JuDGE.jl

JuDGE stands for: Julia Decomposition for Generalized Expansion. Functionally, it is a solver which leverages the syntax of the JuMP modelling language to solve a particular class of capacity expansion problems.

For more details see our working paper: JuDGE.jl: a Julia package for optimizing capacity expansion.

Problem Class / Decomposition

JuDGE solves multi-stage stochastic integer programming problems using Dantzig-Wolfe decomposition. The user must specify a tree that represents the uncertainty of the problem, and at each node define a subproblem that can be a linear or integer program. Further, the expansion variables which link the subproblems must be declared.

JuDGE automatically generates a master problem and performs column generation to converge to an optimal solution.

Requirements

JuDGE requires Julia-1.3+, JuMP and appropriate optimiser(s). For academics, Gurobi / CPLEX provide free academic licenses, otherwise, you can use CBC/Clp or GLPK.

Installation

JuDGE is installed by the Pkg utility provided by Julia. In the Julia REPL, simply make the following function call.

] add "https://github.com/reganbaucke/JuDGE.jl"

Then, in your Julia script, use

using JuDGE

to import the functions from the JuDGE module into the current namespace.

Stochastic Knapsack Example

JuDGE is distributed with an example of a multi-stage stochastic integer programming problem. This is a stochastic knapsack problem with investment. The file knapsack.jl contains the implementation of this problem within the JuDGE framework. There are also a number of variants of this model demonstrating other features of JuDGE.

See the Tutorials for details about this.

Limitations

• The package is currently under development, and while it has been tested on a

number of example problems, there may be some issues that have yet to be addressed.

• Support for decisions to reduce capacity is not fully tested.

Bugs

Please raise an issue if you experience an error while using JuDGE.