OnlineLearningDRSupplement
Repository containing supplementary data and code for "Online Learning for Network Constrained Demand Response Pricing in Distribution Systems" by Robert Mieth and Yury Dvorkin.
Installation instructions:
The optimization model was implemented by using JuMP and auxiliary packages in the Julia programming language. Additionally, we used Mosek 8.1.0.63 in our numerical experiments. Mosek is a commercial solver which must be installed and licensed. The solver was chosen for its specific features for semi-definite programming. For more information on solvers, see the JuMP documentation.
The experiments require Julia 0.6 and the following Julia packages:
- JuMP0.18.3
- Mosek0.8.4
- Distributions0.15.0
- CSV0.2.5
- DataFrames0.11.7
You should force the use of particular versions of these Julia packages with
julia> Pkg.pin("JuMP", v"0.18.3")
julia> Pkg.pin("Mosek", v"0.8.4")
julia> Pkg.pin("Distributions", v"0.15.0")
julia> Pkg.pin("CSV", v"0.2.5")
julia> Pkg.pin("DataFrames", v"0.11.7")
Running the code:
The code for the experiments is contained in the run_dronlinelearning.jl file as well as the auxiliary files in the src directory.
The run_dronlinelearning.jl file is the main file that loads the data and performs the iterative algorithm based on the setting specified in a case file. The file input.jl defines necessary data types and functions to load an prepare the network data, model_definitions.jl contains the JuMP formulations, test_power_flow.jl contains the methods for feasibility checks and tools.jl contains some auxiliary functions.
You can run a testcase provided in cases/testcase.jl by executing:
julia run_dronlinelearning.jl cases/testcase.jl
If no specific case argument is provided, the script will run the testcase per default. The casefile is also a julia file and is heavily commented. It allows to easily parametrize the experiment. Note, that a single timestep may take up to several seconds in computation.