gs
    case57
    case6ww
    case9
    case9
    case9Q
    caseformat
    """
    # Add .py directory to access file:
    unshift!(PyVector(pyimport("sys")["path"]), dirname(@__FILE__))

    # Load casefile data:
    pycall(pyimport(caseName)[caseName], Dict{ASCIIString,PyAny})

    # Remove .py directory to leave path as it was:
    shift!(PyVector(pyimport("sys")["path"]));
end

end

A few notes:

• I am looking forward to the next Julia release, which will display the function docstring (the text between the triple double quotation marks) when the user asks for it.

• One gotcha: @pyimport defers to Python’s sys.path, which will not contain the folder case-only by default (see this Github issue). This is why I use the following lines of Julia code to prepend sys.path with the appropriate directory and remove it afterwards:

unshift!(PyVector(pyimport("sys")["path"]), dirname(@__FILE__))
shift!(PyVector(pyimport("sys")["path"]));

• There are two ways to use PyCall. Macros (like @pyimport) accept symbols as arguments, while functions (like pyimport()) accept strings. Macros are more convenient during interactive use, but it is more intuitive to pass a string to the final loadcase() function than to pass a symbol.

• The object returned by pycall(pyimport(caseName)[caseName]) is a Python dictionary with string keys and float values. To convert this object to its Julia analog, I passed the following type description to pycall(): Dict{ASCIIString,PyAny}. This converts the Python dictionary to a Julia Dict with string keys and inferred value types. (PyAny tells PyCall to infer the type, which I know will either be Float64 or Array{Float64}.)

• After some debugging, I pushed my Julia module to Github as an unregistered Julia package. Anyone can use Power.jl as follows:

Pkg.clone("https://github.com/kersulis/Power.jl.git")
using Power
case118 = loadcase("case118")

The finished product gets the job done, but has a significant drawback. It is unreasonable for such a simple data import package to require its users to have Python installed. In my next post, I develop a new package called “MatpowerCases” with no non-Julia dependencies.