SDDP.jl Documentation

SDDP.jl is a package for solving large multistage convex stochastic optimization problems using stochastic dual dynamic programming. In this manual, we're going to assume a reasonable amount of background knowledge about stochastic optimization, the SDDP algorithm, Julia, and JuMP.

Note

If you don't have that background, you may want to brush up on some Readings. Part I of my thesis may also be useful as it contains a primer on how to formulate multistage stochastic optimization problems (Chapter One), as well as an introduction and literature review of the SDDP algorithm (Chapter Two).

Getting started

This package is unregistered so you will need to Pkg.clone it as follows:

Pkg.clone("https://github.com/odow/SDDP.jl.git")

If you want to use the parallel features of SDDP.jl, you should start Julia with some worker processes (julia -p N), or add by running julia> addprocs(N) in a running Julia session.

Once you've got SDDP.jl installed, you should read some tutorials, beginning with Tutorial One: first steps.

Citing SDDP.jl

If you use SDDP.jl, we ask that you please cite the following paper:

@article{dowson_sddp.jl,
	title = {{SDDP}.jl: a {Julia} package for stochastic dual dynamic programming},
	url = {http://www.optimization-online.org/DB_HTML/2017/12/6388.html},
	journal = {Optimization Online},
	author = {Dowson, Oscar and Kapelevich, Lea},
	year = {2017}
}

FAQ

Q. How do I make the constraint coefficients random?

A. Due to the design of JuMP, it's difficult to efficiently modify constraint coefficients. Therefore, you can only vary the right hand-side of a constraint using the @rhsnoise macro.

As a work around, we suggest you either reformulate the model so the uncertainty appears in the RHS, or model the uncertainty as a Markov process. Tutorial Four: Markovian policy graphs explains how to implement this. You might also want to take a look at the asset management example to see an example of this. Make sure you keep in mind that a new value function is built at each Markov state which increases the computation time and memory requirements.